Abstract
Introduction:
The impact of clinical trial participation on overall survival is unclear. We hypothesized that enrollment in a therapeutic drug clinical trial is associated with longer overall survival in patients with metastatic non-small cell lung cancer (NSCLC).
Patients and Methods:
We linked electronic medical record and Washington State cancer registry data to identify patients with metastatic NSCLC diagnosed between 1/1/2007–12/31/2015 that received treatment at a National Cancer Institute-designated cancer center. The exposure was trial enrollment. The primary outcome was overall survival, defined as the date of second line treatment initiation to date of death or last follow-up. We used a conditional landmark analysis starting at the date of second line treatment initiation and propensity scores with inverse probability of treatment weighting to estimate the association between trial enrollment and survival.
Results:
Of 215 patients, 40 (19%) participated in a second line trial. Trial participants were more likely to be never smokers (45% vs. 27%), have a good performance status (88% vs. 77%) and have EGFR (48% vs. 14%) and ALK mutations (8% vs. 5%) than non-participants. Trial participants had similar overall survival to non-participants (HR 1.05; 95% CI: 0.72, 1.53; p=0.81) after adjusting for sociodemographic and disease characteristics.
Conclusion:
Accounting for the immortal time bias and selection bias, trial participation does not appear detrimental to survival. This finding may be reassuring to patients and supports programs and policies to improve clinical trial access.
Keywords: Lung cancer, Clinical trial, Health outcomes, Health services research, Survival analysis
MicroAbstract
It is unclear whether clinical trial participation impacts survival in patients with cancer. Of 215 patients with advanced non-small cell lung cancer diagnosed between 1/1/2007–12/31/2015, 19% participated in a second line trial. Overall survival was similar for trial participants and non-participants, accounting for demographic and disease-related differences between these groups. This finding supports programs and policies to improve trial access.
Introduction
Participation in oncology clinical trials has historically been low at 2–8% due to lack of trial availability, strict eligibility criteria and other structural or patient-level barriers1–5. Trial participation is commonly assumed to have a positive effect on patient care through early access to new medications or treatment regimens, enhanced monitoring and supportive care, or changes in patient or physician behavior due to the feeling of being observed6. This perception could be inflated by trial eligibility criteria that tend to favor healthier patients and by publication bias favoring trials that show a positive effect7. It is unclear whether access to experimental agents or contextual changes in care translate into improved survival among trial participants as compared to non-participants receiving usual care.
The association between clinical trial participation and survival has been studied in a variety of tumor types7–11, including small cell lung cancer12–15 and non-small cell lung cancer (NSCLC)16–18. There has not been a clear consensus, with results split between either no difference in survival or improved survival associated with trial participation7. Understanding the impact of trial participation on outcomes like survival could help clinicians better inform their patients about the trade-offs of trial enrollment. Our goal was to compare the overall survival between trial and non-trial enrollees among a cohort of patients with metastatic NSCLC treated at a single large academic center, using a contemporary dataset with detailed patient and trial level characteristics. We hypothesized that trial participation is associated with longer overall survival. Our study is unique from prior studies in its use of a conditional landmark analysis and propensity scores with inverse probability of treatment weighting to account for immortal time bias and selection bias19, 20, its integration of electronic medical record data with high-quality sociodemographic registry data, and its focus on purely therapeutic drug trials.
Materials and Methods
Patients
We retrospectively reviewed the electronic medical records (EMRs) of 528 patients with de novo or recurrent metastatic NSCLC diagnosed between 1/1/2007 and 12/31/2015 that received treatment at the Seattle Cancer Care Alliance (SCCA) in Seattle, Washington (Figure 1). The SCCA is a National Cancer Institute-designated cancer center that serves a predominantly Caucasian middle-class population from Washington State, Alaska, Idaho, Montana and Wyoming. Inclusion criteria included cytologically or histologically proven NSCLC, diagnosis between 1/1/2007 and 12/31/2015, documentation of metastatic disease at any time, receipt of at least one dose of one or more anti-neoplastic agents within 180 days of confirmation of metastatic disease, and first-line therapy delivered at the Seattle Cancer Care Alliance. Exclusion criteria included active secondary malignancy (defined as receipt of active treatment for any other cancer, diagnosis of any other metastatic cancer, or diagnosis of any other cancer that had been treated with curative intent within three years from diagnosis of metastatic NSCLC), death within 60 days of diagnosis, referral to hospice within 90 days of diagnosis, and first enrollment in a therapeutic drug clinical trial in the first, third, fourth or fifth lines of systemic therapy. This study was approved by the Fred Hutchinson Institutional Review Board.
Figure 1. Consort Diagram Showing Development of Study Population.
We assessed a total of 528 patients with metastatic non-small cell lung cancer diagnosed between 1/1/2007 and 12/31/2015 who received palliative first line systemic therapy at the Seattle Cancer Care Alliance for eligibility. After applying inclusion and exclusion criteria and setting a conditional landmark at the initiation of second line therapy, the total study population was 215 patients.
We abstracted patient sociodemographic characteristics, smoking history, Eastern Cooperative Oncology Group (ECOG) performance status (if available within 60 days of metastatic disease), tumor histology, date of metastatic disease confirmation, epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK) mutation status, presence of brain metastases within 60 days after metastatic disease, systemic treatment history up to five lines, therapeutic drug clinical trial enrollment and characteristics of those trials directly from EMRs. We obtained additional sociodemographic characteristics including patient race, Spanish/Hispanic ethnicity, marital status, census-based median household income, the Area Deprivation Index (ADI)21 at the Washington State level and Rural Urban Commuting Area (RUCA) codes22 from the Washington State Cancer Surveillance System (CSS) database. CSS is part of the National Cancer Institute Surveillance, Epidemiology, and End Results Program and collects incidence, treatment and follow-up data on all new cancer cases in 13 counties in western Washington State23. All background characteristics were defined at the time of metastatic disease confirmation by pathology, imaging or clinician assessment, whichever came first. We obtained death dates from the CSS database complemented by medical records, censoring patients that were alive at end of follow-up (3/7/2019).
Study Design and Treatment
This is a retrospective cohort study. The study exposure was first therapeutic drug clinical trial participation in the second line of therapy for metastatic NSCLC. The primary outcome was overall survival, defined as the time from date of second line treatment initiation to the date of death from any cause or date of last follow-up.
Statistical Analysis
The statistical analysis was performed using Stata statistical software (version 16.1, StataCorp). No variables exceeded 13% missingness, and missing data were reclassified as unknown for the purposes of analysis. We calculated descriptive statistics including means, medians, standard deviations (SD) and 95% confidence intervals (CI) for continuous variables and proportions for categorical variables. Confounding covariates were considered background variables associated with both the trial exposure and the survival outcome and were defined a priori. These were defined as age, gender, race, Spanish/Hispanic ethnicity, marital status, RUCA residence, state-ranked ADI, smoking history, ECOG performance status, histology, EGFR/ALK mutation status and brain metastases and all were incorporated into the propensity score.
Our dataset included the start and stop dates of the first five lines of systemic therapy for each patient and the time of first trial start varied significantly among trial participants. This introduced the potential for immortal time bias in our analysis. Immortal time is a span of time where a study participant is under observation but has not yet initiated the study exposure (a clinical trial) and thus could not have experienced the study outcome (death). The participant had to remain free of the study outcome until the start of the exposure in order to be classified as exposed—that is, the participant had to survive to enroll in a clinical trial19, 24. One analytic method to account for this bias is a conditional landmark analysis. Applying a landmark resets the start of study follow-up to the time of the landmark and each study participant is classified as exposed or unexposed based of their status at the landmark19. We applied a conditional landmark at the date of second line treatment initiation, selecting this time because most trial participants first enrolled in a trial in the second line and this would allow for the largest sample size. Any immortal time incurred prior to start of second line treatment was excluded from our survival analysis with this approach.
Kaplan-Meier survival curves were generated for both the entire study population and stratified by trial participation status (yes vs. no). To further address selection bias, we calculated propensity scores, which are the probability that a study participant will receive the study exposure based on his/her underlying characteristics. We weighted each propensity score by the inverse probability of receiving the treatment (exposure) that the study participant received, followed by stabilization to account for study participants with a very low probability of receiving treatment20. This analytic approach allowed us to maintain our sample size, unlike a propensity score matching approach. The inverse probability of treatment weighted propensity scores were incorporated into a Cox model to assess the association between trial participation and survival. A p-value of <0.05 was considered statistically significant.
Results
Background Characteristics
Among the 215 eligible patients, the mean age at confirmation of metastatic disease was 62 years, 115 (53%) were female, 158 (73%) were white, 144 (67%) were married and 165 (77%) resided in a metropolitan area. Most were former or current smokers (59% and 11%, respectively), had non-squamous tumor histology (93%) and had a favorable ECOG performance status of 0–1 (79%). Forty-four patients (20%) had an EGFR mutation, while 11 (5%) had an ALK rearrangement (Table 1).
Table 1.
Study population background characteristics
| Background characteristic | Study population (n=215) |
|---|---|
| Age at confirmation of metastatic disease, mean (standard deviation), years | 62 (11) |
| Female gender, number (%) | 115 (53) |
| Race, number (%) | |
| White | 158 (73) |
| Black | 6 (3) |
| American Indian/Alaskan Native | 4 (2) |
| Asian | 40 (19) |
| Other or Unknown | 7 (3) |
| Spanish/Hispanic origin, number (%) | 8 (4) |
| Married, number (%) | 144 (67) |
| Median income by zip code, median (interquartile range), dollars | 73,406 (58,641, 90,802) |
| State-ranked ADI, number (%) | |
| < 7 | 152 (71) |
| ≥ 7 | 35 (16) |
| Unknown | 28 (13) |
| RUCA residence, number (%) | |
| Metropolitan | 165 (77) |
| Micropolitan | 8 (4) |
| Town/Rural | 15 (7) |
| Unknown | 27 (13) |
| Smoking history, number (%) | |
| Never or Unknown | 66 (31) |
| Former | 126 (59) |
| Current | 23 (11) |
| ECOG score, number (%) | |
| 0–1 | 170 (79) |
| ≥ 2 | 28 (13) |
| Not documented | 17 (8) |
| Non-squamous tumor histology, number (%) | 201 (93) |
| Presence of a targetable mutation, number (%) | |
| EGFR positive | 44 (20) |
| ALK positive | 11 (5) |
| Brain metastasis detected within 60 days of metastatic disease, number (%) | 61 (28) |
Abbreviations used: ADI=Area Deprivation Index, RUCA=Rural Urban Commuting Area, ECOG=Eastern Cooperative Oncology Group, EGFR=epidermal growth factor receptor, ALK=anaplastic lymphoma kinase
Of the entire study population, 40 patients (19%) participated in their first therapeutic drug clinical trial during the second line of therapy. The remaining 175 patients never participated in a trial. Trial participants were more likely to be female (65% vs. 51%), white (82% vs. 71%), married (78% vs. 65%), never smokers (45% vs. 27%) and have an ECOG performance status of 0–1 (88% vs. 77%) compared to non-participants (Table 2). They also resided in areas with higher median household incomes ($77,948 vs. $72,150) and were less likely to live in socioeconomically disadvantaged areas, identified as state-ranked ADI ≥ 7 (10% vs. 18%). Trial participants were more likely to have EGFR (48% vs. 14%) and ALK mutations (8% vs. 5%) and were less likely to have brain metastases (15% vs. 31%).
Table 2.
Background characteristics by clinical trial participation
| Background characteristic | Non-clinical trial participant (n=175) | Clinical trial participant (n=40) |
|---|---|---|
| Age at confirmation of metastatic disease, mean (standard deviation), years | 62 (11) | 62 (9) |
| Female gender, number (%) | 89 (51) | 26 (65) |
| Race, number (%) | ||
| White | 125 (71) | 33 (82) |
| Black | 6 (3) | 0 (0) |
| American Indian/Alaskan Native | 4 (2) | 0 (0) |
| Asian | 33 (19) | 7 (18) |
| Other or Unknown | 7 (4) | 0 (0) |
| Spanish/Hispanic origin, number (%) | 7 (4) | 1 (3) |
| Married, number (%) | 113 (65) | 31 (78) |
| Median income by zip code, median (interquartile range), dollars | 72,150 (56,520, 89,527) | 77,948 (62,485, 94,063) |
| State-ranked ADI, number (%) | ||
| < 7 | 120 (68) | 32 (80) |
| ≥ 7 | 31 (18) | 4 (10) |
| Unknown | 24 (14) | 4 (10) |
| RUCA residence, number (%) | ||
| Metropolitan | 132 (75) | 33 (83) |
| Micropolitan | 7 (4) | 1 (3) |
| Town/Rural | 13 (7) | 2 (5) |
| Unknown | 23 (13) | 4 (10) |
| Smoking history, number (%) | ||
| Never or Unknown | 48 (27) | 18 (45) |
| Former | 105 (60) | 21 (53) |
| Current | 22 (13) | 1 (3) |
| ECOG score, number (%) | ||
| 0–1 | 135 (77) | 35 (88) |
| ≥ 2 | 26 (15) | 2 (5) |
| Not documented | 14 (8) | 3 (8) |
| Non-squamous tumor histology, number (%) | 164 (94) | 37 (93) |
| Presence of a targetable mutation, number (%) | ||
| EGFR positive | 25 (14) | 19 (48) |
| ALK positive | 8 (5) | 3 (8) |
| Brain metastasis detected within 60 days of metastatic disease, number (%) | 55 (31) | 6 (15) |
Abbreviations used: ADI=Area Deprivation Index, RUCA=Rural Urban Commuting Area, ECOG=Eastern Cooperative Oncology Group, EGFR=epidermal growth factor receptor, ALK=anaplastic lymphoma kinase
Treatment and Trial Characteristics
Clinical trial participants had a higher median number of days from metastatic disease to second line therapy (169 vs. 113 days) and from first line discontinuation to second line treatment start (22 vs. 11 days) compared to non-participants (Table 3). Both groups received a mean of 3 lines of therapy. Trial participants were more likely to receive combination chemotherapy and an anti-vascular endothelial growth factor (VEGF) agent (15% vs. 10%), an EGFR tyrosine kinase inhibitor (TKI; 60% vs. 54%), an ALK TKI (10% vs. 6%), a programmed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) checkpoint inhibitor (43% vs. 12%) or another targeted therapy (38% vs. 4%) during any line of therapy.
Table 3.
Treatment characteristics by clinical trial participation
| Treatment characteristic | Non-clinical trial participant (n=175) | Clinical trial participant (n=40) |
|---|---|---|
| Number of days until second line therapy, median (interquartile range) | 113 (84, 223) | 169 (117,337) |
| Number of days from first line stop to second line therapy start, median (interquartile range) | 11 (7, 16) | 1 22 (14, 30) |
| Number of therapeutic lines, mean (standard deviation) | ||
| Total | 3 (1) | 3 (1) |
| Excluding clinical trials | - | 2 (1) |
| Therapeutic class used at any time during treatment, number (%) | ||
| Chemotherapy | 159 (91) | 34 (85) |
| Chemotherapy + VEGF agent | 18 (10) | 6 (15) |
| EGFR TKI | 94 (54) | 24 (60) |
| ALK TKI | 11 (6) | 4 (10) |
| PD-1/PD-L1 checkpoint inhibitor | 21 (12) | 17 (43) |
| Other targeted therapy | 7 (4) | 15 (38) |
| Trial participation characteristic | ||
| Enrolled in > 1 trial, number (%) | - | 10 (25) |
| Participated in ≥ 1 phase I/II trial, number (%) | - | 34 (85) |
| Participated in ≥ 1 phase III trial, number (%) | - | 8 (20) |
| Participated in ≥ 1 randomized controlled trial, number (%) | - | 10 (25) |
| Participated in ≥ 1 trial of a therapeutic agent later approved by the FDA, number (%) | - | 17 (43) |
Abbreviations used: VEGF=vascular endothelial growth factor, EGFR=epidermal growth factor receptor, ALK=anaplastic lymphoma kinase, PD-1=programmed cell death protein 1, PD-L1=programmed death-ligand 1, FDA=Food and Drug Administration
Of the 40 patients that participated in a clinical trial, 10 (25%) enrolled in more than one trial (Table 2). Thirty-four participants enrolled in at least one phase I/II trial (85%). Ten patients (25%) participated in at least one randomized controlled trial, and 17 patients (43%) participated in at least one trial of a therapeutic agent later approved by the FDA.
Overall Survival by Trial Participation
Among the entire study population, median overall survival from the date of second line therapy initiation was 1.0 years (95% confidence interval (CI): 0.8, 1.2 years). Median overall survival for trial participants was 1.4 years (95% CI: 0.9, 1.9 years) as compared to 0.9 years (95% CI: 0.8, 1.1 years) among non-participants. After accounting for immortal time bias and selection bias, trial participants had a 5% higher risk of death compared to non-participants (hazard ratio (HR) 1.05; 95% CI: 0.72, 1.53; p=0.81) (Figure 2).
Figure 2. Kaplan-Meier Survival Estimates by Clinical Trial Participation.
Median overall survival for non-participants was 0.9 years (95% confidence interval (CI): 0.8, 1.1 years) and for trial participants was 1.4 years (95% CI: 0.9, 1.9 years). After accounting for immortal time bias and selection bias, trial participants had a 5% higher risk of death compared to non-participants (adjusted hazard ratio 1.05; 95% CI: 0.72, 1.53; p=0.81).
Discussion
In this single center, retrospective cohort study of EMRs and registry-derived data, clinical trial participation was not associated with an overall survival benefit in patients with metastatic NSCLC. This finding is consistent with one of the two recent studies in patients with advanced NSCLC16, 17, as well as with a larger study evaluating this question more broadly across randomized cooperative group trials for various tumor types11. Abu-Hejleh et al. found no association between clinical trial participation and overall survival in 815 patients with advanced NSCLC treated between September 2003 and June 2005 (HR 1.05, p=0.81)16. Arrieta et al. found that overall survival was significantly longer for patients that participated in any clinical trial in a cohort of 1,042 patients with advanced NSCLC treated between January 2007 and December 2014 (HR 0.74, p=0.004)17. Neither study focused exclusively on therapeutic drug trials, and Arrieta et al. had a larger percentage of trial participants (71.7%) than expected based on the current literature regarding trial participation in patients with cancer. Importantly, trial participation was considered a binary variable and time of trial enrollment was not factored into either analysis, potentially introducing immortal time bias24–26. Failure to account for this bias has consistently resulted in lower hazard ratios26, which could incorrectly create the impression of a survival effect or could impact the magnitude of an effect. Unger et al. found that trial participation was associated with better survival for patients with poor-prognosis cancers when looking at 5,190 patients enrolled in 21 large cooperative group phase III trials between 1987 and 2007, as compared to a cohort of non-trial patients selected from the Surveillance, Epidemiology, and End Results (SEER) registry11. However, this survival improvement only extended for one year after trial participation and there was no survival benefit associated with trial participation among patients with better prognosis cancers, leading investigators to hypothesize that the short-term survival benefit was related to selection of healthier patients for trials. Peppercorn et al. conducted a systematic review of 24 studies across cancer types and found that approximately half of these studies suggested a survival benefit with trial participation. However, they concluded that the quality of the evidence was poor largely due to methodological issues such as failure to adequately account for potential confounding covariates7.
There are several potential reasons why clinical trial participation was not associated with improved survival in our study. First, it is possible that our study was underpowered to detect a difference in survival. Second, while our study focused on therapeutic drug clinical trials, the heterogeneity of trial designs and investigational drug classes (e.g., targeted therapies versus immune checkpoint inhibitors) may have obscured a difference. Third, the therapeutic assignments of participants on randomized clinical trials are unknown, and it is likely that many of these patients were assigned to the control or standard of care arms. Fourth, while trial patients disproportionately had access to therapeutic advances in lung cancer such as targeted therapies and PD-1/PD-L1 checkpoint inhibitors, not all access to these agents occurred through a trial among the entire study population. Receipt of these agents off-trial may have reduced a difference in survival between trial participants and non-participants. Finally, this study involves a time of rapid advances in lung cancer. Many promising therapies—including pemetrexed maintenance, first- and second-line EGFR and ALK TKIs, and second-line PD-1 checkpoint inhibitors—received FDA approval during this time period and were rapidly incorporated into standard of care at our institution27. It is possible that this obscured any potential survival benefit associated with early access to these agents through a clinical trial at our site.
This study contributes several new insights to the existing literature evaluating the benefits of clinical trial participation. It is impossible to record every physician- and patient-level factor that contributes to trial enrollment. However, the clinically detailed information on patient characteristics, socioeconomic indicators and disease characteristics in this dataset has either not been captured or adjusted for in many prior analyses of this question7. Our dataset also had detailed information on treatment initiation and discontinuation for the first five lines of systemic therapy. This allowed us to account for immortal time bias by performing a conditional landmark analysis. We considered other statistical methods to account for immortal time, including an extended Cox model with the trial exposure as a time-dependent covariate19 and propensity score matching with non-trial patients assigned an index date corresponding to date of trial start in their matched trial patient28. The former would have allowed for a larger study population, but it was unclear whether standard risk adjustment for confounding covariates in this analysis would be adequate to address selection bias. The latter considerably reduced the study population, making interpretation of results difficult due to lack of power or generalizability of our findings20. Another advantage of our dataset was detailed patient-level information on therapeutic agents, allowing us to assess the classes of agents received between trial participants and those that did not participate in a trial. With the caveats of an observational study, our results may assist oncologists with discussions about the trade-offs of clinical trial enrollment for patients with metastatic NSCLC.
A primary limitation of this study is that its observational design does not allow for causal inference, and the results can only suggest the lack of an association between trial participation and survival. Because patients were not randomized to participate or not participate in a clinical trial, the possibility of selection bias prevents firm conclusions about the effect of trial participation on survival. We attempted to account for this with the combination of a conditional landmark analysis and application of propensity scores. However, these methods only partially address selection bias and it is impossible to capture and adjust for the many intangible elements underlying patient selection for enrollment in clinical trials. Another important limitation of our dataset is that the therapeutic assignments for the patients enrolled in randomized trials are unknown. It is likely that many of the participants in randomized trials were assigned to control arms rather than receiving the newer therapies being tested. The reason for discontinuation of trial participation was also not captured in the data. Further data on reasons for discontinuation—whether personal preference, intolerable adverse effects, disease progression or death—might provide additional insight into our results. Furthermore, the nature of monitoring and supportive care differs across clinical trials and it was not possible to capture or adjust for the impact of this.
Conclusion
In summary, patients with metastatic NSCLC that participated in a therapeutic drug clinical trial experienced similar survival to patients that did not participate. These findings may provide reassurance to patients and oncologists that trial participation does not appear detrimental to survival and that clinical trial participation should still be encouraged when appropriate. Furthermore, they should encourage partnerships between trial sponsors, oncology practices and insurance plans to improve access to clinical trials and sustain innovation in oncologic drug development.
Clinical Practice Points.
Clinical trial participation in patients with cancer has historically been low, but there is growing interest in identifying and addressing barriers to participation. Understanding the impact of trial participation on outcomes like survival could help clinicians to better inform their patients about the trade-offs of trial enrollment. Studies to date have not reached a consensus on whether trial participation is associated with overall survival, and many of these studies did not appropriately account for sources of bias with their research methods. We sought to more definitively answer this question with a retrospective cohort study of 215 patients with metastatic non-small cell lung cancer diagnosed between 1/1/2007–12/31/2015 and treated at an academic oncology center. Forty patients (19%) enrolled in a second line clinical trial, and trial participants were more likely than non-participants to have favorable prognostic characteristics like a good performance status or the presence of an underlying driver mutation. After accounting for demographic and disease-related differences between these two groups, we found that trial participants had similar overall survival to non-participants (HR 1.05; 95% CI: 0.72, 1.53; p=0.81). This finding may reassure patients and clinicians that trial participation is not detrimental to survival and that clinical trials should still be considered during care where appropriate. It also supports the development of programs and policies to improve access to clinical trials.
Acknowledgements
The authors would like to thank Emily Silgard, MS for technical support in developing the study dataset, and the Fred Hutchinson Biostatistics Resource for assistance with statistical software code.
Funding Statement
This work was supported by Seattle Cancer Care Alliance Thoracic Oncology Research (THOR) [grant number not applicable]; and the National Cancer Institute [grant number T32CA009515, Dr. Cristina Merkhofer only]. Neither funding source had a role in the study design, the collection, analysis and interpretation of data, in the writing of this manuscript or in the decision to submit this manuscript for publication.
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.
Conflict of Interest Statement
The authors declare no conflicts of interest. (If you need disclosures, they are as follows. None of these represent a conflict of interest. Keith D. Eaton has clinical trial research support from Mirati Therapeutics. Renato G. Martins has clinical trial research support from AstraZeneca, Merck & Company, Inc., Pfizer and Roche. Scott D. Ramsey has the following disclosures: Employment with Flatiron Health; Consulting or Advisory Roles with Bayer Corporation, Bristol-Myers Squibb, AstraZeneca, Merck & Company, Inc., GRAIL, Pfizer, Seattle Genetics, Biovica and Genentech; Research Funding from Bayer Corporation, Bristol-Myers Squibb and Microsoft Corporation; Travel, Accommodations, Expenses from Bayer Schering Pharma, Bristol-Myers Squibb, Flatiron Health, Bayer and GRAIL.)
References
- 1.Murthy VH, Krumholz HM, Gross CP. Participation in cancer clinical trials: race-, sex-, and age-based disparities. JAMA. 2004;291:2720–2726. [DOI] [PubMed] [Google Scholar]
- 2.Sateren WB, Trimble EL, Abrams J, et al. How sociodemographics, presence of oncology specialists, and hospital cancer programs affect accrual to cancer treatment trials. J Clin Oncol. 2002;20:2109–2117. [DOI] [PubMed] [Google Scholar]
- 3.English RALY and Giffin RB Transforming Clinical Research in the United States: Challenges and Opportunities, Workshop Summary: Institute of Medicine of the National Academies; 2010. [PubMed] [Google Scholar]
- 4.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;111:245–255. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Minasian LM, Unger JM. What Keeps Patients Out of Clinical Trials? JCO Oncol Pract. 2020;16:125–127. [DOI] [PubMed] [Google Scholar]
- 6.Braunholtz DA, Edwards SJ, Lilford RJ. Are randomized clinical trials good for us (in the short term)? Evidence for a "trial effect". J Clin Epidemiol. 2001;54:217–224. [DOI] [PubMed] [Google Scholar]
- 7.Peppercorn JM, Weeks JC, Cook EF, Joffe S. Comparison of outcomes in cancer patients treated within and outside clinical trials: conceptual framework and structured review. Lancet. 2004;363:263–270. [DOI] [PubMed] [Google Scholar]
- 8.Sorbye H, Pfeiffer P, Cavalli-Bjorkman N, et al. Clinical trial enrollment, patient characteristics, and survival differences in prospectively registered metastatic colorectal cancer patients. Cancer. 2009;115:4679–4687. [DOI] [PubMed] [Google Scholar]
- 9.Robinson WR, Ritter J, Rogers AS, Tedjarati S, Lieberenz C. Clinical trial participation is associated with improved outcome in women with ovarian cancer. Int J Gynecol Cancer. 2009;19:124–128. [DOI] [PubMed] [Google Scholar]
- 10.Roy P, Vaughan Hudson G, Vaughan Hudson B, Esteve J, Swerdlow AJ. Long-term survival in Hodgkin's disease patients. A comparison of relative survival in patients in trials and those recorded in population-based cancer registries. Eur J Cancer. 2000;36:384–389. [DOI] [PubMed] [Google Scholar]
- 11.Unger JM, Barlow WE, Martin DP, et al. Comparison of survival outcomes among cancer patients treated in and out of clinical trials. J Natl Cancer Inst. 2014;106:dju002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Burgers JA, Arance A, Ashcroft L, Hodgetts J, Lomax L, Thatcher N. Identical chemotherapy schedules given on and off trial protocol in small cell lung cancer: response and survival results. Br J Cancer. 2002;87:562–566. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Cottin V, Arpin D, Lasset C, et al. Small-cell lung cancer: Patients included in clinical trials are not representative of the patient population as a whole. Ann Oncol. 1999;10:809–815. [DOI] [PubMed] [Google Scholar]
- 14.Quoix E, Finkelstein H, Wolkove N, Kreisman H. Treatment of small-cell lung cancer on protocol: potential bias of results. J Clin Oncol. 1986;4:1314–1320. [DOI] [PubMed] [Google Scholar]
- 15.Schea RAP P; Allen PK; Komaki R; Cox JD Limited-stage small-cell lung cancer: patient survival after combined chemotherapy and radiation therapy with and without treatment protocols. Radiology. 1995;197:859–862. [DOI] [PubMed] [Google Scholar]
- 16.Abu-Hejleh T, Chrischilles EA, Halfdanarson TR, et al. The Effect of Receiving Treatment Within a Clinical Trial Setting on Survival and Quality of Care Perception in Advanced Stage Non-Small Cell Lung Cancer. Am J Clin Oncol. 2016;39:126–131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Arrieta O, Carmona A, Ramirez-Tirado LA, et al. Survival of Patients with Advanced Non-Small Cell Lung Cancer Enrolled in Clinical Trials. Oncology. 2016;91:185–193. [DOI] [PubMed] [Google Scholar]
- 18.Davis S, Wright PW, Schulman SF, et al. Participants in prospective, randomized clinical trials for resected non-small cell lung cancer have improved survival compared with nonparticipants in such trials. Cancer. 1985;56:1710–1718. [DOI] [PubMed] [Google Scholar]
- 19.Giobbie-Hurder A, Gelber RD, Regan MM. Challenges of guarantee-time bias. J Clin Oncol. 2013;31:2963–2969. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Austin PC. An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies. Multivariate Behav Res. 2011;46:399–424. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.University of Wisconsin School of Medicine and Public Health: 2015. Area Deprivation Index v2.0. https://www.neighborhoodatlas.medicine.wisc.edu/. Last accessed 8/1/2020.
- 22.United States Department of Agriculture: Economic Research Service: Rural-Urban Commuting Area Codes, 10/24/2019 update. https://www.ers.usda.gov/data-products/rural-urbancommuting-area-codes.aspx. Last accessed 8/1/2020.
- 23.Fred Hutchinson Cancer Research Center: Cancer Surveillance System (CSS). https://www.fredhutch.org/en/research/divisions/public-health-sciences-division/research/epidemiology/cancer-surveillance-system.html. Last accessed 8/1/2020.
- 24.Suissa S. Immortal time bias in pharmaco-epidemiology. Am J Epidemiol. 2008;167:492–499. [DOI] [PubMed] [Google Scholar]
- 25.Beyersmann J, Gastmeier P, Wolkewitz M, Schumacher M. An easy mathematical proof showed that time-dependent bias inevitably leads to biased effect estimation. J Clin Epidemiol. 2008;61:1216–1221. [DOI] [PubMed] [Google Scholar]
- 26.Beyersmann J, Wolkewitz M, Schumacher M. The impact of time-dependent bias in proportional hazards modelling. Stat Med. 2008;27:6439–6454. [DOI] [PubMed] [Google Scholar]
- 27.U.S. Food & Drug Administration: Hematology/Oncology (Cancer) Approvals & Safety Notifications, 12/19/2016 update. http://wayback.archive-it.org/7993/20170111064250/http://www.fda.gov/Drugs/InformationOnDrugs/ApprovedDrugs/ucm279174.htm. Last accessed 8/1/2020.
- 28.McDougall JA, Bansal A, Goulart BH, et al. The Clinical and Economic Impacts of Skeletal-Related Events Among Medicare Enrollees With Prostate Cancer Metastatic to Bone. Oncologist. 2016;21:320–326. [DOI] [PMC free article] [PubMed] [Google Scholar]