MD+DI Online is part of the Informa Markets Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.


Articles from 2002 In February

Parkinson's Brain Implant Approved

Originally Published MDDI February 2002


James Dickinson

Activa Parkinson's Control Therapy uses two surgically implanted medical devices, which are similar to pacemakers.

FDA in January approved expanded labeling for a Medtronic deep-brain stimulator, the Activa Parkinson's Control System originally approved in 1997. Based on new clinical studies, the new labeling provides for use of the device in both sides of the brain to help control symptoms of advanced Parkinson's disease that can't be adequately controlled with medication.

The 1997 approval was for use in one side of the brain to help control tremors on one side of the body.

The Activa system, FDA said, consists of electrodes that are implanted into the brain and connected by leads under the skin to a pulse generator implanted in the abdomen or chest. The pulse generator sends a constant stream of tiny electrical pulses to the brain, blocking tremors. When the device is implanted in both sides of the brain, two separate systems are used.

To turn the stimulator on and off, FDA said, the patient holds a magnet over the pulse generator. The generator must be replaced every three to five years, the life of the battery.

Some 160 patients with advanced Parkinson's disease were enrolled at 18 medical centers in the United States, Canada, Australia, and Europe. The device was implanted bilaterally in 134 patients. The implant procedures were done simultaneously or in stages. The patients were followed for approximately one year.

During the study, FDA said, patients were evaluated for a variety of Parkinson's symptoms using a test for total motor skills, including hand movement, leg agility, facial expression, rigidity, tremor, and speech. They were tested with the Activa system turned on, both with and without medication. Total motor skills improved in nearly half the patients taking medication and approximately 90% of the patients not taking medication.

According to FDA, nearly all of the 160 patients enrolled in the study experienced one or more adverse events. During the entire study, 7.5% had bleeding into the brain; 11% had device-related infection; 10% had paralysis, and 8% had weakness. Some 37% of the adverse events were related to the Activa system. Six percent of the device-related adverse events were serious and ongoing, including a worsening of motor impairment and other Parkinson's symptoms.

Copyright ©2002 Medical Device & Diagnostic Industry

FDA Seeks Comments on SUD Guidance Document

Originally Published MDDI February 2002


FDA is welcoming comments and suggestions concerning a guidance document the agency may issue on the labeling of reprocessed single-use devices (SUDs).

FDA has not yet determined whether or not it will actually publish this guidance document, however. The notice was published in the December 20th, 2001, edition of the Federal Register.

The agency is soliciting comments that specifically address whether the name of the original manufacturer of the device should be removed from a reprocessed SUD and, in addition, whether the reprocessor's name should be placed on the device.

The deadline for written or electronically submitted comments is March 20, 2002.

Copyright ©2002 Medical Device & Diagnostic Industry

Bluetooth: The Future of Wireless Medical Technology?

Originally Published MDDI February 2002


A new technology that borrows from telemetry, IrDA, and 802.11 is set to continue the wireless trend.

William E. Saltzstein

Medical device designers now have several wireless options for their products. No single wireless technology meets all design goals and addresses all the issues presented to designers. For every design project that includes wireless technology, decisions must be made to identify most the appropriate technologies for that application and device-use model. For many applications, Bluetooth wireless technology will be the most appropriate choice.

Bluetooth, while it is certainly not the universal solution to all wireless needs, addresses many of the performance requirements specific to medical applications, and it is a particularly good fit in use models demanding high mobility, long battery life, and no infrastructure support. Bluetooth does more than simply eliminate cables; it provides access to a wide range of standard devices and communications options, including the formation of small networks. Bluetooth enables additional communications links by providing access to wide-area networking through cellular phone data communications as well as traditional Ethernet local-area networks.

The Industrial, Scientific, and Medical (ISM) radio-frequency (RF) band within which Bluetooth operates is shared by several wireless technologies. Medical devices must undergo immunity testing for compatibility with other ISM devices. Because Bluetooth uses the lowest transmit power of the wireless technologies, it has the lowest probability of interference. Many companies have examined coexistence issues among ISM technologies and concluded that there is little cause for concern in actual use.


Bluetooth is the most recent in a string of available wireless technologies. IrDA (overseen by the Infrared Device Association), telemetry, 802.11, and home RF all preceded it. Each technology has its own particular advantages and disadvantages, and Bluetooth borrows from each of them.

Bluetooth originated as a cable replacement technology primarily targeted at personal, portable computing and communications equipment, such as personal digital assistants, cell phones, and laptop computers. The technology makes possible not only replacement of simple point-to-point cabling (with minimal impact in cost and power), but also the quick and simple creation of small ad hoc networks of devices, or piconets.

The Bluetooth specification was created jointly by Ericsson, IBM, Intel, Nokia, and Toshiba and was named after the 10th- century Danish Viking king Harald "Bluetooth" Blatand, who united the warring factions of Denmark and Norway. A primary goal of the Bluetooth specification is to create a truly international standard to be implemented identically worldwide. So far, this goal is being met.

Today, Bluetooth technology has a following of more than 2500 companies. The initial founders have been joined by such companies as 3COM, Lucent, Microsoft, and Motorola. Products incorporating Bluetooth wireless technology are now shipping from many companies. Last year, the Institute of Electrical and Electronics Engineers (IEEE) adopted Bluetooth as the basis for its personal-area network (PAN) standard 802.15.


As specified by Bluetooth, the basic radio to be used transmits 10 m in open air (although current implementations operate well at significantly better range), with an optional power level to allow 100-m operation. With a total bandwidth of 1 Mb/sec, the radio is designed for moderate speed and a theoretical 720 Kb/sec payload, dividing up the bandwidth among the devices using data and voice channels. The technology supports eight devices on a piconet.

The Bluetooth radio was designed for immunity to noise and for ease of implementation on silicon, with major baseband portions implemented in either hardware or firmware to optimize cost, power, and size. Implementations that incorporate the entire radio-frequency and baseband processing section on a single chip—and have few external components—are already shipping.

To achieve robust connections, Bluetooth employs three key techniques: frequency hopping, adaptive power control, and the transmission of short data packets. The Bluetooth protocol (for data) automatically retransmits corrupt data packets that have—because the pseudorandom hopping sequence is designed to maximize frequency spacing between sequential channel hops—most likely hopped away from the interfering source.

Like other ISM radio technologies, Bluetooth operates in the 2.4–2.485-GHz radio band. Its radio meets the power and spectral emissions specifications defined by ETSI ETS 300–328 in Europe and FCC under CFR 47 Part 15 in the United States using the following set of parameters:

  • Frequency hopping, spread spectrum 79 channels, 1600 hops per second.
  • Gaussian frequency shift keying modulation with a 1-m symbol-per-second rate or 1 Mb/sec data rate.
  • 83.5 MHz of spectrum divided into 1-MHz channels.
  • Symbol timing accuracy ±20 ppm timing (when synchronized).
  • Power control based on received signal strength intensity feedback from the receiving device (Class 1 requirement).
  • 0 dBm (1 mW) without power control (Class 3, 10-m range).
  • 20 dBm (100-mW) with power control (Class 1, 100-m range).
  • Bit error rate of 0.1%, receiver sensitivity of –70 dBm.


Figure 1. Examples of Bluetooth interfaces.
Interfaces occur in Bluetooth at several different levels (see Figure 1). Blue RF is an interface designed to allow reuse and interoperability of the Bluetooth radios. Although not formally part of the Bluetooth standard, it has been adopted by many of the silicon suppliers as the digital interface to the radio subsystem.

Baseband processing is responsible for channel coding, decoding, and low-level timing control and management of the link for single data-packet transfers. It decodes and encodes packets of data and provides the "care and feeding" of the RF transceiver sections.

The link manager is responsible for managing the link control states used to establish and maintain the connection between the master device and its slaves, and it includes:

  • Inquiry/inquiry scan (used to find other devices; the baseband part of service discovery).
  • Page/page scan (used to establish connections to a chosen device).
  • Connections: the active states and the power-saving states of hold, sniff, and park.

The host control interface (HCI) is a hardware and software interface specified to allow modular implementations of the lower-level hardware and baseband. Architecturally, HCI allows the burden of the protocol stack to be borne by the host device or computer to lower total implementation cost. Its hardware interface is specified as either USB or UART. (Using USB usually achieves the highest data rates.)

The low-level protocol is responsible for packetization, multiplexing, and demultiplexing packets for the higher-level protocols, and it maintains order in the piconet; this interface is essentially Bluetooth's "traffic cop."

Upper-Level Protocols. Upper-level protocols such as RFCOMM and service discovery protocol (SDP) provide high-level functionality upon which profiles are built. RFCOMM emulates full "hand-shaking" serial ports and is used in profiles that support basic connections between pairs of devices. SDP is described later in this article.

Transmission Channels. Two types of transmission channels are defined in Bluetooth: asynchronous communications link (ACL) and synchronous connection oriented (SCO). ACL channels are used for data communications and are set up between every two Bluetooth devices for use in connection management. ACL is a packet-switched transmission method that provides error detection, forward error correction, packet tracking (numbering), and packet retransmission. It provides entirely error-free links to the application; they will either give good data or keep trying until signaling that the link is broken. (The retransmission of data in the presence of interference increases latency and slows down net data rates.)

Used for the transmission of voice data, SCO is a circuit-switched transmission method that provides known latency and assured transmission rates, and it features error detection and forward error correction. Each SCO channel is given time slots that are predetermined in the transmission sequence, with a maximum of three SCO channels permitted in each piconet. Little bandwidth is left for ACL data when all three SCO channels are being used.

SCO and ACL channels allow selection of both packet length and the amount of forward error correction. These parameters are often automatically controlled and depend on the amount of interference immunity and data throughput needs.

Piconets and Scatternets. A Bluetooth piconet consists of at least a master and a slave; this pairing is defined as a point-to-point connection. A full piconet consists of one master and up to seven slaves. The master controls all timing including the clock and hopping sequence (to which all slaves synchronize). Each master will have slightly different clock (skew) and hopping sequences, which are based on the master's device address (a 48-bit IEEE address). These differences allow for multiple piconets to be established and used in the same physical space. A Bluetooth master is responsible for controlling all data traffic in a piconet. All data transmission goes through the master in a star network topology.

The Bluetooth specification defines the ability to exchange the master and slave relationship between two devices: a desirable feature for implementation of LAN access points. It also allows a device to be both a master on one piconet and a slave on another, or to be a slave on more than one piconet if the hardware and baseband implementations support it.

A scatternet is formed from two or more piconets that share a common member. This shared member may either be a slave on both piconets, or a master on one and a slave on another. Each of these configurations has architectural trade-offs, and the current Bluetooth specification (1.1) does not define the preferred one, nor does it completely define the operation of scatternets.

Profiles. Collections of features and functions required to perform a particular operation are called profiles. The basic, required profile gives all devices some level of interoperability—the ability to function automatically with other devices with little user intervention. This collection of functions is called the Generic Access Profile.

Additional, optional profiles depend upon the application requirements and implementation details. Below is a list of the profiles currently specified. Under development or in the release process are profiles for printing, hands-free, and audiovisual applications; human interfaces; and others. The Version 1.1 profiles are

  • Generic Access—implemented in all devices.
  • Service Discovery—allows devices to determine capabilities of other devices.
  • Cordless Telephony—covers functions for cordless telephone handsets, both audio and dialing.
  • Intercom—enables voice connections and calling between two devices.
  • Serial Port—comprises the functions and methods required for establishing a virtual serial connection between two devices; it is used in many of the higher-level profiles.
  • Headset—sets forth the functions needed to implement a hands-free headset for cell phones and computers.
  • Dial-Up Networking—covers the functions and methods required to establish a remote Internet connection.
  • FAX—specifies wireless send and receive signals for faxing.
  • LAN Access—enables Bluetooth to be used as the transport to a standard local-area network (LAN).
  • Generic Object Exchange (OBEX)—specifies how to transmit high-level objects, such as files. The basis for the following three profiles, OBEX was developed for IrDA and allows Bluetooth to be used by software applications developed for IrDA.
  • Object Push (OBEX based)—transmits named objects containing data.
  • File Transfer (OBEX based)—transfers files between devices.
  • Synchronization (OBEX based)—synchronizes applications on computers, personal digital assistants, and cell phones.


Bluetooth has been designed to facilitate setup of small groups of devices. Two key features, service discovery and ad hoc network support, also play major roles in the design, and security features protect the privacy of communications.

Service Discovery Protocol. Service discovery protocol (SDP) features allow for automatic recognition and configuration between two devices of different types and from different manufacturers.

There are two types of discovery within the Bluetooth specification. Device discovery allows one device to query devices within range and acquire, in turn, key information about their general capabilities. This key information includes full address, human-readable name, and general device type (cell phone, laptop personal computer, headset).

Service discovery enables a device to learn the details of supported profiles and to actually browse those profiles to find out how to access certain features. The service discovery concept allows for even more information and access methods to be exchanged.

Ad Hoc Networks. Ad hoc networking is the capability to quickly establish and dissolve small groups of devices with very little user involvement and no permanent address assignment. Several devices establish one network and retain the relationship only for the desired time of interaction. If security is desired, users can type in passwords or personal identification numbers (PINs) for bonding and encryption.

Security. Bluetooth supports several security features, depending on the application and user requirements. These features range from the protection against eavesdropping inherent to the frequency-hopping spread-spectrum technology, to the use of keys or PIN and password combinations. With the use of PINs (alphanumeric strings of up to 16 characters), the 128-bit SAFER+ encryption algorithm is used to create very strong security and encryption between devices. More security can be added at the application level if desired. This implementation should meet the requirements of systems that must comply with the Health Insurance Portability and Accountability Act of 1996 privacy standards.


Bluetooth technology opens up many possibilities for medical devices to communicate and facilitate healthcare providers' access to essential information. Several studies are being conducted, many of them in Europe, to evaluate the technology as a method for synchronizing personal devices and transferring data in hospitals. At least one medical device has been listed on the Bluetooth Web site as a Bluetooth-approved product. Other devices developed by two companies have been demonstrated, and are serving as feasibility prototypes for Bluetooth implementation. These include a patient-worn pulse oximeter and a portable patient monitor.

Medical Information. Use of handheld and mobile computing technology in the medical field has progressed steadily. Doctors routinely access medical databases for drug dosages and interactions. Home-health nurses keep information on their patients for field use. Many other applications are allowing practitioners to make better-informed decisions.

Bluetooth wireless technology will enable medical devices to connect more easily with each other and to information access points provided by LANs. Bluetooth's low power, low cost, and few configuration requirements will be traded off against 802.11b's higher speed and easy enterprise integration. The use model and environment for each individual application will determine the best choice for wireless data technology.

Medical Devices. Other exciting possibilities exist in the field of medical device communications and networking. Bluetooth clearly provides an advantage for highly mobile battery-powered devices, and it opens up the possibility for these devices to connect to a LAN or wide-area network. Input, output, and storage can be removed from individual devices and centralized to increase flexibility and reduce the devices' cost, size, and power. Devices also can talk to each other within a patient-area network, which will make them less prone to medical errors.

The challenges to integrating Bluetooth in medical devices are discussed in greater detail below, but they are the same as for any RF technology used in medicine. Bluetooth has the advantage of being an adopted and controlled industry standard with an extensive qualification, verification, and approval process to ensure compliance with that specification.

Implementation. When a device manufacturer decides to incorporate Bluetooth into its medical product, it must proceed thoughtfully. Perhaps the easiest way to get started is by using an external adapter to prototype the device design, gain a greater understanding of issues associated with the implementation, and collect valuable user and marketing feedback. This approach could also shorten time to market.

When the prototype is up and running and a manufacturer is able to navigate the make-versus-buy decisions, it will need to consider several issues as it formulates its product and project plans.

The $5 Solution. There has been much discussion in the media about the so-called $5 Bluetooth solution, and some companies claim to have achieved this dollar amount per unit for high-volume shipments in 2002.

The industry goal of $5 includes only the RF and baseband signal processing functions. To implement the complete Bluetooth functionality from the baseboard layers to profiles and application requires external components (at least an antenna or board space for an onboard implementation, and sometimes other discrete components), as well as RAM, ROM, and processor bandwidth. In addition, all of the current Bluetooth software stacks that implement more than the headset profile require an operating system.

The $5 goal assumes that a device, such as a cell phone, already has enough of the above resources and an operating system. It assumes a royalty-free implementation of the Bluetooth software stack, which is not the case unless a device manufacturer implements the rather complex specification. The goal also assumes production volumes in the millions of units per year.

Unfortunately, very few of the above conditions are met for medical devices, so implementations in more-realistic lower volumes are unlikely to meet that cost goal. Realistic estimates for medical device applications for the next year range from $25 to $70 and would require significant development resources. Alternatives requiring little or no development cost more but yield quick time to market and may be the best alternative for low-volume production runs, or until customer acceptance justifies highly integrated solutions.


Most medical device manufacturers have extensive experience, usually negative, with RF. Traditionally, a great amount of both time and money are spent trying to reduce RF emissions and susceptibility. Bluetooth technology transmits and receives RF signals, which require quite a different skill set.


Fortunately, many excellent designs are available for Bluetooth that include all of the RF components, thus eliminating this part of the design task. The best Bluetooth RF transceiver will function poorly, however, if the design and location of the antenna are neglected. The antenna needs to be located in or on the device to limit RF shadowing outside shielded cases and components. It also needs to be designed with the appropriate gain and radiation pattern for a particular application and its placement in the device. Standard 2.4-GHz antennas are available from several suppliers and offer many options for placement, pattern, gain, and physical mounting. Board-mount versions are available if the enclosure is radio-transparent and the radiation pattern is appropriate for the application. It is important to note that plastic material selection can make a difference for embedded antennas since these materials vary in their transparency at 2.4 GHz.

Software Support for Host Stack. Many medical device manufacturers will not need or be able to afford a highly integrated solution that eliminates the HCI and merges the baseband and host processing for their devices. In most situations, designers will use a module that incorporates the baseband and RF, as well as a Bluetooth software stack in their host software. Depending on the type and number of profiles supported (i.e., the amount of functionality required), a medical device's Bluetooth software stack will require:

  • 32–64 KB of code space.
  • 16–32 KB of RAM.
  • Approximately 5 millions of instructions per second of processor bandwidth.

(These approximations represent an average of several suppliers'specifications.)

RF Issues. The three RF questions most commonly asked about Bluetooth and medical devices are: How will it affect the human body? How well will it work with other medical devices and products used in the home and hospital? How well will it work with other technologies on the ISM band?

Human body exposure. Several agencies regulate RF exposure levels, and papers have been written on the topic of RF's effects on the human body. Many of the studies that have been done for 802.11b are also applicable to Bluetooth.

Testing done to date shows no harm from RF levels used in 802.11b and Bluetooth. Most body-worn or portable devices will likely be designed as Class 3 Bluetooth devices (1 mW), which would have significantly less power than most of the transmitters tested (at 100 mW) and are even less likely to have any deleterious effects.

Medical device interference and sensitivity. The ISM band is significantly outside the frequency range of naturally occurring biological signals; therefore, the potential risk of interference to that signal is low and would occur from effects on the components in a particular device design. Because Bluetooth is a very low power transmitter, it is unlikely to interfere. Most medical devices, including implantable devices, must be tested for susceptibility in the ISM band.

Microwave ovens and some other devices that may be found in medical settings use the ISM frequency band. Microwave ovens have been found to pose no interference problems, and other devices using ISM-frequency radiation are well shielded to prevent interference with other devices.

Coexistence in the ISM band. Medical device companies and regulatory agencies, because of their experiences with hospital telemetry systems, have a heightened sensitivity to the issues of interference and coexistence. Their concerns are valid because the hospitals they supply might already have installed wireless LANs, and multiple ISM radios might exist in other environments as well.

Testing performed by several companies has shown that although performance degradation occurs as a result of interference, the real-world effects of interference are manageable by current applications. More work is necessary by the wireless industry to eliminate this concern, but so far interference does not seem to be a significant issue in most models.

For the future, the technologies specification groups (the Bluetooth SIG and the IEEE 802.15) are working toward interoperability solutions that would eliminate the RF issues entirely.


Three important regulatory approvals for wireless medical devices in the United States are those of Bluetooth, FDA, and FCC. Bluetooth approval is new for medical device manufacturers, and devices without wireless components have historically been exempt from FCC approvals in most cases. The FDA approval process is challenging, especially when new technologies are being introduced.

Bluetooth Qualification. The Bluetooth qualification process is well-defined in SIG documentation. Bluetooth technology maintains a strict enforcement policy. For companies to gain free license to the intellectual property that the SIG members have contributed, the products must pass the appropriate qualification process before they are sold (see sidebar). Once qualified, the product may then display the Bluetooth brand, logos, and labeling.

The qualification process is meant to guarantee customers and users a high level of interoperability between devices. It is taken quite seriously and appears to be effective, to judge from a sampling of products that meet the current Version 1.1 specification. (Earlier versions had interoperability problems, which were addressed in 1.1; manufacturers should be wary of products qualified to earlier specification revisions.)

The Bluetooth qualification program allows components to be qualified to various levels. Hardware and software components can be qualified so that a device manufacturer can integrate them into a device without having to repeat the portion of the testing the component has passed. This means that if a device manufacturer integrates an entire qualified OEM module into a device, that manufacturer need not do any additional testing to be qualified for Bluetooth. Other combinations of qualified components have different sets of implications, so it is important to understand what qualification work will be required when evaluating whether to make or buy.

As of this writing, there are five recognized Bluetooth qualification bodies in the United States and 26 worldwide responsible for reviewing all the test results and documentation. In addition, there is one recognized Bluetooth qualification test facility in the United States and four worldwide that can do complete testing and submission. Many compliance testing houses have or will have the capability to perform testing and create packages for submission to these bodies in the future.

FDA. Depending on the implementation and use of Bluetooth, there are several approaches to gaining approval from FDA and international organizations. Readers should note that these are potential paths only and must be reviewed and approved by individual manufacturers' regulatory affairs staff.

If Bluetooth wireless technology is added to a device as an external adapter with no modifications, such as replacement of an existing RS-232 cable, and the device is approved for data communications capabilities through that port to external equipment, it may be possible to use the regulatory approval approach taken by many device companies with external modems. If there is no change in labeling or intended use it may be reasonable simply to use what is called a "letter to file." The Bluetooth adapter may be considered a wireless modem in this case, assuming that all other appropriate industry approvals and qualifications have been completed.

The regulatory agency should be informed that Bluetooth wireless technology is part of a detailed industry standard with tightly enforced qualification and compliance testing that is required for listing of a product. Documentation from the Bluetooth qualification test facility and Bluetooth qualification body provide evidence of this compliance verification. The published specification and test methods documents are tightly controlled by Bluetooth SIG Inc., a U.S. corporation established for that purpose.

If Bluetooth is incorporated internally within a device, submission of a 510(k) application to FDA is required, especially because software and hardware modifications are likely to be involved. It is reasonable to assume that such 510(k) submissions have been made or will be made in the near future. The key, as with all 510(k) submissions, is to identify predicate devices and technologies to use in the submission.

Wireless and telemetry devices have been shipping for decades in the United States and worldwide. Devices using wireless technologies are currently being marketed on the wireless medical telemetry system band in the United States that uses frequency-hopping spread-spectrum technologies as well as traditional frequency modulation technology. Potential predicate devices using 802.11 frequency hopping technology in the ISM band have been cleared by FDA and are currently being marketed and distributed.

Approved and well-tested components that have a good track record as external adaptors will be easier to analyze separately, and convincing regulatory agencies that they are safe and reliable will be less of a struggle. The external approach will also allow testing and gathering of clinical and customer usage data to be included with a submission.

FCC. FCC testing and approval are fairly standard prac- tice in the telecommunications industry, but FCC has not regulated medical devices to date. Once Bluetooth is added to a device, however, FCC testing becomes mandatory because the device becomes an intentional radiator. Bluetooth RF modules have been designed for FCC compliance, which should allow them to pass the appropriate testing if the modules are properly integrated into products.

A potential risk to FCC compliance for designs using pretested components comes from internal systems' emissions that are coupled into the Bluetooth RF and are reradiated by the Bluetooth transmitter and antenna. FCC testing can be performed by the same test houses currently used for IEC emissions and susceptibility testing for medical devices.


Bluetooth wireless technology poses new questions about which parts of the technology should be purchased and which parts should be designed in-house. Companies should carefully weigh the Bluetooth requirements against the system components in terms of resources, including both hardware and software. Bluetooth qualification and FCC regulation must be considered in the project plan. The level of in-house expertise should be considered as well.

It is easy to add up individual components on a parts-cost basis and arrive at a low-price solution, but employing such a method can produce a disastrous outcome when the real costs of engineering and qualification are fully realized. Purchasing a higher-priced, highly integrated solution might save substantial nonrecurring-expense costs and reduce time to market, which offsets the higher price for an entire product generation. Manufacturers should be careful to understand the real development costs and schedule, and how the two can affect revenue calculations.

It is clear that medical devices and medical information management products will benefit from the advantages provided by Bluetooth wireless technology. And although Bluetooth certainly presents new challenges to the medical device industry, the benefits stand to be quite significant in both current and future use models for medical products.


Bray, Jennifer, and Sturman, Charles F. Bluetooth Connect without Cables, 2nd ed. Englewood Cliffs, NJ: Prentice-Hall PTR, 2002.

William E. Saltzstein is president and founder of Code Blue Communications, a company based in Redmond, WA, that has been developing Bluetooth technology.

Copyright ©2002 Medical Device & Diagnostic Industry

Radiofrequency ID Company Eyes Medical Error Market

Originally Published MDDI February 2002


Erik Swain

Although its magnitude is disputed, no one denies that the United States healthcare system faces a serious problem with medical errors. Estimates of the cost due to medication errors alone range from $76 billion to more than $177 billion a year. Among the companies who see this problem as a ripe market for their technologies is Avante International, which specializes in radiofrequency identification.

According to company president Kevin Chung, Avante had not considered entering the healthcare market before seeing the astronomical sums lost on medical errors. The company then used its technology to develop a patient-tracking system called Positive Patient Medication Matching. The system shares what Chung calls the "inherent advantage" of radiofrequency identification (RFID) over bar codes, which is that a direct line of sight is not required when scanning.

A traditional drawback for RFID, however, has been its substantial surface-area requirements. For this reason, most applications of RFID in healthcare have been on transport packaging rather than primary packaging. What sets Avante's RFID system apart from others, says Chung, is that it will work on very small vials and syringes without any risk to the code's integrity.

Another drawback of this technology has been security. RFID is a writeable code and can be changed after it leaves the factory or warehouse. In response, Avante created a "relational check code" (RCC) method. Like all RFID, Avante's has a permanent code that is burned in with a specific identification along with a writeable portion. RCC adds a third code that combines the other two using encryption. This third code cannot be copied.

The system runs on a universal platform with open architecture, which Chung hopes will meet the global needs of drug and device companies. It uses an ISO-standard frequency of 13.56 MHz, and will be able to interact with existing drug-interaction and other databases.

The system's antenna, in the form of a small shelf, is designed to be placed next to the patient in a hospital setting. The caregiver puts the patient's tag, the prescription's tag, and the tagged medications on the shelf. The system will then signal whether this is the right medication for the right patient.

Avante's first push will be for military and VA hospitals. Applications for the civilian sector may still be a "few years off," according to Chung.

On December 3, 2001, FDA published a proposed regulation that would require bar codes on all drug and biological products to reduce the number of medication errors. Avante believes that the proposal, despite its specification of "bar codes," leaves room for the RFID system.

Copyright ©2002 Medical Device & Diagnostic Industry

Calculating Equivalent Time for Use in Determining the Lethality of EtO Sterilization Processes

Since its introduction more than 40 years ago, EtO processing has proven very effective for the sterilization of medical products.1 Approximately 45% of the medical devices handled by contract sterilizers in the United States today are treated using an EtO process, and it remains the method of choice for products constructed from materials that are incompatible with high temperatures or radiation.

More complex than either thermal or radiation processing, EtO sterilization does present challenges to its users. Not only does it require extensive postprocess testing of products for residuals, its effects are difficult to express mathematically because the rate of microbial lethality is affected by three primary process parameters: EtO concentration, temperature, and relative humidity (RH). Methods of integrating the effects of these factors were lacking until recently, when one approach was published in this magazine (Alfredo Rodriguez et al., September 2001 MD&DI, p. 100).2 This article presents another approach, developed over the last year by researchers at Biotest Laboratories Inc. (Minneapolis), SGM Biotech Inc. (Bozeman, MT), and STS Inc. (Rush, NY). Direct and comprehensive, this work led to the achievement of simple yet broadly applicable integration methods for use in process situations.

Although many factors complicate the EtO sterilization process, including EtO absorption, product material effects, the existence of microenvironments in the process chamber, and the achievement and maintenance of steady-state conditions, this article focuses only on the three primary parameters.

The effects of these factors were studied at the researchers' laboratories using biological indicator evaluator resistometer (BIER) vessels. Designed for the calibration of biological indicators (BIs), these vessels operate within the narrow control windows defined in the domestic and international standards on EtO sterilization.3,4 The interrelated problems of calculating integrated lethality and comparing dissimilar process variables were addressed by applying known principles that have been derived from other sterilization processes.

It is hoped that the use of the equations presented here, in combination with such new technologies as the direct chamber infrared (IR) method for measuring EtO concentrations, will enable more precise integration of the complete EtO sterilization cycle than has been possible.5 The goal in developing this mathematical model for sterilization lethality was threefold:

  1. To enable EtO processors to predict the accurate D-values and related process times necessary to achieve a sterility assurance level (SAL) of ¾ 10–6.
  2. To provide the tools necessary to optimize process cycles by reducing cycle times and/or EtO concentrations, thereby minimizing EtO residual levels and outgassing quarantine times and maximizing process throughput.
  3. To provide a method for quantifying integrated lethality, which will allow the industry to move toward parametric release.


It has been demonstrated that EtO concentration and temperature are independent variables and that the rate of lethality increases as either is increased. Although some researchers have reported a temperature-dependent plateau effect for EtO concentration, others have not confirmed the phenomenon.6,7 In the work reported here, data at 54°C over tested concentrations between 300 and 750 mg/L did not support the magnitude of the reported temperature-dependent plateau. In the temperature range of 27°–60°C (80°–140°F), low and high RH levels are known to reduce the effectiveness of EtO processes. The generally accepted RH control limits are ~30 and ~90%. Within that range, humidity can be considered a constant; below ~30% RH, sterilization efficacy decreases markedly, while above ~90% the potential for H2O ondensation increases. Tests conducted during this study used an RH of 60%.

The following classically accepted formulas for the application of resistance studies to process calculations are the basis for the mathematical model presented here:


where Uf is equivalent exposure time, N0 is the initial spore population, and Nf is the final spore population.


where U full process is the equivalent time for the full sterilization process.

SAL = 10 log N0–SLR      (3)


Figure 1. The relationship between D-value, SLR, and SAL. The lethality delivered in any sterilization cycle, SLR is calculated as the log of the initial population (N0) minus the log of the final population (NF). The SAL, the probability of a viable microorganism being present on a product unit after sterilization, is calculated using Equation 3.

An annex to ISO standard 11135 identifies a method for calculating process D-values, which represent the dose or time at steady state required to reduce a microbial population by 90% or 1 log10.8,9 Unfortunately, the document provides little guidance to assist users in estimating the equivalent time (U) required for such calculations. In the extreme, use of the actual exposure time (which begins after steady-state pressure has been achieved) rather than equivalent time may lead to a gross underestimation of a process's D-value and concomitant overestimation of the SLR and underestimation of the SAL. The method of D-value calculation is irrelevant to this problem. Whenever equivalent time is underestimated for D-value calculations, the result will be the same. Figure 1 demonstrates the relationship between D-value, SLR, and SAL at steady state when microbial inactivation follows a straight-line log-linear relationship.

EtO process D-value calculations have been used primarily in BIER vessel studies, where the time to steady state approaches zero and the equivalent exposure time approaches the actual exposure time. However, applying any D-value calculation method to EtO systems used for actual product sterilization is inappropriate because standard-sized process chambers do not produce square wave cycles and substantial lethality is generated during both their charge (gas injection) phase and gas evacuation phase (which do not fall within the exposure time). For steam or dry-heat sterilization systems, integrated lethality or lag correction factors can be applied.9 However, for EtO systems, no such standard methodology exists. This situation accounts for the popularity of the AAMI overkill validation technique and the equivalent ISO half-cycle method, neither of which require calculations of D-valve, SLR, or SAL.8

If the actual exposure time is used in Equation 1 rather than equivalent exposure time, then as the exposure time approaches zero when log N0 – log Nf is some positive number, then the D-value also approaches zero; subsequently, SLR approaches infinity and SAL approaches
10¥. While no one would suggest that a D-value would equal zero, this extreme example demonstrates the dangers of underestimating the equivalent time. Adding some arbitrary number to increase U does not allow the individual responsible for process validation to know whether U and, therefore, the D-value, are still being underestimated, and if so to what extent. Overestimating the equivalent time for the full process will similarly result in an overestimation of SLR and an underestimation of SAL, although the percentage error will generally be less than when U is underestimated for D-value calculations, because the latter error is multiplicative.

Figure 2. The EtO sterilization cycle.

EtO Lethality. Table I presents results from sterilization validations conducted at Biotest for a variety of medical products. These validations used an exposure time of zero minutes, yet resulted in few or no positive BIs, which is not surprising if one understands the concept of accumulated lethality. The table also includes estimated equivalent times for these zero-minute exposures, related D-values, and full-process-cycle SALs. The EtO sterilization cycle being validated in most of this testing is depicted in Figure 2. For this process, the lethality attributed to EtO begins with the injection of the gas into the process chamber. Whether pure EtO is used, as in the process shown, or a gas mixture (such as CO2/EtO or EtO and an HCFC diluent), lethality increases as the concentration increases, and the concentration increase is proportional to the pressure rise in the chamber.11 For processes with well-controlled pressure ramp-up rates, EtO concentration changes also are proportional to time during gas injection and evacuation. A cycle's exposure-time phase starts when the control pressure has been achieved, which occurs after gas injection is completed. It should be noted that, in practice, absorption, microenvironments, diffusion, and chemical reactions that consume the gas can slow the development of steady-state EtO concentrations in some process chamber locations or product areas.

Product Type Positive BIs/Total BIs Calculated U (min) Calculated D-Value Calculated Full-Cycle SAL Full Cycle ProcessExposure Time (hr)
Introducer, delivery forceps, catheter 1/20 Tha
1/20 SCb
24.65 3.34 Th
3.38 SC
1 x 10­66 Th
1 x 10­66 SC
Occluder delivery system 6/20 Th
10/20 SC
24.15 3.75 Th
3.92 SC
<1 10­87 Th
<1 x 10­72 SC
Tubing sets and scopes 5/20 Th 24.15 3.69 <1 x 10­72 4
Cannula 2/20 Stripsc 36.9 5.29 1 x 10­74 4
Catheters, introducers 22/44 Th 47.75 7.56 1 x 10­32 2.5
Rotor blade 17/20 Mpsd 24.95 4.16 <1 x 10­51 4
Suture anchor 17/20 Mps 25.25 4.00 1 x 10­54 4
Compass tips and magnets 15/20 Th 25.7 4.07 1 x 10­52 4
Clamp covers, loops, brush, boots 3/20 SC 24.65 3.63 1 x 10­60 4
Optical fiber 0/20 SC 24.15 < 3.31 <1 x 10­74 4
Sensor, probe, wire, etc. 0/20 SC 24.65 < 3.38 <1 x 10­65 4
Orthopedic implant product line including bone- harvesting device 0/80 Strips
0/80 SC
5/40 IPe
44.9 < 6.18 IP <1 x 10­33 IP 4
Unassembled bone-harvesting device
0/20 strips; 0/20 SC
< 6.19 Strips
<1 x 10­32 Strips
Injectable polymer system
1/19 strips
1 x 10­52
a Th = 1.5-in. single-strand cotton thread inoculated with >1x106 B. subtilis (SGM Biotech).
b SC = Self-contained test, >1x106 B. subtilis (SGM Biotech).
c Strips = Paper strips, >1x106 B. subtilis (SGM Biotech).
d Mps = Mini paper-strip BIs, 2x10mm, >1x106 B.subtilis (NamSA).
e IP = Inoculated product from a spore suspension.
Table I. Zero-minute exposure data and calculations for equivalent time (U), D-value, and SAL for validation testing of various medical products. SAL was calculated by the Stumbo, Murphy, and Cochran method.10

For the cycle in Figure 2, the EtO gas injection time is 11 minutes and the exhaust time is 16 minutes, which are common times in EtO processing. An 11-minute nitrogen (N2) overlay immediately follows the EtO injection phase; hence EtO concentration is at its maximum during that period. Data such as these can be converted to equivalent time for D-value, SLR, and SAL calculations using the mathematical model described below. The technique is based on lethality rate (LR), which can be expressed either as a rate function with units of Dlog N per minute at specified conditions or as the reciprocal of the D-value.

Figure 3. Log/linear plot of D-values versus EtO concentration.

Parameters affecting EtO sterilization have been studied extensively and numerous investigators have shown that microbial D-values decline as EtO concentration increases.1,7,12–16 These observations were verified by recent studies at SGM Biotech in which sets of BIs were tested at EtO concentrations of 300, 450, 600, and 750 mg/L. Three lots each of standard spore strips and the company's self-contained EZ Test were tested in an EtO BIER vessel from Joslyn Valve (Macedon, NY). Comparative D-values from the studies are listed in Table II. Test results also are graphed on a log10/linear plot in Figure 3, which indicates there were reasonable straight-line fits with R2 values of 0.9695 and 0.9909 for spore strips and the self-contained test, respectively. However, Figure 4, which is a linear/linear plot of both D-values and lethality versus EtO concentration for spore strips, depicts a more useful relationship.

Lot No. Log Spore Population BI Type Ethylene Oxide Concentration (mg/L)
300 450 600
G-92P 6.531 Self-contained test 5.8 4.2 3.6 2.8
G-103P 6.322 5.6 4.2 3.2 2.8
G-105 6.255 5.2 4.0 3.2 2.6
Average NA 5.5 4.1 3.3 2.7
BSUB-235 6.398 Paper strips 6.7 4.3 3.5 2.9
BSUB-244P 7.0 6.2 4.4 3.4 2.8
BSUB-249P 6.398 6.1 4.1 3.4 2.8
Average NA 6.3 4.3 3.4 2.8
Table II. Comparative D-values at four EtO concentrations calculated using the Spearman-Karber method.17 These test results are also shown graphically in Figures 3 and 4.
Figure 4. Linear/linear plot of D-value and lethality versus EtO concentration.

The Microsoft Excel program for best fit of data predicts that, when plotted against EtO concentration (C), the D-value predicts a parabolic curve. As C approaches zero, then D will approach infinity. Logically it follows that EtO-associated lethality (1/D) must approach zero as C approaches zero, creating an intersection on the lethality rate plot at x = 0, y = 0; where D = 1/LR, as LR approaches zero, then D approaches infinity, which also is predicted by the plot of D, which is asymptotic in both directions, or hyperbolic. Thus a linear/linear plot of the lethality rate allows a simple approach to calculating equivalent time if temperature is considered to be constant:

LR ~ C, or LR = kC      (4)

where k is the rate constant. The equation can also be expressed

and so on. Solving for LR,

Since the D-value is a reciprocal of the lethality rate, the equation can also be used to solve for D:


which simplifies to:

LR may also be used to derive Dlog N as a function of time (t) where:


log No – log Nf = kC Dt

To calculate accumulated lethality at a constant temperature (T1), each increment is multiplied by the time at that increment, which is expressed in the summation formula:


Figure 5. Cumulative lethality (SLR) of standard versus ramped BIER cycles, at constant temperature.

Testing the Models. As the next step in this research project, testable hypotheses based on the theoretical application of these concepts were developed. That is, that at a constant temperature, lethality can be calculated based on time-weighted EtO concentrations, and equivalent time at steady state can be approximated for those phases of sterilization where concentration changes are proportional to time by dividing real time by two. Testing was performed at STS using a BIER vessel that could be ramped up over 10 equal pressure steps of 60.6 mg/L. Adjustment time between steps was 5–6 seconds for a total injection time of 50–60 seconds, which is the same as that used for a standard cycle performed using the same BIER unit. The exhaust phases for the ramped and standard cycles were also equal. The hold time at each step for the ramped cycles was varied to provide a time-weighted average EtO concentration that was equal to that of a standard cycle. For example, to simulate a 30-minute standard cycle, a ramped cycle was performed using hold times of 6 minutes at each of the first 9 steps and 3 minutes at the last (606-mg/L) step. This provided a time-weighted average of 18,810 mgxmin/L the same as achieved with a standard 30-minute cycle in that BIER unit, with no adjustment for charge and discharge phases. This relationship is shown in Figure 5.

Five lots of BIs, representing three different manufacturers and three different BI types, were processed using comparable ramped and standard BIER cycles. Three of the lots were standard BI strips, each from a different manufacturer. These strips were all cultured in the same lot of Trypticase soy broth from BBL Inc. (Syracuse, NY). Media for the other two lots, which were self-contained EZ Test and Releasat BIs from SGM Biotech, are part of the BI systems. Standard conditions for the test cycles were a constant temperature of 54° ± 1°C, EtO concentration of 600 ± 30 mg/L, and RH of 60 ± 10%. The fraction-negative results of these tests are given in Table III, and the resulting D-value calculations are listed in Table IV. As the tables indicate, the results for the two methods of gas exposure were extremely close and certainly within the ±20% recommended by USP XXIV for verification of D-values.18 Furthermore, no strict bias was evident suggesting a random variation. These data support formulas 4 and 5.

BI Type Cycle No. Log No. EtO BIER Cycle
15 min 20 min 22.5 min 25 min 30 min
Std. Ramped Std. Ramped Std. Ramped Std. Ramped Std. Ramped
No. of positive BIs out of 10 total
EZ Test 1 6.380 10 10 9 4 10 6 0 0 0 0
2 00 00 10 8 10 9 0 0 0 0
Releasat 1 6.176 10 10 2 8 0 3 0 0 0 0
2 00 00 3 9 3 3 0 0 0 0
Strips 1 6.301 10 10 4 10 7 10 1 3 0 0
2 00 00 10 10 9 9 2 0 0 0
Strips 1 6.342 10 10 10 9 4 10 0 2 0 0
2 00 00 10 10 9 10 3 0 0 0
Strips 1 6.447 10 10 9 10 8 10 7 1 0 0
2 00 00 10 10 10 10 4 0 0 0
Table III. BI fraction/negative results from testing using comparable ramped and standard BIER vessel cycles.
BI Type Log No. EtO BIER Cycle
Std. Ramped Ramped, % Deviation from Std.
D-Value 0
EZ Test
Releasat 6.176 3.10 3.39 +9
Strips 6.301 3.48 3.69 +6
Strips 6.342 3.53 3.41 ­4
Strips 6.447 3.62 3.23 ­12
Average NA 3.42 3.35 ­3
Table IV. Calculated D-values for tests using comparable ramped and standard BIER vessel cycles. The Stumbo, Murphy, and Cochran calculation method was used.10 This D-value was recalculated assuming one negative at 22.5 minute and one positive at 25 minutes because all units were positive at 22.5 minutes, even though one negative was observed at 20 minutes and all units were negative at 25 minutes. This D-value was also recalculated assuming one positive at 25 minutes to assure comparable treatment of the standard and ramped cycles for this BI type.

Accounting for Temperature Variations. The work described thus far focused on EtO lethality at a constant temperature. A technique for calculating the effect of temperature variations on D-values—known as the z-value effect—has been described for steam and dry-heat applications.19 To account for such an effect on EtO sterilization, the researchers involved in this study consulted a range of sources.2,6,7,11–15 One reported a theoretical lower limit of Q10 = 1.8 for EtO sterilization,14 but a consensus seems to have evolved for a nominal Q10 value of 2. (This means that a 10°C change would affect lethality by a factor of 2.) Thus a Q10 value of 2 was used for a set of temperature-related tests along with a z value of 33.2°C, which was calculated using the relationship z = 10°C/log10 Q. This value was intermediate between a recently suggested z value of 36°C and an older recommendation of 29.4°C.15,9 After the test results (which are described below) were reviewed, however, it became apparent that the best choice of z to fit the experimental data is 29°C, which is essentially the result for a Q10 value of 2.21 and very close to the calculated value of 29.4°C suggested by another previous study.11

Because they are independent variables, a reference EtO concentration (Cref) and temperature (Tref) can be used to calculate the equivalent time for various temperatures as follows:



For example, using z = 29°C, if the exposure time (t) is 40 minutes, the temperature (T) is 40°C, and the concentration (C) is 300 mg/L, the equivalent time at Cref = 600 mg/L and Tref = 50°C is 9 minutes:

In addition, because D ~ U, the above equation also can be used to address D-value:


To determine accumulated equivalent time where conditions are changing for EtO concentration and/or temperature, a summation equation can be applied:


Dref Derivation BI Lot No. 235 BI Lot No. 244 BI Lot No. 249
DC600°, T54 Empirical 3.5 3.4 3.4
DC750°, T40 Empirical 8.5 7.8 8.6
Calculated 8.5 8.3 8.3
DC600°, T45 Empirical 7.4 6.6 7.1
Calculated 7.2 6.9 6.9
DC300°, T60 Empirical 4.7 4.4 4.3
Calculated 4.3 4.2 4.2
Table V. Empirical and calculated D-values for three lots of BIs processed at various EtO concentrations and temperatures. D-values were calculated using the Spearman-Kaber method.17

To test the applicability of Equation 9, BIs from three lots were processed using various EtO concentrations and temperatures at SGM Biotech. The test conditions were: (1) 60°C, 300 mg/L EtO, and 60% RH; (2) 40°C, 750 mg/L EtO, and 60% RH; and (3) 45°C, 600 mg/L EtO, and 60% RH. The empirical D-value results from these tests are shown in Table V, along with the D-values that were calculated from the original lot certifications at the standard BIER conditions of 54°C, 600 mg/L EtO, and 60% RH. In all cases, the empirical and calculated values agreed closely.

Consideration of the Rodriguez Article. Because it addressed the same issue as the research reported here, the article by Rodriguez et al. was considered carefully.2 One point raised was that the Rodriguez equation for calculating survivors after the first time increment (Dt) is not obvious since the natural log (ln) is used:


However, by deferring to the first-order theory of inactivation20 and comparing this equation with Equation 6 above, it was determined that the empirically determined rate constants (k) for the two studies (designated R for Rodriquez and M for Mosley) are related in the following way:


Therefore, it was concluded that the two equations are equivalent. Based on the rate constant relationship, it was also determined that Equation 6 correlates to the Rodriguez equation for the relationship among D, k, and C:



Another issue that was considered involved the final integration equation in the Rodriguez article.


The current authors agree that the equation is mathematically correct but believe it is unusable in the given format because the EtO concentrations presented in the Rodriguez article cannot be defined by any reasonable equation. In addition, no solution for Cn was given, although n = 1 was suggested. However, in the model validation section that followed, a summation formula was used in the software work-sheet that transforms to Equation 10 in this article. This final transformation essentially completes a mathematical proof of the two approaches. It was concluded that the independent work reported here confirms the equations proposed by Rodriguez and vice versa.


When using the general formulas for calculating process lethality (Equations 1–3), it is critical that equivalent time is not underestimated in Equation 1 or overestimated in Equation 2 because of the effect on SAL. However, the use of approximations or estimations is unavoidable, and a large error is always worse than a small one. Depending on the metrology systems available to particular users, several approaches to maximizing the usefulness of these equations are possible.

For all of the equations, the reference temperatures, concentrations, D-values, and LR values should be for the target control conditions. In addition, whether it results from absorption, chemical reactions, or penetration or diffusion lags, any delay in achieving the expected EtO concentration within the load or product will lead to a calculated U that is greater than the actual U. Therefore, if estimations of product EtO concentration are based on theoretical levels related to EtO pressure rise, equivalent time will be overestimated. On the other hand, a nitrogen overlay may produce a very short period of higher-than-predicted EtO concentrations in the product, and if this phenomenon is not considered the equivalent time will be underestimated.

Total Integration. Ideally, metrology systems capable of monitoring EtO concentrations within the load and the product would provide the most accurate and complete data. However, the available techniques for monitoring product EtO concentrations involve the active withdrawal of samples into a gas chromatograph or IR analytical system, and the act of sampling changes the concentration at the location of interest. The next best approach is to monitor the chamber EtO concentration and combine that data with product temperature data. Inputting these data incrementally would allow the data acquisition system to calculate cumulative equivalent time for the full process, as well as for the shorter cycles used in D-value calculations. Product release criteria could then be established based on total equivalent time as established in the validation.

Estimated Integration. EtO concentrations can be estimated based on the pressure differential in the chamber during the various sterilization cycle phases. These phases include EtO injection time, nitrogen overlay time (where appropriate), exposure time, and the time from the start of exhaust to the inflection point. EtO concentration will decrease during the exposure-time phase unless a pure EtO cycle is used and makeup gas charges are performed to maintain pressure throughout the period. Because this decrease is due to the preferential absorption of EtO by the product materials, regardless of which other gases are present in the chamber, the changes in concentration can be estimated. Decreases in EtO concentration should be directly proportional to decreases in pressure differential. In the case of gas mixtures, any makeup charges will contribute the proportional amount of EtO present in the mixture to the overall EtO concentration in the chamber.

Simple Approaches. If a system for determining changes in chamber EtO concentration is not available, a reasonable over-estimation of equivalent time can be made for use in Equation 1 by applying the following formula (or sections thereof as appropriate):


This approach can also be used to estimate U without regard to changes in temperature. During the early stages of the process cycle, temperatures are below the target level. If increasing temperature is taken into account, the estimated equivalent time will be lower than if temperature is treated as if it were at steady state throughout the cycle. With regard to both EtO concentration and temperature, therefore, the overestimates that result from using this simplified approach will result in the use of full process cycles that are longer than necessary to achieve the required SAL of ¾ 10–6.

In contrast, when equivalent time is used in Equation 2 for the full process, it is critical to ensure it is not overestimated. Obviously, the more actual temperature and EtO concentration data that can be included in the calculations, the better, but any justifiable estimating method that leads to overestimates of D-value and SAL is acceptable. A common approach for estimating U for the full process is to disregard the equivalent time contributed by the cycle phases prior to and following the exposure-time phase. The equivalent time contributed during those phases then becomes an added safety factor. Another, more-accurate approach would be to double the equivalent time calculated for a half cycle by including the EtO injection, nitrogen overlay, and evacuation time periods.

Figure 6. EtO sterilization lethality at two locations where temperatures are 10°C apart.

The Boundaries Approach. It has also been suggested that the use of boundaries may be appropriate for estimating equivalent time for EtO sterilization.21 One approach is to consider the variations in temperature within the chamber. Figure 6 plots the expected change in log N over time during an EtO process where both EtO concentration and temperature are increasing. The concave curves seen in the figure represent the rate of microbial lethality at two locations in the chamber that differ in temperature by 10°C. Assuming a Q10 of 2, the rate of microbial population decrease will differ by a factor of 2 between these two locations. The higher-temperature boundary can be used to estimate equivalent time for D-value calculations and the lower-temperature boundary to estimate it for determining the full process SLR.

It should be noted that the Pflug and Holcomb method for D- value calculations may be inappropriate in some situations.9 One of its primary advantages is the accurate determination of mean time until sterility. For processes where all-zero or only fraction-negative results are obtained for the shortest possible EtO exposure cycle, the method cannot be used. In such situations the Stumbo, Murphy, and Cochran method for lethality calculations is recommended.10


The equations presented here, along with other recently published formulas on accumulated and integrated lethality for EtO process sterilization, have been developed to assist practitioners involved in validation and routine product release. When temperature data are used in conjunction with direct EtO concentration measurements, the prospects for moving toward parametric release are enhanced. This approach will also normally lead to shorter cycles than have been possible using older methodologies or in cases where equivalent time for the full process has been overestimated.


The authors wish to thank Irving Pflug, Carl Bruch, Duane Hass, and Richard Holcomb for their helpful suggestions and for reviewing portions of this manuscript.


1. CW Bruch, "Gaseous Sterilization," Annual Review of Microbiology 61 (1961): 245–262.
2. AC Rodriguez et al., "Calculating Accumulated Lethality and Survivorship in EtO Sterilization Processes," Medical Device & Diagnostic Industry 23, no. 9 (2001): 100–107.
3. BIER/EO Gas Vessels, ANSI/AAMI ST44:1992 (Arlington, VA: Association for the Advancement of Medical Instrumentation [AAMI], 1992).
4. Sterilization of Health Care Products—Biological Indicators—Part 2: Biological Indicators for Ethylene Oxide Sterilization, ISO 11138-2:1994 (Geneva: International Organization for Standardization, 1994).
5. PG Smith, "Continuous Monitoring of EtO Concentrations during Sterilization," Medical Device & Diagnostic Industry 23, no. 2 (2001): 80–88.
6. DJ Burgess and RR Reich, "Industrial Ethylene Oxide Sterilization," in Sterilization Technology: A Practical Guide for Manufacturers and Users of Health Care Products, ed. RF Morrissey and GB Phillips (New York: Van Nostrand Reinhold, 1993), 152–195.
7. K Kereluk et al., "Microbiological Aspects of Ethylene Oxide Sterilization," Applied Microbiology 19 (1970): 157–162.
8. Medical Devices—Validation and Routine Control of Ethylene Oxide Sterilization, ISO 11135:1994 (Geneva: International Organization for Standardization, 1994), 11–16.
9. IJ Pflug, RG Holcomb, and MM Gomez, "Thermal Destruction of Miroorganisms," in Disinfection, Sterilization, and Preservation, ed. S Block (Philadelphia: Lippincott, Williams & Wilkins, 2001), 79–129.
10. CR Stumbo, JR Murphy, and J Cochran, "Nature of Thermal Death Time Curves for P.A. 3679 and Clostridium botulinum," Food Technology 4 (1950): 321–326.
11. TS Lui et al., "Dichlorodifluoromethane—Ethylene Oxide Sterilization as a Sterilant at Elevated Temperatures," Food Technology 22 (1968): 86–89.
12. JR Gillis, "Cycle Development—Microbial Challenge Systems," in Industrial Ethylene Oxide Sterilization of Medical Devices—Process Design, Validation, Routine Sterilization, AAMI Technological Assessment Report No. 1-81 (Arlington, VA: AAMI, 1981), 21–22.
13. JR Gillis, "Ethylene Oxide Sterilization," in Validation of Aseptic Pharmaceutical Processes, ed. FJ Carleton and JP Agalloco (New York: Marcel Dekker, 1986), 357–376.
14. RR Ernst, "Ethylene Oxide Gaseous Sterilization for Industrial Applications," in Industrial Sterilization, International Symposium, Amsterdam 1972, ed. GB Phillips and WS Miller (Durham, NC: Duke University Press, 1973), 181–208.
15. CW Bruch "Ethylene Oxide Sterilization—Technology and Regulation," in Industrial Ethylene Oxide Sterilization of Medical Devices—Process Design, Validation, Routine Sterilization, AAMI Technological Assessment Report No. 1–81 (Arlington, VA: AAMI, 1981), 3–5.
16. LJ Joslyn, "Gaseous Chemical Sterilization," in Disinfection, Sterilization, and Preservation, ed. SS Block (Philadelphia: Lippincott, Williams & Wilkins, 2001), 337–359.
17. RG Holcomb and IJ Pflug, "The Spearman-Karber Method of Analyzing Quantal Assay Microbial Destruction Data," in Selected Papers on the Michobiology and Engineering of Sterilization Processes, ed. Irving J. Pflug (Minneapolis: Enviromental Sterilization Laboratory, 1988), 83–100.
18. "Biological Indicator for Ethylene Oxide Sterilization, Paper Strip," in US Pharmacopeia XXIV (Rockville, MD: United States Pharmacopeial Convention, 1999), 231–232.
19. IJ Pflug, Microbiology and Engineering of Sterilization Processes, 10th ed. (Minneapolis: Environmental Sterilization Services, 1999).
20. O Rahn, Injury and Death of Bacteria by Chemical Agents, Biodynamica Monograph No. 3, ed. BJ Luyet (Normandy, MO: Biodynamica, 1945), 9–41.
21. IJ Pflug and R Holcomb, personal communication with the authors. October 23, 2001.

Illustration by Bek Shakirov

Copyright ©2002 Medical Device & Diagnostic Industry

Avoiding Warning Letters

Originally Published MDDI February 2002


John Lincoln

How can my company avoid receiving an FDA warning letter after we have been issued Form 483?

FDA-regulated companies frequently receive warning letters after the completion of a GMP compliance audit and the issuance of the FDA Form 483. Many of these instances could have been avoided. Obviously, the best way to avoid receiving a letter is to be in compliance with defined regulations. But even if 483 observations are received, there are methods to reduce the chance of receiving a follow-up warning letter.

Such methods may include:

  • Developing an active company compliance culture, and conveying that attitude to the investigator.
  • Correcting most observations—if not all—while the FDA investigator is still on-site. Ensuring that any remaining issues are in the process of being addressed, and that this is communicated to the investigator in writing before or during the close-out meeting—and before FDA's establishment inspection report is completed.
  • Ensuring that senior management is actively involved in the initial meeting, during the audit, in the final meeting, and in all subsequent communication with the agency.
  • Ensuring that all communication with the investigator and the agency is professional, that issues are satisfactorily addressed, and that proof of that is provided, including copies of the final, signed-off documents.
  • Seeing that promised action is taken and documented, and that copies are provided to the agency.
  • Ensuring that responses are timely. It is advisable to communicate frequently (monthly or quarterly).
  • Ensuring that the initial response is immediate, indicating the completed corrective action and the timelines for action in process.

There are some problems for which nothing can prevent the issuance of a Warning Letter. These would include reoccurring observations for the same problem, promised action not taken, operating with a lack of control, a poor past-compliance history, etc.

Sometimes the attitude senior people convey to the investigator can almost guarantee a warning letter. A few wrong responses: "It costs too much to implement that"; "Senior management is too busy"; "That's the responsibility of our QA manager"; and "We don't know (and aren't going to quickly find out)." Of course, if the company conveys the impression that it's out of control, or that management is not fully involved, that compliance is only a middle-management responsibility, then no matter what the response, a warning letter will be forthcoming.

If in doubt about the strategy to pursue, there are professional organizations and consultants who have been down this road before and can assist in developing an action plan and timeline.

What if your company has already received a warning letter? The points above still apply. An immediate effort should be made to respond to each issue on the 483 and in the warning letter. Tie each response to the specific observation—by number, reference, or quote.

Always, the best defense is avoidance through active compliance. But if a letter is received, immediate and periodic communication with the agency, supported with proof of compliance, will minimize the negative affects of regulatory action. If this is done while bringing the company quickly back into compliance, also it may provide the RA/QA function with new-found credibility and support from senior management.

John Lincoln is a principal at J. E. Lincoln and Associates, a consulting company serving industries regulated by FDA. Lincoln specializes in project management, product-to-market issues, defect reduction, cycle time reduction, ERP, quality management systems, and regulatory affairs.

Copyright ©2002 Medical Device & Diagnostic Industry

How to Plan Life Tests for Optimum Cost-Benefit

Originally Published MDDI February 2002


Understanding the basic principles behind all the statistical wizardry and jargon of life testing can lead to better results.

Tom Clifford and Vanessa Koutroupas

Product development often requires accelerated-aging or fatigue-life testing. In a typical test, samples are put into a test chamber under carefully controlled conditions and inspected periodically for failures. The pattern of failures over time establishes the life of the population under those conditions. This finding is then used to evaluate the fitness of the product for a particular application.

Managing the test process can be highly stressful. Test specimens are typically rare and costly prototypes, testing facilities and inspections are expensive, and the tests are time-consuming. You usually get only one shot at the test, and the results can determine the fate of the product line, or the company. Your selection of the test variables is always an uncomfortable balance of cost, time, testing resources, availability of samples, and the "goodness" of the final result.

The deliverable will be some metric describing the failure distribution of the population. You are asked to get the most accurate metric (i.e., decision information) as rapidly as possible for your testing dollar. You must select the number of samples, the inspection frequency, and the method of analysis, and decide whether the test can be stopped before all samples have failed.

Typical tests reported in the literature are not much help. They turn out to be sadly imprecise in terms of statistical credibility. This paper describes the effects of the testing variables on the quality of the resulting metrics. It shows how to use Monte Carlo analyses to determine these effects with proper statistical confidence. Costs are also considered, and guidelines are provided to arrive at the most cost-effective choice of testing variables.

The goal is to encourage you to evaluate your testing options as well as the reported data, using available software and the methods and reasoning described herein. Examples are drawn from solder-joint accelerated-aging tests, but the statistics and results will apply to most accelerated-fatigue-life tests, including, in our experience, aging of plastic packaging.1,2


Two types of tests are common.3 One is the pass-fail test. This is the simpler type, asking only whether the samples passed a specific threshold. Pass-fail testing will be discussed briefly towards the end of this article. The primary focus here is on the more-complicated second type, which seeks to define the distribution of failure times of the population being sampled and to show the level of confidence in this finding. The test engineer uses this information to compare one product with another, to check process stability or product uniformity, to quantify a material change, or to provide reliability numbers (e.g., what percent of the population will fail at X cycles).

What's a good metric for this second type of test? Certainly not great swooping full-color Weibull curves of asserted failure data points on some sort of probability axis. That's pretentious, probably misleading, and impractical for making decisions. Similarly, because lifetimes are never linear, don't plot anything, straight-line or curved, on linear paper, and don't report averages.

In most cases, you need only to determine and report a couple of metrics: one showing where most samples fail, and another indicating the breadth of the distribution. For thermal-cycle life tests, the median, or F50, is a good central metric. That's when 50% of the population will have failed. Other central metrics are available: Weibull fans, for instance, favor the "eta," which is when 63% fail. We prefer rounder numbers, however, especially a simple one like F50. (Note that some texts call F50 something different, like N50. However, we will reserve N to indicate sample size.)

Now for the second necessary metric: how to indicate the breadth? Weibull-curve fans would suggest that we use the "beta," an esoteric parameter indicating the slope of a straight two-parameter Weibull-log line. That looks very impressive, but it is not very useful.

For the second metric, we instead suggest F1, which is when 1% of the population will have failed. That's understandable to anyone. And the ratio of F50:F1 defines the slope of the Weibull curve as accurately as beta does. Moreover, anyone can plot a straight line on a Weibull plot, using those two points, and can read off F0.1 or F90, or any other value. (Note that projecting to the parts-per-million level below F0.1 requires some expert help, possibly using a three-parameter Weibull and thoughtful exploration of types and tests of distributions.)3

Figure 1. Typical Weibull plot, showing 95% confidence intervals.

Figure 1 shows a typical Weibull plot of eight samples, drawn from a population where eta =1000 and beta = 6. While impressive, all you really need to do is report the metrics. The basic metrics can be read off this plot: the F50 is ~870, and F1 is ~310. If other failure-probability points are needed, anyone can plot these two points on blank log-Weibull paper, draw a straight line, and read the other metrics directly from the graph. In this case, F90 is ~1200, and F10 is ~550, for example. (Note that this plot also shows the confidence intervals, discussed below.)

Distributions, of course, can be narrow or broad. When failure times clump together, we say the distribution is narrow. When failure times are spread way out (some early, some very late), we say the distribution is broad. The slope of a straight line on these cumulative probability curves is an indicator of the breadth of the distribution. Because of the way these curves are constructed, a steep slope indicates a narrow distribution, and a shallow slope indicates a broad distribution.

For Weibull distributions, the parameter beta describes the slope of the curve. A higher beta means a steep slope and narrow distribution. Narrow distributions from the literature typically are beta 8 to 12. These are a treat to deal with. Broader distributions, which show very early failures as well as samples that keep hanging on, are typically beta 2 to 6. These present special challenges.


The quality of your results is determined long before the data start rolling in. Quality is built into the test plan: it is not determined or discovered after the test. The term quality does not mean how well the data satisfy your boss's expectations, or how smoothly the plotted points line up. Quality means how well the metric describes the underlying population.

Assuming nothing goes awry during testing, your initial selection of testing factors will determine the quality of your resulting metrics. And more importantly, you can know, long before you start getting data, what that quality will be. You can set up a quick-and-dirty test to generate ballpark values, or a more elaborate test to discriminate between populations that may be very similar. You are in control. Solid numerical measures of goodness are available for making these decisions. And just what is a proper measure of goodness, i.e., what measure shows how well the metric describes the population?

Classic statistics offers a useful measure: the confidence interval (CI). This is a calculated function of the breadth of the population and the number of samples. Commercially available software, such as Reliasoft Weibull5++, will automatically provide the CI.

A small CI says you are confident the population metric is somewhere within a tight interval; that is, you know your distribution pretty well. A large CI means that you know only that the population metric is somewhere within a large interval; that is, you are confident that the true population is somewhere in that ballpark, but you don't know the underlying population very well.

If you want to know a population precisely, run the test so that the CI will be small. Accordingly, our measure of goodness will be the CI. Each metric gets its own software-calculated CI.

Report the F50 and its CI, and the F1 and its CI. This is essentially all you need. These data describe the best estimate of the underlying population, as well as a measure of the uncertainty about those metrics.

By the way, mathematicians have provided a whole range of CIs. We can select 99, 95, or 50% CI, or whatever we want. Most engineers use 95% CI, which is what we'll use. That means we are 95% confident that the true population metric is within that interval. High-risk products, of course, might demand a 99% CI.

In Figure 1, where the F50 is ~870, the 95% CI around that point is 640 to 1050. This means that we can be 95% confident that the true F50 of the underlying population is somewhere between 640 and 1050. The 95% CI around F1.0 is ~60 to ~500.

CI is strongly affected by the selection of test variables. For example, testing many samples will tighten the CI around any metric, and permit us to maintain that we are confident that the population metric is within a very tight interval of uncertainty. To permit relative comparisons, we'll use a ratio of the CI to the metric, and we'll call that the uncertainty, or 100(CI/F50), in percent. For example, if the metric is F50 = 870, and the CI is 840 to 900, you could be confident the F50 of the population would be between 840 and 900. That's a small interval. The calculation is 100(60/870) = 7%.

By contrast, if your metric is F50 = 870, and the CI is 640 to 1050 (410 cycles), as shown in the figure, that means you are confident only that the F50 falls somewhere between 640 and 1050. That's 100(410/870) = 47%. That's not so good. The CI provides this measure of goodness, and is also useful in deciding whether one population is really different from another. If the CIs do not overlap, you can be confident they are different. If the CIs overlap, the samples are very likely from the same population, and the populations are very likely to be identical.


What elements will affect uncertainty? All of them—some more than others. The more samples, the better your knowledge of the population. Frequent inspection is good, as is letting all samples fail. If your distribution happens to be narrow, better yet. What if there are several components on the same board, all failing at different rates? What test plan will allow you to resolve small differences between populations A and B? You need facts. You need to know how each testing element affects the quality of the resulting metrics.

How do we learn about something? Try it and see. In statistics, this approach is called the Monte Carlo method. To study a distribution, take random samples from a population and analyze them. Do this several times and see what happens. This approach is based on two fundamental concepts: you can learn about a population by drawing random samples from it, and every random sample is just as valid as any other. The Monte Carlo method lets you look at the quality of the metrics you will encounter under different test schemes: how many to test, how often to inspect, how long to let them run before stopping, and so forth.

For example, suppose you want to know what sample size is needed to describe the F50 of a population. First select the distribution you're interested in. Typically, you know something about what you're testing—approximately when samples will start failing and when most will fail. From that rough idea, you can select the distribution you'll use for the Monte Carlo run. Draw a sample set N = 5. Calculate F50. Draw another five; calculate F50. Do this many times. You'll discover that the F50s will vary widely. Now draw several sets of N = 20. Calculate F50s. The resulting F50s will be nearly identical. Using Monte Carlo analysis, you have thus demonstrated that a larger sample size provides a more accurate measure of the population. Pick another distribution (for failures that happen sooner or happen later), and do the exercise again. You'll soon start seeing the effect of sample size.

We'll use this Monte Carlo method to look at all the important testing variables, exploring populations with different distributions. We'll base our analysis on Weibull distributions, and will use F50 as the metric. Note that log-normal distributions behave similarly. The measure of quality is the uncertainty, as calculated above. A large uncertainty means the metric is not very indicative of the underlying population. A small uncertainty means the metric is less uncertain, i.e., a better metric.

Table I is a very simple example of Monte Carlo runs. One dramatic and telling observation is that perfectly valid samples can look very different from their underlying population. This is especially true for small sample sizes. The N = 5 samples, from a population of beta = 8, show betas ranging from 6 to 14. You can see that randomly chosen small handfuls of data will produce widely varying and equally valid metrics. In contrast, choosing larger handfuls of data will get you closer every time.

Three Trials at N = 5
Three Trials at N = 20
Sample etaa
Sample betaa
CI around F50b
CI/F50, as %c
CI around F1b
CI/F1, as %c
acalculated by Realiasoft Weibull5++
bread from the probability plot
ccalculated by the author

Table I. Example of a simple Monte Carlo run: beta = 8, eta = 2000, continuous inspection, test not suspended. Data points were selected randomly from the population by Reliasoft software.

You now have a mathematically valid tool for your comparison decisions: a certain test scenario will result in a knowable metric quality. In this one example, if you decide to test 20 samples, your uncertainty around F50 will be ~12%. If you choose to test only five samples, your uncertainty will be ~40%. Similarly, uncertainties around F1 will be ~40% and ~125%, respectively. This simulation can reveal much about the relationship between sample and underlying population. It also provides valid data sets, to provide experience with expected variations and outriders.

Figure 2. Effect of sample size on uncertainty.

Effect of Sample Size. A basic principle of statistics is the need to test many samples to properly characterize your population. Figure 2 confirms this for a simple test where eta = 2000, with broad (beta = 4) and narrow (beta = 16) distributions and where samples are continuously inspected, and the test is not suspended. It's clear that a sample size of five (N = 5) results in two to three times greater uncertainty compared to N = 20. A sample size of 20 is almost as good as 100, which is as good as you'll get. Note that in every case F1s are more uncertain than are F50s.

Note also that narrow distributions are much more desirable. The uncertainty of any metric, at any sample size, will always be better in a narrow distribution (beta =12 to 16) than in a broad one (beta = 4). This makes sense: From a tight distribution, any specimen is going to behave much like any other specimen, so it matters less how many you test. Tailor your test plan to fit your expected distribution, but start with eta = 6 to 10 and use it until you know better.

Effect of Inspection Frequency. Typically, samples are inspected periodically. You will know that the sample failed between X and Y cycles, but you won't know exactly when. Reassuringly, when properly treated, this sort of data can be quite good, even if you have as few as 4 to 6 inspections during the population's failure span. Also, inspection intervals need not be the same throughout the test. Good software or manual methods can account for variations.

Figure 3. Effect of inspection frequency.

If you expect the span of failures to be ~500 to ~2000 cycles, check every 100 cycles. Then check every 200 to 400 cycles. Don't worry if a particular inspection shows many failures having occurred within one interval. That can be dealt with using the right software and knowledge. Note that there are devices that can tell you exactly when the failure occurs (two of these in the authors' experience include circuit analyzers to continuously monitor fatigue-induced electrical opens, and pressure-decay switches to monitor leaks in plastics packaging). These devices provide a continuous record, such that you know exactly when a specimen fails, However, you do not need continuous data to get good results. Figure 3 shows some representative case results.

Effect of Test Completion. By stopping the test before the last sample fails, you can save lots of valuable time and make good use of the data that are available. Suspending a test can affect its cost and feasibility.

Figure 4. Effect of suspension and sample size for two distributions.

Suspending a test when 60 to 80% of samples have failed can save time and money without seriously affecting the quality of the results. Sample size is important, and good software or manual methods are necessary. Figure 4 shows some examples of the quality penalty incurred by suspending a test. It doesn't hurt as much as you might think.

While the quality impact can be relatively slight, the time savings benefit can be substantial. In a test with an eta = 2000, beta = 8, you would expect completion at about 4000 cycles. At 15 cycles per day, that's 8 months. Suspending at 60% saves about 2000 cycles, or 4 months. Most importantly, you can see and evaluate any quality penalty and time-to-market benefit long before finishing the test.

Effect of the Data Distribution. Failure distributions can be broad (some samples fail early, others very late) or narrow (all samples fail around the same time). Most are Weibull shaped, others are log-normal shaped.3

For example, the most uniform solder-joint populations, under tightly controlled test conditions, typically fail in a Weibull shape, beta = 10 to 12. Sloppier workmanship control or several overlapping failure modes (such as those typically encountered with the newest microelectronics prototypes) will result in a broad distribution, beta = 2 to 4.

For complex situations, the distributions will be log-normal, or even bimodal. It is important to note that you cannot tell for sure just by looking at the straightness of a line of plotted points from a small sample whether the population is Weibull or log-normal if your sample size is 50 or less.

For most useful sample sizes, i.e., N = ~20 or less, plots on either type of paper will look ragged. That's to be expected. An apparently ragged line doesn't usually mean that the underlying distribution is not consistent with the type of graph paper you are using, that you have a bimodal distribution, that you have outriders, or that anything is wrong with the data. Monte Carlo experience teaches that perfectly valid random-sample sets will often plot raggedly.

If you are concerned that you do not know whether you have log-normal or Weibull data, be reassured that it doesn't much matter. You'll get about the same central metrics, whether you treat the failure points as Weibull or as log-normal. Good software can easily analyze your data either way. F50 and F1 are relatively insensitive to which shape you think you have.

In the above examples, Weibull data treated as Weibull give you 1400 and 700 for F50 and F1, respectively. The same data points treated as if they were log-normal give you 1450 and 830. The log-normal data treated as log-normal give you 1500 and 820. The same data points treated as Weibull give you 1600 and 680 cycles.

Selection of the most appropriate type of distribution can be an important exercise, but possibly unnecessary in an industry with a mature database of straight lines on Weibull paper. As always, data quality is always better when the distribution happens to be tight (large Weibull eta). Again, metrics at the lower end of the curve, at F.01 and below, warrant special attention.


Manual methods of data reduction can be effective. Use Weibull-log paper unless a mass of test data plot straighter on log-normal paper. Plot each continuous-data failure point, run a best-fit straight line, and read off the F50 and F1. For N = ~10, your metrics will be within 10% of what you'd get from good software. For N = >20, any difference in F50 estimates is trivial. However, for N = 5, the F50 will be 15 to 25% off and the F1 will be even worse. The real problem is the small sample size, not the method.

Note also that suspended data, plotted properly by hand, will very closely match the software's straight-line curve fit. Intervalized data can also be manually plotted, with an excellent match to software results. The trick—and this is essential— is to plot the failure point between the time the sample was OK and the time it was found to have failed. If several samples fail within that interval, spread out the failures uniformly within the interval. This is a bias-free method. It is mathematically and logically incorrect to graph points at the end of the interval.

Note also that attempting to calculate the CI manually will probably be futile. This conclusion about manual-graphing adequacy can be readily confirmed by a few Monte Carlo trials comparing manual results with software results. Any difference is overshadowed by sample-size effects.


Not as simple as one might think, pass-fail tests certainly don't describe the underlying population. For borderline populations, sample size can be a very important factor, and, in fact, it can be a tool to try to accomplish a biased result.

If you are devious and want to prove that your borderline population passes, test only a few samples, hoping that your sample set happens to not contain any that fail at the early end of the distribution. You might fool the customer. Worse yet, you will fool yourself.

Figure 5. Effect of sample size on the likelihood of detecting a "<1000 failure" in a pass/fail test.

Figure 5, a Monte Carlo run on a typical (beta = 8, eta = 1500) distribution, is revealing. Test five samples, and there is only a 20% chance of detecting a failure if you stop the test at your threshold of 1000 cycles. However, from that same population, if you had tested 20 samples, you would probably have encountered at least one failure.

Note that in all the above discussions, we never use the phrase "good enough." Proper selection of testing variables depends on such things as the available samples, time, test racks, inspection resources, and particularly the required level of certainty. If all you need is ballpark accuracy, do the test fast and cheap. Conversely, to demonstrate the reliability of a critical device, you should run the test with more samples, more-frequent inspections, no suspension, and so forth. You should base your decision on data from Monte Carlo runs.


It's possible to estimate the confidence intervals around reported metrics, but only if the actual data are presented. Unfortunately—and perhaps revealingly—they are usually not reported.

In a hypothetical case with a sample size of 10 (beta 6, eta 1000, inspected every 100 cycles) using Monte Carlo analysis, the F50 uncertainty will be 47% (between ~450 and ~950 cycles), and the F10 uncertainty will be 82%. Even worse, when you see a sample size of five, assume that any mid-range metric will be at least 200% off, either way. Conversely, if you see a sample size of 30 or more, a tight beta (8–12), some persuasive evidence of frequent or continuous inspection, no suspension, and proper data reduction, you can be 95% confident that the true population F50 will be within perhaps 5% of the reported sample F50.

If there are no data points, assume there's a reason. A colorful line on a pretty graph with no points and no backup data is a clear invitation to purchase snake oil. Fully reported data can be readily analyzed to determine their credibility.


Figure 6. Effect of sample size, suspension, and inspection interval on the quality of the F50 metric.

All this statistical wizardry is not enough. Money is the name of the game. Samples and inspections cost money. But you really need a certain level of quality in your resulting metrics. What should you do? Use lots of samples but suspend the test sooner? Save on samples but inspect more frequently? Inspect less often but let the test go to completion? Use lots of samples and suspend half-way through?

Table II and Figure 6 describe several hypothetical test scenarios. The costs should be understandable, as the statistics come from Monte Carlo runs. Table II summarizes a few hypothetical planning cases, where costs have been assigned to some of the major testing elements.

Test Number
Inspection Interval (cycles)
Percentage Completed
Number of Weeks
Number of Inspections
Sample Cost ($1000s)
Inspection Cost ($1000s)
Facilities Cost ($1000s)
Total Cost ($1000s)
Data Uncertainty (CI/F50, %)
N = 8
N = 24

Table II. Costed examples of test plans. Elapsed time is 100 cycles per week; test is 100% complete in 30 weeks.

For this exercise we assumed a common distribution (eta = 2000, beta = 8). Assume that samples cost $500 each, inspections cost $20 each, and testing costs $250 per week. You can see the quantitative effect of backing off on your inspection frequency (save money, lose a bit of accuracy), letting the test go to completion (it'll cost time and money, but you'll gain some accuracy), or starting with more samples (more initial expense, but more testing options, much better quality, and time savings).

Plotting these data in Figure 6 provides a visually intuitive sense of the sort of trends and trades that can be prepared and discussed using this technique. Using this approach, you can make sweeping plans or explore subtle details, made more rational and accurate by the effort you put into estimating the costs of the test elements and the statistical cases chosen. If samples are rare or very expensive, you can show how to compensate (more frequent inspections and less suspension). If time is all-important, you can determine how many samples you need and how soon you can suspend the test. If ballpark numbers are OK, you can show how inexpensively and quickly you can obtain that. For any test setup, you can know what you'll lose by shutting down early to save time.


Life tests can be treated like any other task, using comfortable principles and tools. Good software and Monte Carlo analysis can help you plan an optimum test, and can help you evaluate the life-test data that are reported in the literature.


The authors acknowledge the support of the life-test personnel at Lockheed Martin Sunnyvale, particularly Bob Haugh, software; Robert Hill, design; and Grant Inouye, reliability. Technical inputs from Dave Groebel of Reliasoft are also gratefully acknowledged.


1. "Life Tests: Get the Biggest Bang for the Buck," Printed Circuits Europe, 22, Fall 2000.
2. "Thermal-Cycle Life Tests: Help!" Circuits Assembly, September 2000.
3. NIST/Sematech Engineering Statistics Handbook, sect. 8.2 (Austin, TX: Sematech).

Copyright ©2002 Medical Device & Diagnostic Industry

The Qualification Side of Bluetooth Implementation

Originally Published MDDI February 2002


Bluetooth wireless technology is royalty free for Bluetooth SIG (Special Interest Group) members, but there is a hitch: products offered for sale must complete the appropriate testing and have their test results reviewed and approved by a Bluetooth qualification body. The Bluetooth name explicitly implies interoperability and can be used only on qualified products, which is meant to be a benefit for manufacturers, integrators, and end customers. If the Bluetooth intellectual property were sold without qualification, the license would not be in effect and the offending company would be infringing on certain patents. (In addition, Bluetooth is a registered trademark owned by Bluetooth SIG Inc.)

While the complete Bluetooth testing process can be quite long and expensive for a fully custom implementation and is complicated even for products incorporating module or stack components, several test houses provide such services as well as other compliance testing both in the United States and abroad.

Qualified components carry their level of approval with them and do not require retesting. This means that if an approved module incorporating the RF and baseband sections is used, those portions of the approval obtained by the supplying company automatically apply to the product into which the module is incorporated. Only additional stack or profile work above that purchased would require testing.

When product-level Bluetooth devices are used in other products, such as an approved PCMCIA card and its associated Windows software, no additional approvals are required if the new product is to be sold to an end customer; the integrating company need not even join the SIG (it is free, but it requires a legal commitment to the free license for the Bluetooth intellectual property). In another example, if a company were to use a device like the serial port adapter from ConnectBlue (Malmo, Sweden) in a product, the end product would immediately carry the complete Bluetooth approvals; the company would be able to use the Bluetooth name and ship compliant products immediately.

Products approved to the 1.1 specification.

Shown below is a graph of the products approved to the current Bluetooth 1.1 specification (the author recommends only use of 1.1 components and products). It shows the types of products approved as of this writing. Products represent such end-user devices as phones, printers, and medical devices (only one has been approved to date). The HW and SW components are chips and modules, and stacks not intended for end customer use. Development tools are usually kits and (along with demo products) are intended to provide a demonstration of the technology or of a supplier's implementation of the technology.

Copyright ©2002 Medical Device & Diagnostic Industry

Putting Teeth into FDA Warning Letters

Originally Published MDDI February 2002


A recent HHS directive seeks to bring about a rational, risk-based approach to enforcement.

James G. Dickinson

Keeping FDA's Hands off Internal Audits | FDA's Device Expertise at Risk

How seriously do you regard an FDA warning letter? Does it electrify your company and send the board into shock? For years, Washington lawyers who specialize in FDA matters have been lamenting the failure of FDA warning letters to be taken seriously by companies that receive them.

In November, Health and Human Services (HHS) deputy secretary Claude A. Allen, acting in the absence of a permanent FDA commissioner, directed the agency to stop issuing warning letters and so-called "untitled" letters that have not first been cleared for legal sufficiency and policy by the FDA chief counsel, Daniel E. Troy, a new Bush administration appointee.

This immediately provoked an outcry that the Bush crowd was turning the FDA watchdog into a lapdog of industry.

But the man who is credited with having invented these kinds of letters, former chief counsel (during the Nixon administration)—and a candidate for commissioner—Peter Barton Hutt, applauded Allen's move.

"I think it's terrific good news," he told this writer. "Warning letters are chaos now. They are ignored by everybody."

When he started regulatory letters, the precursors of today's warning letters, they gave recipients only two choices—prompt compliance or federal court. "People took them seriously," Hutt said. "Then they gradually began to slide, and FDA began to let more and more people write them. Then, in the late 1980s, FDA took the position that these were so trivial that you could ignore them. . . . If someone sued, FDA said, 'Well, we don't know whether this is our position or not.' Nobody pays any attention to them anymore."

But, devalued though they are from Hutt's era, warning letters have their defenders. Criticism of Allen's action moved the deputy secretary to take the unusual step of answering those defenders. He told me that his motive was not to "chill the impact of FDA enforcement," as some critics had charged.

"What we're talking about," he said, "is having a rational, reasonable, risk-based practice of enforcement. We agree that there are appropriate places for using warnings to industry to let them know what it is that we will be looking at, what we want to enforce.

"But it's also important that we have a risk-based approach, and that means we must lay out our compliance priorities, the outcomes we are looking for, and how we hope to achieve them."

Allen said his directive stemmed from a recent request from HHS Secretary Tommy Thompson. In the continuing absence of a permanent FDA commissioner, Thompson asked him and other senior HHS officials to work very closely with FDA to bring about some "reasoned, thoughtful changes" that would increase the agency's impact in fulfilling its mission."

In the course of this, Allen said he had noticed "some letters come through, and it just raised questions, like, 'What is this?' In the course of a meeting that we had, I raised the question as to what these things were and why they were going out in such a hodgepodge fashion on issues that really have broad implications but may or may not be the highest priority for the agency or the department."

In his directive to acting commissioner Bernard A. Schwetz, Allen related what happened next: "At a recent meeting with the secretary on FDA issues, considerable discussion was directed toward the practice of different components of the agency sending out warning letters and so-called 'untitled' letters that have not necessarily been first reviewed and approved by the FDA's Office of Chief Counsel (OCC)."

This "unleashing of the field" was an innovation of former commissioner David A. Kessler, partly as a means of ratcheting up enforcement and partly as a response to field complaints about long delays in getting FDA clearance for warning letters.

Allen insisted that he does not want his directive to result in such clearance delays now. Indeed, his directive states: "I have urged OCC to review these let-ters expeditiously."

This is all part of building FDA into "the agency of the 21st century," Allen said. The HHS leadership believes the directive will end the current cycle of untitled letters, warning letters, and "recidivist" letters, and in their place lay out a clear road map for industry to follow, with risk-based priorities and standards established by FDA. In doing this, Allen said, "We want to work very closely with our constituencies and stakeholders who are impacted by our decisions. . . . We are looking for FDA to lay out priorities, and enforce those priorities."

As for how FDA will find the resources to do this, Allen said these "may already exist" within the department. He noted that FDA lawyers already "work very closely" with the HHS Office of General Counsel. "We will walk through that as needed," he said.

Keeping FDA's Hands off Internal Audits

Nature abhors vacuums, and governments don't like them either. Consider the regulatory vacuum or no-man's-land that has existed between inspections done by FDA and those done by a company on itself (known as "internal audits").

For years, FDA's position has been to informally encourage internal audits as a good way of keeping companies voluntarily focused on compliance with good manufacturing practices—the theory being that less will be found noncompliant at self-auditing companies when the next FDA inspection comes around. As part of its encouragement, FDA has informally kept its regulatory hands off such internal audits.

That seemed to change in November, however, when FDA released a warning letter it had written to a Florida medical device manufacturer. The letter indicated that the agency had used one of the firm's internal audits as a basis for regulatory citation. Also in November, the Washington law firm of Hogan & Hartson—apparently by coincidence—formally raised the issue of the "hands-off audits" policy in a letter to FDA on behalf of an unnamed client.

In its warning letter to Medical Device Technologies, dated October 29, 2001, FDA complained that a three-week-old internal audit it found during an inspection at the company's Gainesville, FL, facility had failed to identify quality system regulation deficiencies that were documented easily enough by an FDA investigator. Medical Device Technologies makes catheters and needle sets.

How FDA got the firm's internal audit has not been made public. All a company official would say was that the audit was not provided to FDA. According to the warning letter, other issues raised during the inspection included the company's failure to ensure that quality requirements were being met by its suppliers and contractors. For example, FDA said that the firm's purchasing control procedures did not include all specifications for vendor-supplied devices and that the company failed to require periodic supplier audits. Also, FDA said the firm did not adequately address complaints related to a vendor-supplied product that eventually resulted in a recall.

In its apparently unrelated comments to FDA, Hogan & Hartson told the agency it needs to confirm that a recent draft guidance, Biological Product Deviation Reporting for Licensed Manufacturers of Biological Products Other Than Blood and Blood Components, will not be used to require manufacturers to produce internal audit reports and other audits conducted by clients and contractors that may include possible CGMP violations.

The law firm said the use of these documents would "negate the intent of audit programs or otherwise create a disincentive for manufacturers to be as comprehensive as possible in their scope." According to the law firm, internal audit findings are "relatively insignificant and do not raise concerns about the quality, purity, effectiveness, or safety of the product." Other findings may point out weaknesses or raise potential issues, the letter said, both of which are company methods of demonstrating concern, interest, and diligence in controlling quality and preventing errors. Requiring such documents during inspection may discourage companies from performing thorough self-audits.

FDA's Device Expertise at Risk

Cultural and technological pressures are overwhelming FDA's Center for Devices and Radiological Health and major changes are necessary if science is to continue to play a fundamental role in the center, says a November 16 report by a CDRH external review subcommittee, titled Science at Work in CDRH.

The committee said the center is at a pivotal place in its history—with more than 30% of its staff eligible for retirement; a "woefully inadequate" budget allocation for recruitment, staff training, and development; and an eroding infrastructure in such areas as information systems, laboratory equipment, and training.

"The subcommittee believes that CDRH scientists may become regulators without sufficient scientific training" if the current trends continue for the center, the report stated. Charged by FDA's science board with identifying how science is used within the center, the committee issued its findings based on a review of agency documents and a series of interviews with staff, management, industry, and other relevant parties.

"Significant" changes to CDRH staff recruitment, training, and retention are vital if the center is to remain scientifically competent, the report says. "Additionally, the existing expertise will not be the same expertise that is needed for new technologies. In the discussion with industry representatives, a common theme was that the breadth of existing scientific experience is not sufficient for the future."

The current level of support does not provide CDRH's scientific staff with enough opportunity to remain current within their area of expertise, the report adds, and it says questions exist as to the adequacy of some scientific reviews by scientists hired decades ago who have had little opportunity or time for continued training in their fields.

Additionally, the committee commented on apparent gaps in scientific expertise within the agency. Gaps in neurology, behavioral sciences, and information technology exist, as well as a lack of "human factors expertise and software specialists," the report noted. These gaps point to a need for the center to reach out to other agencies in order to have access to specialized scientific expertise.

Too much emphasis is being placed on the timeline aspect of the center's charge, the report says, and not enough on its long-term needs. For example, some guidance documents are 10 years old and out of date and there are no mechanisms for a systematic review and updating process. Also, management does not appear to have any staffing plan or evaluation of what technical positions are needed to support the center's goals.

Further, "use of outside experts is limited by organizational barriers, budgets, and time constraints, by concerns about confidentiality and conflicts, and by legal requirements for action within restricted time windows. . . . Even when CDRH has expertise within the organization as a whole, the responsible individuals within the Office of Device Evaluation are not necessarily aware that such expertise is present, because they have no detailed database (electronic or otherwise) as a catalog."

The report makes a number of recommendations to the center, including outsourcing functions while still maintaining oversight; establishing an electronic database for liaison functions and an internal and external expertise inventory; assessing the current breadth of expertise and developing a long-term strategic staffing and recruitment plan; and expanding its outreach and scientific interactions with industry and universities. This last component is to be achieved through visitor programs and professional development forums to exchange information between FDA staff, industry, and academia, with an emphasis on scientific fields that will yield new medical devices within the next five years.

Copyright ©2002 Medical Device & Diagnostic Industry

Elastomer compound offers a fast-curing alternative to moisture curing

. A UV curable elastomer compound from Master Bond Inc. (Hackensack, NJ; 201/343-8983) offers an alternative to conventional moisture curing silicone materials for use-in-place seals, gaskets, bonding, and protective coatings. UV21 consists of a one-component, 100% reactive composition. Curing is achieved almost immediately when exposed to UV light of 320–365 nm and radiation dosages of 20–40 W/cm2. Average temperature range is from –65° to 250°F. Providing electrical insulation unaffected by exposure to moisture, UV21 improves resistance to most chemicals and adheres to glass and plastics, including polycarbonates and acrylics.