Association of Protocol Design‐Related Factors With Enrollment Duration in Clinical Trials

ABSTRACT

Patient enrollment is a critical factor in the success of clinical trials; however, many trials fail to recruit participants within a planned timeline. Protocol design is a key barrier to efficient recruitment. However, few studies have quantitatively evaluated the association between protocol design and enrollment duration across various disease areas and time periods. We conducted a cross‐sectional study using data from ClinicalTrials.gov and another related database. Industry‐sponsored phase 2 and 3 drug trials completed between January 2017 and September 2024 were included. Key variables related to the protocol design were extracted from trial protocols and database entries. The enrollment duration was estimated using the trial start dates, primary completion dates, and outcome time frames. Multiple linear regression analyses were conducted to identify factors associated with enrollment duration, including comparisons between the pre‐ and post‐COVID‐19 pandemic periods. In total, 1286 trials met the inclusion criteria. Multiple linear regression identified three variables significantly associated with a longer enrollment duration: number of original enrollments (p = 0.023), number of inclusion criteria (p < 0.001), and days of assessment (p = 0.013). Before the pandemic, the number of assessment days was the primary factor, while after the pandemic, the number of inclusion criteria had the strongest association. This study demonstrated that the enrollment duration was significantly influenced by protocol design features. To improve recruitment efficiency, careful consideration of the inclusion criteria, assessment burden, and enrollment targets is essential during trial planning. Causal relationships should be interpreted cautiously and further examined in future research.

Study Highlights

• What is the current knowledge on the topic?

  • ○
    In clinical trials, protocol design has been recognized as a potential barrier to successful participant enrollment. Previous studies have reported the associations between certain design elements and recruitment performance; however, many have been limited to specific therapeutic areas or developmental phases. Furthermore, few studies have quantitatively and cross‐sectionally assessed how protocol design affects enrollment duration.
• What question did this study address?

  • ○
    This study evaluated the relationship between enrollment duration and protocol design elements using data from industry‐sponsored drug trials registered at ClinicalTrials.gov. Enrollment duration was estimated based on the registered start date, primary completion date, and time frame of the primary outcome. To enable a more comprehensive assessment of the protocol design, variables were obtained not only from structured registry data but also through manual extraction from individual trial protocols.
• What does this study add to our knowledge?

  • ○
    This cross‐sectional study offers a comprehensive analysis of the protocol elements associated with the duration of participant enrollment in industry‐sponsored drug trials. It revealed that the number of original enrollments, number of inclusion criteria, and days of assessment were significantly associated with the enrollment duration.
• How might this change clinical pharmacology or translational science?

  • ○
    The identified elements are available at the planning stages of the clinical trial. Early evaluation of these factors in relation to the anticipated recruitment capacity may help set more realistic enrollment timelines and reduce the risk of enrollment delays or failure.

1. Introduction

Patient enrollment remains a major challenge in clinical trials, with 19% of trials reportedly failing to recruit a sufficient number of participants [1]. Additionally, only approximately 50% of trials complete enrollment within the planned timeline [2, 3, 4]. The median cost of a single trial has been estimated at $19 million [5], and enrollment delays can prolong trial duration and consequently increase development costs. Inadequate sample sizes not only increase the risk of trial failure but also limit the patients' access to novel therapies.

Recent efforts have focused on identifying factors that improve the success rate of clinical trials, with protocol design highlighted as one potential barrier to patient enrollment. Specifically, unnecessary eligibility criteria and excessive data collection may hinder recruitment, and recommendations have been made to address these aspects [6, 7, 8, 9, 10].

Although prior studies have examined the relationship between patient enrollment and protocol design, many have been limited to specific disease areas, development phases, or selected aspects of protocol design [11, 12, 13, 14]. Furthermore, few studies have quantitatively assessed the impact of protocol design on the duration of patient enrollment (hereafter, enrollment duration), and none have comprehensively evaluated this relationship.

The coronavirus disease 2019 (COVID‐19) pandemic has introduced additional complexity to the clinical trial landscape. It may have disrupted trial operations, altered protocol designs, and affected recruitment activities, underscoring the need to examine potential differences between the pre‐ and post‐COVID‐19 pandemic periods.

In this study, several inclusion criteria were established to ensure a homogeneous and comparable study population. We restricted our analysis to industry‐sponsored clinical trials, as prior studies have shown that such trials generally have shorter enrollment durations than investigator‐initiated trials [15]. Furthermore, clinical trials involving drugs account for a large proportion of all trials [16] and are thus considered representative for general analysis. Clinical trials involving pediatric patients were excluded, as they face significant enrollment challenges owing to ethical and procedural complexities [17]. Finally, we focused only on phase 2 and 3 trials, as phase 1 trials often involve healthy volunteers [18] and differ fundamentally in enrollment dynamics.

Therefore, this study aimed to comprehensively evaluate the relationship between enrollment duration and protocol design using data from ClinicalTrials.gov and the Clinical Trials Transformation Initiative (CTTI) Aggregate Analysis of ClinicalTrials.gov (AACT) database. We also examined changes in protocol design and enrollment status between the pre‐ and post‐COVID‐19 pandemic, thereby providing insights into future trial designs.

2. Methods

2.1. Study Design and Data Sources

This cross‐sectional study analyzed data extracted from ClinicalTrials.gov [19] and the AACT database [20]. This study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.

ClinicalTrials.gov is one of the largest clinical trial databases globally, managed by the National Library of Medicine at the National Institutes of Health and includes trials conducted by both public and private sponsors worldwide. The AACT database, provided by the Clinical Trials Transformation Initiative, organizes and structures the ClinicalTrials.gov data to facilitate analysis.

This study used publicly available data from these databases, and therefore did not require approval from an institutional review board or ethics committee.

2.2. Data Extraction

A snapshot version of the AACT database as of September 30, 2024, was downloaded for analysis. Trials were eligible if participants were adults or older adults, phase was 2 or 3, study type was interventional, overall status was completed or terminated, lead sponsor was industry, intervention type was drug or biological, and a protocol document was available.

Trials in the pre‐COVID‐19 pandemic period were defined as those with a start date on or after January 18, 2017, and a completion date on or before March 31, 2020. Trials in the post‐COVID‐19 pandemic period were defined as those with a start date on or after April 1, 2020, and a completion date on or before September 30, 2024. January 18, 2017, was used as the lower bound for trial start dates, as it marked the beginning of the mandatory registration of start dates on ClinicalTrials.gov.

The protocols of eligible trials were downloaded from ClinicalTrials.gov. As the protocol version implemented during the enrollment period could not be consistently identified, a standardized extraction procedure was used. Information was extracted from a single protocol version when available and from the most recent available protocol when multiple versions were available.

2.3. Variables of Interest

The primary outcome was enrollment duration, estimated using the trial start date, primary completion date, and time frame of the primary outcome listed in the database (Figure 1). On ClinicalTrials.gov, the start date is defined as the date when the first participant is enrolled, whereas the primary completion date is defined as the date when the final participant completes the primary outcome assessment. As the interval between the enrollment of the last participant and the primary completion date is expected to be largely driven by the primary outcome time frame, the enrollment duration was estimated as the period from the start date to the primary completion date minus the primary outcome time frame. When multiple time frames were reported for a trial's primary outcome, the longest duration was adopted.

FIGURE 1.

Estimation of the enrollment duration. Enrollment duration was estimated using the trial start date, primary completion date, and the time frame of the primary outcome listed in the database. The definitions of the start date and primary completion date according to ClinicalTrials.gov are described in the footnotes below. Because the interval between enrollment of the last participant and the primary completion date is largely determined by the primary outcome time frame, enrollment duration was estimated as the period from the start date to the primary completion date minus the primary outcome time frame. ^aPrimary Completion Date: The date on which the last participant in a clinical study was examined or received an intervention to collect final data for the primary outcome measure. ^bStart Date: The actual date on which the first participant was enrolled in the clinical study. ^cPrimary outcome time frame: The duration of the primary outcome measure.

For protocol design‐related variables, selection was guided by a hypothesis‐driven framework grounded in the CTTI Quality by Design project—Critical to Quality (CTQ) factors principles document [21]. In this report, “protocol design” refers to structured trial characteristics that define eligibility criteria, assessment number and visits, and enrollment targets, based on information consistently available in ClinicalTrials.gov and publicly available protocol documents. From the AACT database, we extracted protocol design elements relevant to the CTQ framework that were consistently available as structured variables. Qualitative or narrative elements (e.g., operational descriptions that could not be standardized across trials) were not included, as they could not be reliably and uniformly analyzed across studies. Accordingly, the following variables were extracted: number of secondary outcomes, other outcomes, number of intervention arms, randomization, masking, and placebo. Since other outcomes were often marked as not applicable, we categorized them as not applicable or ≥ 1. The distribution of the number of intervention arms was reviewed, and the trials were classified into three categories: one arm, two arms, and three or more arms. The following variables were manually extracted from the protocols: number of inclusion criteria, number of exclusion criteria, participation duration, number of assessments, and days of assessment. The number of assessments was the total count of distinct assessments or examinations specified in the trial protocol. Days of assessment were defined as the total number of distinct calendar days on which any trial‐related assessments or evaluations were scheduled. The details of the extraction rules are provided in Table S1.

The number of original enrollments (target sample size) was manually extracted from the ClinicalTrials.gov.

In addition, we included the trial phase and disease area as adjustment variables because they are considered major factors influencing both protocol design and enrollment duration. In the AACT database, each trial's disease condition was coded using the Medical Subject Headings (MeSH) terms. Following previous studies, disease areas were categorized as neoplasms, respiratory tract diseases, infections, endocrine system diseases, cardiovascular diseases, nervous system diseases, autoimmune diseases, and other diseases [22]. If a trial's disease conditions spanned multiple areas, one representative category was selected based on protocol content. Neoplasms constituted the largest proportion of trials included in this study and were therefore selected as the reference category, allowing for stable and interpretable comparisons across disease areas.

2.4. Statistical Analysis

From the perspective of estimating and analyzing enrollment duration, trials that met the following exclusion criteria were omitted from the analysis: disease conditions not coded with MeSH terms; trials covering multiple phases (e.g., phase 1/2); rollover trials with parent studies; trials with multiple enrollment periods (e.g., interim analyses); trials lacking a duration‐based description in the primary endpoint time frame; trials with estimated enrollment durations of less than 1 day; and trials from which no protocol‐design‐related variables could be manually extracted.

Descriptive statistics were calculated for all variables. Univariable and multiple linear regression analyses were conducted to assess the relationship between each variable and enrollment duration. Continuous variables with skewness ≥ 2 were log‐transformed. Multiple linear regression was performed using only the variables that were significant in the univariable analysis (p < 0.05) and had a variance inflation factor (VIF) < 10, both to reduce the risk of model overfitting and multicollinearity and to improve model interpretability given the number of candidate variables.

Statistical significance was set at p < 0.05. Missing values were excluded from the analysis. All analyses were conducted using R version 4.4.2 (R Core Team, Vienna, Austria) and relevant packages.

3. Results

The data extraction process is illustrated in Figure 2. From the AACT database snapshot as of September 30, 2024, 1623 industry‐sponsored drug trials were identified based on the predefined inclusion criteria. After excluding 337 trials, a total of 1286 trials were included across the overall period. Of these, 737 were conducted during the pre‐COVID‐19 pandemic period, and 549 were conducted during the post‐COVID‐19 pandemic period.

FIGURE 2.

Flowchart of the selection and exclusion process for trials. *Trials were selected according to the following criteria: Age group set to “adult” and “older adult”; trial phase set to “phase 2” and “phase 3”; study type set to “interventional”; overall status set to “completed” and “terminated”; lead sponsor set to “industry”; intervention type set to “drug” and “biological”; and availability of “protocol” among the provided documents. Trials in the pre‐COVID‐19 pandemic period were defined as those with a start date on or after January 18, 2017, and a completion date on or before March 31, 2020. Trials in the post‐COVID‐19 pandemic period were defined as those with a start date on or after April 1, 2020, and a completion date on or before September 30, 2024.

Descriptive statistics for variables other than disease areas across the study periods are presented in Table 1. The median (IQR) for the number of original enrollments was 160 (69–360) for the overall period, 167 (60–375) for the pre‐COVID‐19 pandemic period, and 156 (80–331) for the post‐COVID‐19 pandemic period. The median (IQR) for the number of inclusion criteria was 9 (7–11) across all periods. For the number of exclusion criteria, the medians were 22 (17–28) for the overall period, 22 (17–27) for the pre‐COVID‐19 pandemic period, and 22 (17–28) for the post‐COVID‐19 pandemic period. Regarding other outcomes, 91.8% were categorized as not applicable and 8.2% as ≥ 1 in the overall period; 90.2% vs. 9.8% in the pre‐COVID‐19 pandemic period; and 94.0% vs. 6.0% in the post‐COVID‐19 pandemic period. Regarding the number of intervention arms during the overall period, 12.1%, 49.7%, and 38.4% had one, two, and three or more arms, respectively, with two‐arm trials comprising approximately half. This trend was consistent in both the pre‐ (46.5%) and post‐COVID‐19 pandemic periods (53.9%).

TABLE 1.

Descriptive statistics for variables.

	Overall period			Pre‐COVID‐19 pandemic period			Post‐COVID‐19 pandemic period
	n	(%)	Median (IQR)	n	(%)	Median (IQR)	n	(%)	Median (IQR)
Number of original enrollments	1285	—	160 (69–360)	736	—	167 (60–375)	549	—	156 (80–331)
Number of secondary outcomes	1157	—	7 (3–13)	647	—	6 (3–14)	510	—	7 (4–13)
Other outcomes	1286	—	—	737	—	—	549	—	—
Not applicable	1181	91.8	—	665	90.2	—	516	94.0	—
≥ 1	105	8.2	—	72	9.8	—	33	6.0	—
Number of inclusion criteria	1259	—	9 (7–11)	724	—	9 (7–11)	535	—	9 (7–11)
Number of exclusion criteria	1251	—	22 (17–28)	722	—	22 (17–27)	529	—	22 (17–28)
Participation duration, days	1135	—	169 (78–343)	651	—	168 (80–316)	484	—	176 (72–362)
Number of assessments	1081	—	28 (23–36)	628	—	28 (23–36)	453	—	28 (23–36)
Days of assessment	1134	—	10 (7–16)	659	—	10 (7–16)	475	—	11 (8–17)
Number of intervention arms	1286	—	—	737	—	—	549	—	—
1 arm	155	12.1	—	94	12.8	—	61	11.1	—
2 arms	639	49.7	—	343	46.5	—	296	53.9	—
≥ 3 arms	492	38.3	—	300	40.7	—	192	35.0	—
Randomization	1286	—	—	737	—	—	549	—	—
Yes	1089	84.7	—	615	83.4	—	474	86.3	—
No	197	15.3	—	122	16.6	—	75	13.7	—
Masking	1286	—	—	737	—	—	549	—	—
Yes	963	74.9	—	540	73.3	—	423	77.0	—
No	323	25.1	—	197	26.7	—	126	23.0	—
Placebo	1286	—	—	737	—	—	549	—	—
Yes	738	57.4	—	398	54.0	—	340	61.9	—
No	548	42.6	—	339	46.0	—	209	38.1	—
Trial phase	1286	—	—	737	—	—	549	—	—
Phase 2	769	59.8	—	420	57.0	—	349	63.6	—
Phase 3	517	40.2	—	317	43.0	—	200	36.4	—

Abbreviation: IQR, interquartile range.

Table 2 shows the enrollment duration for all trials and by disease areas across the overall, pre‐COVID‐19 pandemic, and post‐COVID‐19 pandemic periods. For all trials, the median enrollment durations were 261 days (IQR: 151–424) in the overall period, 265 days (155–438) in the pre‐COVID‐19 pandemic period, and 258 days (147–409) in the post‐COVID‐19 pandemic period. When examined descriptively based on median values, enrollment duration appeared to vary across disease areas between the pre‐ and post‐pandemic periods. For example, trials on infections tended to have shorter median enrollment durations in the post‐pandemic period, whereas trials on endocrine system diseases and nervous system diseases tended to have longer median enrollment durations. Of note, these observations are descriptive only and were not formally tested for statistical significance.

TABLE 2.

Enrollment duration of all trials and each disease area.

	Overall period			Pre‐COVID‐19 pandemic period			Post‐COVID‐19 pandemic period
	n	(%)	Days, median (IQR)	n	(%)	Days, median (IQR)	n	(%)	Days, median (IQR)
All trials	1286	—	261 (151–424)	737	—	265 (155–438)	549	—	258 (147–409)
Disease area
Neoplasms	136	10.6	177 (32–338)	77	10.4	211 (55–377)	59	10.7	120 (26–234)
Respiratory tract diseases	109	8.5	215 (128–362)	56	7.6	222 (124–349)	53	9.7	200 (128–390)
Infections	190	14.8	238 (137–392)	78	10.6	304 (177–469)	112	20.4	217 (127–312)
Endocrine system diseases	96	7.5	249 (167–416)	74	10.0	235 (166–408)	22	4.0	333 (202–423)
Cardiovascular diseases	84	6.5	310 (215–474)	53	7.2	265 (164–477)	31	5.6	332 (250–469)
Nervous system diseases	138	10.7	331 (198–495)	86	11.7	295 (186–494)	52	9.5	351 (205–510)
Autoimmune diseases	118	9.2	344 (223–475)	70	9.5	339 (221–474)	48	8.7	372 (223–477)
Other diseases	415	32.3	267 (159–416)	243	33.0	256 (142–413)	172	31.3	280 (169–425)

Abbreviation: IQR, interquartile range.

The results of the univariable and multiple linear regression analyses for enrollment duration are shown in Tables 3, 4, 5. To account for skewed distributions, the number of original enrollments, number of secondary outcomes, participation duration and days of assessment were log‐transformed prior to regression analyses. Univariable analyses were performed for all the protocol design‐related and adjustment variables. Multiple linear regression was subsequently conducted using variables that showed a significant association with enrollment duration (p < 0.05) in the univariable analyses. In the multiple linear regression for the overall period (Table 3), enrollment duration was significantly associated with the number of original enrollments (adjusted effect coefficient [aEC] = 13, p = 0.023); number of inclusion criteria (aEC = 6.6, p < 0.001); and days of assessment (aEC = 32, p = 0.013). Specifically, holding other variables constant, each additional inclusion criterion was associated with an approximately 6.6‐day longer enrollment duration. For log‐transformed variables, a 10% increase in the number of original enrollments was associated with an approximately 1.2‐day longer enrollment duration, whereas a 10% increase in days of assessment was associated with an approximately 3.0‐day longer enrollment duration.

TABLE 3.

Univariable and multiple linear regression analyses with enrollment duration (overall period).

	Estimated coefficient	Univariable			Adjusted estimated coefficient	Multiple
		95% CI		p
		95% CI				p
Lower limit	Upper limit	Lower limit	Upper limit
Number of original enrollments a	17	8.2	26	< 0.001	13	1.9	25	0.023
Number of secondary outcomes a	−4.4	−16	6.9	0.44	—
Other outcomes
≥ 1 (vs. Not applicable)	25	−14	64	0.21	—
Number of inclusion criteria	4.8	1.8	7.8	0.002	6.6	2.8	10	< 0.001
Number of exclusion criteria	2.9	1.6	4.1	< 0.001	0.17	−1.6	1.9	0.84
Participation duration a	28	16	39	< 0.001	4.6	−13	22	0.60
Number of assessments	2.9	1.7	4.0	< 0.001	1.3	−0.22	2.8	0.094
Days of assessment a	44	26	62	< 0.001	32	6.7	56	0.013
Number of intervention arms
1 arm	Ref				Ref
2 arms	74	40	108	< 0.001	−25	−109	59	0.56
≥ 3 arms	85	50	120	< 0.001	−38	−123	46	0.37
Randomization
Yes (vs. No)	72	42	101	< 0.001	50	−37	137	0.26
Masking
Yes (vs. No)	67	42	91	< 0.001	29	−21	79	0.26
Placebo
Yes (vs. No)	43	22	65	< 0.001	8.2	−23	40	0.61
Trial phase
Phase 3 (vs. Phase 2)	14	−8.1	35	0.22	—
Disease area
Respiratory tract diseases (vs neoplasms)	39	−9.4	87	0.11	−88	−161	−15	0.018
Infections (vs neoplasms)	51	9.0	94	0.017	−69	−138	0.84	0.053
Endocrine system diseases (vs neoplasms)	66	16	117	0.01	−86	−162	−11	0.025
Cardiovascular diseases (vs neoplasms)	116	64	169	< 0.001	12	−65	90	0.76
Nervous system diseases (vs neoplasms)	120	74	166	< 0.001	−5.7	−78	67	0.88
Autoimmune diseases (vs neoplasms)	134	87	182	< 0.001	−1.9	−75	72	0.96
Other diseases (vs neoplasms)	80	42	117	< 0.001	−36	−101	30	0.29

Abbreviation: CI, confidence interval.

^a Logarithmic transformation was adopted.

TABLE 4.

Univariable and multiple linear regression analyses with enrollment duration (pre‐COVID‐19 pandemic period).

	Estimated coefficient	Univariable			Adjusted estimated coefficient	Multiple
		95% CI		p
		95% CI				p
Lower limit	Upper limit	Lower limit	Upper limit
Number of original enrollments a	18	6.9	30	0.002	12	−2.3	27	0.097
Number of secondary outcomes a	3.1	−12	18	0.68	—
Other outcomes
≥ 1 (vs. Not applicable)	29	−19	78	0.24	—
Number of inclusion criteria	2.5	−1.6	6.7	0.23	—
Number of exclusion criteria	1.6	−0.3	3.4	0.10	—
Participation duration a	29	13	44	< 0.001	−1.2	−25	23	0.92
Number of assessments	2.4	0.9	3.9	0.002	0.97	−0.89	2.8	0.30
Days of assessment a	61	37	86	< 0.001	66	30	101	< 0.001
Number of intervention arms
1 arm	Ref				Ref
2 arms	82	37	127	< 0.001	5.9	−95	107	0.91
≥ 3 arms	90	44	136	< 0.001	10	−91	112	0.84
Randomization
Yes (vs. No)	69	31	108	< 0.001	36	−70	142	0.51
Masking
Yes (vs. No)	60	28	92	< 0.001	20	−45	85	0.54
Placebo
Yes (vs. No)	35	5.8	64	0.019	13	−27	54	0.53
Trial phase
Phase 3 (vs Phase 2)	8.6	−21	38	0.56	—
Disease area
Respiratory tract diseases (vs neoplasms)	−3.2	−72	65	0.93	−96	−197	4.7	0.062
Infections (vs neoplasms)	71	8.5	134	0.026	−37	−133	59	0.45
Endocrine system diseases (vs neoplasms)	23	−41	86	0.48	−139	−238	−40	0.006
Cardiovascular diseases (vs neoplasms)	76	6.7	146	0.032	−5.5	−109	98	0.92
Nervous system diseases (vs neoplasms)	73	12	134	0.018	−39	−138	59	0.43
Autoimmune diseases (vs neoplasms)	92	28	156	0.005	−26	−125	74	0.61
Other diseases (vs neoplasms)	41	−9.8	92	0.11	−56	−145	34	0.22

Abbreviation: CI, confidence interval.

^a Logarithmic transformation was adopted.

TABLE 5.

Univariable and multiple linear regression analyses with enrollment duration (post‐COVID‐19 pandemic period).

	Estimated coefficient	Univariable			Adjusted estimated coefficient	Multiple
		95% CI		p
		95% CI				p
Lower limit	Upper limit	Lower limit	Upper limit
Number of original enrollments a	15	1.4	29	0.031	14	−3.9	33	0.12
Number of secondary outcomes a	−17	−35	0.85	0.062	—
Other outcomes
≥ 1 (vs. Not applicable)	11	−56	78	0.74	—
Number of inclusion criteria	7.7	3.3	12	< 0.001	11	5	17	< 0.001
Number of exclusion criteria	4.2	2.5	5.9	< 0.001	0.070	−2.5	2.7	0.96
Participation duration a	27	8.7	44	0.004	4.4	−18	27	0.70
Number of assessments	3.5	1.8	5.2	< 0.001	1.4	−0.92	3.8	0.23
Days of assessment a	20	−7.8	49	0.15	—
Number of intervention arms
1 arm	Ref				Ref
2 arms	65	13	117	0.014	−45	−184	94	0.53
≥ 3 arms	77	23	131	0.006	−87	−228	53	0.22
Randomization
Yes (vs. No)	78	32	123	< 0.001	80	−64	224	0.27
Masking
Yes (vs. No)	78	41	115	< 0.001	14	−61	89	0.72
Placebo
Yes (vs. No)	59	26	91	< 0.001	13	−37	63	0.61
Trial phase
Phase 3 (vs. Phase 2)	19	−14	52	0.26	—
Disease area
Respiratory tract diseases (vs neoplasms)	92	25	160	0.007	−42	−148	63	0.43
Infections (vs neoplasms)	59	1.3	116	0.045	−37	−138	64	0.48
Endocrine system diseases (vs neoplasms)	143	54	232	0.002	26	−85	157	0.56
Cardiovascular diseases (vs neoplasms)	171	92	250	< 0.001	69	−43	182	0.23
Nervous system diseases (vs neoplasms)	185	117	253	< 0.001	78	−27	184	0.15
Autoimmune diseases (vs neoplasms)	190	121	259	< 0.001	64	−44	172	0.24
Other diseases (vs neoplasms)	130	77	184	< 0.001	24	−71	119	0.62

Abbreviation: CI, confidence interval.

^a Logarithmic transformation was adopted.

Separate multiple linear regressions were conducted for the pre‐ and post‐COVID‐19 pandemic periods, with results shown in Tables 4 and 5, respectively. During the pre‐COVID‐19 pandemic period, only the number of days of assessment (aEC = 66, p < 0.001) was identified as a significant factor. Conversely, in the post‐COVID‐19 pandemic period, only the number of inclusion criteria (aEC = 11, p < 0.001) remained significant.

4. Discussion

This cross‐sectional study utilized data from ClinicalTrials.gov to estimate enrollment duration and evaluate its association with the protocol design. The results identified three protocol design‐related factors—number of original enrollments, number of inclusion criteria, and days of assessment—as significantly associated with enrollment duration in industry‐sponsored drug trials.

To the best of our knowledge, only one previous study has assessed the relationship between protocol design and enrollment duration; however, the study evaluated individual protocol design elements separately and did not identify specific design elements independently associated with enrollment duration [23]. Compared with that study, the present study has several novel aspects. First, the earlier study did not identify the specific protocol design factors affecting enrollment duration, whereas our study identified three such variables. Second, the use of ClinicalTrials.gov enabled analyses across a broader sample size and a wider range of disease areas. Third, our study incorporated a new comparison between the pre‐ and post‐COVID‐19 pandemic periods.

The observed association between a higher number of inclusion criteria and a longer enrollment duration indicates that trials with more stringent eligibility requirements tended to have fewer eligible patients. Similarly, previous reports have shown that stricter eligibility criteria are often associated with a smaller pool of potential participants [24, 25, 26, 27]. However, these findings are descriptive and do not establish causal relationships.

Similarly, an association between a higher number of original enrollments and longer enrollment duration was observed in this study. Although no prior studies have directly evaluated this relationship, earlier research has reported that larger sample size targets are often linked to under‐enrollment [28]. Larger enrollment targets are frequently accompanied by an increased number of participating countries or sites [29, 30], however, in this study, enrollment duration tended to be longer for trials with higher original enrollment targets.

More stringent eligibility criteria can reduce the pool of candidate patients and increase the risk of recruitment delays. However, narrowing the eligibility criteria allows trials to focus on a more specific patient population, which may enhance the clarity of treatment effects. This can lead to greater differences between groups in outcomes and, as a result, improve the likelihood of detecting statistically significant effects even with smaller sample sizes [31]. Another study similarly suggested that narrower eligibility criteria may enhance statistical power and reduce required sample sizes [24]. Therefore, optimizing the balance between the number of inclusion criteria and the target sample size is crucial to maximize recruitment efficiency.

We also observed that a higher number of days of assessment was associated with a longer enrollment duration. Increased assessment days have also been associated with longer enrollment durations in prior reports, which suggested that frequent hospital visits and greater time commitments may pose psychological and logistical burdens for patients [32, 33].

To explore potential differences before and after the COVID‐19 pandemic, we conducted exploratory stratified analyses comparing the pre‐ and post‐pandemic periods; these analyses were not intended to establish causal effects. In the pre‐COVID‐19 pandemic period, the number of assessment days appeared to be the factor most strongly associated with enrollment duration. During this period, many clinical trial assessments were conducted on‐site at medical facilities [34, 35, 36], which may have contributed to the observed association between assessment intensity and longer enrollment duration.

In the post‐COVID‐19 pandemic period, the number of inclusion criteria appeared to be the dominant factor. Analysis by disease area across the pre‐ and post‐pandemic periods suggested that trials in endocrine system diseases and nervous system diseases tended to have an increase in inclusion criteria after the pandemic, whereas trials in respiratory tract diseases and infections tended to have a decrease (Table S2). These observations are descriptive and should be interpreted cautiously. While the findings may reflect shifts in trial conduct or patient recruitment patterns, causal conclusions cannot be drawn. It is possible that factors such as avoidance of medical visits or stricter eligibility criteria contributed to these descriptive trends [37, 38, 39], but this could not be formally tested.

In this study, we identified three protocol design elements associated with enrollment duration. Whether these factors are also related to overall trial duration or the time from trial initiation to drug approval represents an important and relevant question. To our knowledge, few studies have directly evaluated these relationships. One prior study reported that greater protocol complexity was associated with longer trial execution time, conceptually supporting the notion that protocol design may influence time‐related outcomes [40]. However, the study did not evaluate individual design elements (such as the number of inclusion criteria or days of assessment) independently, nor did it directly analyze time to regulatory approval. Accordingly, future research is warranted to examine not only enrollment duration but also overall trial duration and time to approval in relation to specific protocol design elements.

However, this study has some limitations. First, as it is observational in nature, it does not establish a causal relationship between the protocol design and enrollment duration. Second, the use of ClinicalTrials.gov may limit the comprehensiveness and generalizability of the findings, as trials not registered on the platform or without publicly available protocol PDFs were excluded, and some registry fields were missing or inconsistent, possibly affecting data completeness and accuracy [41, 42]. Third, owing to the cross‐sectional and registry‐based nature of the study, analyses were restricted to quantitative protocol variables that were consistently available across trials. Consequently, we could not assess qualitative aspects of protocol design, such as clinical severity, invasiveness, or operational burden. Additionally, the protocol version implemented during the enrollment period could not be consistently identified, and extraction based on the most recent protocol may not fully reflect amendments introduced during recruitment, including changes intended to improve enrollment. Some protocol amendments primarily involve clarifications or administrative adjustments rather than substantive changes to core design elements [43], although the impact of amendments on enrollment cannot be fully excluded. Fourth, although multiple linear regression was used to describe associations between protocol variables and enrollment duration, variable selections for multivariable models was constrained to those significant in univariable analyses, which may not fully account for all potential confounders. Model fit statistics (e.g., R ²) were not reported because the analyses were intended to be descriptive rather than predictive. While a VIF threshold was applied to mitigate multicollinearity, residual correlation between variables cannot be completely excluded. Fifth, rare diseases may pose unique recruitment challenges; however, because rare‐disease trials could not be reliably identified using registry coding, we were unable to determine their distribution across disease categories. Finally, the primary outcome, enrollment duration, was estimated and may have differed from the actual duration. In addition, heterogeneity or imprecision in the primary outcome time frames reported on ClinicalTrials.gov may introduce non‐differential measurement error in the estimated enrollment duration.

In conclusion, this observational study using public databases revealed that the number of original enrollments, number of inclusion criteria, and days of assessment were significantly associated with enrollment duration. These factors critically influence recruitment effectiveness and should be evaluated early in the trial planning stage to align with site resources and anticipated recruitment capabilities. This may improve the realism of enrollment timelines and reduce the risk of recruitment failure. Causal relationships between protocol design and enrollment duration should be interpreted cautiously and warrant further investigation in future studies.

Author Contributions

H.T., M.S., Y.N., and T.N. wrote the manuscript. H.T., M.S., Y.N., N.H., T.S., and S.S. designed the research. H.T., M.S., and Y.N. performed the research. H.T., M.S., and Y.N. analyzed the data.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: cts70529‐sup‐0001‐supinfo.docx.