OUR CONSIDERATIONS We began our project to improve upon the more traditional control banding systems used in pharmaceuticals with two key questions: first, could exposure modeling for inhalation exposure be integrated into our in-company exposure assessment approaches? And second, could some of the functionality provided by existing control banding systems—especially the ability for nonexperts to qualitatively identify appropriate engineering control measures—be maintained?

We approached these questions by first surveying available exposure modeling tools. Among the considerations important to us were the scientific basis of a model, the degree of validation with real-world exposure monitoring data, applicability across various types of exposure scenarios, acceptability under regulatory frameworks, and—crucially—whether a model and its underlying algorithm were publicly published and available. Modeling tools of interest included the commercially available Stoffenmanager, ECETOC’s Targeted Risk Assessment (TRA), the Advanced REACH Tool (ART), and AIHA’s IHMOD spreadsheet, all of which are available online. ART interested us the most because it was a more sophisticated tier 2 model and the development of its algorithm is published in the scientific literature along with several validation and model comparison studies. Its application in the pharmaceutical industry was of particular relevance to our project.

THE ADVANCED REACH TOOL ART is a mechanistic model based on a source-receptor approach that aims to describe the transport of a hazardous substance in air from the source to the receptor. It considers several independent modifying factors, including the substance emission potential, activity emission potential, localized controls, segregation, personal enclosure, surface contamination, and dispersion. While chemical and physical laws, along with empirical data from literature, were used in establishing the model, it’s important to note that its development also included an element of expert judgment with peer review by leading experts from industry, research institutes, and public authorities. The original algorithm was calibrated with more than 2,000 occupational exposure measurements and has been further validated and compared in several follow-up publications. In general, validation studies indicated sufficient conservatism in modeled exposure estimates compared to occupational exposure measurements when higher decision statistics and confidence intervals were selected (the 90th percentile and 90 percent confidence interval, for example), and a general tendency to overestimate lower exposures while underestimating higher exposures. One validation study comparing ART estimates to data from the pharmaceutical industry indicated a tendency toward underestimating exposures, although it was based on a relatively small dataset. (Readers who wish to understand more about ART’s development and validation can find further reading in the box below.)

The existing ART algorithm only allows for the assessment of vapors, mists, and dusts, and estimates exposure over a task-based duration, expressing results in mg/m3. Since the model generates an underlying exposure distribution for each estimate, it also allows for the selection of a decision statistic (for example, the 95th percentile) along with confidence intervals. Since exposure estimates are based on the task, ART further allows calculation of time-weighted average estimates based on the assumption of zero exposure over the remainder of the working shift or by combining several exposure activities into one estimate. Finally, ART allows the integration of occupational exposure data through Bayesian inference to refine and improve accuracy of modeled exposure estimates.

ART is publicly available online.

Since ExCMO is based on ART’s algorithm, it was important to verify that we had correctly reconstructed the original algorithm by comparing results from the two tools using identical input parameters. We established through internal validation runs to a degree we judged appropriate that ExCMO could be considered an accurate replication of ART. However, ExCMO departs from the original ART in some important ways specific to our intended application:

Increased usability. In ExCMO, all exposure scenario inputs are organized on a single input page. When using the original ART, users must navigate through various pages, which we felt could potentially lead to poor adoption. The simplified design of ExCMO is intended to increase the speed and ease with which an exposure scenario can be assessed.

Default modeling inputs. ExCMO features certain model inputs as defaults to reduce the number of inputs required per assessment. In this case, the underlying ART algorithm was not adapted; for ExCMO’s intended application as a qualitative tool in pharmaceutical manufacturing, it made sense to set defaults for specific input requirements. (One example has to do with the input for drop height. When assessing transfer of powders, ART includes an option for powders falling with a drop height exceeding 0.5 meters. This is unlikely in our industry since best practices for powder handling in pharmaceutical operations dictate drop heights of less than 0.5 meters. Therefore, ExCMO by default calculates exposure estimates based on a drop height of less than 0.5 meters.)

Maintaining control-banding-like output. ExCMO’s IHCS pathway was developed based on concepts borrowed from control banding, specifically in expressing the results of modeling in terms of recommended engineering control measures. Local control options available in the ART algorithm are ordered based on assigned control efficiency multipliers and grouped into distinct categories to which we assigned further descriptors (for example, Control Strategy 2, described as “open handling,” is named “Local Exhaust Ventilation” in ExCMO).

Additional engineering control options. ExCMO has two additional control options that represent common control systems utilized in pharmaceuticals: “horizontal/laminar down flow booth (enclosing source) with screen and glove ports” and “closed restricted access barrier system.” We used our professional judgment to assign control efficiency multipliers based on similarities to other control options available in ART. For example, we applied the ART multiplier for “low specification glovebox” to the “closed restricted access barrier system” in ExCMO.

Like the original ART, ExCMO also has the ability to integrate exposure measurements through Bayesian integration. This function allows real data from a similar exposure scenario to be incorporated into the modeling estimate.

ADVANTAGES AND LIMITATIONS From our perspective, some advantages of using the exposure modeling approach stem from a comparison to historic control banding systems. For instance, the ART algorithm allows model predictions to be sensitive to the percentage of hazardous material present in a mixture, a typical scenario we encounter when assessing exposures in pharmaceutical operations. Since this sensitivity was not possible with control banding, control recommendations were potentially overly conservative. Another example has to do with the ability to apply the time-weighted average concept in predicting required engineering controls. Whenever an exposure scenario is consistent in duration and there is high confidence that no further exposure will occur during the remainder of the workday, exposure modeling enables the selection of control options that are more appropriate to the risk. These examples demonstrate the potential for ExCMO to reduce unnecessary conservatism compared to previous control banding systems. A further advantage over control banding is the tool’s ability to estimate exposures for products suspended or dissolved in non-volatile liquids.

In our view, exposure modeling is often discussed in academic literature, but less often systematically applied in industry. In addition, several validation studies indicate differing conclusions on model accuracy. Because of this, one limitation for practicing industrial hygienists likely relates to the level of understanding and trust in the outputs that models generate; however, we believe this is solved through framing existing tools like ExCMO as qualitative tools for the time being. A further limitation associated with the use of ExCMO is that the tool requires a valid OEL since the OEL is the key parameter used to determine recommended engineering control measures or for the purposes of estimating compliance. When an OEL does not exist, we recommend applying the NIOSH occupational exposure banding process to estimate OEL ranges that may be applicable. It’s also important to consider the need for training to ensure that assessors have an appropriate understanding of the meaning of different model input options to generate reasonable exposure estimates.

ExCMO is a success at our companies because it simplifies a fairly complicated exposure model into a user-friendly tool for use by our environmental health and safety professionals—most of whom are not certified industrial hygienists. The tool also provides a unified way of assessing our organizations’ chemical exposures. Our stakeholders appreciate that ExCMO is a more sophisticated type of assessment that produces an output they can understand, especially for engineering controls, and workers like seeing the results directly. One challenge for us has been overcoming the learning curve for the primary users of ExCMO, particularly teaching them how and when to apply different inputs for exposure scenarios. This requires training, both in terms of a basic theoretic understanding and practical application.

OUR EXPERIENCE ExCMO is an in-company tool intended to enhance qualitative exposure assessment approaches, including the identification of risk-appropriate engineering controls. The tool incorporates several industry- and company-specific settings; however, ExCMO is in principle based on the scientific work of the exposure scientists and organizations involved in developing ART.

We have incorporated an internal company-specific version of ExCMO into our standard exposure assessment process for qualitative assessment. One of our companies uses it as a standardized approach to estimate exposure risk before deciding when and where to conduct exposure monitoring. At the other, all company sites perform exposure assessments for active pharmaceutical ingredients (APIs) using ExCMO, which allows the company to obtain a uniform view of the level of control for all APIs across its sites. ExCMO is also our companies’ recommended tool for new engineering projects to identify what level of control is required based on the hazards associated with different substances; this helps ensure the appropriate level of control is in place from the beginning of each project.

The ExCMO tool is publicly available online. We encourage readers to test it out, keeping in mind its limitations and the understanding that exposure models can generate uncertain estimates. Remember also that ExCMO is based on ART, which was originally designed for all industries. We made changes to better align ExCMO with pharmaceutical operations, but it is by no means a tool only for our industry.

Acknowledgment: The authors wish to thank the Center for Primary Care and Public Health (Unisanté) in Lausanne, Switzerland, and in particular Dr. Nenad Savic for their collaboration in building the ExCMO tool.

Send feedback to The Synergist.