Abstract
Background
The occurrence of clinical trial termination before planned completion (or “early termination” [ET]) remains poorly characterized in the breast cancer field. This study investigates the reasons, determinants, and temporal patterns of ET in breast cancer trials.
Methods
We conducted a cross-sectional study using the Aggregate Analysis of ClinicalTrials.gov (AACT) database, including breast cancer trials initiated between 2001 and 2025. Reasons for ET were identified and classified into predefined categories. Factors associated with ET were evaluated using multivariable logistic regression, and temporal patterns were assessed using competing risk analysis and a multivariable Fine-Gray subdistribution hazard model.
Results
A total of 5251 trials were included. “Poor accrual” was the most common reason for ET (39.2%), followed by “sponsor decision or strategic reasons” (22.4%) and “operational or management issues” (13.4%). “Poor accrual” was the leading cause of ET among phase 2 and 3 trials, academic-sponsored studies, and single-center trials, whereas “sponsor decision or strategic reasons” predominated among phase 1 trials, industry-sponsored studies, and trials conducted globally. In multivariable logistic regression analyses, phase 2 trials had a higher likelihood of ET (OR = 1.48; P < .001), whereas multicenter studies (OR = 0.70; P < .001) and trials with non-treatment purposes (OR = 0.61; P < .001) had a lower likelihood of ET. Similar patterns were observed in the Fine-Gray model.
Conclusions
Early termination of breast cancer trials is a heterogeneous phenomenon, with distinct patterns based on study characteristics. Tailored strategies to improve trial feasibility, participant recruitment, and operational planning may reduce the risk of ET.
Keywords: Early termination, Premature discontinuation, Breast cancer, Clinical trials
Graphical abstract
1. Introduction
Over the past decade, breast cancer has been one of the most dynamic areas of cancer research, with major advances in the molecular characterization of the disease and the development of new treatment strategies, including targeted therapies, immunotherapy, and drug-delivery technologies [1]. These advancements have contributed to a rapidly expanding drug development landscape, with an increasing number of clinical trials evaluating novel therapeutic strategies in different disease settings. Between 2015 and 2025, over 9000 breast cancer clinical trials were registered on ClinicalTrials.gov, with approximately 70% classified as phase 2 or 3 interventional studies [2]. In addition to this quantitative increase, clinical trial designs have become more complex, often applying adaptive methodologies, translational endpoints, and molecular selection strategies, resulting in more personalized treatment approaches and more rapid integration of scientific discoveries into clinical practice [3].
Clinical trials require substantial resources, including financial investments, organizational infrastructure, and patient participation [4]. In this context, termination of clinical trials before planned completion (or “early termination”) generates considerable challenges for the scientific community and healthcare systems, potentially compromising clinically meaningful evidence, reducing interpretability, wasting resources, and raising important ethical concerns for enrolled participants [5]. However, early termination can be scientifically or ethically justified in specific situations, such as unacceptable toxicity, therapeutic futility, or early demonstration of benefit [6]. Although previous studies have investigated factors potentially associated with trial discontinuation, the available evidence remains limited and heterogeneous, with most studies focusing on specific tumor types, geographic regions, or drug classes [[7], [8], [9]]. Notably, to our knowledge, a comprehensive evaluation of the occurrence of early termination in breast cancer clinical trials is currently lacking.
2. Methods
We conducted a cross-sectional study to evaluate the reasons, the determinants, and the temporal patterns of early termination in breast cancer clinical trials. The information used in this study was retrieved from ClinicalTrials.gov, a registry provided by the U.S. National Library of Medicine reporting privately and publicly funded clinical studies conducted globally [10]. Specifically, ClinicalTrials.gov was queried using the Aggregate Analysis of ClinicalTrials.gov (AACT), a publicly available database that contains protocol and result data elements about studies registered in ClinicalTrials.gov [11]. The study was conducted in accordance with the “Strengthening the Reporting of Observational Studies in Epidemiology” (STROBE) reporting guidelines [12].
2.1. Study selection and data extraction
On January 23, 2026, we searched the AACT database for all interventional clinical trials initiated between January 1, 2001, and December 31, 2025, using a search strategy based on the keywords “mammary” and “breast”. We included clinical trials that enrolled only breast cancer patients and had available recruitment status information. As in previous publications, the definition of “early termination” was determined based on the recruitment status reported in ClinicalTrials.gov [9,[13], [14], [15]]. Particularly, clinical trials labeled “terminated” or “withdrawn” were considered as early terminated, whereas those labeled “completed” were considered as studies that reached their planned completion. All remaining recruitment statuses were collectively classified as “active” studies (Table S1).
For each clinical trial, the following variables available in ClinicalTrials.gov were retrieved and/or recoded according to predefined categories: NCT number, start date, completion date, phase, number of enrolled participants, primary purpose, sponsor type, number (single-center vs. multicenter), and geographical region of recruiting centers, and the reasons for early termination (Table S1). As in a previous publication, reasons for termination were recoded into eight predefined categories: “poor accrual”, “safety or tolerability issues”, “operational or management issues”, “sponsor decision or strategic reasons”, “regulatory or ethical issues”, “scientific or study design issues”, “inefficacy or interim results”, and “other or unspecified” [13]. Reasons for early termination were independently categorized by two groups of investigators using predefined categories. Any discordant classifications were reviewed by a third investigator and resolved by consensus. The level of agreement between authors for the classification of early termination reasons was quantified using Cohen's kappa coefficient [16].
2.2. Statistical analysis
Trial characteristics were summarized using absolute frequencies and percentages for categorical variables, while continuous variables were reported as medians with their corresponding interquartile ranges. Comparisons between categorical variables were performed using Pearson's Chi-squared test, whereas continuous variables were compared with the Wilcoxon rank-sum test. Reasons for early termination were first evaluated across the overall 25-year study period and subsequently analyzed according to key study characteristics, including study duration categorized using a 24-month cutoff based on existing literature [13]. Given the limited number of trials with non-treatment objectives, the primary purpose was further dichotomized as “treatment” versus “non-treatment”.
Subsequently, factors potentially associated with early termination were investigated using logistic regression analysis. Univariable logistic models were first fitted for each variable of interest. All pre-specified variables (study phase, recruiting centers, primary purpose, sponsor, and geographical region) were included in the multivariable models, and no selection was performed based on univariable significance. Results were reported as odds ratios (OR) with 95% confidence intervals (CI). Collinearity among covariates included in the multivariable model was evaluated using variance inflation factors (VIFs). Discrimination of the multivariable logistic model was evaluated using the receiver operating characteristic curve (AUC) and internally validated using bootstrap resampling (200 resamples), with optimism-corrected estimates reported. Calibration was evaluated using a calibration plot, calibration slope, and intercept. Overall accuracy was summarized using the Brier score.
To evaluate the temporal patterns of early termination, we performed a competing risks analysis. Particularly, clinical trials labeled as “terminated” were considered the event of interest. Studies with “completed” status were treated as competing events. Trials with “active” status were considered censored observations. Studies labeled as “withdrawn” were excluded from the competing risk analysis because no patients had been enrolled. Time-to-event was defined as the interval from the study start date, as recorded in ClinicalTrials.gov, to the completion date for completed trials, the termination date for terminated trials, or the data lock cut-off date of January 23, 2026, for active studies, which were treated as censored observations (Table S1). Initially, cumulative incidence curves were generated for the overall cohort of clinical trials. Furthermore, stratified cumulative incidence curves were constructed according to key trial characteristics. Subsequently, a multivariable Fine–Gray model was fitted, and subdistribution hazard ratios (sHRs) with corresponding 95% CI were reported. Collinearity among covariates included in the multivariable model was evaluated using VIFs. Discrimination of the Fine-Gray model was assessed using the time-dependent area under the AUC at predefined time points, with optimism-corrected estimates obtained through bootstrap cross-validation (200 resamples). Calibration was evaluated through calibration plots at predefined time points.
A complete-case analysis was performed for each model, including only clinical trials with complete data on all variables. No imputation methods were applied for missing data. The number of studies included in each analysis is reported in the corresponding tables. Although there was no a priori hypothesis, a two-sided p-value < 0.05 was considered statistically significant. The Bonferroni correction was applied to account for multiple comparisons, with statistical significance defined at an adjusted p-value < 0.05. All statistical analyses and graphical representations were performed using RStudio: Integrated Development Environment for R (version 4.5.2 - Posit software, PBC - Boston, MA, USA), using the packages reported in Table S2.
3. Results
We identified a cohort of 5251 clinical trials, including 2663 (50.7%) “completed”, 1054 (20.1%) “terminated”, and 1534 (29.2%) “active” studies (Table S3).
3.1. Classification and distribution of reasons for early termination
Inter-rater agreement for classifying early termination reasons was high, as indicated by a Cohen's kappa coefficient of 0.80 (95% CI 0.77 to 0.83) (Fig. S1). Over the 25-year study period, “poor accrual” was the most frequent cause of early termination (39.2%), followed by “sponsor decision or strategic reasons” (22.4%) and “operational or management issues” (13.4%) (Table S4). Although the distribution of reasons for early termination did not differ significantly between studies terminated before and after the 24-month cutoff (Table S5, Fig. S2), significant differences emerged across several study characteristics. In terms of study phase, “sponsor decision or strategic reasons” predominated in phase 1 trials (38.7%), occurring more frequently than in phase 2 and 3 studies (17.9% and 19.6%; P < .001). Conversely, “poor accrual” was the leading cause of termination in phase 2 and 3 trials (43.2% and 40.2%), with higher proportions than in phase 1 studies (25.6%; P < .001) (Fig. 1, Table S6, Fig. S3). Regarding study purpose, “poor accrual” was the most frequent cause of early termination in both treatment and non-treatment studies, although it was more common among the latter (37.5% vs. 51.9%; P < .001). In contrast, “sponsor decision or strategic reasons” occurred more frequently in treatment-oriented studies (24.4% vs. 7.5%; P < .001) (Table S7, Fig. S4). Significant differences were also observed according to geographical region, sponsor type, and study organization. “Poor accrual” predominated among studies conducted in Europe, the United States, and Asia, with the highest proportion observed in Europe (55.2% vs. 43.1% and 47.1% respectively; P < .001), whereas “sponsor decision or strategic reasons” was the leading cause among global studies (48.9%; P < .001) (Table S8, Fig. S5). “Poor accrual” was also the leading cause of termination among academic-sponsored trials (52.4%), with a higher proportion than government- or industry-sponsored studies (38.6% and 16.0%; P < .001). Conversely, “sponsor decision or strategic reasons” predominated among industry-sponsored trials (51.2%), occurring substantially more frequently than in academic- and government-sponsored studies (7.2% and 8.8%; P < .001) (Table S9, Fig. S6). Finally, “poor accrual” predominated among single-center trials and was more common than in multicenter studies (49.3% vs. 33.7%; P < .001). In contrast, “sponsor decision or strategic reasons” and “inefficacy or interim results” were more frequently reported in multicenter trials (33.0% and 11.4%) than in single-center studies (11.3% and 4.7%; P < .001) (Table S10).
Fig. 1.
Reasons for early termination of clinical trials according to study phase.
3.2. Factors associated with early termination
Significant differences were observed between “terminated” and “completed” trials across multiple study characteristics (Table 1). Terminated studies were more often phase 2 trials (67.2% vs. 55.2%) and less frequently phase 3 trials (11.2% vs. 20.4%; P < .001) than completed studies. In addition, terminated studies had a higher proportion of treatment-oriented studies (88.8% vs. 84.0%; P = .001) and were more commonly conducted at a single center (48.5% vs. 40.6%; P < .001) compared to completed trials. No significant differences were found by sponsor type (P = .14). In terms of geographical distribution, terminated studies were more often conducted in the United States (56.3% vs. 42.6%) and less frequently in Asia (7.9% vs. 11.5%), Europe (17.1% vs. 20.3%), and global settings (14.5% vs. 19.3%; P < .001) compared to completed studies (Table 1). Multivariable logistic regression analysis demonstrated that phase 2 trials had a significantly higher probability of early termination (OR = 1.48, 95% CI 1.23-1.80; P < .001) compared to phase 1 trials. In contrast, multicenter trials showed a lower probability of premature termination (OR = 0.70, 95% CI 0.58-0.84; P < .001) than single-center trials. Regarding study purpose, non-treatment studies had a significantly lower probability of early termination (OR = 0.61, 95% CI 0.48-0.78; P < .001) compared to treatment-oriented studies. Finally, using studies conducted in the United States as the reference, studies conducted in Asia (OR = 0.53, 95% CI 0.40-0.70; P < .001), Europe (OR = 0.69, 95% CI 0.55-0.85; P < .001), global settings (OR = 0.71, 95% CI 0.55-0.92; P = .010), and other regions (OR = 0.59, 95% CI 0.40-0.85; P = .005) all showed a significantly lower probability of premature termination (Table 2). No multicollinearity was detected (Table S11). Model AUC was 0.60 (optimism-corrected 95% CI 0.58-0.62), with fair calibration and a Brier score of 0.19 (optimism-corrected 95% CI 0.18-0.20) (Fig. S8).
Table 1.
Characteristics of “completed” and “terminated” clinical trials.
| Trial characteristics | Completed (n = 2663) | Terminated (n = 1054) | Adjusted P |
|---|---|---|---|
| Phase, No. (%) | <0.001 | ||
| 1 | 651 (24.4%) | 228 (21.6%) | |
| 2 | 1470 (55.2%) | 708 (67.2%) | |
| 3 | 542 (20.4%) | 118 (11.2%) | |
| No. of participants, median (IQR) | 61 (30-152) | 12 (1-39) | <0.001 |
| Missing | 35 | 2 | |
| Primary purpose, No. (%) | 0.001 | ||
| Treatment | 2223 (84.0%) | 934 (88.8%) | |
| Non-Treatment | 423 (16.0%) | 118 (11.2%) | |
| Missing | 17 | 2 | |
| Recruiting centers, No. (%) | <0.001 | ||
| Single-center | 1035 (40.6%) | 451 (48.5%) | |
| Multicenter | 1514 (59.4%) | 479 (51.5%) | |
| Missing | 114 | 124 | |
| Sponsor, No. (%) | 0.14 | ||
| Academic | 1289 (48.4%) | 556 (52.8%) | |
| Government | 207 (7.8%) | 71 (6.7%) | |
| Industry | 1058 (39.7%) | 373 (35.4%) | |
| Other | 109 (4.1%) | 54 (5.1%) | |
| Geographical region, No. (%) | <0.001 | ||
| United States | 1107 (42.6%) | 561 (56.3%) | |
| Asia | 299 (11.5%) | 79 (7.9%) | |
| Europe | 527 (20.3%) | 170 (17.1%) | |
| Global | 502 (19.3%) | 144 (14.5%) | |
| Othera | 164 (6.3%) | 42 (4.2%) | |
| Missing | 64 | 58 |
Statistically significant results are shown in bold (P < .05). P values adjusted using Bonferroni correction (k = 6).
Africa, North America, Oceania, and South America were combined into a single category because of their small sample size.
Table 2.
Uni- and multivariable logistic regression models of factors associated with the occurrence of early termination.
| Trial characteristics | Univariable analyses |
Multivariable analyses |
||||
|---|---|---|---|---|---|---|
| Events/Total (%) | OR (95% CI) | P | Events/Total (%) | OR (95% CI) | P | |
| Phase | ||||||
| 1 | 228/879 (25.9%) | (Referent) | 199/827 (24.1%) | (Referent) | ||
| 2 | 708/2178 (32.5%) | 1.38 (1.16 to 1.64) | <0.001 | 621/2024 (30.7%) | 1.48 (1.23 to 1.80) | <0.001 |
| 3 | 118/660 (17.9%) | 0.62 (0.48 to 0.80) | <0.001 | 108/610 (17.7%) | 0.90 (0.68 to 1.19) | 0.47 |
| Recruiting centers | ||||||
| Single-center | 451/1486 (30.3%) | (Referent) | 451/1473 (30.6%) | (Referent) | ||
| Multi-center | 479/1993 (24.0%) | 0.73 (0.62 to 0.84) | <0.001 | 477/1988 (24.0%) | 0.70 (0.58 to 0.84) | <0.001 |
| Primary purpose | ||||||
| Treatment | 934/3157 (29.6%) | (Referent) | 825/2950 (28.0%) | (Referent) | ||
| Non-Treatment | 118/541 (21.8%) | 0.66 (0.53 to 0.82) | <0.001 | 103/511 (20.2%) | 0.61 (0.48 to 0.78) | <0.001 |
| Sponsor | ||||||
| Academic | 556/1845 (30.1%) | (Referent) | 490/1727 (28.4%) | (Referent) | ||
| Government | 71/278 (25.5%) | 0.80 (0.59 to 1.05) | 0.12 | 59/255 (23.1%) | 0.76 (0.55 to 1.04) | 0.10 |
| Industry | 373/1431 (26.1%) | 0.82 (0.70 to 0.95) | 0.01 | 327/1320 (24.8%) | 1.08 (0.88 to 1.32) | 0.47 |
| Other | 54/163 (33.1%) | 1.15 (0.81 to 1.61) | 0.43 | 52/159 (32.7%) | 1.10 (0.77 to 1.56) | 0.60 |
| Geographical region | ||||||
| United States | 561/1668 (33.6%) | (Referent) | 511/1592 (32.1%) | (Referent) | ||
| Asia | 79/378 (20.9%) | 0.52 (0.40 to 0.68) | <0.001 | 78/377 (20.7%) | 0.53 (0.40 to 0.70) | <0.001 |
| Europe | 170/697 (24.4%) | 0.64 (0.52 to 0.78) | <0.001 | 164/683 (24.0%) | 0.69 (0.55 to 0.85) | <0.001 |
| Global | 144/646 (22.3%) | 0.57 (0.46 to 0.70) | <0.001 | 136/615 (22.1%) | 0.71 (0.55 to 0.92) | 0.01 |
| Othera | 42/206 (20.4%) | 0.51 (0.35 to 0.71) | <0.001 | 39/194 (20.1%) | 0.59 (0.40 to 0.85) | 0.005 |
Abbreviations: CI = Confidence Interval, OR = Odds Ratio.
Statistically significant results are shown in bold (P < .05).
Africa, North America, Oceania, and South America were combined into a single category because of their small sample size.
3.3. Temporal patterns in the cumulative incidence of early termination
Competing risks analysis demonstrated that study completion was the most frequent outcome, with cumulative incidence steadily increasing after study initiation. In contrast, the cumulative incidence of early termination increased primarily during the initial years following study initiation and subsequently stabilized (Fig. 2). Stratified analyses identified distinct temporal patterns of early termination according to study characteristics. By study phase, phase 2 trials had the highest cumulative incidence of early termination over the observation period, whereas phase 3 trials consistently had the lowest cumulative incidence (Fig. S9). In terms of geographical regions, studies conducted in the United States had the highest cumulative incidence of early termination, whereas those conducted in Asia had the lowest incidence throughout the observation period (Fig. S10). Regarding sponsor type, industry-sponsored studies experienced a more rapid increase in cumulative incidence of early termination during the initial years after study initiation, while academic- and industry-sponsored studies converged to similar cumulative incidences over time; government-sponsored studies maintained the lowest cumulative incidence overall (Fig. S11). Single-center studies consistently showed a higher cumulative incidence of early termination than did multicenter studies throughout the observation period (Fig. S12). In the multivariable Fine-Gray model, phase 2 trials had a significantly higher risk of early termination (sHR = 1.39, 95% CI 1.16-1.67; P < .001) compared to phase 1 trials. Multicenter trials had a lower risk of premature termination (sHR = 0.66, 0.56-0.78; P < .001) than single-center trials, and non-treatment studies had a lower risk (sHR = 0.75, 0.59-0.95; P = .02) than treatment-oriented studies. Industry-sponsored trials had a significantly higher risk of early termination (sHR = 1.46, 1.20-1.77; P < .001) compared to academically sponsored studies. Finally, using studies conducted in the United States as the reference, studies in Asia (sHR = 0.41, 0.31-0.55; P < .001), Europe (sHR = 0.81, 0.66-0.99; P = .04), global settings (sHR = 0.76, 0.59-0.97; P = .03), and other regions (sHR = 0.66, 0.45-0.95; P = .03) had significantly lower risks of premature termination (Table 3). No multicollinearity was detected (Table S11). For years 1-10, the model AUC ranged from 0.59 to 0.63 (Table S12), with fair calibration (Fig. S13).
Fig. 2.
Cumulative incidence of early termination and study completion over time.
Table 3.
Uni- and multivariable Fine–Gray models of factors associated with the occurrence of early termination.
| Trial characteristics | Univariable analyses |
Multivariable analyses |
||||
|---|---|---|---|---|---|---|
| Events/Total | sHR (95% CI) | P | Events/Total | sHR (95% CI) | P | |
| Phase | ||||||
| 1 | 166/1089 | (Referent) | 164/1066 | (Referent) | ||
| 2 | 541/2848 | 1.27 (1.06 to 1.51) | 0.008 | 525/2768 | 1.39 (1.16 to 1.67) | <0.001 |
| 3 | 89/962 | 0.58 (0.45 to 0.74) | <0.001 | 86/918 | 0.77 (0.58 to 1.02) | 0.06 |
| Recruiting centers | ||||||
| Single-center | 348/1965 | (Referent) | 348/1952 | (Referent) | ||
| Multi-center | 429/2805 | 0.80 (0.70 to 0.93) | 0.003 | 427/2800 | 0.66 (0.56 to 0.78) | <0.001 |
| Primary purpose | ||||||
| Treatment | 706/4228 | (Referent) | 690/4117 | (Referent) | ||
| Non-Treatment | 88/653 | 0.78 (0.62 to 0.97) | 0.03 | 85/635 | 0.75 (0.59 to 0.95) | 0.02 |
| Sponsor | ||||||
| Academic | 407/2518 | (Referent) | 402/2462 | (Referent) | ||
| Government | 41/296 | 0.74 (0.54 to 1.01) | 0.06 | 39/283 | 0.79 (0.57 to 1.10) | 0.16 |
| Industry | 305/1848 | 1.08 (0.93 to 1.25) | 0.33 | 291/1772 | 1.46 (1.20 to 1.77) | <0.001 |
| Other | 43/237 | 1.16 (0.86 to 1.58) | 0.34 | 43/235 | 1.04 (0.76 to 1.42) | 0.79 |
| Geographical region | ||||||
| United States | 422/2041 | (Referent) | 421/2016 | (Referent) | ||
| Asia | 58/793 | 0.46 (0.35 to 0.60) | <0.001 | 58/793 | 0.41 (0.31 to 0.55) | <0.001 |
| Europe | 138/828 | 0.79 (0.65 to 0.95) | 0.01 | 134/816 | 0.81 (0.66 to 0.99) | 0.04 |
| Global | 136/907 | 0.72 (0.60 to 0.88) | <0.001 | 131/879 | 0.76 (0.59 to 0.97) | 0.03 |
| Othera | 31/257 | 0.58 (0.40 to 0.84) | 0.004 | 31/248 | 0.66 (0.45 to 0.95) | 0.03 |
Abbreviations: CI = Confidence Interval, sHR = Subdistribution Hazard Ratio.
Statistically significant results are shown in bold (P < .05).
Africa, North America, Oceania, and South America were combined into a single category because of their small sample size.
4. Discussion
In this study, we systematically evaluated breast cancer clinical trials registered on ClinicalTrials.gov over a 25-year period to investigate, for the first time, the reasons, the determinants, and the temporal patterns associated with early termination. Our results showed that “poor accrual” was the leading cause of early termination, followed by “sponsor decision or strategic reasons” and “operational or management issues”. In addition, the large sample of clinical trials included in this study and the application of multiple statistical analyses using several key study characteristics allowed us to demonstrate that early termination is a heterogeneous phenomenon, characterized by different patterns related to the methodological, organizational, and strategic characteristics of clinical trials.
The most significant finding from our study is the predominant role of “poor accrual”, which likely reflects the growing complexity of the breast cancer clinical research landscape. In recent years, the progressive biological fragmentation of the disease, the introduction of biomarker-driven strategies, and the rapid increase in the number of active clinical trials [2] have likely increased competition for recruiting eligible patients. In this context, an unrealistic estimate of enrollment potential can quickly lead to accrual difficulties and subsequent study discontinuation. Interestingly, “poor accrual” was the leading cause for early termination in late-phase trials, academically sponsored studies, and single-center trials, suggesting that these settings may be more vulnerable to recruitment challenges. Single-center studies may face greater difficulties due to limited access to collaborative networks, dedicated infrastructure, and sufficiently large patient populations, whereas academic studies may have fewer financial and organizational resources to support recruitment activities and trial conduct. Collectively, these factors may increase susceptibility to accrual-related discontinuation. Notably, “poor accrual” was also the leading cause of early termination in both treatment and non-treatment studies, underscoring the broad impact of recruitment challenges across the breast cancer clinical research landscape.
A distinct pattern was observed for studies terminated due to “sponsor decision or strategic reasons”, which was the leading cause of early termination for early-phase trials, industry-sponsored studies, and globally conducted trials. These results suggest that such trials may be subject to dynamic and strategic decision-making processes, likely associated with the rapidly changing oncology therapeutic landscape and the need to optimize clinical development programs. These observations align with our previous analysis of cancer clinical trials across different tumor types, in which “sponsor decision or strategic reasons” was the leading cause of early termination in phase 1 studies, while “poor accrual” was the predominant reason for termination in phase 2 and 3 trials [13]. Collectively, these results highlight the importance of strategic and portfolio-related considerations in early-phase drug development [17]. In addition, the higher prevalence of “sponsor decisions or strategic reasons” in global and multicenter trials may suggest that, in these settings, strategic and portfolio-related considerations play a relatively greater role in trial discontinuation. One possible explanation is that broader recruitment networks and greater enrollment capacity may reduce the impact of accrual-related challenges, making other reasons for early termination more prominent.
Furthermore, our results showed that early termination due to “operational and management issues” was significantly less common in industry-sponsored trials compared to academic- or government-sponsored studies, likely reflecting the greater availability of centralized infrastructure, regulatory support, and operational coordination. Notably, “operational or management issues” were more frequent in single-center studies. Although such issues might be expected to occur more frequently in multicenter studies, this finding suggest that local organizational limitations are a critical factor, especially in trials conducted in less structured environments. Factors such as administrative support, regulatory compliance, data management, and the availability of dedicated staff can substantially influence the sustainability of a clinical trial.
Beyond the descriptive analyses, the results of both the multivariable logistic regression and competing risks analysis further refined these results, identifying phase 2 studies as particularly vulnerable to early termination. The increased risk associated with phase 2 trials is significant, as this stage is a critical transition in oncology drug development, during which substantial resources have already been invested, and decisions regarding further clinical development are often made. In contrast, multicenter studies were consistently associated with a lower risk of early termination, likely due to broader recruitment networks, enhanced operational support, and improved access to eligible patient populations. Trials with non-treatment purposes also demonstrated a lower likelihood of early termination, suggesting that likely their generally reduced logistical and regulatory complexity may facilitate study completion. Competing risks analysis indicated that the cumulative incidence of early termination increased most rapidly during the initial years following study initiation, highlighting that many feasibility, recruitment, and operational challenges occur early in the clinical trial life cycle.
Overall, our results highlight that strategies to reduce early termination should be tailored to the specific characteristics of each clinical trial, rather than applying a uniform approach. In studies where “poor accrual” was identified as the predominant cause, efforts should focus on improving trial feasibility before activation, implementing realistic recruitment projections, and strengthening referral networks. Existing evidence demonstrates that structural barriers, such as limited local availability of clinical trials and restrictive eligibility criteria, are major obstacles to trial participation and often outweigh patient refusal [18]. Expanding trial availability through decentralized research infrastructures and adopting broader eligibility criteria, when scientifically justified, may enhance patient access and reduce accrual-related study discontinuation. Additionally, a more patient-centered trial design by involving patient advocates in protocol development may improve trial enrollment. In contrast, in studies where “sponsor decision or strategic reasons” was the leading cause of early termination, closer collaboration between investigators and sponsors may facilitate the identification of unmet clinical needs and support more efficient strategic decision-making throughout the trial. Continuous feasibility reassessment and adaptive portfolio management may further reduce unnecessary study discontinuation.
Several limitations should be acknowledged. First, data were obtained exclusively from ClinicalTrials.gov, which may not fully represent all clinical trials conducted globally. Second, information reported in the registry is self-reported by investigators and sponsors, which can lead to incomplete, inaccurate, or outdated data. In particular, the reasons for early termination are reported as free-text entries and may be heterogeneous, incompletely described, or subject to misclassification [19]. Nevertheless, ClinicalTrials.gov remains one of the most comprehensive and widely used sources for clinical trial research, enabling the evaluation of large cohorts of studies across several settings and time periods [19]. Finally, the large sample size spanning 25 years and the high inter-rater agreement in classifying reasons for early termination improve the reliability of our analysis.
5. Conclusion
Early termination of breast cancer trials is a heterogeneous phenomenon, exhibiting distinct patterns based on study characteristics. While “poor accrual” was the leading cause of early termination among phase 2 and 3 trials, academic-sponsored studies, and single-center trials, “sponsor decision or strategic reasons” predominated among phase 1 trials, industry-sponsored studies, and trials conducted globally. Implementing tailored strategies to enhance trial feasibility, participant recruitment, and operational planning may reduce the risk of early termination.
CRediT authorship contribution statement
Giovanni Maria Iannantuono: Conceptualization, Investigation, Methodology, Visualization, Writing – original draft, Writing – review & editing. Gennaro Daniele: Methodology, Supervision, Writing – review & editing. Luca Mastrantoni: Formal analysis, Methodology, Visualization, Writing – original draft, Writing – review & editing. Diana Giannarelli: Methodology, Writing – review & editing. Dario Trapani: Writing – review & editing. Bishal Gyawali: Writing – review & editing. Stefano Sganga: Investigation, Writing – review & editing. Elena Giudice: Investigation, Writing – review & editing. Silvia Riondino: Writing – review & editing. Eriseld Krasniqi: Writing – review & editing. Serena Capici: Writing – review & editing. Francesca Fulvia Pepe: Writing – review & editing. Rosalba Torrisi: Writing – review & editing. Andrea Spinazzola: Investigation, Writing – review & editing. Antonio Bassolino: Investigation, Writing – review & editing. Arianna Meacci: Investigation, Writing – review & editing. Khashayar Yazdanpanahardakani: Investigation, Writing – review & editing. Letizia Padula: Investigation, Writing – review & editing. Roberto Rosenfeld: Investigation, Writing – review & editing. Luisa Carbognin: Writing – review & editing. Domenica Lorusso: Writing – review & editing. Giuseppe Curigliano: Writing – review & editing. James L. Gulley: Writing – review & editing. Alessandra Fabi: Writing – review & editing. Emilio Bria: Writing – review & editing. Mario Roselli: Writing – review & editing. Charalampos S. Floudas: Writing – review & editing. Marina Elena Cazzaniga: Methodology, Supervision, Writing – review & editing.
Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Ethics in publishing
This study did not involve human participants, and therefore no institutional review board approval or informed consent was required.
Declaration of generative AI and AI-assisted technologies in the manuscript preparation process
During the preparation of this work, the authors used ChatGPT (OpenAI) exclusively to support the visual rendering and stylistic development of the graphical abstract. All scientific content, including the selection and organization of information, wording, data, and interpretation, was determined solely by the authors. The final graphical abstract was reviewed and verified by the authors, who take full responsibility for its content.
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Declaration of competing interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests.
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.breast.2026.104931.
Contributor Information
Giovanni Maria Iannantuono, Email: giovannimaria.iannantuono@irccs-sangerardo.it.
Gennaro Daniele, Email: gennaro.daniele@policlinicogemelli.it.
Luca Mastrantoni, Email: l.mastrantoni@hotmail.it.
Diana Giannarelli, Email: diana.giannarelli@policlinicogemelli.it.
Dario Trapani, Email: dario.trapani@ieo.it.
Bishal Gyawali, Email: gyawali.bishal@queensu.ca.
Stefano Sganga, Email: stefano.sganga@guest.policlinicogemelli.it.
Elena Giudice, Email: elena.giudice@sanpiox.humanitas.it.
Silvia Riondino, Email: silvia.riondino@uniroma2.it.
Eriseld Krasniqi, Email: eriseld.krasniqi@ifo.it.
Serena Capici, Email: serena.capici@irccs-sangerardo.it.
Francesca Fulvia Pepe, Email: francescafulvia.pepe@irccs-sangerardo.it.
Rosalba Torrisi, Email: rosalbamaria.torrisi@irccs-sangerardo.it.
Andrea Spinazzola, Email: andrea.spinazzola@guest.policlinicogemelli.it.
Antonio Bassolino, Email: antonio.bassolino@guest.policlinicogemelli.it.
Arianna Meacci, Email: arianna.meacci@guest.policlinicogemelli.it.
Khashayar Yazdanpanahardakani, Email: k.yazdanpanaharda@campus.unimib.it.
Letizia Padula, Email: letizia.padula@irccs-sangerardo.it.
Roberto Rosenfeld, Email: roberto.rosenfeld88@gmail.com.
Luisa Carbognin, Email: luisa.carbognin@policlinicogemelli.it.
Domenica Lorusso, Email: domenica.lorusso@hunimed.eu.
Giuseppe Curigliano, Email: giuseppe.curigliano@ieo.it.
James L. Gulley, Email: james.gulley@nih.gov.
Alessandra Fabi, Email: alessandra.fabi@policlinicogemelli.it.
Emilio Bria, Email: emilio.bria@policlinicogemelli.it.
Mario Roselli, Email: mario.roselli@uniroma2.it.
Charalampos S. Floudas, Email: charalampos.floudas@nih.gov.
Marina Elena Cazzaniga, Email: marinaelena.cazzaniga@irccs-sangerardo.it.
Appendix A. Supplementary data
The following are the Supplementary data to this article:
References
- 1.Xiong X., Zheng L.-W., Ding Y., et al. Breast cancer: pathogenesis and treatments. Signal Transduct Targeted Ther. 2025;10:49. doi: 10.1038/s41392-024-02108-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Dynarowicz K., Bartusik-Aebisher D., Czech S., Kawczyk-Krupka A., Aebisher D. A decade of innovation in breast cancer (2015-2025): a comprehensive review of clinical trials, targeted therapies and molecular perspectives. Cancers (Basel) 2026;18:361. doi: 10.3390/cancers18030361. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Duan X.-P., Qin B.-D., Jiao X.-D., Liu K., Wang Z., Zang Y.-S. New clinical trial design in precision medicine: discovery, development and direction. Signal Transduct Targeted Ther. 2024;9:57. doi: 10.1038/s41392-024-01760-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Spreafico A., Hansen A.R., Razak A.R.A., Bedard P.L., Siu L.L. The future of clinical trial design in oncology. Cancer Discov. 2021;11:822–837. doi: 10.1158/2159-8290.CD-20-1301. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Williams R.J., Tse T., DiPiazza K., Zarin D.A. Terminated trials in the ClinicalTrials.gov results database: evaluation of availability of primary outcome data and reasons for termination. PLoS One. 2015;10 doi: 10.1371/journal.pone.0127242. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Stensland K.D., McBride R.B., Latif A., et al. Adult cancer clinical trials that fail to complete: an epidemic? J Natl Cancer Inst. 2014;106 doi: 10.1093/jnci/dju229. [DOI] [PubMed] [Google Scholar]
- 7.Alhajahjeh A., Hmeidan M., Elatrsh M., et al. Understanding the termination of urologic cancer clinical trials: insights and challenges. JCO Glob Oncol. 2024;10 doi: 10.1200/GO.23.00349. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Khunger M., Rakshit S., Hernandez A.V., et al. Premature clinical trial discontinuation in the era of immune checkpoint inhibitors. Oncologist. 2018;23:1494–1499. doi: 10.1634/theoncologist.2018-0003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zhang Z., Yin J., Yue Y., Su Y., Jiang H. Assessing clinical trial failure risk factors and reasons in gastric cancer. BMC Gastroenterol. 2022;22:496. doi: 10.1186/s12876-022-02592-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Home | ClinicalTrials.gov. https://clinicaltrials.gov/(accessed April 27, 2026).
- 11.AACT database | clinical trials transformation initiative. https://aact.ctti-clinicaltrials.org/(accessed December 5, 2025).
- 12.von Elm E., Altman D.G., Egger M., Pocock S.J., Gøtzsche P.C., Vandenbroucke J.P. Strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ. 2007;335:806–808. doi: 10.1136/bmj.39335.541782.AD. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Iannantuono G.M., Sganga S., Giovagnoli T., et al. Temporal patterns and factors associated with early termination in phase 1 clinical trials for solid tumors. J Natl Cancer Inst. 2026 doi: 10.1093/jnci/djag091. [DOI] [PubMed] [Google Scholar]
- 14.Stensland K.D., DePorto K., Ryan J., et al. Estimating the rate and reasons of clinical trial failure in urologic oncology. Urol Oncol. 2021;39:154–160. doi: 10.1016/j.urolonc.2020.10.070. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Shah H.A., Mishra A., Gouzoulis M.J., Ben-Shalom N., D'Amico R.S. Analysis of factors leading to early termination in glioblastoma-related clinical trials. J Neuro Oncol. 2022;158:489–495. doi: 10.1007/s11060-022-04039-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.McHugh M.L. Interrater reliability: the kappa statistic. Biochem Med. 2012;22:276–282. [PMC free article] [PubMed] [Google Scholar]
- 17.Arap W., Ganesan S., Pasqualini R. Premature termination of early-phase cancer clinical trials: ‘it's the economy, stupid!’. J Natl Cancer Inst. 2026 doi: 10.1093/jnci/djag130. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Unger J.M., Vaidya R., Hershman D.L., Minasian L.M., Fleury M.E. Systematic review and meta-analysis of the magnitude of structural, clinical, and physician and patient barriers to cancer clinical trial participation. J Natl Cancer Inst. 2019;111:245–255. doi: 10.1093/jnci/djy221. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Tse T., Fain K.M., Zarin D.A. How to avoid common problems when using ClinicalTrials.gov in research: 10 issues to consider. BMJ. 2018;361 doi: 10.1136/bmj.k1452. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.