Originally Published MDDI April 2002

STERILIZATION

D-values for several ranges of expected BIER-vessel performance have been calculated and compared with a common BI-certificate value.

Commercially available biological indicators (BIs) are of critical importance to both the healthcare providers and device manufacturers that perform EtO sterilization.^1–3 Because it is more easily controlled, radiation sterilization lends itself to parametric release based on chemical dosimeters.⁴ For steam sterilization, healthcare providers generally use commercially available BIs, but manufacturers of sterile products may take several other approaches to validation and routine release.⁵ Naturally occurring bioburden, BIs derived from resistant bioburden isolates, and commercially available BIs⁶ are commonly used for validation, while BIs or parametric release can be used for routine release.

However, parametric release is not commonplace with EtO sterilization processes, and most routine release testing is performed with commercially available BIs. In addition, commercial BIs are used to establish microbial-lethality relationships in nearly all EtO process validations.

When using BIs for these purposes, device manufacturers are dependent on the certifications performed by the BI manufacturers to assure that their products will meet the appropriate quality system regulation requirements. Such certifications are conducted in specialized test vessels known as biological indicator evaluator resistometers, or BIER vessels.^7,8 The test requirements for BIs are specified by the U.S. Pharmacopeia, ISO standards, and European norms.^9,10 Because the certifications are so important, some users attempt to verify the data presented in the certificate, but doing so can be challenging.

Not only does verifying certificate values require hundreds of samples and a BIER vessel, but the performance of different BIER vessels is likely to vary, even though they are required to operate within the precise limits defined by the AAMI and ISO standards.^3,7 Specifically, BIER vessels for EtO must comply with the following steady-state control limits: an operating temperature of 54° ± 1°C, an EtO concentration of 600 ± 30 mg/L, and a relative humidity (RH) of 60 ± 10%. BIER vessels are designed to produce "square-wave" control conditions, as depicted in Figure 1. This design is intended to provide come-up or gas charge times of zero in addition to nearly constant conditions at steady-state control levels. Gas charge and discharge times must be < 1 minute each to comply with the EtO BIER standards. The ability of a given vessel to remain within the target parameters is a function of instrument design, calibration, and preventative maintenance, as well as other aspects of quality system controls. Therefore, the confirmatory evaluation testing of BI resistance characteristics must be considered in light of operational BIER precision.

After reviewing the history of this issue, this article presents a mathematical model for predicting the variability of D-values based on BIER-vessel precision. The use of this model in lethality determination was described by the author in this magazine previously.¹¹ Some possible solutions for tightening BIER reproducibility are also discussed.

THE AAMI ROUND-ROBIN

AAMI conducted a round-robin test of BIER-vessel reproducibility coordinated by the FDA laboratory in Minneapolis and published in 1990.¹² The purpose of that study was to assess whether the medical industry would be able to comply with the then-new USP limits for D-value reproducibility using available BIER-vessel equipment.

The scope of the program included both steam and EtO BIER vessels and participation was unrestricted; that is, any company having a vessel that met the protocol requirements and willing to perform the rigorous test could take part. A majority of the vessels included in the study were manufactured by Joslyn Valve (Macedon, NY).

Results were analyzed statistically at the FDA lab to determine the primary factors affecting outcomes. Although media variability was part of the experimental design for the steam BIER portion of the study, all participants in the EtO BIER testing received presterilized soybean casein digest media from a single lot from BBL (Cockeysville, MD).

Based on the test data submitted by each of the 17 participants in the EtO portion of the study, four different D-value determinations were made for each participant, for a total of 68 individual D-value determinations and 17 mean D-values (Table I). The mean D-value was 2.50 minutes with a standard deviation of 0.23 minutes. The round-robin results also showed standard deviations of 0.022–0.202 minutes for individual BIER units, with an average of 0.100. As expected, interlab variability is greater than intralab and of more concern for the industry because the former is the source of confirmatory test results.

In the study report, the observed standard deviations were converted to a percentage of the observed mean D-values as follows:

The section on EtO BIERs concluded, "Although the ±18% variability achieved for all vessels at two standard deviations suggests that the ±20% specified by the USP is a good choice for release of production lots by manufacturers of BIs, a larger variability may be expected when various companies use the BIs with differing media. Thus it appears that repeating the ±20% of labeled D-value will be difficult to achieve throughout the user population."¹²

A review of data from the AAMI round-robin study reveals some of the possible effects of BIER-vessel imprecision:

Some device industry EtO users concluded from the report results that the Joslyn vessels had been proven superior because their operating precision was somewhat narrower than the totality of units tested, as measured by the BI results. However, the report itself did not make that conclusion, and, in fact, it is impossible to do so from the data presented. There was no data to confirm that Joslyn units were either better or worse than other designs at achieving and maintaining the desired target conditions. The sample size, or number of units of a particular design, would also influence measured variability.

What can be rationally concluded is that multiple units with a single product design would be expected to perform with more precision than a multiplicity of designs, given that the same levels of design control, calibration, and maintenance had been applied to all designs and vessels.

In this case, the designs varied considerably across all BIERs in the test program, and many were one-of-a-kind vessels. In addition, the Joslyn units were the only EtO BIER vessels widely marketed at that time. It should also be noted that precision is not an indicator of accuracy, and reliance on a single design type may not be scientifically prudent, lacking proof of its superiority.

ESTIMATING BIER PRECISION MATHEMATICALLY

A recently developed mathematical model makes it possible to predict the variability of BIER vessel precision.¹¹ The expected D-values when a BIER vessel is operating at the extreme limits of the acceptable ranges can be calculated using the equation:

where C is the standard EtO concentration and C_ref is a concentration at the extreme limit of the range, T is the standard temperature and T_ref is a temperature at the extreme limit of the range, and z is the temperature required to diminish the value of D by 90%. The use of 29°C for z is based on experimental data reported earlier.¹¹ For example, assuming a standard D-value of 3.5 minutes at 54°C and 600 mg/L, one can calculate the expected D-values at the operational extremes of C_ref = 630 mg/L and T_ref = 55°C as follows:

In investigating this approach, D-values for several ranges of expected performance were calculated based on various criteria. These results are presented in Table II, along with the percent deviation from the chosen standard D-value of 3.5 minutes, which is a common BI-certificate value. The choice of a standard D-value is irrelevant to the determination of percentage D-value variability, however, because that will be the same regardless of the D-value used in the calculations.

The concentration and temperature ranges specified in the AAMI standard for BIER operating parameters—600 ± 30 mg/L and 54° ± 1°C—were considered first. The range, based on the acceptable temperature-calibration tolerance of ±0.5°C, was calculated next, and finally the EtO gas tolerance claimed by AlliedSignal/Honeywell for Oxyfume 2002 of 10 ± 0.5% by weight was factored into the calculations. Subsequently, these same factors were applied to the AAMI round-robin test protocol ranges, where the temperature limits were 54° ± 2°C (per USP XXI). BIER vessels do not typically use pure EtO gas.

It should be noted that most BIER vessels are controlled by pressure. If EtO users do not request certifications of each tank's concentration from their gas vendor, or if they do not apply the accepted deviation from the target percentage to subsequent pressure control levels, then this level of uncertainty must be added to calculations of normal BIER performance variability. Even if tank certifications are supplied, users will rarely be informed about the tolerance of the measurement instruments used for the certification. Hence, this would be an additional unknown that would have to be factored into the equation in order to determine the expected limits of variability.

BIER vessel control systems typically provide active increases and passive decreases, which result in a skewed sine-wave control output. While no system controls temperature and pressure at the exact level specified, operation at the extremes is unlikely unless there is a problem with calibration or design. In other words, although the majority of BIER control systems will not maintain the temperature at exactly 54°C, they will control it in a very tight range around that specification. The same is true for pressure.

Thus, it is safe to predict that any BIER vessel in reasonable control would be able to operate at ±80% of the acceptable range limits. If the acceptable temperature range is 54° ± 1°C, one vessel in reasonable control could be expected to operate with a mean temperature as high as 54.8°C, while another might operate as low as 53.2°C. Table III recalculates the values from Table II based on an 80% limit of the BIER temperature control ranges. The same value (80%) was applied to the gas concentrations.

Equation 2 can also be used to determine the relative contribution to BIER-vessel variability of temperature and EtO concentration by performing two calculations, the first using the target concentration and a temperature extreme, and the second using the target temperature and a concentration extreme. For example, using the target EtO concentration of 600 mg/L and a T_ref of 53°C and then the target temperature of 54°C and a C_ref of 570 mg/L will yield the following D-values:

and

For those ranges, it is clear that temperature makes the greater contribution to variability.

Similarly, if the additional potential variability from the calibration tolerance for temperature (±0.5°C) and a 5% variance for the gas concentration from the supplier are factored into the calculations, the D-values can be determined as follows:

and

Inherent in the design of BIER-vessel operation is the concept of instantaneous or "zero" come-up. Although steam BIER-unit standards require < 10 seconds,⁸ EtO BIER units are allowed < 60 seconds.⁷ The same time of < 60 seconds is permitted for the discharge phase. When these times are considered in light of the recently published information on equivalent time for EtO and lag-correction factors,^11,13–15 two points become apparent: (1) this may be another variation in precision between different EtO vessels, and (2) failure to account for this time creates an additional error not considered in previous discussions. The magnitude of the error can now be estimated by calculating a D U correction factor (DU_cf), where U is equivalent time. If one considers the change in EtO concentration to occur at constant temperature and at a constant pressure rate change from zero mg/L to the target concentration over a 60-second period, then a reasonable estimation of D U_cf is t/2 as depicted in Figure 2, or (60 seconds)/2 = 30 seconds. Because the same error occurs both during charge and discharge phases, the total error is ~60 seconds, or 1 minute.

If a particular vessel performs these phases in less than 60 seconds, then the correction for that unit would be Dt/2 for each phase, where t = time of charge or discharge. (Note: Should the pressure ramp rate change during gas charge and appear asymptotic, then the error would be greater than 30 seconds during the charge phase (Figure 2) and would be less than 30 seconds during the discharge phase.) In short, each EtO BIER exposure of some time t_i, should have the DU_cf added to calculate equivalent time, or U. Table IV shows calculations of D-values for three lots of paper-strip BIs before and after adding a DU_cf of 1 minute. This correction to the D-value (DD) can be approximated by using a transformation of the Pflug and Holcomb formula for D-value calculations:

where N₀ = starting population, which is assumed to be 10⁶.

Depending on the original D-value and assuming values are normally rounded to the nearest 0.1 minute, this could result in an error of 0.2 minutes. Obviously, the percent error depends on the original D-value, but for values of 2.5–3.5 minutes the potential error is 6.4–4.6%, respectively, for DD = 0.16 minutes. In addition, if specialty BIs are being produced, the percent error increases as the population decreases. Therefore, the DD error becomes 0.31 minutes for a specialty BI with a population of 10³, which correlates to 12.4–8.9% error for D-values of 2.5–3.5, respectively.

Reducing the charge and discharge times for EtO BIER vessels is not recommended. Problems associated with microenvironments or gas compression and potential change of state for H₂O may be exacerbated by more rapid charge times. Using either the approach of adding a DU_cf, which is essentially a lag correction factor, or integrated lethality, will overcome the deficiency.

DISCUSSION AND RECOMMENDATIONS

The report on the AAMI round-robin study, which was published more than 10 years ago, suggested that the ±20% limit for BI D-value variability was reasonable as release criteria for BI manufacturers, but not as a mechanism for confirming the certified values.¹² Based on the predictions of BIER-vessel variability presented here, a ±20% limit for confirmation testing between two different BIER vessels still cannot be accomplished consistently using current technology.

This finding does not account for other affecting variables, such as culture media and incubation temperature, which would make such confirmation even less probable. It should be noted that although USP XXIV changed the temperature limits from 54°± 2°C in 2000, directly referencing the AAMI BIER-vessel standard limits of 54° ± 1°C, this had little impact because BI manufacturers were already in compliance with the more stringent AAMI standard.^7,8

One researcher believes that the precision of intralab D-value determinations has been enhanced by recent changes in the applicable standards, specifically, by the ISO requirement for 20 replicates per cycle during fraction/negative tests and the USP requirement for < 75% of the D-value as the difference between successive exposure cycles.^16–18

If there is a need for tighter limits of reproducibility, then one possible solution is to tighten the EtO BIER operating-temperature limits to 54° ± 0.5°C and the calibration tolerance to ±0.2°C. This latter value is in keeping with another researcher's recommendation regarding lag correction factors for thermal applications.¹⁹

Of course, maintaining control within stricter tolerance limits would require more-frequent calibrations of the BIER vessels' temperature-monitoring system and a thorough understanding of applicable variables.²⁰ Temperature control and consistency across the chamber is not as dynamic in EtO systems as it is in saturated-steam systems, because the latter have the advantage of the latent heat of moisture condensation to mediate thermal transfer. In EtO BIER systems, heat transfer will be affected by conduction and convection.¹³ Therefore, no change in operating limits should be implemented until it has been shown that the proposed limits are reasonably achievable given the systems' limitations.

With regard to EtO concentration limits, a sensor could be installed in the BIER-vessel chamber to monitor gas concentrations. This approach would replace the ±30-mg/L variability in the gas specification as well as the variation in concentration from the gas supplier with a single variable, the direct measurement tolerance. Again, more-frequent calibration of the BIER system would be required to maintain measurement precision.

A seemingly simpler approach would be for EtO BIER users to input the certified tank concentration data supplied by the gas vendor into their system for use in pressure control. However, this method places even more reliance on the accuracy of the certification from the gas supplier than current practice does.

In addition, it does not account for the fact that EtO gas concentrations change as a tank approaches empty, which can have a significant effect on BI test results. For example, one set of data in the AAMI round-robin study was eliminated because all of the BIs tested positive; a subsequent gas analysis showed EtO concentrations in the tank used in the testing that yielded those results were much lower than expected.

Another way to account for changes in EtO concentration and/or temperature is to calculate accumulated equivalent time using actual incremental temperature and EtO concentration data in the following equation:

Where U is the equivalent time at incremental time t, t_T is time at some temperature T, z is 29°C, T_i is temperature at incremental time t, T_ref is 54°C, C_i is EtO concentration at incremental time t, and C_ref is 600 mg/L. This approach actually simplifies the issues regarding BIER-vessel operating tolerances, because all variations are accounted for by using empirical data to calculate equivalent time. The accuracy of the equation then becomes a primary concern, as does instrument calibration.

Excluded here are the direct potential errors of pressure tolerances for operation or calibration. This is a complex area, and cycle or instrument design may control EtO concentration either by pressure differential or ranges about the set point for each cycle phase. These different approaches can result in different levels of error, and it is unnecessary to explore the various combinations to make the point.

In addition, the system design may or may not include these variables in the ±30-mg/L EtO concentration limit. If these variables are not included in this limit, then the pressure errors would be additive to those errors already discussed for gas concentration.

CONCLUSION

The equations presented here and in an earlier article were developed to fill gaps in the medical industry's knowledge of EtO sterilization lethality, with the goal of enabling cycle optimization and, eventually, parametric release of sterile devices. In the meantime, routine release and process validation rely on the use of commercial BIs, which are certified by their manufacturers in BIER-vessel tests. Therefore, the recognized variability in the operational precision of these vessels remains an important concern, which can be addressed in part by the estimation method and data in this article.

ACKNOWLEDGMENTS

The author wishes to acknowledge the reviews and helpful suggestions of John Gillis, Carl Bruch, and James Whitbourne.

REFERENCES

Gregg A. Mosley is president of Biotest Laboratories Inc. in Minneapolis.

Illustration by Bek Shakirov

Copyright ©2002 Medical Device & Diagnostic Industry