Abstract
Background/Aims:
One in five cancer clinical trials fails with another third failing to meet enrollment goals. Prior efforts to improve enrollment focus on patient facing interventions, but geographic factors such as regional cancer incidence may doom trials before they even begin. For these reasons, we examined associations of regional prostate cancer incidence with trial termination, and identified scientifically-underserved areas where future trials might thrive.
Methods:
We merged US phase 2–3 prostate cancer clinical trial data from ClinicalTrials.gov with prostate cancer incidence data from statecancerprofiles.cancer.gov. We matched trial information from 293 closed and 560 active trials with incidence data for 2,947 counties. Using multivariable logistic regression, we identified associations with trial termination. We identified ‘scientifically-underserved’ counties with the highest cancer incidence quintile (>61 annual cases) but lowest active trials quintile (0 or 1 trial).
Results:
Of 293 closed trials, one in three was terminated (n=96, 32.8%). On multivariable analysis, only lower regional prostate cancer incidence was associated with higher likelihood of premature trial termination (OR 0.98, 95% CI [0.96–0.99] for every 100 cases, p=0.03). We identified 188 counties with >61 annual prostate cancer cases but 0 or 1 active trials, indicating potential scientifically-underserved areas.
Conclusions:
In this novel study, we found prostate cancer trials in areas with low prostate cancer incidence were more likely to fail. We also identified scientifically-underserved areas where trials might thrive. Our findings provide a more nuanced understanding of clinical trial feasibility and upstream opportunities for improvement.
Keywords: clinical trials, prostate cancer, quality improvement, access to care
INTRODUCTION
Patient enrollment in cancer clinical trials helps ensure standard of care, at a minimum, and is recommended by national and international cancer organizations. Yet, less than 8% of patients with cancer enroll in a trial.(1–3) In fact, one in three completed cancer trials will not reach target enrollment, and one in five terminate before reaching endpoints, usually due to enrollment issues.(4–7) Improving enrollment in cancer clinical trials has promise to benefit patients and science.
However, recruitment interventions to improve enrollment have had mixed results leaving enrollment a critical unsolved problem.(8) One challenge is a tendency to focus enrollment efforts downstream after a trial has already started.(3) Less effort has been focused on improving trial design to optimize enrollment prior to launch, resulting in trials that may be mismatched with local needs. In other words, it’s possible trials are set up to fail (i.e., terminate) due to low regional cancer prevalence.(3) It’s also possible there are ‘scientifically-underserved’ areas with high cancer prevalence yet few trials creating missed opportunities to support enrollment.(2,7,9) With an estimated 8% of all cancer patients needing to enroll on current cancer trials to meet enrollment goals, better understanding the role of geography in optimizing trial enrollment and avoiding trial failure can be considered a duty of the oncology community.(10)
For these reasons, we used a national online registry of clinical trials and geographic cancer incidence data to examine the extent to which regional cancer incidence was associated with clinical trial termination for the most common male malignancy, prostate cancer. This novel study also identified scientifically-underserved areas where trials might thrive. Our findings provide a more nuanced understanding of prostate cancer clinical trial feasibility and upstream opportunities for enrollment improvement.
METHODS
Study population
We conducted a retrospective cohort study of prostate cancer clinical trials through a unique combination of three distinct, publicly available data sources. First, we queried ClinicalTrials.gov for all US phase 2 and 3 prostate cancer clinical trials from January 2007 through December 2020. Next, we extracted county and Health Service Area (HSA) level prostate cancer incidence data from the National Cancer Institute and Centers for Disease Control and Prevention available through statecancerprofiles.cancer.gov. Finally, we extracted hospital referral region (HRR) prostatectomy rates from the Dartmouth Health Atlas to inform our sensitivity analyses described below. This resulted in a total of 293 prostate cancer clinical trials completed or terminated during the study period for our primary analysis. For the secondary analysis of trial distribution, we included 560 prostate cancer clinical trials with status of “Active,” “Recruiting,” or “Enrolling by invitation.”
Clinical trial and geographic characteristics
ClinicalTrials.gov is an online registry with required registration for all US-based clinical trials.(11) Each registered trial includes trial characteristics, enrollment information, trial status, and ZIP codes for participating trial sites. We used Python (Supplement) to extract trial information and calculate anticipated accrual for prostate cancer trials using a previously published algorithm.(6) Specifically, we extracted trial phase, intervention type (e.g., medical, surgical, radiation therapy), sponsor, multicenter status, anticipated accrual, and completion status (completed, terminated) from ClinicalTrials.gov. We did not assign exclusive sponsors or intervention types to each trial; a trial could have multiple types of intervention or sponsors and each type was considered separately in our models.
For each trial, we identified all associated trial sites. Using each site’s ZIP code, we matched sites to a corresponding county and Health Service Area (HSA) using a database of county, HSA, and ZIP codes (Supplement). We assigned the highest incidence count from any site as the trial’s incidence rate. For our secondary analysis, we rank-ordered counties by the number of active prostate cancer clinical trials and calculated quintiles based on county-level clinical trial volume for counties with at least one active trial (1: 0–1 trials, 2: 2–3 trials, 3: 4–5 trials, 4: 6–10 trials, 5: >10 trials).
Prostate cancer incidence by geographic region
The National Cancer Institute and Centers for Disease Control and Prevention publish county-level cancer incidence and mortality data at statecancerprofiles.cancer.gov, reported as both an absolute count and population-adjusted averaged over a five-year period (2013–2017 for this study).(12) We considered counties or HSAs noted as “3 or fewer cases” to have 3 annual cases. Of note, two states (Kansas and Minnesota) do not report county-level incidence data and were excluded from corresponding analyses. We then rank-ordered county-level incidence data for all counties, and created quintiles based on average annual prostate cancer cases: 1: ≤6 cases, 2: 7–13 cases, 3: 14–25 cases, 4: 26–61 cases, 5: >61 cases).
For our primary analysis of the association of cancer incidence with trial failure, we used health service area (HSA) as the geographic unit of interest for prostate cancer incidence. The HSA is a geographic unit defined by cluster analysis of where residents obtain routine hospital care.(13) We selected this geographic unit to reflect hospital use patterns that may not necessarily be captured only by county-level data. Incidence and mortality data for each HSA are directly downloadable from statecancerprofiles.cancer.gov. Additionally, we performed sensitivity analyses using US counties as the geographic unit in analysis of cancer incidence and trial failure.
We used absolute count in our analyses as an absolute number of participants are required for a clinical trial to reach adequate sample size, independent of the surrounding population. A county with 10 annual prostate cancer cases could not feasibly enroll more than 10 prostate cancer patients onto trials regardless of total population size, and representing this as a high rate could be misleading for trial planning purposes. For example, a rate of 1 case / 1,000 population represents 8,000 potentially eligible patients in New York City, but only 120 potentially eligible patients in Ann Arbor (population estimates from US Census data, census.gov). We did perform a sensitivity analysis considering population-adjusted prostate cancer rates instead of absolute count in a separate model to adjust for other potential population-based confounders.
Scientifically-underserved counties
To better understand scientifically-underserved areas, we created county-level maps using Python to identify counties with no or few registered active clinical trial sites, and mismatches between incidence and number of active trial sites. Specifically, we identified counties with the highest quintile of prostate cancer incidence but the lowest quintile of active prostate cancer clinical trials indicating scientifically-underserved areas.
Study outcomes
Our primary outcome was trial termination, defined in ClinicalTrials.gov by “terminated” overall status. Our secondary outcome was the number of scientifically-underserved counties.
Statistical analyses
We compared characteristics of terminated trials to those of completed trials with a specific focus on relationships to highest HSA-level prostate cancer incidence. We compared continuous variables (prostate cancer incidence, anticipated accrual) between groups using the Kruskal-Wallis test. We compared categorical variables (trial phase, intervention type, sponsor type, multicenter status) using the Chi-squared test. Next, we used multivariable logistic regression to identify associations with trial termination, using trial phase, intervention type, sponsor type, multicenter status, anticipated accrual, and HSA-level cancer incidence as covariates.
Sensitivity analyses
One underappreciated aspect we sought to investigate through sensitivity analysis was whether prostate cancer clinical trial enrollment might be impacted by other indicators of cancer care access and treatment. As a surrogate, we assessed rates of localized prostate cancer treatment. To do this, we used the Dartmouth Health Atlas, which collects claims data from the Centers for Medicare and Medicaid Services, the US Census, the American Hospital Association, the American Medical Association, and the National Center for Health Statistics.(14) As such, prostatectomy rates are presented as number of prostatectomy cases performed per 1,000 Medicare beneficiaries in a given hospital referral region (HRR). Next, we matched site ZIP codes to HRRs using the Dartmouth Health Atlas crosswalk file.(15)
In addition to our primary analysis using HSA-level prostate cancer incidence, we also evaluated the association of county-level incidence on trial failure to assess if this additional level of granularity affected identified associations. We similarly extracted county-level incidence for each trial and assigned the highest county-level incidence for each trial. We re-ran our multivariable logistic regression model with the county-level prostate cancer incidence instead of HSA-level incidence. Additionally, we calculated the cumulative theoretical reach of each trial by summing the county-level prostate cancer incidence from all trial sites for each trial. We then re-ran the multivariable logistic regression with total incidence instead of highest incidence.
We set the probability of type I error at 0.05, and all testing was two-sided. We performed statistical analyses in R. This study was deemed not regulated by our Institutional Review Board.
RESULTS
We identified 293 completed or terminated and 560 active prostate cancer clinical trials taking place in 2,947 US counties with available incidence and mortality data (Table 1). One in three prostate cancer clinical trials was designated as terminated (n=96, 33%). There were more terminated trials with “other” sponsors (79% vs. 60%, p <0.01), but other trial characteristics were similar between terminated and completed trials. For example, a similar percentage of trials were phase 3 (7.3 and 7.6%, p=1), the distribution of trial start years was similar (p=0.97), and enrollment goals were similar for terminated vs. completed trials (median goal 50 patients, p=0.66).
Table 1.
Characteristics of prostate cancer clinical trials registered in ClinicalTrials.gov from 2007 to 2019.
| Characteristic | Overall (n=293) | Completed (n=197, 67%) | Terminated (n=96, 33%) | p-value |
|---|---|---|---|---|
| Start year (%) | 0.97 | |||
| 2007 | 48 (16.4) | 32 (16.2) | 16 (16.7) | |
| 2008 | 36 (12.3) | 24 (12.2) | 12 (12.5) | |
| 2009 | 39 (13.3) | 25 (12.7) | 14 (14.6) | |
| 2010 | 33 (11.3) | 22 (11.2) | 11 (11.5) | |
| 2011 | 34 (11.6) | 21 (10.6) | 13 (13.5) | |
| 2012 | 30 (10.2) | 22 (11.2) | 8 ( 8.3) | |
| 2013 | 29 (9.9) | 18 (9.1) | 11 (11.5) | |
| 2014 | 15 (5.1) | 11 (5.6) | 4 (4.2) | |
| 2015 | 16 (5.5) | 11 (5.6) | 5 (5.2) | |
| 2016 | 8 (2.7) | 7 (3.6) | 1 (1.0) | |
| 2017 | 4 (1.4) | 3 (1.5) | 1 (1.0) | |
| 2019 | 1 (0.3) | 1 (0.5) | 0 (0.0) | |
| Phase | 1 | |||
| Phase 2 (%) | 271 (92.5) | 182 (92.4) | 89 (92.7) | |
| Phase 3 (%) | 22 (7.5) | 15 (7.6) | 7 (7.3) | |
| Multicenter trial | 0.18 | |||
| Single center (%) | 159 (54.3) | 101 (51.3) | 58 (60.4) | |
| Multicenter (%) | 134 (45.7) | 96 (48.7) | 38 (39.6) | |
| Intervention type | ||||
| Biological (%) | 29 ( 9.9) | 22 (11.2) | 7 (7.3) | 0.40 |
| Drug (%) | 246 (84.0) | 160 (81.2) | 86 (89.6) | 0.10 |
| Device (%) | 3 (1.0) | 2 (1.0) | 1 (1.0) | 1 |
| Radiation (%) | 18 (6.1) | 14 (7.1) | 4 (4.2) | 0.47 |
| Sponsor type | ||||
| Other (%) | 194 (66.2) | 118 (59.9) | 76 (79.2) | <0.01 |
| Industry (%) | 143 (48.8) | 93 (47.2) | 50 (52.1) | 0.51 |
| NIH (%) | 84 (28.7) | 58 (29.4) | 26 (27.1) | 0.78 |
| Other US Fed. (%) | 6 (2.0) | 5 (2.5) | 1 (1.0) | 0.68 |
| Anticipated accrual median [interquartile range]) | 50 [33–90] | 50 [34–90] | 50 [30–87] | 0.66 |
Associations with Trial Termination
On univariable analyses, we found terminated trials were associated with lower unadjusted HSA-level prostate cancer incidence compared to completed trials (terminated, median 1,981, interquartile range [IQR] 811–2,898 annual cases vs. completed, median 2,376, IQR 1,222–4,442 annual cases, p = 0.02). This association diminished after population adjustment (terminated 110.9 cases per 100,000 men vs. completed 119.4 cases per 100,000 men, p=0.10). However, terminated trials had lower HSA-level prostate cancer mortality compared to completed trials (median 272 [IQR 127–486] vs. 337 [IQR 176–700] annual deaths, p = 0.02).
Similarly, county-level prostate cancer incidence was lower in terminated trials compared to completed trials (terminated, median 1,008, IQR 401–1,816 vs. completed, median 1,098, IQR 622–3,068 annual cases, p = 0.03). The total case count potentially reached by each trial was also significantly lower for terminated trials compared to completed trials, (terminated, median 1,158, IQR 401–3041 total potential annual cases vs. completed, median 1,626, IQR 634–4,442 annual cases, p = 0.045). However, we found hospital referral region (HRR) prostatectomy rates were similar for terminated vs. completed trials (0.96 vs. 1.01 prostatectomy cases per 1,000 Medicare beneficiaries, p=0.17).
On multivariable analysis, lower HSA-level prostate cancer incidence was associated with premature trial termination (OR 0.98, 95% CI [0.96–0.99] for every 100 incident cases, p = 0.03). In this model, multicenter vs. single center trial, anticipated accrual, intervention type, and sponsor type were not associated with trial termination. In a sensitivity analyses, using county-level instead of HSA-level incidence identified a similar association (OR 0.97, 95% CI [0.95–0.99] for every 100 incident cases, p = 0.03). Using population-adjusted rates for prostate cancer incidence instead of absolute count diminished the association with premature trial termination (OR 0.99, 95% CI [0.97–1.01]). (Data in Supplement)
The total count of incident cases reachable by a trial was not associated with trial termination in a similar model (OR 0.99, 95% CI [0.99–1.00] per 100 cases, p = 0.47). We additionally ran sensitivity analyses removing the ‘multicenter’ variable from the model, and including an interaction term between ‘multicenter’ and the total count. None of these additional models produced significantly different results, though inclusion of the interaction term did amplify the magnitude of the apparent association of multicenter trials with lower likelihood of trial termination (OR 0.53, 95% CI [0.26–1.08], p = 0.08). (Data in Supplement)
Identifying Scientifically Underserved Counties
Of 2,947 USA counties, 899 (28.5%) had at least 1 completed or terminated clinical trial, and 764 (25.9%) had at least one active clinical trial. Considering only counties with at least one active clinical trial, the median number of unique active prostate cancer trials per county was 5 trials (IQR 2–9 trials). The lowest quintile of active prostate cancer clinical trials was 1 or 0 clinical trials per county. The median average annual prostate cancer incidence count was 18 cases per county (IQR 8–46 cases). The highest quintile had an average annual prostate cancer incidence count of >61 prostate cancer cases per county. There were 188 counties with >61 average annual prostate cancer cases but 1 (n=57) or 0 (n=131) active clinical trials, indicating potential scientifically-underserved areas. No counties with the highest quintile of active prostate cancer trials were in the bottom third of prostate cancer incidence, and only 3 counties had the highest quintile of prostate cancer trials but were below median prostate cancer incidence.
DISCUSSION
We found lower regional prostate cancer incidence was associated with higher likelihood of trial failure. In fact, regional prostate cancer incidence was the only variable in our model significantly associated with trial completion. Our findings highlight the upstream importance of assessing feasibility of cancer clinical trial enrollment taking regional cancer incidence into consideration, at least in part, as a potential threat to enrollment and source of trial failure. Further, we identified many areas with high prostate cancer incidence yet few clinical trials suggesting scientifically-underserved areas whose populations may benefit from access to clinical trials, and that could be targeted to improve enrollment to new or existing trials. Taken together, our findings provide potentially actionable insights into addressing the critical unsolved challenges facing clinical trial enrollment.
Our study expands prior work exploring associations between clinical trial completion and adequacy of enrollment. While studies have found associations between successful trial completion and trial characteristics (e.g., multicenter or later stage trials),(7,16) the underlying etiology of poor enrollment has been poorly characterized across the phases of clinical trial implementation. Our results suggest cancer incidence in clinical trial regions may contribute to the downstream likelihood of trial success adding another factor to this growing body of research and quality improvement.
Limited access to clinical trials because of geography is not necessarily surprising. For example, roughly half of patients with cancer do not have a clinical trial site within a reasonable driving distance of their home.(17) However, our study expands on these findings, linking regional cancer incidence to trial success. We highlight areas with many prostate cancer cases but few trials, suggesting areas in which a trial may be more successful by accessing more patients with cancer. While 8% of all cancer patients would need to enroll on clinical trials to fill all open spots,(2,10) these estimates do not explicitly link clinical trial site locations to potential patients, adding to the novelty of our findings.
Another implication of our findings is the concept of ‘scientifically-underserved’ patient populations. Enrollment on a clinical trial is considered the best possible management of a patient with cancer by many, including National Comprehensive Cancer Network guidelines.(18) Independent of the effect on trial success, our work highlights regions of the country with significant disconnects between local incidence and investment for patients with cancer. Further work is needed to assess patterns of care for these patients, and our identified regions should be investigated for adequacy of research and clinical trial access. Notably, we did not find oversaturation of trials in areas with lower prostate cancer incidence. However, assessing how many trials is appropriate for the incidence in a given region is unclear. Further work is needed to evaluate if such a threshold exists, and must incorporate not only crude incidence but also trial eligibility criteria and local contextual factors.
Finally, we hypothesized areas with more use of health resources may have better trial success rates. For prostate cancer, higher rates of radical prostatectomy may reflect more healthcare resource availability. These populations may also be more likely to seek and accept interventions, leading in theory to greater acceptance of clinical trial protocols. We hypothesized more radical prostatectomy cases would correlate with higher rates of clinical trial success, but we found no such association. It is possible that more prostate cancer trials are for advanced or metastatic disease, i.e., relying on patients ineligible for localized treatment, resulting in no association of prostatectomy with trial success. This hypothesis is supported by our finding that trials in areas with fewer annual prostate cancer deaths, suggesting less prevalence of advanced disease, failed more frequently.
Other trial-related or regional resources may also factor into trial success rates. From a regional standpoint, we found that adjusting for county population attenuated the effect of cancer incidence on trial success. One hypothesis is areas with more resources perform trials better. Alternatively, more resources may be preferentially spent on trials in areas with larger populations. Although not statistically significant, we did identify a sizeable point estimate of association of multisite trials with success independent of total case reach, suggesting there is something about being a multisite trial that leads to greater success. In theory, there may be more invested into multisite trials than single site trials. These larger investments may diminish the benefits of areas with smaller total populations but similar absolute cancer incidence rates. If this is the case, it is possible that smaller resource investments in targeted areas may be more efficient in increasing trial enrollment. Our analysis highlights this disconnect, but is limited in capacity to explain the root cause of this disconnect between absolute and population adjusted effects on trial completion. Further study evaluating this etiology is necessary for understanding the disparity and designing targeted interventions to bridge the completion gap.
We note several study limitations. First, we relied on trial registration data in ClinicalTrials.gov for assessment of clinical trial characteristics. While trial registration is required on ClinicalTrials.gov for US-based trials, and is suitable as a starting point for identifying potential target areas for further trial investigation, we cannot fully assess the completeness of trial site registration.(19) Second, we estimated cancer incidence using large registries assuming this a surrogate for cancer prevalence and a larger eligible patient population. We do not have cancer stages, comorbidities, or other criteria relevant to clinical trial enrollment. Nonetheless, we would not expect wide variation in clinical disease characteristics and stage across county-level data. Third, it is possible that some areas with many clinical trials serve different populations and are not competing, or that areas with high cancer incidence do not have many eligible patients. Regardless, issues with competing clinical trials have face-validity and our novel approach raises important considerations.
It is also possible that different sites within the same county serve different populations, and measurement of HSA and county-level populations may not be granular enough to accurately reflect individual trial site populations. We used different geographic units of analysis (health service area, county, and hospital referral region) with the intention of accounting for some of these differences, but our results were similar between models. Additionally, referral patterns do not necessarily follow county lines, so our results are only an approximation of disconnects, and may not apply in all regions. Further work is needed to identify differences in local contexts related to regions. For example, a private hospital and city hospital within the same county likely serve different populations, perhaps with different disease characteristics. While our analysis begins the process of identifying geographic disparities for clinical trials, future studies could highlight differential access to trials among these different types of institutions even apparently collocated within the same geographic region. Such site-specific analyses could also address disparities in trial resources and enrollment at both the site and individual patient level for historically underrepresented groups and facilities.
Finally, we relied on the highest incidence rate for our analyses. It is possible that trials with many sites could successfully enroll with only a few patients from each of many sites, as opposed to many patients from a single site. We did assess total reach of each trial by including total count of cases for all counties with a trial site, but this does not account for site-related factors, competing trials, or differential enrollment rates for various sites. We did find a univariable association with higher total incidence for completed vs. terminated trials, but this did not hold on multivariable analysis. This should be further explored in future analyses with more specific trial site-level data.
These limitations notwithstanding, our findings emphasize the need for consideration of clinical trial feasibility, including assessing the pool of potentially eligible patients, during the planning phase of clinical trials. Moving forward, incorporating new considerations into clinical trials analysis can provide new approaches to clinical trial improvement. By identifying underlying barriers to efficient trial conduct and developing suggested improvement interventions, clinical trials can be made more efficient and new treatments can be brought more quickly and effectively into patient care. Further exploration of our findings, including granularity in trial eligibility criteria, population characteristics and comorbidities, and competing trial sites, will allow for further development of improved clinical trial site planning for new and existing trials.
CONCLUSIONS
Prostate cancer clinical trials in areas with lower prostate cancer incidence were more likely to fail than trials in areas with higher prostate cancer incidence. We identified nearly 200 counties with many prostate cancer cases but few clinical trial sites, potential targets for future clinical trial sites. Future efforts designing clinical trials should consider regional cancer incidence in feasibility assessments, and poorly enrolling trials consider opening additional trial sites in scientifically-underserved areas to attempt improved enrollment.
Supplementary Material
Figure 1.
Number of unique active clinical trials per county (county level data suppressed for Minnesota and Kansas)
Figure 2.
Counties with highest quintile of prostate cancer cases, but lowest quintile of active clinical trials.
Table 2.
Multivariable logistic regression for associations with prostate cancer clinical trial termination
| Variable | Adjusted Odds Ratio (aOR) | 95% CI | p value |
|---|---|---|---|
| Highest HSA-level prostate cancer incidence (100 cases) | 0.98 | 0.96–0.99 | 0.03 |
| Trial Phase | |||
| Phase 2 | Reference | ||
| Phase 3 | 0.76 | 0.22–2.36 | 0.65 |
| Multicenter trial | |||
| Single center | Reference | ||
| Multicenter | 0.90 | 0.48–1.69 | 0.75 |
| Anticipated accrual | 1.00 | 0.997–1.002 | 0.99 |
| Drug trial | 1.90 | 0.80–5.01 | 0.17 |
| Radiation trial | 0.92 | 0.23–3.10 | 0.90 |
| Procedure trial | 1.23 | 0.42–3.35 | 0.69 |
| Industry sponsor | 1.38 | 0.73–2.64 | 0.32 |
| NIH sponsor | 0.60 | 0.28–1.25 | 0.18 |
| Other US Federal sponsor | 0.55 | 0.03–4.26 | 0.61 |
| Other sponsor | 1.87 | 0.85–4.13 | 0.12 |
ABBREVIATIONS
- CMS
Centers for Medicare and Medicaid Services
- HRR
Hospital Referral Region
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Supplemental Figure. Average annual incident prostate cancer cases 2013–2017 from statecancerprofiles.cancer.gov.
Declaration of interests
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References
- 1.Gross CP, Murthy V, Li Y, Kaluzny AD, Krumholz HM. Cancer Trial Enrollment After State-Mandated Reimbursement. JNCI Journal of the National Cancer Institute 2004. Jul 21;96(14):1063–9. [DOI] [PubMed] [Google Scholar]
- 2.Murthy VH, Krumholz HM, Gross CP. Participation in Cancer Clinical Trials: Race-, Sex-, and Age-Based Disparities. JAMA 2004. Jun 9;291(22):2720. [DOI] [PubMed] [Google Scholar]
- 3.Unger JM, Vaidya R, Hershman DL, Minasian LM, Fleury ME. 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. 01;111(3):245–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Stensland KD, DePorto K, Ryan J, Kaffenberger S, Reinstatler LS, Galsky M, et al. Estimating the rate and reasons of clinical trial failure in urologic oncology. Urologic Oncology: Seminars and Original Investigations 2020. Nov;S1078143920305603. [DOI] [PMC free article] [PubMed]
- 5.Khunger M, Rakshit S, Hernandez AV, Pasupuleti V, Glass K, Galsky MD, et al. Premature Clinical Trial Discontinuation in the Era of Immune Checkpoint Inhibitors. Oncologist 2018;23(12):1494–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Stensland K, Kaffenberger S, Canes D, Galsky M, Skolarus T, Moinzadeh A. Assessing Genitourinary Cancer Clinical Trial Accrual Sufficiency Using Archived Trial Data. JCO Clin Cancer Inform 2020. Jul;4:614–22. [DOI] [PubMed] [Google Scholar]
- 7.Stensland KD, McBride RB, Latif A, Wisnivesky J, Hendricks R, Roper N, et al. Adult cancer clinical trials that fail to complete: an epidemic? J Natl Cancer Inst 2014. Sep;106(9). [DOI] [PubMed] [Google Scholar]
- 8.Treweek S, Pitkethly M, Cook J, Fraser C, Mitchell E, Sullivan F, et al. Strategies to improve recruitment to randomised trials. Cochrane Database Syst Rev 2018. 22;2:MR000013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Pasalic D, Tang C, Jagsi R, Fuller CD, Koong AC, Ludmir EB. Association of Industry Sponsorship With Cancer Clinical Trial Accrual. JAMA Oncol 2020. Aug 20; [DOI] [PMC free article] [PubMed]
- 10.Tran G, Harker M, Chiswell K, Unger JM, Fleury ME, Hirsch B, et al. Feasibility of Cancer Clinical Trial Enrollment Goals Based on Cancer Incidence. JCO Clin Cancer Inform 2020;4:35–49. [DOI] [PubMed] [Google Scholar]
- 11.Zarin DA, Tse T, Williams RJ, Carr S. Trial Reporting in ClinicalTrials.gov - The Final Rule. N Engl J Med 2016. Nov 17;375(20):1998–2004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.NIH/CDC. About State Cancer Profiles [Internet]. StateCancerProfiles.cancer.gov. Available from: https://statecancerprofiles.cancer.gov/about/
- 13.Pickle LW. Atlas of United States mortality Hyattsville, Md: National Center for Health Statistics, Centers for Disease Control and Prevention, U.S. Dept. of Health and Human Services; 1996. (DHHS publication). [Google Scholar]
- 14.Dartmouth Health Atlas. Research Methods FAQ [Internet]. Dartmouth Health Atlas [cited 2021. Jan 7]. Available from: https://www.dartmouthatlas.org/faq/#research-methods-faq
- 15.Supplemental Research Data: ZIP Code Crosswalks [Internet]. Supplemental Research Data [cited 2021. Jan 7]. Available from: https://atlasdata.dartmouth.edu/downloads/supplemental#crosswalks
- 16.Amstutz A, Schandelmaier S, Frei R, Surina J, Agarwal A, Olu KK, et al. Discontinuation and non-publication of randomised clinical trials supported by the main public funding body in Switzerland: a retrospective cohort study. BMJ Open 2017. Aug 1;7(7):e016216. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Galsky MD, Stensland KD, McBride RB, Latif A, Moshier E, Oh WK, et al. Geographic accessibility to clinical trials for advanced cancer in the United States. JAMA Intern Med 2015. Feb;175(2):293–5. [DOI] [PubMed] [Google Scholar]
- 18.National Comprehensive Cancer Network. Bladder Cancer (Version 6.2020) [Internet] [cited 2020. Dec 24]. Available from: https://www.nccn.org/professionals/physician_gls/pdf/bladder.pdf
- 19.Zarin DA, Fain KM, Dobbins HD, Tse T, Williams RJ. 10-Year Update on Study Results Submitted to ClinicalTrials.gov. N Engl J Med 2019. 14;381(20):1966–74. [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.