How do Clinical Research Coordinators learn Good Clinical Practice? A mixed methods study of factors that predict uptake of knowledge

Abstract

Background

Good Clinical Practice (GCP) is an international standard for the design and conduct of clinical trials to ensure ethical and scientific integrity. Recent NIH policy mandates GCP training for all investigators and staff involved in NIH funded clinical trials, yet approaches to GCP training vary widely. There is limited data on GCP knowledge among the clinical trial workforce and no evidence regarding effective methods to learn GCP.

Methods

We used an exploratory sequential mixed methods design. We conducted 18 exploratory qualitative interviews with clinical research coordinators to help inform the development of the quantitative survey. We then administered a validated 32-item, multiple choice test of GCP knowledge with a survey of work and training experiences to 625 clinical research coordinators at three academic medical centers in the United States. Variables that were significantly associated with GCP knowledge were entered into a multiple regression analysis to identify unique predictors of GCP knowledge. We controlled for verbal-numerical reasoning and learning orientation.

Results

During qualitative interviews, clinical research coordinators reported that formal GCP training had value but they simultaneously emphasized the importance of experience, day to day practice, and observing colleagues and mentors as essential to supplement formal training. In our quantitative survey, five variables predicted a total of 22% of variance in GCP knowledge scores: Years of experience as a Clinical Research Coordinator, working on diverse types of trials, supporting industry-funded trials, being certified in clinical research coordination, and aggregated hours of online and face-to-face training (in that order).

Conclusions

The duration and richness of experience as a Clinical Research Coordinator were the strongest predictors of GCP knowledge; a finding consistent with our exploratory qualitative interview results. Our findings suggest formal online and face-to-face training has a minimal influence on GCP knowledge. The type of training – whether online or face to face – does not make a significant difference in GCP knowledge scores. Much of the variance in GCP knowledge remains unexplained, calling for further research in this area.

Background and Aims

Good Clinical Practice (GCP), first developed by the International Conference for Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use in 1996, is an international standard for the design, conduct, and reporting of clinical trials to ensure ethical and scientific integrity.¹ The principles of GCP have been adopted into law in various countries and in Europe as the EU Clinical Trials Directive.² In the United States, the Food and Drug Administration uses GCP as guidance rather than formal law.³ Compliance with GCP is meant to protect the rights and safety of participants, ensure that trials are conducted with rigor, meet regulatory requirements, and safeguard the integrity of data. These objectives must be achieved to acquire trustworthy evidence regarding the risks and benefits of treatments and interventions, while protecting participants and preserving public trust in research.⁴ In this sense, GCP is an essential element in realizing the overall social benefits that clinical trials can provide. Methot et al. aptly describe GCP as a “system of shared responsibilities.”⁵

Clinical trials have become increasingly complex often involving multiple sites across different countries as well as complex trial designs, protocols, and interventions. As a consequence, clinical trials are subject to heavy regulations, governance, and monitoring rules and are among the most regulated activities globally.^5–10 According to the Declaration of Helsinki, research must be carried out by individuals with appropriate training and qualifications in clinical research.¹¹ GCP training is often mandated, making GCP the “bible of trial conduct” for clinical trial personnel.¹² Recent NIH policy, effective March 2017, mandates GCP training for all investigators and staff involved in NIH funded clinical trials.¹³ According to the NIH, GCP training should be refreshed every 3 years and is defined as a “class or course, academic training program, or certification”.¹³

The goal of GCP training is to ensure team members have the same basic foundational knowledge, yet there are no consistent or formalized educational requirements that define an appropriate level of qualification.^{9, 14} Training requirements tend to be general and contain few details on what key criteria should be included.¹⁰ Approaches to GCP training include online modules that average 45–60 minutes or longer, face-to-face trainings from one time only to multiple sessions, and formal certification programs offered through professional organizations that can take more than a year to complete.^{9, 10} GCP trainings are offered by professional organizations, industry, government, and Clinical and Translational Science Award (CTSA) hubs.¹⁴

While principal investigators are ultimately responsible for all ethical and regulatory aspects of a clinical trial, clinical research coordinators are integral to the clinical trial research enterprise.¹⁵ Clinical research coordinators work closely with principal investigators and are involved in developing protocols, writing consent forms, recruitment, informed consent, and data management. Over time, the role of clinical research coordinators requires increasing autonomy, accountability and responsibility.¹⁶ Clinical research coordinators are also the main point of contact for subjects during a trial, and act as liaisons between study subjects and principal investigators.¹⁵ Thus, they are heavily involved in the day to day aspects of clinical trials and ensuring GCP requirements are met.

There have been several recent efforts to standardize requirements for GCP training and identify core competencies to address gaps in current training.^{9, 10, 14, 17} The National Center for Advancing Translational Science sponsored the “Enhancing Clinical Research Professionals’ Training and Qualification Project” which consisted of a consortium of all CTSAs, and aimed to establish a “foundation of GCP training expectations for all clinical trials work.”⁹ This process included establishing a Joint Taskforce for Clinical Trial Competency that identified 8 core competencies for principal investigators and clinical research coordinators, with 51 corresponding specific sub-competencies across the 8 competency domains.^{10, 14} CTSA hubs are now implementing mandatory GCP training for all personnel involved in NIH funded clinical trials, to be refreshed at least every 3 years, but institutions can choose which training programs to offer as long as they meet the minimum requirements laid out by the “Enhancing Clinical Research Professionals’ Training and Qualification Project”.⁹

While training in GCP is considered necessary for responsible and ethical management of clinical trials, there is limited data on GCP knowledge among the clinical trial workforce, or evidence about the most effective means to learn GCP.^{10, 18, 19} In light of the recent NIH mandate, and increasing attention regarding assessment and standardization of GCP training, this study examined GCP knowledge among clinical research coordinators working in the U.S., and explored the association of knowledge with work experience, face-to-face and online training, certification, and other clinical research coordinator characteristics.

The aims of this mixed methods study were: 1) to explore through qualitative methods clinical research coordinators’ perspectives and experiences regarding learning GCP and, 2) to use testing and quantitative survey methods informed by our qualitative findings to identify what factors predict knowledge of GCP among clinical research coordinators. Such information was meant to inform the evaluation of current approaches to GCP training.

We focused on clinical research coordinators because they play an essential role in the day to day conduct of clinical trials including ensuring consistent data collection, informed consent, retention of trial participants, and encouraging compliance with regulations.^{7, 20}

Methods

Study design

This exploratory sequential mixed methods study included qualitative interviews with clinical research coordinators at a research-intensive academic medical center in the U.S. to explore their experiences and perspectives about learning GCP.²¹ This data informed development of our quantitative survey. Following this preliminary exploratory study, data were collected from clinical research coordinators using a cross-sectional survey and a newly developed test of GCP knowledge at three research-intensive medical centers that have CTSAs.

Ethical review

This project was approved as an Exempt study by the IRB at each of the three participating CTSA sites. At two sites, participants were presented with the informed consent document electronically and were asked to indicate consent before proceeding to the online survey and GCP test. A third site did not require a consent form as completion of the survey was considered consent to participate. Participants were informed that data would be confidential, and it would not be possible to identify any individual taking part.

Qualitative interviews

We conducted 18 semi-structured qualitative telephone interviews with clinical research coordinators (75% female) of varying experience levels (1–32 years) who had completed online GCP training modules. In light of the limited data regarding GCP knowledge, interviews explored prior experience and satisfaction with formal online GCP training, informal methods used to learn GCP outside of formal training, what clinical research coordinators think are the most effective methods for learning GCP, and their top recommendation to a new clinical research coordinator for how to learn GCP. The purpose of the qualitative interviews was to inform development of the quantitative survey. The emphasis individuals placed on experience and practice informed the work experience section of the 28-item survey to capture a breadth of variables about work experience that might predict GCP knowledge.

See Supplemental Materials for a full description of methods, qualitative findings, and the interview guide.

Quantitative cross-sectional survey and GCP knowledge test

Sample and procedure

Criterion sampling was used to enroll clinical research coordinators for drug, device, or biologic trials with a range of experience at three academic medical centers with CTSAs. We sought a large and diverse enough sample to explore a variety of predictor variables. Power analysis indicated that assuming a power of .90 and an adjusted alpha of .01, assuming we ran up to 5 models, even a modest effect (Cohen’s f² = .05) could be detected in a multiple regression model with 14 predictors in a sample size of 616. Thus, we aimed to obtain a sample size of 600.

Email distribution lists of clinical research coordinators (and sometimes other clinical research staff because clinical research coordinators job titles may differ) employed at each site were used to send recruitment messages. A representative at each institutions’ CTSA sent an email indicating the opportunity to participate in an online survey examining experiences and prior GCP training followed by GCP knowledge questions. Participants were told the study would take 30–40 minutes. The emails included a link to the online study administered via Qualtrics. Participants were instructed to complete the survey only once. The survey was administered with forced-choice to ensure participants responded to all parts.

Across sites, 2415 emails were sent to clinical research staff. Three follow up emails were sent as reminders at approximately one-week intervals. To confirm eligibility, the first questions of the survey screened to ensure the potential respondent were clinical research coordinators and that they worked in drug, device, or biomedical intervention clinical trials involving human subjects. Survey participants were compensated $20 upon completion of the survey, and were eligible to win one of three raffle prizes of $100 per site. Our response rate was approximately 33% of all eligible individuals. We estimated our response of 1884 eligible individuals by adjusting the total number of emails sent across all sites (N=2415) by the percentage of people who responded that they were not eligible because they were not clinical research coordinators (22% overall).

Survey instruments and GCP knowledge test

The survey instruments included: (a) a 28-item survey of work and training experience (and basic demographics) developed by our research team and informed by qualitative interview findings, (b) a 32-item GCP knowledge test developed by our research team, (c) an existing 5-item learning orientation scale,²² and (d) an existing 13-item, timed measure of verbal-numerical reasoning;²³ instruments were completed by respondents in this order.

The survey of clinical research coordinator work and training experience asked participants to characterize their prior and current work experience coordinating trials. Before administering the survey, we conducted four cognitive interviews with clinical research coordinators to confirm item clarity, and revised items as necessary. Years of work experience as a clinical research coordinator was a continuous variable. Participants selected “all that apply” for types of trials coordinated (drug, device, biologic, or other), coordinator responsibilities (recruitment/enrollment, informed consent, data collection, data entry, regulatory activities, supervising others), and experience, if any, with different types of audits (FDA, institution, study sponsor). These responses were summed to create 3 variables: number of trial types coordinated, number of coordinator responsibilities, and number of types of audits. Number of principal investigators supported in current role and number of trials coordinated in the last 12 months were ordered categorical responses. Finally, 3 dichotomous (yes-no) questions asked participants to indicate if they coordinated federally-funded trials, industry funded trials, and if they had experience serving on an IRB.

Next, participants characterized their formal education and training experience. Respondents indicated their academic degrees and formal certifications held. They estimated the number of online and face-to-face hours of formal GCP training they had participated in over their career so far. We provided instructions to help guide respondents when making their estimate. Our team has previously used this method to estimate prior training hours among researchers.²⁴ Feedback on the items from the cognitive interviews of the survey items indicated that clinical research coordinators felt they could generate an estimate using approximations such as 10-hour intervals.

The GCP knowledge test included 32 multiple-choice items developed by our team.²⁵ All items consisted of a stem and four response options consisting of one best answer and three plausible distractors. The competency domains reflected on the test included clinical trial operations, study site management, ethical and participant safety considerations, and data management and informatics. The items were written according to best practices for multiple-choice items and demonstrated adequate validity and reliability.²⁶

We employed an existing, validated 5-item learning orientation scale.²² A learning orientation reflects a desire to learn new skills, master new situations and improve competence in work settings.²² It was measured as a control variable, as learning orientation may affect knowledge gained from training and learning opportunities on the job; thus we wanted to control for its potential association with GCP knowledge. Participants indicated their agreement to statements such as, “I enjoy challenging and difficult tasks at work where I learn new skills,” using a 7-point Likert scale from 1 (strongly disagree) to 7 (strongly agree). The Cronbach’s alpha coefficient was .76 in this sample. The scores used in analysis were the average of the 5 items.

We included a verbal-numerical reasoning measure²³ also as a control variable when modeling the relationship of GCP knowledge to experiential and individual variables, as verbal-numerical reasoning could affect learning and knowledge acquisition, in addition to performance on a test-taking task. The 13-item measure presented verbal and numerical problems and tasked respondents with answering as many as possible in 2 minutes (skipping any they wished). The test has been administered to over 500,000 healthy participants in the UK Biobank²³ and has a Cronbach’s alpha coefficient of .62 based on data from approximately 112,000 UK Biobank participants.²⁷ Correct answers are summed, thus the maximum possible score is 13. In this sample, the alpha coefficient was .85.

Data analysis

Qualitative interview data were analyzed using structural coding (See supplemental materials).²⁸

Statistical analysis was performed in IBM SPSS Statistics for Windows, version 24.0, released 2016 (IBM Corp., Armonk, NY, USA). Data were downloaded from Qualtrics and imported into SPSS. The accuracy of the data import was verified before performing analysis. There were no missing data, as force-choice responding format was used. We excluded 3 cases from the dataset because a combination of low completion time (< 10 minutes) and extreme low outlier scores on the GCP test (≤37% correct) suggested insufficient effort responding. Respondents were asked to complete the study only once; but 8 individuals submitted two payment forms available at the end of the anonymous link. These data could not be identified for removal.

Descriptive statistics describe the participants’ work characteristics, education and training, and GCP knowledge test scores. We examined bivariate relationships of all potential predictor variables with GCP knowledge scores, using Pearson’s, Spearman’s, or point-biserial correlations depending on the measurement scale of the variable. We set our p-value for statistical significant at p < .01 for consideration of correlations for inclusion in regression.

Next, we performed multiple regression analysis to identify variables that explained a significant portion of variability in GCP knowledge scores. In a first step of the model, we entered verbal-numeric reasoning scores, learning orientation, and institution as control variables. Variables were candidates for entry as predictors if they were significantly positively associated with GCP scores in bivariate correlations, and were hypothesized to be related (e.g., we did not include race or other demographic variables not hypothesized to explain GCP knowledge). Many variables were significantly correlated with GCP knowledge scores; however, they were also correlated with each other (e.g., number of principal investigators supported and number trials coordinated in last 12 months), presenting the potential for estimation problems created by multicollinearity. Thus, when two potential predictor variables were highly inter-correlated (greater than .50), we selected the one with the highest correlation with GCP knowledge.

Although we wished to compare the distinct effects of online and face-to-face training, we ran four models to examine the stability of the estimates with different predictors in the model, and to address our concern about including both types of training in the same model given the high correlation (rho = .41, p<.001) of online and face-to-face training. In the first model (Model A), online and face-to-face training were both entered as separate predictors; in the second (Model B), only face-to-face training was entered; in the third (Model C), only online training was included; and in the fourth (Model D), total hours of combined (online and face-to-face) training was entered. We examined the standardized beta weights because our aim was to understand the relative importance of each predictor. Finally, we ran a model with no training predictors to compare to Model D. Comparison of the R-squared value for the full model with training to the no-training model allowed us to specify the proportion of variance in GCP knowledge explained by training.

Results

Our qualitative interview findings (Supplemental Materials) indicate that the majority of clinical research coordinators did not regard formal training as the most effective method for learning GCP. While individuals did find value in formal training, it was primarily described as providing a refresher, reminder, or baseline knowledge. All individuals emphasized the importance of experiential learning to supplement online training, and the need to get hands on practical experience, observe and ask questions of others, and talk to colleagues.

Responses from 625 clinical research coordinators (Institution I N = 285; Institution II N = 184; Institution III N = 156) were included in the quantitative survey analysis. The demographic characteristics of the sample are provided in Table 1.

Table 1.

Sample Demographics

Age
18–25	12% (76)
26–34	34% (213)
35–44	24% (147)
45–54	17% (104)
55–64	12% (76)
65+	1% (9)
Gender
Female	87% (546)
Male	13% (79)
Race
White	76% (477)
African American	12% (73)
Asian	8% (47)
American Indian or Alaska Native	<1% (2)
Multiple racial categories	3% (21)
Other	1% (5)

The mean GCP knowledge score was 24.8 (SD 3.8), with a range of 7 – 32. The median score was 25 out of 32.

Table 2 presents the descriptive statistics for work experience and characteristics and bivariate associations with GCP knowledge scores. On average, clinical research coordinators had 7 years of experience (SD = 6.6). They coordinated both industry (74%) and federally-funded (72%) studies. While 78% of respondents worked on drug trials, they also coordinated other types of trials including device (35%) and biologic/vaccine trials (26%). Most (83%) were supporting the trials of two or more principal investigators. Clinical research coordinators were responsible for multiple aspects of trial coordination: recruitment, consent, data collection and entry were among the most common activities. Respondents reported having undergone institutional (62%), sponsor (58%), and FDA (21%) initiated audits. A small proportion (5%) have served on an Institutional Review Board (IRB). Years of experience had the strongest association with GCP scores, followed by number of trial types coordinated and industry funding.

Table 2.

Work Experience

		Correlation with GCP score
Years of Experience, mean ± SD	6.92 ± 6.63	0.34, p = 0.001
Trial Types Coordinated (possible range 1–4), mean ± SD	1.75 ± .80	0.28, p = 0.001
Drug Studies	78% (484)†
Device Studies	35% (218)
Biologics/Vaccine Studies	26% (163)
“Other” Studiesa	37% (228)
Coordinator Responsibilities (possible range 1–7), mean ± SD	4.70 ± 1.45	0.13, p = 0.001
Recruitment/Enrollment	85% (528)†
Informed Consent	85% (529)
Data Collection	91% (567)
Data Entry/Management	87% (544)
Regulatory Activities	67% (419)
Supervising Others	48% (298)
“Other”b	9% (55)
Types of Audits Received (possible range 0–4), mean ± SD	1.44 ± 1.03	0.17, p = 0.001
Institution	62% (385)†
Study Sponsor	58% (360)
FDA	21% (131)
“Other”c	4% (25)
Number PIs Supporting		0.16, p = 0.001
1	17% (104)
2–3	39% (242)
4–5	24% (147)
6 or more	21% (132)
Number Trials Coordinated in Last 12 Months		0.22, p = 0.001
1–4	43% (271)
5–9	30% (185)
10–15	12% (77)
16–20	5% (33)
21–30	4% (22)
31+	6% (37)
Federally funded		0.10, p = 0.012
Yes	72% (448)
No	28% (177)
Industry funded		0.23, p = 0.001
Yes	74% (263)
No	26% (162)
Service on an IRB Service		0.12, p = 0.002
Yes	5% (32)
No	95% (593)

^aCommon responses: behavioral studies and observational studies

^bCommon responses: assistance with proposals, data analysis, manuscripts, budgeting matters, and protocol development

^cCommon responses: self-audits, VA, DOD, and IRB audits.

^†“Select all that apply” responses; thus total > 625.

As shown in Table 3, the majority (72%) of clinical research coordinators did not hold a formal certification; 28% were trained nurses. Half had bachelor degrees, and 34% held a master’s degree. A majority reported having 0–10 (45%) or 11–20 hours (22%) of face-to-face GCP training, and 0–10 (41%) and 0–20 (26%) hours of online GCP training. Several education and training variables were associated with GCP scores: certification, hours of online training, hours of face-to-face training all had positive correlations of similar magnitudes.

Table 3.

Education and Formal Training

		Correlation with GCP score
Highest Degree Held		0.03, p = 0.394
High School Diploma or GED	2% (14)
Associate’s Degree	7% (44)
Bachelor’s Degree	50% (312)
Master’s Degree	34% (211)
Doctoral Degree	4% (24)
Other	3% (20)
CRC Certifieda		0.23, p = 0.001
Yes	29% (178)
No	72% (447)
Licensed or Registered Nurse		0.12, p = 0.003
Yes	28% (176)
No	72% (449)
Hours Face-to-Face Training		0.21, p = 0.001
0–10	45% (282)
11–20	22% (138)
21–30	10% (60)
31–40	4% (27)
41–50	4% (22)
51–60	4% (23)
61–90	3% (19)
91–120	2% (11)
121–150	2% (11)
151+	5% (32)
Hours Online Training		0.24, p = 0.001
0–10	41% (258)
11–20	26% (161)
21–30	12% (76)
31–40	7% (45)
41–50	3% (19)
51–60	4% (27)
61–90	2% (14)
91–120	1% (7)
121–150	1% (6)
151+	2% (12)
Reading trade journals		.06, p = 0.130
Weekly or monthly	18% (111)
Few times per year or less	82% (514)

^aParticipants reported their type of certification, selecting “all that apply” (n = 203 certifications reported): ACRP 9% (57); SoCRA 14% (88); Undergraduate CRC certification or degree 4% (23); Graduate certification or degree 10% (6); Other formal training 5% (29).

Verbal-numerical reasoning scores had a mean of 5.6 (SD = 1.6) with a range of 0 – 11 out of a possible 13. This is consistent with the mean of 5.99 (SD = 2.16) reported for approximately 160,000 UK biobank participants (Lyall et al. 2016). Verbal-numerical reasoning was not associated with higher GCP test scores (r = .03, p = .47). The mean learning orientation score was 6.0 (SD = .70) out of a possible score of 7, with a range of 1 to 7. Learning orientation was not correlated with GCP test scores (r = .01, p = .80).

The results of the regression shown in Table 4 indicate that years of experience explained the most variance in GCP scores. Coordinating diverse types of trials, coordinating industry funded trials, and being certified were also statistically significant predictors. Hours of online and of face-to-face training were correlated, and thus, when both are entered into a regression analysis, neither significantly predicts GCP knowledge. However, face-to-face training alone did significantly predict a small amount of variance, and the combined total number of hours of training in GCP (in Model D) explained a slightly greater amount of variance in GCP scores, similar in magnitude to certification, but proportion online was not a significant predictor. Certification and total training remain significant even after accounting for years of experience and richness of experience, which were the strongest predictors. Yet, comparison of the R-squared value for Model D (F(11, 613) = 16.08, p < .001, R² = 0.224) to a no-training model, (F(9, 615) = 18.89, p < .001, R² = .217), indicates that adding training to the model accounts for less than 1% of the variance in GCP knowledge scores.

Table 4.

Regression Analysis Predicting GCP Scores

Variable	Model A: Hours F2F and Online		Model B: Hours F2F only		Model C: Hours Online only		Model D: Total Hours composite with proportion online
Step 1	β	p	β	p	β	p	β	p
Institution	−0.03	0.457	−0.03	0.457	−0.03	0.457	−0.03	0.457
Learning orientation	0.01	0.846	0.01	0.846	0.01	0.846	0.01	0.846
VN reasoning	0.03	0.472	0.03	0.472	0.03	0.472	0.03	0.472
Step 2
Institution	−0.06	0.115	−0.07	0.058	−0.05	0.156	−0.06	0.138
Learning orientation	−0.02	0.628	−0.02	0.680	−0.02	0.687	−0.02	0.621
VN reasoning	0.08	0.041	0.07	0.043	0.07	0.047	0.07	0.042
Years of experience	0.27	0.001	0.27	0.001	0.27	0.001	0.27	0.001
Industry funded	0.13	0.001	0.14	0.001	0.13	0.001	0.13	0.001
Number trial types	0.18	0.001	0.18	0.001	0.18	0.001	0.18	0.001
Number trials 12 months	0.04	0.372	0.04	0.323	0.03	0.426	0.04	0.395
Number types of audits	−0.03	0.400	−0.03	0.412	−0.03	0.448	−0.03	0.405
Certified	0.12	0.003	0.12	0.003	0.13	0.001	0.12	0.003
Total Hours F2F	0.06	0.127	0.08	0.032
Total Hours Online	0.04	0.305			0.07	0.067
Total Hours Training							0.10	0.016
Proportion Online							0.01	0.729
R2	0.22		0.22		0.22		0.22
F-value, p-value	16.06, 0.001		17.56, 0.001		17.40, 0.001		16.08, 0.001

β = Standardized Beta coefficient; p = p-value; VN = verbal-numerical. Bold variables are those with the largest β’s (≥0.10); thus they explain the most variance in the models.

Discussion

The exploratory qualitative interview findings indicated that experience, day to day practice, observing peers and colleagues, and having mentors are essential elements to learning GCP. Individuals described the benefits of formal training primarily in terms of an initial introduction to GCP as a new employee, a refresher or reminder of GCP, or as necessary for being in regulatory compliance rather than as their main knowledge source for how they learned to put GCP into practice. This led us to include diverse questions regarding prior experience, type of work environment, and responsibilities.

Our quantitative analysis revealed that years of experience as a clinical research coordinator was the strongest predictor of GCP knowledge. Additional significant predictors include, in descending order, working on diverse types of trials, working on industry funded trials, holding a formal certification, and total hours of training. Training was associated with GCP knowledge, but the effect was small. Our findings suggest that the type of training—whether online or face-to-face—does not make a clinically significant difference in GCP knowledge scores; although face-to-face training performed slightly better in one regression model, it was highly correlated with hours of online training, and the total number of hours clearly explained more variance in levels of GCP knowledge. Overall, the study indicates that the effect of training is minimal, accounting for less than 1% of the variance, even when online and face-to-face training hours are aggregated.

Our findings suggest that learning on the job through day-to-day experience and exposure to diverse clinical research settings (through exposure to more trial types and industry) and rigorous training through formal certifications or degrees contribute to GCP knowledge more than interventions such as online GCP modules or face-to-face GCP training. Clinical research coordinators who work on industry funded trials might be expected to know more (e.g., about FDA processes) and may have more resources available to them (e.g., compliance manuals). As expected, verbal-numerical reasoning was associated with GCP knowledge; the magnitude of this effect was similar to that of training. Thus, we would recommend future studies take into account reasoning when examining GCP knowledge. The learning orientation scale may not have been significant in the models due to range restriction. It is likely that people attracted to this type of work in research share a motivation to learn new things. Thus, it is likely unnecessary to consider this variable in future studies.

Our findings support the Joint Taskforce for Clinical Trial Competency suggestion that online training is insufficient on its own and should be supplemented by additional training and education.¹⁹ Online and face-to-face training modules likely provide a baseline of knowledge, and may be especially important for those who are new to the job. Nevertheless, day-to-day experience and learning on the job, combined with exposure to a diverse work environment through multiple types of trials, are essential to gaining GCP knowledge. This is analogous to training in medicine: While classroom education provides a useful foundation, the practice of medicine is learned largely through clinical experience in diverse, supervised settings.

Our study has several limitations. First, we used a regression model to understand the factors that predict GCP knowledge. This enables us to determine what factors are associated with GCP knowledge and, in combination with our qualitative data, to build a theory of how clinical research coordinators learn GCP. But we were unable to use an experimental (randomized) design or even a quasi-experimental (pre- and post-testing design). In part, this was due to the fact that GCP training became mandatory shortly before our study commenced; it is now a standard part of clinical research coordinator onboarding. However, our design had the advantage of controlling for diverse individual factors (such as verbal-numerical reasoning and learning orientation), and enabled us to explore the association of a wide variety of variables with GCP knowledge. Second, our study relied upon self-reported learning experiences and behaviors. The fact that the relationships between these self-reported behaviors and GCP knowledge were consistently in the directions hypothesized offers some evidence of the validity of this approach, but we did not have a way to verify the information about learning experiences and behaviors. Third, our study was limited to clinical research coordinators working in large academic medical centers with CTSAs, which provide a lot of training opportunities and hire individuals with relatively high levels of education. It is possible that our findings might be different with those working in a larger variety of clinical research settings. Finally, our multiple regression model explained only 22% of the total variance in GCP knowledge scores indicating that other factors are at work and further research is needed.

The present study contributes to the clinical research community’s understanding of some key contributors to GCP knowledge among clinical research coordinators.

The GCP test items have not been published because publishing items and answer keys diminishes their value for testing purposes. The authors will make the GCP Knowledge Test available to those who wish to assess GCP training programs or the knowledge of clinical research associates. The test and scoring guide may be requested on the research team’s test service webpage: [url will be inserted following peer review].