This article is more than 5 years old. Information may no longer be current.
Model predicts low accrual rates for clinical trials
Researchers have identified several trial-level and treatment-related characteristics of National Clinical Trials Network-sponsored trials that are associated with low accrual rates.
Based on these factors, researchers developed a model that predicts risk for low accrual.
Nearly 25% of publicly sponsored clinical trials fail to enroll enough participants, according to study background.
“Clinical trials that do not achieve sufficient accrual are frequently unable to inform clinical practice or benefit patients,” Caroline S. Bennette, MPH, PhD, of the pharmaceutical outcomes research and policy program at University of Washington in Seattle, and colleagues wrote. “Moreover, low-accruing clinical trials often represent a waste of scarce human and economic resources.”
Bennnette and colleagues examined data from 787 phase 2 or 3 National Clinical Trials Network (NCTN) cancer clinical trials launched between 2000 and 2011 to determine factors associated with low trial accrual. From those data, the researchers developed a predictive model for low accrual — defined as trials that either closed with or were accruing less than 50% of their target cohorts. An additional 46 trials that opened in 2012 or 2013 were used for prospective validation of the model.
Eighteen percent (n = 145) of the trials closed or were accruing at less than 50% of the target 3 years or more following the trial launch.
Trial-level factors associated with low accrual included trial launch in settings of higher competition (median, 4.4 vs. 2.9 trials per 10,000 eligible patients per year), evaluation of conditions with lower annual incidence (median, 64,990 vs. 76,100 patients), requirement of larger enrollment fraction (0.068% vs. 0.031% of eligible patient population each year) and a nonmetastatic setting (nonmetastatic vs. metastatic setting, 23% vs. 16%). Treatment-related factors associated with low accrual included evaluation of multimodality therapy (48% vs. 41%), surgery (19% vs. 16%) or radiation (32% vs. 20%).
Based on these findings, the researchers created a multivariable logistic model to predict low accrual that included 12 factors.
These factors included the number of competing trials per 10,000 eligible patients each year (OR = 1.88; 95% CI, 1.32-2.68); phase 3 vs. phase 2 (OR = 1.86; 95% CI, 1.03-3.37); enrollment as a percentage of the eligible population; targeted therapy (OR = 0.57; 95% CI, 0.36-0.89); radiation therapy (OR = 1.81; 95% CI, 1.16-2.81); annual incidence of clinical condition per 10,000 patients (OR = 0.99; 95% CI, 0.97-1.01); tissue sample requirement to determine eligibility (OR = 1.26; 95% CI, 0.84-1.87); investigational new drug (OR = 0.34; 95% CI 0.17-0.69); metastatic setting (OR = 1.46; 95% CI, 0.82-2.58); sample size, per 100 patients (OR = 0.95; 95% CI, 0.91-0.99); more than one condition evaluated (OR = 1.98; 95% CI, 1.26-3.11); common solid cancer compared with liquid or rare solid cancers (OR = 2.32; 95% CI, 1.31-4.1); and the interaction term (OR = 2.47; 95% CI, 0.84-7.25).
The model appeared to exhibit good discriminatory performance when applied internally to the trials launched between 2000 and 2011 (boot-strap corrected area under the curve [AUC] = 0.73; 95% CI, 0.69-0.78) and when applied to trials launched between 2012 and 2013 (AUC = 0.73; 95% CI, 0.55-0.92).
“Systematically considering the overall influence of these factors could aid in the design and prioritization of future clinical trials,” the researchers wrote. “[This can] improve selection, support and completion of publicly funded cancer clinical trials.”
Derek Raghavan
In an accompanying editorial, Derek Raghavan, MD, PhD, FACP, FRACP, FASCO, president of Levine Cancer Institute at Carolinas HealthCare System and HemOnc Today’s chief medical editor for oncology, discussed whether this research places the “right emphasis” on improving clinical trial accrual.
“The real issue is still the lack of patients involved in cancer trials, irrespective of whether they are conducted by cooperative groups, private or single-institution entities or the pharmaceutical industry,” Raghavan wrote. “… In addition to the predictors of low accrual that relate to specifics of design, there are barriers to enrollment, such as availability of appropriate protocols, performance status and comorbidities, insurance and fiscal issues, disparities of cancer care and distance from cancer center.”
Rather than avoiding trial factors associated with low accrual, innovative strategies are needed to rethink the design of trials, Raghavan added.
“In the broad context of health care delivery, there has been a dramatic shift from volume-drive to value-driven constructs, and this movement has found its way into oncology, with the genesis of programs to reduce profligate waste and structure attempts to place mathematical algorithms into the definition of value,” he wrote. “It is time to apply these concepts to trial design and to conceive studies that focus on real value with substantial increments in outcome, rather than focusing solely on P values, Forest plots and waterfall plots as (false?) measures of progress.” – by Anthony SanFilippo
Disclosure: The study was funded in part by grants from the Agency for Healthcare Research and Quality and the National Science Foundation. Raghavan reports no relevant financial disclosures.