Pilot study seeks to validate new and more granular outsourcing model classifications in differentiating performance across custom contract-service approaches.
Recent research published on the market for contract clinical services reveals robust growth. An analysis of the top 10 contract research organizations (CROs) conducted between 2017 and 2021 reflects 12% growth in total revenue with a 17% growth rate observed from 2020 and 2021.1 The top 10 CROs generated approximately $34 billion in revenue in 2022, representing an estimated 69% of total revenue spent on contract clinical services globally. Other estimates of the total global CRO services market report $76.6 billion for 2023 with expected growth of 10.7% CAGR to $127.3 billion by 2028.2
During the last 10 years two primary outsourcing models—full-service (FS) and functional service provider (FSP)—have characterized the predominant approaches and have been the topic of much debate.3 Sponsors using full-service model outsourcing engage CROs to support a wide range of clinical trial activities; sponsors using functional service models outsource specific services for a given function (such as medical writing or clinical monitoring).4
There are conflicting, largely anecdotal reports, on which model—FS or FSP—delivers better performance. Recent articles in the trade press suggest that there has been a shift toward FSP model usage.3,4 A recent article in Applied Clinical Trials states that large biopharma company use of FSP outsourcing is growing at more than 13% annually.4
Wide variation in outsourcing practices suggests that FS and FSP distinctions are inadequate to capture the range of outsourcing models utilized today.5 Wide variation is observed within companies by therapeutic area, department, phase, and geographic area.6 A variety of relationship structures have been used, including preferred provider, fee-for-service or transactional outsourcing, “hybrid full-service,” compound or program-based, and sole-source outsourcing.5,6 Company use of a combination of outsourcing approaches is prevalent even within a single department.6,7 In step with rising protocol complexity and customization during the past two decades, outsourcing has also become more customized and hybridized.
FS and FSP outsourcing models at this time are often misunderstood. Biopharmaceutical companies, especially large ones, are typically fragmented and engage in complex projects with new collaborations continually forming. Industry mergers and acquisitions also result in varied simultaneous outsourcing practices.8,9 Given the lack of clear-cut or easily defined models, there is a strong need for new terminology and definitions to optimize CRO service use.8
Recently, Tufts Center for the Study of Drug Development (CSDD) and ICON plc proposed a new framework to differentiate sourcing models based on type of contract services.8 Rather than using traditional definitions this novel approach uses the CRO’s infrastructure and processes for clinical trial conduct resulting in mixed models and combinations of sourcing models. The development of this classification is based on the following factors: outsourcing orientation, infrastructure support, and accountability or contracting structure.8
The framework accounts for outsourcing model variation and combinations currently in use and includes characteristics of managed services (including full service and single or multi-service), mixed models (or blended); functional continuum (embedded and FSP) and internal pharma managed or In-house models.
In an effort to validate this new framework and assess the relationship between more granular outsourcing models and clinical trial performance, Tufts CSDD and ICON plc collaborated on a pilot study. ICON’s Partner of Choice (POCs) companies contributed data from their pivotal trials supporting recent products approved by FDA.
Tufts CSDD and ICON plc collaborated on the development of a data collection worksheet. ICON’s POC companies indicated the primary outsourcing model for each clinical trial (main model), as well as the model used for each of six individual functions: study start-up, project management, clinical monitoring, data management, statistics and programming, and medical writing.
The worksheet also asked several other questions:
Using its proprietary dataset on pivotal trials, Tufts CSDD populated the data collection worksheets for ICON’s member companies with data on pivotal trials supporting products approved by the FDA between 2010 and 2020. Tufts CSDD data included NCT number, official clinical trial title, additional study IDs, therapeutic area, approval year of the product investigated, and the sponsor of the product. The Tufts CSDD dataset also included data on pivotal trial scope—the trial duration (in months), number of trial centers, number of countries, number of participants, number of endpoints, and number of eligibility criteria.
The worksheets also provided a model key that respondents could use to identify the outsourcing models used.
An additional clarifying definition was distributed to respondents via email and is provided below:
Data collection worksheets were distributed in mid-December 2022 and respondents were given until mid-March 2023 to complete them. Worksheets were then merged into a single dataset, including the additional variables from the Tufts CSDD pivotal trial dataset.
Frequencies were calculated for the dataset, including phase, therapeutic area, main model used, model used for each function, and responses to the additional questions. Main model frequency was calculated by therapeutic area (oncology or non-oncology). Mean, coefficient of variance, and median were calculated for scope variables by main model used, as were frequencies for question responses.
Data was collected in Excel and merged, cleaned, and analyzed using SAS 9.4.
Nine sponsor companies completed the worksheets and provided data on 117 pivotal trials. The dataset was comprised of trials primarily from large pharma (97.4% of trials) and mainly consisted of Phase III trials (80.3%). Endocrinology was the most common therapeutic area (27.4%), followed by oncology (23.1%), with cardiovascular and respiratory diseases the third-most common (12.0% each).
Table 2 lists the frequency of use of each model as the main or primary outsourcing model and for each specific function. The FSP model is by far the most commonly used model with almost half of clinical trials relying on it for study delivery. FSP is also the most common model used for clinical monitoring and data management. Sponsor companies report that study start-up, project management, statistics and programming, and medical writing are most commonly handled In-house (35.9%, 45.2%, 39.1%, and 46.7%, respectively).
For oncology trials, full-service models are most commonly used (29.6%), followed by multi-service (22.2%), and blended approaches (18.5%). Among non-oncology trials, functional service models are the most commonly used (57.8%), followed by blended and in-house (11.1% each).
The frequencies of responses to each question are listed in Table 4. The sponsor was responsible for delivery of the study in almost all cases (96.4%). Sponsors reported that half of clinical trials finished on time with a slightly lower percentage (42.5%) of clinical trials finishing late. Sponsors report that clinical trials rarely finished early (7.6%). All sponsors indicated that the clinical trial was accepted on first review and had no major quality issues. Clinical trials were typically delivered to budget (74.7%), sometimes went over budget (24.1%), and were almost never under budget (1.3%).
Clinical trials relying on the full-service model as the main model may be some of the longest and most complex, according to the scope variables reviewed (see Table 5). Sample sizes for individual models are too small to test, however, the means do give some indication that full-service trials are more than a year longer than other trials, on average. They also consistently have among the highest, if the not the highest, average in each of the other scope variables, except for number of endpoints.
Table 6 contains the frequencies for responses to the questions regarding whether a clinical trial was delivered on time and to budget by main outsourcing model used. For full-service, multi-service, blended, and staff augmentation, “on time” was the most common response. For FSP and in-house, “finished late” was the most common response. Multi-service was the only model where “over budget” was the most common response; for the remaining models, “on budget” was the most common response.
This pilot study demonstrated that the new outsourcing model classifications proposed by Tufts CSDD and ICON plc provide reasonable discrimination between custom outsourcing approaches and capture the many different outsourcing models currently in use.6,7
A major finding in this study is that FSP outsourcing is the primary model in use at this time, representing nearly half of all clinical trials in the sample. It also is the most common model for clinical monitoring and data management. This result is not surprising, given that clinical monitoring and data management require high relative study costs and are exceptionally labor intensive. Clinical monitoring can represent up to 35% to 45% of costs of a clinical trial and use of an FSP model has been reported to reduce costs and improve efficiencies.11 This result suggests a major outsourcing shift: studies conducted three-five years earlier found that full-service outsourcing was the most prevalent model used by 77.3% of sponsor companies and that although the majority of spending on outsourcing went to full-service providers, allocation in dollars was expected to shift toward functional service providers in the near future as FSP usage was more positively perceived by survey respondents.12
Another major finding of this study is that in-house sourcing is the most prevalent model for study start up, project management, statistics and programming, and medical writing. There are, however, many other models being used for these functions, ranging from four to six different models per function.
This study also found that full-service outsourcing was primarily used for oncology clinical trials while functional service provider outsourcing was the predominant model for non-oncology trials. Clinical trials outsourced under a full-service model were also characterized as longer in duration, more complex (e.g., more endpoints and eligibility criteria) and larger in scope (e.g., more countries, investigative sites, and patients enrolled) than those under FSP outsourcing. Pivotal trials outsourced under a full-service model on average took one year longer than other clinical trials. Oncology trials also comprised the second-largest therapeutic area in the sample (23% of total trials). In general, oncology trials have been found to have longer clinical trial durations and involve more countries and investigative sites, more patients per protocol, and generate more data for analyses.10
It is interesting to note that embedded and staff augmentation were the least-reported outsourcing models used. There was no usage of embedded for five of the six functions, with the exception of medical writing, and was never reported as the main model. In the embedded model, also referred to as FSP 2.0, the vendor takes on additional levels of responsibility compared to FSP and augmented models. Perhaps biopharmaceutical companies have not yet adapted to this approach for a majority of functions. In addition, staff augmentation was not used for start-up, project management, or clinical monitoring activities. Project management and start-up were predominantly reported as in-house functions, while clinical monitoring was most commonly reported as following an FSP model. These functions were most likely not a good fit for the staff augmentation model, as the definition describes limits to vendor headcount and a tactical capacity management solution.
A limitation of the study is that FDA acceptance and quality issue questions received unanimous (100%) agreement from respondents with regards to program acceptance on first review by FDA and no known major quality issues that compromised the pivotal trial. Further investigation is required here that could potentially elicit more detail regarding these responses. These questions also have lower response rates than the others, and one could speculate that respondents may have skipped these questions.
In general, the results of this pilot study demonstrate that more granular classification of outsourcing model use is associated with differentiated pivotal trial scope, complexity, and outcome measures. Pivotal trials managed under a full-service outsourcing arrangement, for example, were associated with larger relative scope, longer duration, and higher complexity compared to other outsourcing models assessed. Future research will look to gather a much larger sample of clinical trials, and may look at planned or upcoming trials, as well as outcome and quality measures tied to Tufts CSDD’s more granular outsourcing model classification. In addition, measures of satisfaction with performance, oversight and effectiveness, and relationship quality will be assessed using the proposed framework.
Mary Jo Lamberti, PhD, is Research Associate Professor, Tufts Center for the Study of Drug Development, Tufts University School of Medicine; Zachary Smith, MA, is Senior Data Scientist, Tufts Center for the Study of Drug Development, Tufts University School of Medicine; Maria DiPietro is Senior Vice President, Operational Solutions Architect & Strategy, ICON plc; John Barry is Senior Vice President, Consulting and Strategy Office, ICON plc; Ken Getz, MBA, is Executive Director and Professor, Tufts Center for the Study of Drug Development, Tufts University School of Medicine