Content area
Ethylene oxide (EtO) sterilization validation programs are based on historical data, industry experience, and theoretical knowledge. When designing a specific program, the knowledge and experience of the individuals responsible often determines how accurately these data, experiences, and theories are applied. The assumptions and premises that form the foundation for the validation program are critical to the ultimate safety ofproducts distributed to the marketplace. Demonstrating appropriate microbial lethality is the ultimate purpose of sterilization validation. When using EtO processes, this is usually dependent on the proper use of biological indicators (Bls) in conjunction with the specific product, load, and cycle to be validated. The authors discuss the use of Bls, determination of microbial lethality, process controls, and variables related to these subjects in order to assist validation teams in designing successful validation programs, and/or problem solving when unexpected results are obtained.
Full text
Etbylene oxide (EtO) sterilization validation programs are based on historical data, industry experience, and theoretical knowledge. When designing a specific program, the knowledge and experience of the individuals responsible often determines how accurately these data, experiences, and theories are applied. The assumptions and premises that form the foundation for the validation program are critical to the ultimate safety of products distributed to the marketplace. Demonstrating appropriate microbial lethality is the ultimate purpose of sterilization validation. When using EtO processes, this is usually dependent on the proper use of biological indicators (BIs) in conjunction with the specific product, load, and cycle to be validated. The authors discuss the use of BIs, determination of microbial lethality, process controls, and variables related to these subjects in order to assist validation teams in designing successful validation programs, and/or problem solving when unexpected results are obtained.
(Biomedical Instrumentation & Technology 2005; 39:466-482).
Biological indicators (BIs) are commonly used as monitors of process lethality in ethylene oxide (EtO) sterilization during cycle development, validation, and routine release.1-4 For most industrial applications, some data in the quantal or fraction-negative (FN) region are desired outcomes of the test program. The appropriate times for use of population recovery or FN measurement methods during sterilization studies are shown in Figure 1, where the depicted lethality follows a classic log-linear relationship. It is recognized that the log-linear relationship applies to a monodisperse, homogenous population of a single species of microbe, as clearly explained by Rahn.5 In addition, for the relationship to be log-linear, the log of the population must be plotted against time (t) where the entire exposure period (x-axis; independent variable) must be at steady state conditions or plotted against equivalent time (U), based on methods of integration.6-10 Although log^sub 10^ is used most often, many early publications used In11,12 and the log base was irrelevant to the concepts, as shown in Figure 2, where values of N from various log bases are plotted on a linear scale. Each D-value of 1 minute in Figure 2 creates a 90% reduction in N (population), hence any log transformation plot of the respective populations yields a straight line. Use of FN methods assumes a log-linear relationship, based on first-order kinetics, and applies from the starting population (N^sub 0^, where t or U = 0) through the quantal region of the declination, where the samples used in the calculations are statistical replicates.13 Quantal data are often referred to as "fraction-negative," because for the calculation of Dvalues, the most common methods have established formulae based on the number of negatives14,15 compared with the total number of samples. However, in microbial test reports, it is common practice to list the number of positives over the total (i.e. 6/10 normally means six positives out of 10 total samples). We will use this latter format throughout the text.
The purpose of this publication is to provide guidance regarding the use of FN data from process sterilization test programs. Analysis of such data from validation or cycle development studies often is based on acceptance of the data carte blanche, without consideration of the measurement precision or the process variability. Most often, such tests are applied in (1) selection of a biological master product, (2) comparisons of minimum and maximum loads, (3) process D-value determinations, and (4) calculations of full-cycle process times.
A common practice for determining the biological master product (BMP) from among a family of similar products or a group of dissimilar products follows. Sample sets of each prospective candidate product are seeded with BIs, are placed in a production or simulated production vessel, and are exposed to a cycle time where some of the BIs are expected to be positive when tested for sterility. Once results are obtained in the quantal region, the BMP is selected, based on the product that has produced the highest percentage of positives.16 These results will be referenced as Example No. 1.
* Product A 7+/10
* Product B 5+/10
* Product C 3+/10
* Product D 1+/10
* Product E 0+/10
Many users would deduce that Product A is the most difficult to sterilize and would select this as the BMP based on these results.16 However, such a conclusion is based on many assumptions, and the conclusion is valid only if the assumptions apply to the specific test program used. As we will show, this conclusion is often erroneous. Several factors alone or in combination can affect the interpretation of these results. Affecting factors may be segregated as follows: (1) BI types and preparation, (2) measurement precision of BIs, and (3) control of process parameters.
These factors and how they can impact the results will be discussed. The concept of "control of process parameters" involves the disciplines of microbiology, mathematics, and engineering; dierefore, we have chosen to address this complex subject last. In addition, we identify some experimental design steps that will assist users in avoiding many of the common pitfalls and scientific deficiencies that may prevent a sound sterilization validation program.
BI Types and Preparation
BI Resistance
Graham and Boris19 demonstrated the effects of different packaging materials and different lots of the same packaging materials on measured BI resistance. For the data displayed in Table 1, one lot of inoculated paper strips was packaged in four lots each of either glassine or kraft paper envelopes. Decimal reduction values are shown for these eight final BI lots. The table also lists the theoretical number of positives required per hundred samples in order to produce each of those respective D-values using the most probable number (MPN) technique of Stumbo, Murphy, and Cochran (SMC).14 The final two columns display the range of positives per 100 samples that would produce those D-values at variances of ±20% or ±10%, respectively.
Caputo and Rohn20 demonstrated that use of different carrier materials resulted in different resistances using the same lot of Bacillus atrophaeus spore suspension, as shown in Table 2. As noted, the carrier material made a substantial difference in the measured resistance. Some of these differences are almost certainly due to variations in clumping21 or spore distribution22 resulting from surface variations and the hydrophobic or hydrophilic nature of some carriers. The observation on resistance variance is particularly important when one considers that commercially available spore suspensions typically are characterized for resistance measurements by placement on filter paper in glassine envelopes. Unless the user were to utilize the same lot of the same filter paper and the same lot of glassine envelopes, he or she could not expect the same resistance characteristics. For many process applications, such differences might not be significant or measurable.
Clumping
With gamma sterilization and, generally, with moist heat sterilization, the existence of a clump of organisms does not substantially affect the rate of delivery of sterilizing conditions to the center of the clump. However, in 1961, Bruch22 reported that prolonged resistance was observed for EtO sterilization if "uneven distribution of the microbial spore preparation was present." Gillis and Schmidt provided scanning electron micrographs,21 which demonstrated clumping and spore occlusion as one explanation for prolonged and inconsistent resistance during validation studies using inoculated product.
From a theoretical perspective, we recognize that microbial or organic clumps create a unique challenge for chemical sterilization. This is because the organic matter on the outside of the clump acts as a protective barrier between organisms on the inside of the clump and the chemical sterilizing agent outside the clump. Even when microbes on the outer edges of the clump are inactivated, they still have numerous molecular sites available for reaction with the sterilant chemical. With these chemicals, the sterilizing agent is consumed in each chemical reaction, whether a particular spore is or is not already inactivated. Hence, spore clumps or organic matter clumps containing spores may afford significant protection during chemical sterilization because the chemical sterilant molecules continue to be consumed by the organic matter on the outside of the clump. This affects the sterilant concentration within the clump and the related rate of inactivation of individuals within clumps. This has been demonstrated for bacteria, viruses, and protozoan cysts with a number of typical chemical sterilizing agents, including those generating addition reactions (e.g. EtO and formaldehyde), and oxidizing agents (e.g. ozone [O^sub 3^], hydrogen peroxide [H^sub 2^O^sub 2^], peracetic acid, and chlorine compounds), as well as complexing elements like iodine and bromine.23-29 In regards to large, infrequently occurring clumps of microbes (100 spores or more), a notable difference between the sterilization rates for microbial entities inside the clump, compared with those on the outside, is predicted. Sterilization by EtO requires the presence of both H2O and EtO, therefore, layers of the clump may impact penetration of either or both of these molecules.
This phenomenon is not true of moist heat sterilization, where H2O acts to transfer energy to cellular molecules (including cellular H2O), which generally increases their rotational or vibrational energy states. This increased energy is transferred to adjacent molecules via collision. Hence the effect is not lost, but merely passed along and diluted until irreversible damage to various molecules has occurred. The occurrence of a clump may slightly slow this transfer of energy to microbes within the clump, but in a fully hydrated system transfer, it still occurs very rapidly.
For those using inoculated product challenges, especially if using in-house, prepared spore suspensions, the cleanliness of the spore crop is critical. Historical publications have shown the problems that can be created by carryover of media constituents (organic and inorganic), as well as cellular debris from the sporulation medium.19,30-32
Most Difficult to Sterilize Challenge
The concept of "most difficult challenge"1,2 for products or product locations is a major concern for cycle development of validation programs. Many test protocols are established to determine the most difficult to sterilize product or location; different types of BIs are placed in the different products or locations, and then the relative resistance is compared after exposure in a biological indicator evaluator resistometer (BIER) vessel or some larger, but nonproduction, vessel. One question is, "Can such a test series determine the most difficult to sterilize product?" Usually the answer is no! Only if exactly the same spore lot, on the same carrier, in the same package material and lot were used, and then only if these were tested in the same sterilizer, at the same time, in equivalent locations7-9,12 could one determine the most difficult to sterilize product or product location. Because different lots of spores and different lots or types of packaging materials or carrier materials all affect BI resistance measurements, we recognize that any changes in the BIs used from one product or location to another will affect the comparative results. However, the next question might be, "Does it matter?" The answer to this question may be that it does not matter, but that depends on how the information is to be used.
For instance, if one has determined that each BI/product combination is a reasonable simulation and that it can be defended scientifically, then it does not matter that they are not directly comparable. The biological challenge, then, has become the BI/product combination. The type of BI cannot be changed without producing somewhat unpredictable changes in relative resistance. That is, if one has determined that the BI/product A is the most difficult challenge and a BI strip in glassine was used, then one could expect that if a second BI lot was used that was of the same type from the same manufacturer but had a higher D-value, then the BI/product combination would yield a higher process D-value. However, if one had decided to use direct product inoculation from a liquid suspension as a replacement for the BI strip, the relative results could not be predicted. The new combination could be more or less resistant than the original. And, in fact, the new BI/product A combination might not prove to be the most resistant, compared with the other BI/product combinations used in the original study. Therefore, it is important that the user understand what has and has not been proven, in order for the information to be applied properly. Once a BMP has been selected, the type of BI used in the BI/product combination cannot be changed without affecting the expected relative resistance. The lot or supplier of the original BI type could be changed, and the overall BI resistance in the supplied BI should create a similar shift in the resistance of the BI/product combination. Figure 3 illustrates several different types and sources of BIs.
Consideration of Load Effects and Most Difficult to Sterilize Challenge
In one case, a manufacturer had determined both the BMP and the physical master product (PMP) from an extensive product line. The former was determined to be a small device with a piece of tubing, several inches long with a very small diameter, attached to a terminal balloon; the packaged product weighed a few grams. Comparative testing of different products was conducted in a small research and development vessel using a few samples of each product type. The PMP was a bucket of PVC tubing, some 40 feet long weighing several pounds. The quality department had decided to use the "worst case" and to validate the process cycle by placing BMP samples inside the PMP container, in pallets of the PMP. The result was that a 10^sup -6^ sterility assurance level (SAL) could reasonably be achieved only by using a combined bioburden/BI validation approach that resulted in a 13.5 hour EtO exposure period to produce a 9 log spore log reduction (SLR) for the process. By contrast, when the validation was reconsidered, inoculated products of both types were used in an overkill approach (half-cycle method), with inoculated samples in pallets of like product. The result was that both products demonstrated a 12 log reduction in a 4.5 hour EtO exposure cycle where all other conditions were the same as in the original validation. In their efforts to create a "worst case" challenge, this manufacturer had created an unnecessarily difficult one. The results were higher sterilization costs, higher product residuals, and longer inventory quarantine times.
Oftentimes, planning to inoculate 106 spores in the most difficult to sterilize location, of the most difficult to sterilize product, is both unwise and unnecessary. The product bioburden and its distribution on the product should be considered. For the reasons identified by Bruch22 and Gillis and Schmidt21 (i.e. uneven distribution, occlusion, and clumping), one should avoid direct product inoculation whenever possible, but particularly when high numbers of spores are used (i.e. >10^sup 4^). In any event, direct inoculation usually is not representative of the normal distribution of bioburden on the product, and almost certainly, the total bioburden would both be less than 106 and not all concentrated at the most difficult to sterilize location. Spreading the inoculum over several locations often is preferable to placing the entire inoculum in one location. For direct inoculation, one may want to consider not using more than 10^sup 3^, unless experience has shown higher levels to provide a predictable and nonproblematic challenge.
Measurement Precision of Biological Indicators
Table 3 represents the data from three lots of BI strips in glassine envelopes produced by one BI manufacturer using the same lots of paper and glassine for each. The BI lots were tested approximately two weeks apart in the same BIER vessel and using the same lot of outgrowth media. Ethylene oxide BIER vessels operate within precisely defined operating parameters4,16 and produce square wave charge and discharge processes (Figure 4). Table 3 depicts results that are not exactly the same from this series of replicate tests.
Table 4. Predictions of biological indicator evaluator resistometer (BIER) vessel performance variability, based on extremes of the acceptable ranges.
By analyzing the data, several other conclusions may be made. First, the D-value variation is quite small compared with limits published by US Pharmacopeia (USP)33,34 or analysis of BIER vessel variance noted by Oxborrow35 and Mosley36 for interlab variation. In addition, it has been reported that intralab variance of ±5% to ±10% is reasonably achievable, assuming well-designed quality system controls.7 In considering only those cycle pairs where at least one of the replicate cycles yielded fractional results, we note the following:
* There are 14 such replicate pairs.
* In three replicate pairs, the same results out of 10 total samples were obtained both days: eight replicate pairs differed by one positive, and three differed by two positives. Therefore, when 10 samples were exposed, 78.6% of the time, replicate exposures resulted in a difference of one or more positives, and 21.4% of the time, a difference of two was noted.
* The mean from all 14 cycles of the absolute values of the differences is 1.
* The standard deviation (SD) of these differences is 0.679.
* The mean + 2 SD = 2.368, and the mean + 3 SD = 3.037. This suggests that in order to account for expected differences 95% of the time, a range of ±2.4 positives would be required, and in order to account for expected differences 99.7% of the time, a range of ±3 would need to be applied.
* The latter value correlates to an expected range of 2/10 to 8/10 when the target is the center of the quantal range for 10 samples or five positives.
* At a 95% confidence level, we expect a range of slightly less than three of 10 positives to slightly more than seven of 10 positives.
The number of positives resulting from an exposure dose in the quantal range for a true set of replicate samples is a function of probability alone, as described by Rahn,5 and such variances are expected; therefore, they are not indicative of a lack of precision of the BI. The log-linear declination is not a function of variation of resistance across the population.5,6,12
Returning to Table 1, one can see the range of positives that could be considered significant based on a ±10% or ±20% D-value variation. It is clear that BIs tested under BIER vessel control conditions will produce variance that would cause one to reconsider the conclusion about Product A in Example No. 1. In fact, products A, B, and C have no statistically significant differences, and if the average of positives is considered for the quantal results (16/4 = 4), then product D also would be included in this group. A single set of comparative data is statistically inadequate, especially in light of the fact that the entire subsequent sterilization process validation may be based on these data. The preceding statement does not consider variations in vessel location and other process variables to be considered in the next major section. If all other variables in the design of the experiment (DOE) are well controlled and there is a major difference in displayed resistances, then a single experiment might allow scientifically discriminating results, but as a general rule, multiple exposures are more likely to be required.
Culturing Variables
Several publications have reported on the effects of different types of media and different incubation temperatures on resistance results for both steam and EtO BIs.19,35,37-40 In general; lower temperatures produced more survival from quantal test results (i.e. greater apparent resistance), and the effect of media varied with the type of BI organism. Significant differences were seen for both Geobacillus stearothermophilus and Bacillus atrophaeus. Although these variables will not be discussed further in this article, to produce truly comparable data, one must consider these factors in the DOE. If the growth media is an intrinsic part of the BI test system, as with self-contained BIs, then that is the media that must be used in conjunction with the BI's manufacturer recommendations for growth conditions. If the user is to supply the culture media, then the same lot from the same supplier should be used to produce truly comparable test results.
Control of Process Parameters
When preparing for any comparative resistance testing, one must consider the control of the critical process parameters. For instance, AAMI Technical Information Report (TIR) 20 states, "...conditions are expected to be consistent from cycle to cycle." This statement may be governed by variations as wide as the operating ranges allowed in ISO 11135 and EN 550,1,2 or as wide as may be justified by each individual manufacturer. However, understanding what these variations can produce within a given cycle9 and from one cycle to another are critical in order to appropriately compare BI test results.
Resistance Measurements
The D-value calculations for EtO have been used primarily in BIER vessel studies where "time to steady state" approaches zero and "equivalent exposure time approaches actual exposure time."7,35,36 The problem with applying any D-value method to EtO processes for product sterilization is that production vessels do not produce square wave cycles and do generate substantial lethality during both the sterilant charge phase and primary discharge or sterilant gas evacuation phase, as illustrated in Figures 4, 5, and 6. The sterilant charge phase also may be referred to as the "come-up" or "gas injection" phase.
ISO document 11135, annex B (B.7), lists the method of Pflug and Holcomb13 for calculating process D-values. Unfortunately, the standard provides little guidance to assist users in estimating "equivalent time" for such calculations. In the extreme, use of "exposure time" rather than "equivalent time" may lead to gross underestimations of D-value and concomitant overestimation of process SLR, and hence, underestimates of process SAL.7-9 Figure 1 demonstrates the relationship between D-value, SLR, and SAL (probability of a nonsterile unit, or PNSU) at steady state, where microbial inactivation follows a straight-line log-linear relationship. It is important to note that the method of D-value calculation is irrelevant to the problem. Whenever equivalent time (U) is underestimated for D-value calculations, the result will be the same.
Most would consider the operation of BIER vessels to be very precise, in comparison to process vessels. However, the AAMI round-robin test of industry BIER vessels demonstrated that BIs detected considerable differences in the accuracy, precision, and bias of such vessels.35 Oxborrow et al 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. It appears that repeating the ±20% of labeled D-value will be difficult to achieve throughout the user population." In order to analyze the variables of process control for EtO sterilization, recently developed formulae for integrated lethality and D-value comparisons can be used.7 These formulae allow integration of both varying EtO concentrations and temperatures during both dynamic and the relatively steady state control conditions.
Formulae for Calculations of U and D at Varying Temperatures and EtO Concentrations-
Table 5. Predictions of process EtO vessel performance variability, based on extremes of acceptable ranges.
Using the above formulae, allowed variances in EtO BIER vessel operation have been reviewed and have been shown to impart significant variability to measured resistance.36 These values are comparable to those reported from the AAMI round-robin study35 on BIER vessels. Table 4 demonstrates the expected variation in resistances based on allowed operating ranges for EtO BIER vessels. Such variations are substantially less than those expected for production vessels. As shown in Figure 4, BIER vessels produce square wave controls, minimizing the variations of come-up, steady state control, and discharge. Table 5 shows the possible effect on microbial lethality of such variations, based on some common tolerances for temperature and gas concentrations used in EtO production sterilization cycles, which are much greater than those allowed for BIER vessels.
Production Vessel Location and Accumulated Lethality
The development of formulae for calculations of integrated lethality and equivalent time in EtO sterilization can provide a more thorough analysis of cycle development and validation data based on both physical and biological results.
The results for equivalent time and lethality follow patterns similar to those typical for thermal models where temperatures differ across the load. The determination of SAL for the load should be a function of the lowest SLR determined from thermal mapping during the process validation. As noted by Pflug et al,13 "In the analysis of fraction-negative (FN) data, only those units, n^sub i^, that are true replicates, can be used in the analysis. If, for example, 20 spore-strip biologic indicators are scattered at 20 different locations throughout the autoclave, there may be no true replicates: n is one!" These same principles apply to EtO sterilization where both temperature and EtO concentration may vary.7-10 Figure 5 shows a typical EtO sterilization cycle using a deep vacuum, pure EtO, and an N^sub 2^ overlay gas injection. As depicted in Figure 6, accumulated lethality will be affected at different locations based on temperature variations across the load, whether the differences remain at some minimum ΔT or whether they become asymptotic. These physical measurements can be monitored directly using temperature probes.
In the hypothetical example depicted in Figure 6, four thermocouple load locations are monitored for a process with target conditions of 50°C and 600 mg/L EtO. The data assumes a process D-value of 4 minutes and demonstrates accumulated lethality (SLR). The data from thermocouples 1 to 4 show the effects of temperature differences over the gas injection stage (30 minutes), and the following 50 minutes at four locations: No. 1 starts at 40°C and increases to 50°C, No. 2 starts at 35°C and increases to 45°C, No. 3 starts at 30°C and increases to 40°C, and No. 4 starts at 30°C and increases to 50°C. The first three locations may occur where temperature remains stratified over this exposure period. The rates for L^sub R^, and hence D-value, are different at the three locations over the entire period, and the difference in accumulated lethality (SLR) increases with time. The fourth location develops a difference from location No. 1 of 2.5 logs of SLR, but the rate of L^sub R^ becomes parallel as the temperature reaches the same target of 50°C.
However, EtO concentration can vary within products and across the load for some product/load/cycle combinations.7-10,43,44 These EtO conditions stabilize later in the process cycle, but they cannot be measured directly at small locales by physical monitoring techniques currently available to the industry. Differences in EtO concentrations could make the examples in Figure 6 closer or further apart, depending on the concentration differences. The only measurement system capable of integrating all physical parameters responsible for lethality is a BI. Therefore, the biological results from the validation are critical to overall understanding of lethality across the product load.
One solution to avoid the condition noted by Pflug (n is 1), is to place multiple BIs at each location during the validation. When locations demonstrating slower-to-accumulate lethality are identified, these can be probed with additional BI samples for accurate D-value calculations (e.g. 5-10 per location). Determination of process D-values should be based on the PNSU throughout the entire load, rather than in a specific load location. If lethality is mostly a function of position, then each location will have a separate D-value, and statistical methods based on probability cannot be applied correctly to dissimilar samples using data from all locations as a single set.9
The Importance of H2O in EtO Sterilization
The single most critical parameter in EtO sterilization is H2O, and for predictable and controlled sterilization to occur, H2O must be present at adequate levels. The exact boundaries for both the upper and lower Relative Humidity (RH) limits have not been agreed upon universally. At the lower boundary, values of 30% to 50% RH have been referenced in the literature as necessary for predictable microbial inactivation, whereas 80% to 85% RH is considered reasonable for the upper limit. Once RH is within the acceptable window for a product and process, increases or decreases in RH within this window do not produce measurable changes in microbial resistance. The boundaries are affected by temperature, load materials, and specific cycle dynamics. Both low RH and dew point conditions during EtO exposure phases can produce changes in microbial lethality that are difficult to predict quantitatively. Lower and higher RH levels can produce dramatic and quantum increases in measured microbial resistance, especially for bacterial endospores. At very low RH conditions, sterilization by EtO may not be able to be accomplished in any practical time frame. Additionally, high RH levels for some products can result in damage to the product or packaging.
Gas Stratification in Process Cycles
When use of the old 12:88 (EtO/CFC 12) gas mixture was common, and before preconditioning became widespread, stratification in process sterilization was not uncommon. This generally occurred when the gas mixture encountered cool product and the two gases separated due to static density differences. The heavier CFC 12 molecules would become predominant at lower levels, whereas the lighter EtO would become predominant at higher levels. When preconditioning became commonplace and the times for this phase increased, these stratification effects were largely eliminated because the product encountering the sterilant gas mixture was at higher temperatures. For those cycles using deep vacuums and only pure EtO gas, the stratification effect does not occur because significant levels of other gases are not present.
However, when the use of N^sub 2^ overlays became widely used, a unique form of stratification was detected by colleague Jon Hoogenakker. The use of N^sub 2^ overlays was implemented in some sterilization cycle designs to reduce the possibly of explosion from the EtO gas, which is flammable and explosive in air mixtures from about 3 % to 97%. In order to minimize chances of a spark-induced event, some contract sterilizer companies did not initiate the chamber circulating systems until all the N^sub 2^ had been injected into the chamber. This equated to an exposure time (t) of O minutes although the U could be significant, often 20 to 30 minutes. Because N^sub 2^ is lighter than EtO and is injected last, it tends to produce higher concentrations of N^sub 2^ higher in the vessel, displacing the EtO to the lower levels. Hoogenakker detected what he described as a "watermark" effect during validation studies. At 0-minute exposure cycles, he found all negative BIs below the watermark, fraction results within it, and all positives above it. The studies were repeated with the circulation beginning during the N^sub 2^ injection phase, and all BIs from the same 0-minute cycle were negative regardless of their vertical location in the load. Conditions of gas mixture nonequilibrium also are common immediately after the charge of N^sub 2^, due to compression effects of the different charging gases on any preexisting gases. These conditions are not considered stratification, but still may occur across the load, pallet, boxes, or within product.45,46
Tailing and Apparent Tailing
This subject is too extensive to be covered adequately as a subsection of this manuscript, therefore, the comments herein will be somewhat cursory in nature. We refer readers to a recent publication that discusses the scientific deficiencies in ISO 11135 and EN 550 regarding lag factors9 and the apparent existence of tailing. A true log-linear response requires steady state EtO sterilant conditions. The combination of process vessels, product loads, and the respective required cycles make steady state conditions throughout the exposure of product to EtO impossible.
Tailing is a phenomenon whereby the microbial decay during an exposure cycle results in a concave upward curve, as shown in Figure 7. True tailing in EtO sterilization often is encountered when direct inoculation of products with spore suspensions is used, resulting in clumps.21,22 In addition, inadequate humidity in the load materials and/or the spores on the BIs may produce inordinately high resistance and tailing.31,45 It is sometimes possible to detect tailing using FN testing if one observes sporadic positives that continue over ever-increasing exposure times. The concern is that if tailing has occurred, but SAL levels are extrapolated based on a loglinear line, the SLR for the process may be overestimated and the SAL may be underestimated incorrectly. In other words, the full-cycle process time may be too short, because all process sterilization parameters have not been quantified accurately. Recent publications7-9 allow the industry to overcome this deficiency and have identified errors in some commonly accepted validation methods.
The subject of lag factors has been discussed for thermal sterilization methods,41,42 for steam BIER vessel errors at high temperatures,46 and process EtO cycles.7-10 In EtO sterilization, lethality begins as soon as EtO enters the chamber. The rate of lethality increases as EtO concentration increases and as temperature rises. However, timing of the gas-dwell phase of the cycle does not begin until all EtO gas has entered the chamber and, for some cycles (see Figures 5 and 6), after an N^sub 2^ overlay of gas has been added. As previously noted, this front end of the cycle, as well as the exhausting (terminal) end of the cycle, can impart significant lethality. In fact, for many products, no surviving BIs can be obtained in a 0-minute exposure due to the lethality imparted by these cycle phases. Figure 8 shows the effect of such a lag factor, and how it may result both in apparent tailing and in underestimation of the correct half-cycle exposure period. The latter error can result in failure to accurately calculate the necessary full cycle time as shown.
Note: This assumes pressure rise and decline is steady over the identified EtO introduction and exhaust phases, hence pressure rise and theoretical EtO concentration is directly proportional to time.
This approach always overestimates the lag factor (U^sub 1^), as seen using Formula 1 or 2, because actual temperatures are lower than the target, steady state temperature.
Conclusions and Recommendations
In the beginning, we identified three factors that could affect one's ability to correctly interpret and apply EtO sterilization validation studies: (1) BI types and preparation, (2) measurement precision of BIs, and (3) control of process parameters.
The subsequent conclusions and recommendations may be a function of one or more of these factors.
1. There are many DOE factors to be considered when embarking on an EtO sterilization validation project. Understanding the variables before initiating the project often helps avoid numerous potential problem areas. Previous experience with the same or similar products, processes, sterilizers, and biological monitors will often prove the best reference sources.
2. When designing experiments to compare BI results for different products or product locations, one should consider running the products in a vessel that allows placement of the product samples in boxes of the same products and as similar to the production load shipper boxes as possible. Load considerations may prove essential for selecting the correct BMP and final cycle time. If it is feasible, conduct comparative studies in the production vessel during the fractional cycle(s), and expose all products at the same time and in equivalent positions whenever possible. Once a BI/product challenge is determined, remember the type of BI initially used cannot be changed without producing potentially unpredictable results.
Figure 8. Half cycle method errors when calculations are based on exposure time (t) rather than equivalent time (U).
3. If FN results are not significantly discriminating for selection of the BMP or process challenge device, repeat the exposure and recognize that sometimes one BMP may be statistically no different from alternates. Also, consider supplementing FN data with population enumeration data from fractional exposure cycles. However, if very short exposure cycles produce few or no positive results, then perform cycles that do not reach full EtO concentration levels, such as half-charge cycles. Approximate equivalent process times can be calculated using Formula 1, 2, or 4.
4. Comparison of results from one vessel run to another can prove misleading unless the critical physical process variables are carefully reviewed and differences quantified. Comparisons of exposure times (t) and correlating results can prove to be of little value, even in the same vessel, when variations in temperature, EtO concentration, and humidity levels are ignored.
5. Results today may not be comparable with results from last year without carefully controlling test media, incubator conditions, and BI types and lots. The relative impact of each factor should be considered.
6. In order to comply with ISO 11135 or EN550, Methods A or B, validation process parameters (gas concentration and temperature) must be run at the values that will test the lower limits of the accepted control ranges. Although this approach is not required, it also may be a reasonable strategy for the half-cycle method (Method C). Many existing EtO process cycles are much longer than necessary and fine-tuning the cycle while running at the minimum conditions may not produce a longer or more expensive cycle. This approach of testing the lower control limits will produce greater confidence that products will meet their intended SAL attributes.
7. The single most critical parameter in EtO sterilization is H2O. When unexpected BI failures occur during routine sterilization activities, it may be that possible RH extremes in the load, often affected by seasonal changes, were not accounted for and tested during validation activities.
We would like to thank John Gillis for providing BI results from BIER vessel testing, John Hoogenakker for sharing EtO production vessel stratification information, and Carl Bruch for his review and comments on portions of this document.
References
1. International Organization for Standardization (ISO). Medical Devices-Validation and Routine Control of Ethylene Oxide Sterilization. Geneva, Switzerland: ISO; 1994. ISO 11135.
2. European Committee for Standardization (EN). Sterilization of Medical Devices-Validation and Routine Control of Ethylene Oxide Sterilization. Brussels, Belgium: EN; 1994. EN 550.
3. International Organization for Standardization (ISO). Sterilization of Health Care Products-Biological Indicators-Guidance for the Selection, Use, and Interpretation of Results. Geneva, Switzerland: ISO; 2000. ISO 14161.
4. International Organization for Standardization (ISO). Sterilization of Health Care Products-Biological Indicators-Part 2: Biological Indicators for Ethylene Oxide Sterilization. Geneva, Switzerland: ISO; 1994. ISO 11138-2.
5. Rahn O. Injury and Death of Bacteria by Chemical Agents. Normandy, MO: Biodynamica; 1945. Biodynamica Monograph, No. 3. Republished in: Pflug IJ, ed. Selected Papers on the Microbiology and Engineering of Sterilization Processes. 5th ed. Minneapolis, MN: Environmental Sterilization Laboratory; 1988: 1-16.
6. Pflug IJ. Microbiology and Engineering of Sterilization Processes. 11th ed. Minneapolis, MN: Environmental Sterilization Laboratory; 2003.
7. Mosley GA, Gillis J, Whitbourne J. Calculating equivalent time for use in determining the lethality of EtO sterilization processes. MDDI. 2002;24(2):54-63.
8. Mosley GA, Gillis JR, Whitbourne JE. Formulae for Calculations of Integrated Lethality for EtO Sterilization Processes, Refining the Concepts and Exploring the Applications." Pharm Technol. 2002;26(10):114-134.
9. Mosley G A and Gillis J R. Factors affecting tailing in ethylene oxide sterilization. Part 1: when tailing is an artifact...and scientific deficiencies in ISO 11135 and EN 550. PDA J Pharm Sci Technol. 2004;58(2):81-95.
10. Mosley GA, Gillis JR Krushefski G. Evaluating the formulae for integrated lethality in ethylene oxide sterilization using six different endospore forming strains of bacteria, and comparisons of integrated lethality for ethylene oxide and steam systems. PDA J Pharm Sci Technol. 2005;59(1). 64-86
11. Chick H. An investigation of the laws of disinfection. J Hygiene. 1908;8:92-158.
12. Lee Richard E, Gilbert CA. On the application of the mass law to the process of disinfection-being a contribution to the "Mechanistic Theory" as opposed to the "Vitalistic Theory." J Phys Chem. 1918;22(5):348-372.
13. Pflug IJ, Holcomb RG, Gomez MM. Thermal destruction of microorganisms. In: Block S, ed. Disinfection, Sterilization and Preservation. Philadelphia, PA: Lippincott, Williams & Wilkins; 2001:79-129.
14. Stumbo CR, Murphy JR, Cochran J. Nature of thermal death time curves for PA 3679 and Clostridium botulinum. Food Technol. 1950;4:321-326.
15. Holcomb, RG, Pflug IJ. The Spearman-Karber method of analyzing quantal assay microbial destruction data. In: Pflug IJ, ed. Selected Papers on the Microbiology and Engineering of Sterilization Processes. 5th ed. Minneapolis, MN: Environmental Sterilization Laboratory; 1988;83-100.
16. Booth AF. Sterilization Validation & Routine Operation Handbook-Ethylene Oxide. Technomic Publishing Co Inc; 2000. Lancaster, PA
17. Association for the Advancement of Medical Instrumentation (AAMI). BIERIEO Gas Vessels. Arlington, VA: AAMI; 1992. ANSI/AAMI ST44.
18. Association for the Advancement of Medical Instrumentation (AAMI). Parametric Release for Ethylene Oxide Sterilization. Arlington, VA: AAMI; 2001. AAMI TIR20.
19. Graham GS, Boris CA. Chemical and biological indicators. In: Morrissey RF, Phillips GB, eds. Sterilization Technology: A Practical Guide for Manufacturers and Users of Health Care Products. New York, NY: Van Nostrand & Reinhold; 1993;36-69.
20. Caputo RA, Rohn KJ. The effects of EtO sterilization variables on Bl performance. MDDI. 1982;4(7):37-42. Republished in: Pflug IJ, ed. Selected Papers on the Microbiology and Engineering of Sterilization Processes. 5th ed. Minneapolis, MN: Environmental Sterilization Laboratory; 1988:281-286.
21. Gillis JR, Schmidt WC. Scanning electron microscopy of spores on inoculated product surfaces. MDDI. 1983;5(6):46-49.
22. Bruch CW. Gaseous sterilization. Annu Rev Microbiol. 1961;61:245-262.
23. Haas CN, Karra SB. Kinetics of microbial inactivation by chlorine. Water Res. 1984;18(11):1443-1449.
24. Wickramanayake GB, Sproul OJ. Ozone concentration and temperature effects on disinfection kinetics. Ozone Sci Eng. 1987;10:123-135.
25. Chen YSR, Sproul OJ, Rubin AJ. Inactivation of Naegleria gruberi cysts by chlorine dioxide. Water Res. 1985;19(6):783-789.
26. Rubin AJ, Engel JP, Sproul OJ. Disinfection of amoebic cysts in water with free chlorine. J Water Pollut Control Fed. 1983;55(9):1174-1182.
27. Gard S. Chemical inactivation of viruses. In: The Nature of Viruses. Boston, MA: Little Brown & Co Inc; 1957:123-146.
28. Young DC, Sharp DG. Poliovirus aggregates and their survival in water. Appl Environ Microbiol. 1977; 1:168-177.
29. Berman D, Hoff JC. Inactivation of simian rotavirus SA11 by chlorine, chlorine dioxide and monochloramine. Appl Environ Microbiol. 1984;48(2):317-323.
30. Royce A, Bowler C. Ethylene oxide sterilisation-some experiences and some practical limitations. J Pharm Pharmacol. 1961;13:87T-94T.
31. Doyle JE, Ernst RR. Influence of various pretreatments (carriers, desiccation and relative cleanliness) on the destruction of Bacillus subtilis var niger spores with gaseous ethylene oxide. J Pharm Sci. 1968;57:433.
32. Perkins JJ, Lloyd RS. Recent developments in the sterilization of surgical materials. In: Report of a Symposium. London, UK: Pharmaceutical Press; 1961:76-90.
33. US Pharmacopeia. Biological Indicator for Ethylene Oxide Sterilization, Paper Strip. 24th ed. Rockville, MD: United States Pharmacopeial Convention; 2000.
34. US Pharmacopeia. (55) Biological Indicators-Resistance Performance Tests. 24th ed. Rockville, MD: United States Pharmacopeial Convention; 2000.
35. Oxborrow GS, Twohy CW, Demitrius CA. Determining the variability of BIER vessels for EtO and steam. MDDI. 1990;12(5):78-83.
36. Mosley GA. Estimating the effects of EtO BIER-vessel operating precision on D-value calculations. MDDI. 2002; 24 (4): 46-52.
37. Davis SB, Carls RA, and Gillis JR. Recovery of sublethal sterilization damaged Bacillus spores in various culture media. In: Underkofler LA, Wulf ML, eds. Developments in Industrial Microbiology. Vol. 20. Arlington, VA: Society for Industrial Microbiology; 1979:427-438.
38. Nelson FE. Factors which influence the growth of heat treated bacteria. I. A comparison of four agar media. J Bacteriol. 1943;45:395.
39. Boris C, Graham GS. The effect of recovery medium and test methodology on biological indicators. MDDI. 1985; 10(2):43-48.
40. Shintani H, Sasaki K, Kajiwara Y, ltoh J, Takahashi M, Kokubo M. Validation of D-value by different SCD culture medium manufacturer and/or different SCD culture medium constituent. PDA J Pharm Sci Technol. 2000;54(1):6-12.
41. Pflug IJ. Measuring the thermal resistance of microorganisms: selecting an appropriate test system, correcting for heat-transfer lags, and determining minimum heating times. PDA J Pharm Sci Technol. 2003;57( 3): 60-185.
42. Jaynes JA, Pflug IJ, Harmon LG. Some factors affecting the heating and cooling lags of processed cheese in thermal death time cans. J Dairy Sci. 1961;44(12):2171-2175.
43. Zhu Z, Matthews IP. Process dynamics of EtO sterilization. MDDI. 2003;25(3):66.
44. Mosley GA. Letter to the editor regarding article: Process dynamics of EtO sterilization. MDDI. 2003;25(6):26-27.
45. Kaiser U, Heider D, Gömann J, Junghann U. Kill kinetics study of Bacillus subtilis spores in ethylene oxide sterilisation processes. Zentr Steril. 2002; 10:163-167.
46. Mosley GA, Gillis JR. Operating precision of steam BIER vessels and the interactive effects of varying Z valves on the reproducibility of listed D-values. PDA J Pharm Sci Technol. 2002;56(6):318-331.
From Biotest Laboratories, lnc, Minneapolis, MN (GM); and STERIS Isomedix Services, Mentor, OH (CH).
Address correspondence and reprint requests to Gregg A. Mosley, Biotest Laboratories lnc, Minneapolis, MN 55433 (e-mail: [email protected]).
Gregg A. Mosley is president of Biotest Laboratories, Minneapolis, MN. He is a frequent lecturer and author on sterilization and microbiology.
Clark W. Houghtling is corporate account manager and senior EO technical specialist with Steris Isomedix Services, Mentor, OH. His background includes extensive knowledge in sterilization technology, plant and quality management, finished goods testing, medical device manufacturing, sales and marketing, and regulatory affairs.
Copyright Alliance Communications Group, A Division of Allen Press, Inc. Nov/Dec 2005