Comparison of Different Incubation Conditions for Microbiological Environmental Monitoring

Discussion

In our in situ approach we observed distinct differences in recovery of total aerobic microbial count or mould count between the incubation conditions of interest. Recovery of total aerobic microbial count was most efficient at elevated temperatures, particularly if samples were taken from gowns of personnel (Figure 2A). We considered this observation reasonable—most microorganisms on gowns are rather derived from personnel than the environment, and microorganisms colonizing humans should be adapted to preferentially grow at temperatures close to the skin temperature of approximately 35 °C (9). The observation that total aerobic count from air and inanimate surfaces also showed a trend towards superior recovery at elevated temperatures may seem less straightforward (Figure 1A). Microorganisms naturally occurring in the environment are generally not expected to prefer elevated temperatures for replication because such temperatures are not frequently encountered in a central European climate zone. For instance, microbial recovery from water, which can be a source of microorganisms in a cleanroom (10), was shown to be more efficient at lower temperatures (11). Inhibited growth of moulds at elevated incubation temperatures, and thus less overgrowth of the bacterial colonies, may offer a possible explanation. However, the intermediate readings throughout the incubation period probably mitigated such an effect. Furthermore, the low numbers of recovered moulds suggest that overgrowth had no significant contribution to this observation. Alternatively, the observed preference is probably attributed to the fact that a good proportion of microorganisms in a pharmaceutical cleanroom is actually shed by personnel (12, 13). Further support for this explanation can be derived from a preliminary study we conducted at our site (Figure 5). Whereas the samples of the main study originated from production and packaging rooms that were in operation, samples of the preliminary study were collected in a secluded area without personnel flow. In the preliminary study, no indication for superior recovery of total aerobic microbial count at 30–35 °C as compared to lower incubation temperatures was observable (Figure 5A), whereas such a trend was evident in the main study (Figure 1A). Thus, the presence of personnel and the resulting shedding of human-associated microorganisms in the environment may have been the underlying cause for the superior recovery of total aerobic microbial count from air and inanimate surfaces at elevated temperatures.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Microbial recovery results from a preliminary in situ study. Data was collected in a secluded location without personnel flow. Total incubation time was 5 days and used nutrient medium was TSA. Recovery of total aerobic microorganisms excluding moulds (A) and moulds (B) is shown. Moulds were subtracted from total aerobic microorganisms count because of their untypically high relative abundance in this experiment and thus strong influence on the total aerobic microorganism count result. Mean value and standard error of the mean (SEM) are indicated from at least 170 replicates per incubation condition. Two-tiered incubation means incubation for 3 days at 20–25 °C and for 2 days at 30–35 °C; the initial incubation condition is specified. Significant differences in CFU count were calculated with Kruskal-Wallis test and Dunn's test corrected for multiple comparisons as post-test (ns, not significant; *, P < 0.05; ****, P < 0.0001).

In contrast to total aerobic count, recovery of moulds continuously and drastically decreased with increasing incubation temperatures (Figure 1B and 5B). Also this observation seemed plausible because indoor moulds in a temperature-controlled environment have only weak selective pressure to be able to efficiently grow at temperatures above 25 °C. Initial incubation at 30–35 °C seemed to severely impair mould growth despite subsequent incubation at 20–25 °C (Figure 5B), arguably challenging that incubation with the high temperature first represents a generally superior approach for two-tiered incubation schemes as some may infer from USP <1116> (2). Whether mould recovery could be increased through prolonged incubation at 20–25 °C after being subjected to 30–35 °C should be a topic for further research. The use of the mycological medium SDA conveyed a slight increase in mould recovery compared to TSA. Of interest, in a recent retrospective study in hospital pharmacies and compounding pharmacies, Weissfeld et al. came to the conclusion that the use of mycological media (in this case yeast malt extract agar) is important (14). However, in the study by Weissfeld et al. the mycological medium was incubated at 28 ± 2 °C and the general microbiological medium (TSA supplemented with polysorbate and lecithin) was incubated at 35 ± 2°C as recommended in one of their relevant guidelines, USP <797> (15). Thus, the study does not allow dissecting the contribution of incubation temperature or nutrient medium towards the observed difference in fungal recovery. Our study highlights that the differences observed in the study by Weissfeld et al. may also be explained by a temperature effect rather than a nutrient medium effect (Figures 1B and 5B). Furthermore, our data also indicate that the incubation temperature of 28 ± 2 °C proposed for mycological media in USP <797> (15) may arguably be too high for optimal recovery of environmental moulds.

The differences observed in the in situ study were only poorly reflected in the in vitro study despite application of a strain-specifically paired statistical evaluation approach. We hypothesize that the underlying cause of this observation was the strain selection. Although some of the used microorganisms clearly preferred a particular incubation condition, these preferences leveled out with the microbial selection used for this study. With another selection of microorganisms, however, a different answer may have been obtained. Strain selection bias may probably be the main cause for the very different outcomes of EM validation studies performed at different laboratories and companies. In general, we see four main contributors to strain selection bias.

First, the most prominent effect of incubation conditions on microbial recovery from air and inanimate surfaces was observed for moulds. However, in a well-controlled environment, moulds only represent a small fraction of microorganisms recovered from EM samples (particularly in higher grade cleanrooms in sterile manufacturing) (13). This may elicit a tendency to represent moulds less well in study designs. Our own in vitro study, which had been defined based on microorganisms available in our cryo collection, was strongly affected by this deficiency.

Second, the pharmacopeia-referenced A. brasiliensis mould strain (ATCC 16404) has an obvious preference for elevated incubation temperatures (Figure 3C). As demonstrated in the present study, this preference seems to deviate from the temperature range optimal for recovery of moulds from the environment. Thus, this strain may pose a questionable reference with regards to EM. The other pharmacopoeia-referenced strains which were included in our study did not exhibit clear temperature preferences (not shown). Obviously, most of the pharmacopoeia-referenced strains can efficiently be recovered within a broad temperature range, and as no clear preferences exist, a variety of different incubation conditions could be defined on basis of these results. However, whether this broad temperature tolerance is also representative for the majority of environmental microorganisms cannot be deduced. Thus, it should be carefully considered which conclusions may validly be drawn for the incubation conditions of the routine monitoring program on basis of such laboratory results.

Third, the origin of externally purchased isolates used in a study should be carefully considered. Microbial strains from accredited microorganism collections often represent patient isolates. The prerequisite for such pathogenicity is the ability or even preference to grow at elevated temperatures—but whether such strains are representative for a pharmaceutical manufacturing environment is questionable. Thus, the use of such strains may lead to a bias towards preference of high incubation temperatures, although this may not reflect the preference of most environmental strains. An earlier study claiming that an incubation temperature of 30–35 °C is ideal for fungal detection (6) featured a microbial spectrum in which four of nine strains were patient isolates. This strain selection bias at least partially contributed to a study outcome not supportable by data generated from actual environmental samples.

Finally, in the case of in-house isolates, the incubation condition used for isolation is also a potential source of significant bias. For instance, if the EM program used to date consisted of incubation at an elevated temperature, the sole reliance on isolates obtained at that temperature is problematic. This is due to the fact that microorganisms not able to grow at these elevated temperatures would not have been isolated. Therefore, the isolate selection would be highly biased towards confirmation of the previously used incubation conditions. Our in-house isolate Penicillium sp., for instance, had been isolated at more than 25 °C; thus, it is not surprising that this particular strain could tolerate elevated incubation temperatures (Figure 3B). Accordingly, even if we had included a larger number of mould isolates from our environment in the in vitro study, the significance of these results would remain questionable. Because the incubation temperature routinely applied to that date had been 25–30 °C, there is a very high probability that an in vitro study with moulds isolated from routine EM would confirm these conditions—although this may not necessarily be representative for the actual environment. Similarly, if no mycological media had been used in the past, there is a high probability that a validation applying previously isolated fungi would not point towards a superior recovery with mycological media as compared to the general microbiological growth medium used for isolation—based on the fact that fungi that indeed strongly prefer mycological media had a lower chance to be included in the isolate collection. Therefore, in order to have a strain selection which is less affected by bias, isolation would have to occur using all of the different media and incubation temperatures of interest, which essentially resembles an in situ approach.

As outlined above, introduction of bias in in vitro experiments aiming towards comparison of different EM incubation conditions is difficult to avoid. Therefore, we consider the results of our in situ approach as more meaningful. We also believe that our study may be considered generally representative for manufacturing sites in similar climate zones. Although the exact strain composition at different sites may differ, temperature preferences of the indigenous microorganisms should remain comparable. For instance, microorganisms of human origin may be expected to be better recoverable at temperatures close to the skin temperature (as also supported by the data in Figure 2A and B). Accordingly, if it is assumed that the predominant part of the cleanroom microflora in a particular area is human-derived, one could also reasonably expect superior recovery of environmental microorganisms at elevated incubation temperatures. Our results obtained for an area with personnel flow (Figure 1A), in which such a trend was evident, and an area without personnel flow (Figure 5A), where no correlation between incubation temperature and recovery of total aerobic microorganisms was observable, support this notion. Such factors should, in principle, be independent of the manufacturing site. However, whether a generalized shift towards a preference of higher incubation temperatures (particularly for moulds) would be observed in cleanrooms located in a tropical climate zone may be a question to be further elucidated. The observation that fungal recovery from tropical forest soil was superior at an incubation temperature of 30 °C as compared to 25 °C indicates that this may indeed be the case (16).

In our study design we to a certain extent omitted yeasts, which were only reflected in total aerobic count. The reason for this apparent shortcoming was the expected difficulty to unambiguously differentiate some yeast from bacteria based on colony morphology. The large sample size of our study forbade systematic screening of all recovered colonies with a microscope or use of other identification techniques. In principle, recovery of yeasts from the environment could follow a similar pattern regarding impact of different temperatures and nutrient media as was observed for moulds. This assumption may be supported by the view expressed in several guidelines that recommend that the same incubation conditions and nutrient media should be used for detection of yeasts as well as moulds (1, 3). However, the authors are not aware of studies experimentally addressing that issue for pharmaceutical cleanrooms. An obvious way to elucidate this question would be an assessment of the impact of the different temperatures and nutrient media in an in situ study using SDA and TSA supplemented with antibiotics.

Based on our results, the highest overall recovery from aerobic environments (air, inanimate surfaces) would have been achieved when total aerobic count was obtained from TSA incubated at 30–35 °C and mould count from SDA incubated at 20–25 °C. But also other approaches yielded acceptable results. While the two-tiered incubation strategy had a slight deficiency in recovering total aerobic count, mould detection was very effective. Incubation at an intermediate temperature of 25–30 °C, on the other hand, led to very efficient recovery of total aerobic count, but had certain deficiencies in recovering moulds. However, when defining an incubation strategy for EM samples, it should be emphasized that the use of two plates with different nutrient media is not optimal in every case. For monitoring of an aseptic manufacturing operation, for instance, the double-sampling dictated by use of two nutrient media would lead to a higher number or an extended duration of interventions or manipulations close to open product as compared to a single-plate strategy. This potential concern may be less relevant when production occurs in an isolator or involves use of a restricted access barrier system. Moreover, human-derived bacteria typically represent by far the most common contaminant in well-controlled ISO 4.8/5 environments, whereas fungi can be considered rare (1%) in comparison (13). These considerations further strengthen the argument that monitoring of such areas should not regularly be performed with two media, as the higher product contamination risk through the operator may outweigh the slightly improved capability to detect fungi. Rather, a monitoring strategy that relies on a single nutrient medium incubated under conditions allowing for reasonable detection of bacteria, yeasts, and moulds (i.e., intermediate temperature of 25–30 °C or two-tiered incubation) complemented with periodic use of mycological media may represent a more appropriate approach. It is also worth mentioning that the above rationale supports the recommendation provided in USP <1116> (2) to initially apply the high temperature for two-tiered incubation schemes in order to efficiently recover human-associated bacteria. However, the resulting recovery of moulds seems to be so low (Figure 5B) that the value of the subsequent change to a reduced incubation temperature should be addressed in systematic studies. Eventually, an extension of incubation time at the reduced temperature may ameliorate mould recovery. Also, further work focused on yeasts may additionally support or oppose the necessity of the proposed temperature shift.

Similar to ISO 4.8/5 environments, we believe that a single-plate strategy may also allow for appropriate surveillance of the microbiological state of control of areas with higher microbial burden. In particular, the definition of appropriate alert levels and data trending strategies enable assessment of the microbiological control status irrespective of whether the overall microbial recovery would be higher through the use of two nutrient media. However, in areas where product considered distinctly critical regarding fungal contamination (e.g., some inhalation products) is openly handled, the use of mycological media at defined frequencies should indeed be considered as part of the EM program in a risk-based manner.

In conclusion, we assessed microbial recovery from EM samples in an in situ approach and compared the results to an in vitro study. Our results indicated highest recovery of total aerobic count in areas with personnel flow using a general microbiological growth medium (TSA) incubated at 30–35 °C and highest recovery of moulds with a mycological medium (SDA) incubated at 20–25 °C. However, also single-plate strategies applying a general microbiological growth medium either in a two-temperature incubation scheme starting at 20–25 °C or using an intermediate incubation temperature of 25–30 °C yielded reasonable results. Recovery of moulds was inversely correlated with the applied incubation temperature. Interestingly, subsequent shift from a high incubation temperature to a low incubation temperature did not seem to ameliorate mould recovery with the incubation time applied. We showed that the in situ approach was more powerful in detecting differences between the incubation conditions of interest. Furthermore, we offered explanations why in vitro studies conducted by different laboratories and companies came to conflicting results and sometimes also counterintuitive conclusions. For instance, the assertion that for recovery of environmental fungi in moderate climate zones incubation above 30 °C should be equivalent to incubation at ambient temperature (6) is conclusively rejected by our study. Indeed, care must be taken when designing laboratory-based in vitro studies in order to mitigate bias, and the extent to which such studies allow investigators to draw conclusions for EM samples should be well reflected. In general, we consider our in situ approach as being more meaningful compared to in vitro studies when examining different incubation conditions for EM samples, and we believe that the results may be potentially conferrable to pharmaceutical manufacturing sites in similar climate zones. Therefore, our study should represent a valuable reference for definition of the incubation strategy of microbiological EM samples.