Proportion of trial participants who were 65 years or older recruited during the interval before 2000 (1997-1999) or after 2000 (2001-2003). A, Phase 3 trials; B, phase 1 and 2 trials. Asterisk indicates P<.05.
Gross CP, Wong N, Dubin JA, Mayne ST, Krumholz HM. Enrollment of Older Persons in Cancer Trials After the Medicare Reimbursement Policy Change. Arch Intern Med. 2005;165(13):1514-1520. doi:10.1001/archinte.165.13.1514
Copyright 2005 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2005
The Medicare program was revised in 2000 to authorize payment of routine care costs for patients in clinical trials. This study evaluates the impact of the Medicare reimbursement policy change on enrollment of older patients into cancer trials.
We conducted a longitudinal analysis of the enrollment of older patients in National Cancer Institute–sponsored lung, colorectal, breast, and prostate cancer trials from January 1996 through June 2003. Our primary analysis was restricted to 23 trials that were actively accruing before and after the Medicare change, enrolled more than 100 patients, and had at least 5% of participants older than 65 years. We used multivariate logistic regression to determine whether the proportion of participants who were elderly increased after the policy change.
The primary study sample consisted of 16 135 patients in 23 trials. After adjusting for cancer type, trial exclusion criteria, and patient sociodemographic characteristics in multivariate analysis, no significant change occurred in the representation of older patients after the policy change (adjusted odds of participants being elderly after vs before policy change, 1.13; 95% confidence interval, 0.96-1.32; P = .14). In a secondary analysis of patients who had enrolled in any trial during the study period, no significant change occurred in enrollment of older persons after the policy change among phase 3 or phase 1 and 2 trial participants (P = .84 and P = .07 in adjusted analyses, respectively).
The Medicare trial reimbursement policy was not associated with a significant increase in the enrollment of older patients in cancer trials.
Cancer is an aging-related disease, with older persons accounting for approximately 61% of all new cancer cases and 70% of all cancer deaths.1 As the US population ages, a greater proportion of cancer patients will be older. Unfortunately, representation of older persons (≥65 years) in clinical research has been disproportionately low. Prior work has suggested that older persons represent approximately one third of research participants, despite the fact that they represent approximately two thirds of patients with cancer.2
Despite the well-recognized importance of clinical trial participation, recruitment of older patients continues to be a challenge. Impediments to research participation may include physician, patient, study design, and health system barriers.2- 5 A frequently cited health system barrier has been the lack of insurance coverage for “routine care costs” for patients enrolling in clinical research.6,7 Routine care costs include items or services that would be provided to patients even if they were not enrolled in a trial, including the reasonable and necessary care used to diagnose or treat complications of care.8 Until recently, Medicare and most private insurers did not always pay for routine care costs because of the experimental nature of these treatments, the uncertainty of potential benefits, and the possible risks.9- 12 As a result, investigators could not guarantee that Medicare would pay for the required care and would disclose this information as part of informed consent.6 Accordingly, prospective trial participants have expressed concern about trial-related costs, and some studies cite concerns about lack of reimbursement as one of the primary reasons for their lack of participation.13,14 In a survey of 5980 cancer patients, 20% of patients who were aware of clinical trials did not participate because they believed the cost of the treatment would not be covered by their insurance.15
The extension of Medicare reimbursement was recommended by an Institute of Medicine report in January 2000.6 In response to this report and to concerns that “too few seniors participate in clinical trials, [and that] current Medicare reimbursement policies often discourage seniors from participating,” the Medicare reimbursement policy was revised in June 2000.16,17 The new, highly publicized policy authorized payment for routine care costs associated with federally sponsored and Food and Drug Administration–approved clinical trials.17- 19 Although a recent analysis concluded that state reimbursement mandates had mixed results regarding the enrollment of patients younger than 65 years into cancer trials, the effect of the Medicare policy on enrolling older patients into cancer trials has not been investigated.7 We therefore conducted this study to determine whether the new Medicare reimbursement policy enhanced the enrollment of older persons in ongoing National Cancer Institute (NCI)–sponsored clinical trials.
We conducted an analysis of cancer trial enrollment from January 1996 through June 2003 to assess whether the proportion of participants who were 65 years and older increased after the Medicare policy was changed in 2000. Our study included participants in NCI-sponsored cooperative group breast, colorectal, lung, and prostate cancer trials, the 4 most common causes of cancer-related mortality during the study period.1,20
We used 2 approaches to address the main hypothesis of the study. In the first approach, we restricted our analyses to a specific group of trials. Because prior work has demonstrated substantial variation among clinical trials in the enrollment of older persons, only patients who were enrolled in specific protocols were included in this main analytic sample. This allowed us to account for changes in characteristics of active trials and protocol-specific eligibility criteria over time by including only trials that were actively recruiting both before and after the policy change. We used several criteria to select trials for our analysis. We excluded trials that had enrolled fewer than 100 patients during the study period or that had less than 5% of participants 65 years and older. We hypothesized a priori that trials with less than 5% elderly patients were designed in such a way to limit access and eligibility of older patients in general, and it is unlikely that the policy change would have had much effect on their recruitment. Finally, we restricted the sample to trials with 20% to 80% of patients enrolled before the Medicare reimbursement policy revision.
For the second approach, we analyzed all participants who enrolled in cooperative group trials during the 10 quarter periods before 2000 with that after 2000. We excluded data from the year 2000 (the year of the policy change). The purpose of this analysis was to provide a broad overview of elderly patient enrollment to determine whether the findings from our primary sample of 23 trials were consistent with the experience across the full spectrum of cooperative group trials. We stratified according to phase (phase 3 vs all other phases) because we hypothesized based on recent work that the impact of the reimbursement policy may have been greater among early-phase trial participants.7 the study were excluded.
Clinical trial participant data were obtained from the NCI Cancer Therapy Evaluation Program. The Cancer Therapy Evaluation Program database includes information about all cooperative group trial participants, including patient characteristics (ie, sex, race, age, and ZIP code of residence) and limited protocol information (ie, name, identification number, and phase). We obtained additional protocol information from the NCI Physician Data Query Web site,21 including eligibility criteria such as cancer type, age, performance status, and specified exclusions for hematopoietic, hepatic, renal, cardiovascular, and pulmonary comorbid conditions.
We coded each comorbid condition as a dichotomous variable based on whether any exclusionary criteria were specified. We then added the number of comorbid condition exclusions specified in each trial, with totals ranging from 0 to 5. We classified trials that only permitted persons with scores of 0 to 1 according to the Eastern Cooperative Oncology Group scale or greater than 60% according to the Karnofsky scale as accepting only good performance status or better.1,22,23 We categorized studies that allowed persons with scores up to Eastern Cooperative Oncology Group 3 or Karnofsky scores greater than 20% as “intermediate performance status eligible” protocols. We included trials that did not indicate any performance status restrictions in the “not specified” category.
We evaluated the enrollment of older participants for each time quarter, beginning January 1996 and ending June 2003. Using the individual patient as the unit of analysis, we used a χ2 test to compare the proportion of trial participants who were older before and after the Medicare policy change. We conducted bivariate analyses, dividing the continuous variables into quintiles and then further aggregating into 2 or 3 categories based on the bivariate distributions when appropriate.
For our primary analysis, we conducted a logistic regression in which the dependent variable was “older” (age at enrollment was ≥65 years) and the independent variable was time in quarters from January 1996 to June 2003. The likelihood that participants would be elderly was correlated with the protocol in which they were enrolled. To account for this clustering of patients within protocols, while still allowing for heterogeneity across protocols, we performed logistic regression with (robust) sandwich estimators of variance.24- 26 We included a binary variable for the Medicare policy change date (before the third quarter of calendar year 2000 vs after) in the regression model. Then we incorporated a linear spline, allowing a change in slope (ie, rate of change in percentage of patients who were elderly) at the time of the Medicare policy enactment, the third quarter of 2000. We included indicator variables for sex, race, cancer type, performance status, and comorbidity exclusion criteria. We chose the variables retained in the model using a backward elimination selection method, retaining those with P<.05.
For our secondary analysis of all trial participants, we used logistic regression to assess the probability of participants being elderly before and after the year 2000. We used Huber-White variance estimators to account for clustering of patients according to protocol.24- 26 We performed all statistical analyses using Stata statistical software, version 8.0 (Stata Inc, College Park, Tex).27 Statistical significance was set at P<.05.
Table 1 gives the assembly of the study sample; 23 of 249 trials met all inclusion criteria, and these trials enrolled a total of 16 135 patients. Approximately 39.3% of the 6300 patients enrolled before the Medicare policy change were 65 years or older, whereas only 33.5% of 9835 patients enrolled after the policy change were older (P<.001 for comparison; Table 2). Women were less likely to be elderly than men (27.4% vs 53.1%; P<.001), whereas there was no significant difference between blacks and whites.
Of lung cancer trial participants, 44.3% were elderly, in contrast to 22.6% of breast cancer participants and 70.5% of prostate cancer patients (P<.001 for each pairwise comparison with lung cancer; Table 2). Patients who lived in areas with higher poverty levels or unemployment rates were more likely to be elderly than patients who resided in areas with lower rates of poverty or unemployment (P<.001 for both comparisons). Patients who resided in counties with higher population density, managed care competition, and managed care penetration were less likely to be older. Finally, the number of trial exclusion criteria was inversely related to the participation of older patients, whereas specified performance status was not related.
The relation between patient and trial characteristics and elderly enrollment before and after the Medicare policy change is given in Table 3. The proportion of white patients who were elderly decreased from 39.5% to 34.2% after the policy change (P<.001); similar findings were noted in black participants (42.2% to 31.2% were elderly; P<.001). Although no significant change occurred in the proportion of female trial participants who were elderly, the proportion of male trial participants 65 years or older decreased from 57.3% to 49.3% (P<.001). Enrollment of elderly patients in lung and breast cancer trials did not change after the policy enactment, whereas significant decreases occurred in the proportion of patients with colorectal and prostate cancer who were elderly.
After stratifying protocols according to eligibility criteria, the participation of elderly patients was noted to be either stable or decreased across each stratum. Finally, when enrollment of each of the 23 trials was analyzed independently, only 2 protocols had significant changes in the enrollment of older patients after policy enactment. One of these protocols was a breast cancer protocol (Breast 3), and the proportion of elderly patients significantly decreased (29.7% to 20.9%; P = .02). Conversely, in one prostate cancer protocol (Prostate 2), the enrollment of older persons significantly increased (68.0% to 85.4%; P = .04).
In the multivariate model, after adjusting for race, cancer type, sex, percentage of county population older than 65 years, percentage below the poverty level, comorbid conditions, and performance status, no significant stepwise increase occurred in the proportion of older persons who enrolled in trials after the Medicare reimbursement change (odds ratio, 1.13; P = .14). There was also no significant rate of change (ie, change in slope) of the trend in enrollment of older patients after the policy change (spline variable; P = .23). Other variables that were related to the likelihood of being older included nonwhite race, cancer type (prostate trials had more elderly participants), and other county and trial characteristics (Table 4).
The Figure shows the proportions of trial participants who were 65 years or older before and after the Medicare policy change in 2000 stratified by phase of trial. Overall, no significant difference occurred in the proportion of phase 3 trial participants who were older than 65 years after 2000 (30.3% vs 29.1% elderly, respectively; P = .84). Among the phase 3 trial participants, no significant change occurred in representation of elderly patients for any of the 4 cancer types. After the Medicare policy was enacted in 2000, a nonsignificant trend occurred toward an increase in enrollment of older persons in phase 1 and 2 trials (40.9% to 49.3%; P = .07). The direction of association varied across cancer types; representation of older persons increased among phase 2 lung cancer participants (P = .04) but decreased among prostate cancer participants (P = .04). No significant change occurred in enrollment of older persons among breast or colorectal cancer phase 2 participants after 2000.
In response to concerns about disproportionately low enrollment of Medicare beneficiaries in clinical trials, the Department of Health and Human Services enacted a highly publicized policy change to cover routine costs for older patients who participate in clinical trials.8,16,28 Our major finding is that this policy remedy has failed to increase the representation of older patients in clinical trials. Our findings were consistent when we analyzed large, ongoing trials and also when we broadened our analytic sample to include all phase 3 and phase 1 and 2 studies. Although well intentioned, this administrative remedy and accompanying investment in covering trial-related costs by the federal government failed to provide significant benefit.
The Medicare policy change may have been ineffective for a number of reasons. Trial exclusion criteria represent a major barrier to enrolling older patients. However, we accounted for this factor in our study design by restricting the main analytic sample to trials in which at least 5% of the study sample was elderly to ensure that the protocols did not preclude elderly patients altogether and thereby render the Medicare policy ineffective on that basis. Moreover, we accounted for variation in the study design of the protocols included in our analysis by accounting for clustering of patients within trials in our multivariate analysis. Conversely, trial eligibility criteria may represent one of the reasons why there was no increase in the enrollment of older patients noted in the full sample of 249 trials. If trials are designed to implicitly or explicitly exclude older persons, the reimbursement policies are unlikely to enhance enrollment.
Even among trials for which older patients may be eligible, inadequate awareness of trials can preclude enrollment.13 In a recent analysis of breast cancer trial enrollment, one of the most common reasons why trial-eligible patients were not enrolled was a lack of awareness that trial participation was an option.3 Similarly, older persons and their physicians may not have been aware that the Medicare reimbursement policy had changed. A recent study29 found that although Medicare and other state legislative mandates are providing payment for routine patient care costs in clinical trials, only 3% of institutions (11 of 373 medical facilities, patient advocate groups, and professional or membership societies) provide sufficient information on their Web sites about this cost coverage and other clinical research information.
Physician attitudes toward enrolling older patients in studies may have mitigated the potential impact of the policy because some oncologists exclude patients from trials based on their age alone.4 Oncologists may have low expectations for older patients being able to continue with treatment for the full duration of the study and believe that they do not need such aggressive therapy at their age owing to declining physical functioning.30 Additionally, the increased prevalence of comorbid conditions with age has resulted in physicians being less likely to enroll trial-eligible older patients.3,4
Finally, it is possible that some clinicians have been submitting bills to Medicare for trial participants even before the policy change, and this policy only made explicit and legal what had been done previously without Medicare’s knowledge. Prior work has suggested that physicians have sometimes misrepresented information to insurance companies to obtain reimbursement for their patients’ treatment.31,32 Additionally, a 1997 General Accounting Office report found that Medicare mistakenly reimbursed up to 50% of submitted claims for patient care costs, despite the fact that such reimbursement was contrary to Medicare policy.9 It is also possible that the legislation may not have been implemented at the local level, and state peer review organizations may have varied in their interpretation of the new rules.
There are several important considerations for our analysis. Trial enrollment data were available through June 2003, and it is possible that it may take more than 3 years to see changes in trial enrollment. Our findings were also limited to NCI-sponsored collaborative group trials of 4 selected cancer types, and the Medicare policy may have had a different effect on other types of trials, such as those involving other cancer types or industry-sponsored studies. However, our findings are consistent with prior studies that suggest that reimbursement policies may not be adequate to change patterns of enrollment or other aspects of medical care. Enactment of a California law that requires insurers to reimburse routine costs of care for clinical trials was not associated with a significant increase in enrollment.33 A more comprehensive analysis found that enactment of trial reimbursement mandates in 4 states was not associated with an increase in phase 3 trial enrollment, although phase 2 enrollment increased modestly.7 Other studies34,35 of reimbursement policy changes regarding mammography and colonoscopy also found that there was no increase in use after the policy change. Our main analysis focused on 23 trials to negate the impact of temporal trends in trial design that could also have affected the enrollment of older persons during our study period. However, our finding that enrollment of older persons did not increase significantly in the full sample of 249 trials strongly suggests that the negative findings of our primary analysis are valid and robust.
In conclusion, the Medicare reimbursement policy expansion seemed to have little effect on participation of older persons in NCI-sponsored cooperative group trials. The paucity of older persons in cancer research is well documented, and several studies3,36,37 have elucidated important barriers confronted by older persons. Unfortunately, it appears that although Medicare reimbursement policy changes may be necessary to facilitate enrollment of some patients, they were not sufficient to enhance overall enrollment. Additional measures need to be taken. Tactics to overcome eligibility barriers may include devising protocols that focus exclusively on older patients or on the treatment of individuals with poor performance status and specific comorbid conditions. In addition to developing new methods to enhance the enrollment of older persons in research, our findings also emphasize the importance of assessing the efficacy of legislative interventions. Enacting well-intentioned policies that have high face validity is insufficient; policies should be evaluated with the same scrutiny and empiricism as the underlying science.
Correspondence: Cary P. Gross, MD, Yale University School of Medicine, Primary Care Center, 333 Cedar St, PO Box 208025, New Haven, CT 06520 (email@example.com).
Accepted for Publication: February 28, 2005.
Financial Disclosure: None.
Funding/Support: Dr Gross’s efforts were supported by a Cancer Prevention, Control, and Population Sciences Career Development Award (1K07CA-90402) and the Claude D. Pepper Older Americans Independence Center at Yale University, New Haven, Conn (P30AG21342).