This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Save to my personal folders Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Gross, C. P. Articles by Krumholz, H. M. Search for Related Content PubMed PubMed Citation Articles by Gross, C. P. Articles by Krumholz, H. M.

Enrolling Older Persons in Cancer Trials: The Effect of Sociodemographic, Protocol, and Recruitment Center Characteristics

From the Sections of General Internal Medicine and Cardiovascular Medicine; Department of Medicine, Robert Wood Johnson Clinical Scholars Program; Department of Epidemiology and Public Health; Yale University School of Medicine, New Haven, CT; and Flying Buttress Associates, Charlottesville, VA

Address reprint requests to Cary P. Gross, MD, Yale University School of Medicine, Primary Care Center, 333 Cedar St, PO Box 208025, New Haven, CT 06520; e-mail: cary.gross{at}yale.edu

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors’ Disclosures of... REFERENCES

 
PURPOSE: To determine the effect of patient, protocol, geographic, and institutional factors on enrollment of older persons onto cancer trials.

METHODS: We conducted a cross-sectional analysis of patients enrolled onto National Cancer Institute–sponsored lung, breast, colorectal, and prostate cancer trials during 1996 to 2002. We used a cross-classified logistic multilevel model to examine the associations between patient, hospital, county, and protocol characteristics, and the likelihood of participants being elderly ( 65 years old).

RESULTS: The final study sample consisted of 36,167 patients enrolled onto 33 trials. After accounting for cancer type, only 6% of the variation in elderly enrollment onto cancer trials was at the protocol level. In contrast, more than 55% of the variation in elderly enrollment was attributable to patient level variation. In multivariate analysis, nonwhite patients were significantly less likely to be elderly than whites (odds ratio [OR] for blacks, 0.51; 95% CI, 0.44 to 0.58; and OR for Hispanics, 0.49; 95% CI, 0.40 to 0.59 v whites). Participants living less than 7 miles from their recruitment center were significantly more likely to be elderly (OR, 1.31; 95% CI, 1.24 to 1.38). Among the 910 recruitment centers, the median adjusted proportion of patients who were elderly was 24.9% (interquartile range, 24.0% to 26.9%). There were a significantly higher number of outlier centers ( 20.8% or 29.3% elderly) than would be expected by a normal distribution (68 observed v six expected; P < .0001).

CONCLUSION: Race and proximity to trial enrollment centers were significantly related to age of trial participants after adjusting for protocol factors. Additional work should explore why some recruitment centers were outliers regarding enrollment of older persons.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors’ Disclosures of... REFERENCES

 
Patients who enroll onto cancer research trials tend to be younger than cancer patients in the general population.^1-4 The disparity in age between trial participants and community cancer patients has been attributed to patient sociodemographic characteristics, attitudes of patients and their physicians, as well as protocol eligibility criteria. Race, sex, and socioeconomic status (SES) are among the sociodemographic characteristics that have been postulated to affect the enrollment of older persons.^1,5 The proportion of cancer trial participants who are minorities has declined in recent years, and within each minority group elderly patients are less likely to enroll than their nonelderly counterparts.^1,6 However, it is unclear whether race is a barrier for enrolling older persons after accounting for other patient and protocol factors. Similarly, although sex and SES were associated with overall cancer trial enrollment in recent work, it is unclear whether these factors have a greater impact on the enrollment of older persons.

Protocol eligibility criteria represent a major barrier to enrolling older persons. Although some studies explicitly exclude elderly patients, others exclude patients with specific comorbid conditions. The typical elderly oncology patient has at least three comorbid conditions, and the prevalence of these conditions increases with age.^4,7 Other exclusion criteria such as disease stage and performance status also reduce enrollment of older persons.⁴ However, disease stage and comorbidity are also correlated with race and socioeconomic status.^8,9 Hence, a comprehensive assessment of the impact of patient factors on enrolling older persons must account for the effect of protocols. That is, although the enrollment of older persons can vary substantially across protocols, it is unclear whether this is attributable to characteristics of the protocols themselves or other factors.

In addition to patient and protocol characteristics, it is important to consider trial enrollment in the context of the specific recruitment sites at which patients are being enrolled. Even within a single multi-institutional trial, there may be substantial variability in the enrollment of older persons across recruitment centers, given that physicians may vary in their attitudes about risks involved with trial participation and the appropriateness of enrolling older persons.^10,11 Sites may also vary in the amount of resources available to help older persons access and navigate their institutions. Although several studies have demonstrated variation across institutions in the care and outcomes of older cancer patients in routine clinical practice, the degree to which enrollment of older persons varies across enrollment sites has not been assessed to our knowledge. It is important to determine whether specific recruitment centers can be identified as enrolling more or fewer elderly patients than their counterparts to identify opportunities to enhance the enrollment of older persons.

To disentangle the patient, protocol, and recruitment center factors that relate to the enrollment of older persons onto trials, it is important to consider that patients are grouped within trials, and each trial is only actively enrolling patients at specific centers. These centers may be located in geographic regions and serve patient populations that have a profound influence on the recruitment of older patients independent of trial factors. Hence, appropriate methodologic techniques are required to account for clustering of patients at the protocol, institution, and county level, and the participation of institutions in multiple trials. Although recent work has suggested that analyses that do not account for this clustering phenomenon can lead to biased results, prior analyses of trial enrollment have failed to do so.^4,12

Although patient race, protocol, and recruitment center factors have been postulated to affect the conduct of clinical research in general, it is unclear which factors are independently associated with the age of trial participants.^13-17 We therefore conducted a cross-sectional study of cancer trial participants, using a cross-classified model that enabled us to extend prior analyses of trial participation by examining the components of variation in elderly enrollment according to hospital, county of residence, and protocol. Our specific objectives were to determine the degree to which patient, protocol, and health system factors contribute to the enrollment of elderly patients onto cancer trials, and to identify specific factors that were related to the age of trial participants. Finally, we assessed variation in the enrollment of older persons across trial enrollment sites.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors’ Disclosures of... REFERENCES

 
Data Source and Construction of Variables
Trial enrollment data for 1996 to 2002 were obtained from the National Cancer Institute (NCI) Clinical Trial Evaluation Program database, which contains patient-level information about participants in all NCI-sponsored cooperative group trials. We studied enrollment onto breast, colon, lung, and prostate cancer trials—the four most common causes of cancer-related mortality during the study period.¹⁸ We only included participants who were 30 years old and had documented racial/ethnic group, age, sex, and zip code of residence.

The construction of our study sample required two steps. For each of the four cancer types, we first selected the 10 cooperative group trials that had enrolled the most patients during 1996 to 2002, and also did not have more than 10% of their patients enrolled at an unknown institution or fewer than 5% of their patients who were 65 years of age. We only included trials with at least 5% elderly to ensure that our study of variability in elderly recruitment included only studies to which at least some elderly patients were enrolled. After obtaining information about the 40 candidate protocols (with represented 45,409 of the 86,624 patients enrolled during the study period), we then excluded the seven protocols that were not phase III or had an explicit age exclusion criteria of 70 years of age or younger. Additional exclusions included male breast/female prostate cancer patients, residents of Alaska and Hawaii because of to the unusual geographic characteristics of these states, and selected missing variables (Fig 1).

View larger version (25K): [in this window] [in a new window]   Fig 1. Construction of study sample.

Protocol characteristics. We obtained additional protocol information from the NCI Physician Data Query (PDQ) clinical trial database, which includes detailed protocol exclusion criteria for NCI trials including age, performance status, and specific comorbid conditions.²¹ We grouped performance status inclusion criteria into three mutually exclusive categories. Trials were designated as eligible for only good performance status if they excluded patients with Eastern Cooperative Oncology Group performance score greater than 1 or a Karnofsky performance status less than 60%.²² Trials were designated as poor performance status protocols if they allowed persons with scores Eastern Cooperative Oncology Group performance score 3 or Karnofsky performance status greater than 20%. The remaining protocols, which specified performance status in between these extremes as the minimum allowed for eligibility, were categorized as medium performance status eligible. Protocols that included patients with stage III or higher were categorized as late stage, whereas the remainder were classified as early stage.

Recruitment center/county characteristics. We obtained additional sociodemographic and geographic data from several sources. Because prior work has suggested that managed care can affect trial enrollment, we obtained county managed care penetration and index of competition estimates for 2000 from the Interstudy County Surveyor Database.^23-25 The index of competition is defined as 1 minus the sum of the squares of each managed-care organization’s market share.^26,27 Values range from 0 to 1; more competitive markets have values closer to 1. The proportion of the population without health insurance for each patient’s state of residence was obtained from the Interstudy County Surveyor Database, and additional county characteristics were obtained from the Area Resource File.^23,28

To identify discrete recruitment centers in the Clinical Trial Evaluation Program database, we used a stepwise approach. Institutions with different names that were located in different zip codes were classified as distinct entities. Institutions located within the same zip code but with nonidentical names were potential duplicates. Research assistants then used several data sources, including the Internet and direct telephone contact when necessary, to determine whether these potential duplicate institutions were distinct entities.

Statistical Analysis
Our primary outcome was elderly status, defined as patient age at enrollment of 65 years of age or older. Cancer type and sex were considered as a single categoric variable for most analyses, to account for the fact that some cancer types were perfectly correlated with sex (breast, prostate), and others were not (lung, colorectal). Continuous variables were categorized into deciles, and further reduced to smaller numbers of categories based on inspection of bivariate patterns. We summarized elderly status by each categorized variable, making comparisons across categories using ² tests of independence adjusted for clustering by protocol.²⁹

We used a cross-classified multilevel logistic model to examine the associations between patient, center, county, and protocol characteristics, and the likelihood of elderly enrollment onto cancer trials.³⁰ Although prior work used analytic techniques that required summarizing patient data by protocol, this approach allowed us to compare patient characteristics across protocols.⁴ This approach also allowed us to account for the simultaneous clustering of patients within hospitals, counties, and protocols, while allowing for multiple centers per protocol and multiple protocols per center.

We initially assessed the components of variance for each classification (protocol, hospital, county) using an empty model that contained only error terms to allow for partitioning the variance in proportion elderly according to protocol, hospital, and county. This preliminary model demonstrated that the total variance in the outcome at the county level (0.4% of all variance) was not significantly different from zero (² = 0.012; P = .40); we therefore fixed this error parameter at zero for subsequent models. A second model included only cancer type, to assess the variation of elderly status across protocols independent of cancer type.¹ We assessed colinearity among candidate variables by examining their correlation matrix, excluding whichever variable we judged least reliable from strongly correlated pairs within the full model (in our final model, we retained population density and excluded urban/rural status and total population using this approach).

Using this final model we reported the characteristics of patients, centers, counties, and protocols that were significantly associated with elderly status. The models were estimated using Monte Carlo Markov Chain simulation: 10,000 iterations were initially used per model, and after inspection of convergence trends this number was judged adequate for our estimates. We also used this model to estimate the adjusted rate of elderly enrollment at each recruitment center as well as the discriminatory performance of the model.³¹ The distribution of adjusted percentage of elderly across centers was characterized by median and interquartile range (IQR). Outliers were defined as centers that met the standard definition for mild outliers ( 1.5 multiples of the IQR below the first quartile or above the third quartile of centers). We used Mlwin 2.0 software (Centre for Multilevel Modeling, London, United Kingdom) for estimating the multilevel models. All other calculations were performed using Stata version 8.0 (Stata Corp, College Station, TX).³²

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors’ Disclosures of... REFERENCES

 
Patient Characteristics
The final study sample consisted of 36,167 patients in 33 trials (Table 1). The most common cancer type was breast cancer (19,165 patients in eight protocols), followed by colorectal (9,084 patients in nine protocols), prostate (4,956 patients in eight protocols), and lung (2,962 patients in eight protocols; P < .001). Although only 16.6% of breast cancer trial participants were elderly, nearly 75% of prostate cancer trial participants were elderly. Men with lung cancer were more likely than women to be elderly (42.1% v 39.3%, respectively). Among participants living within 7 miles of their trial enrollment center, 36.2% were elderly, whereas only 29.3% of patients residing further away from the center were elderly. There was also an inverse relationship between SES and age; approximately 26.3% of participants residing in low poverty zip codes were elderly, compared with 33.0% of participants who resided in high poverty areas (P < .001 in bivariate analysis). After accounting for the clustering of patients according to trial, only sex and cancer type remained significantly associated with the likelihood of a participant being elderly (P < .001). After accounting for clustering of patients within trials, the adjusted proportion of male trial participants who were elderly was 50%, compared with 17% of women. Similarly, the elderly accounted for 11.6% of breast cancer patients and 67.3% of prostate cancer patients (adjusted values).

Recruitment Center/County Characteristics
Neither the number of protocols recruiting at a given center nor designation as NCI Cancer Center or Community Clinical Oncology Program were related to the proportion of participants who were elderly after accounting for clustering of patients according to protocols (P > .05). Similarly, county population characteristics such as managed care competition, cancer incidence, population density, proportion older than age 65 years, and the number of physicians per capita were not significantly related to age of trial participants.

Multivariate Analysis
We first determined the degree to which the age of participants was related to the protocol onto which they were enrolled, the recruitment center at which they were enrolled, or additional patient factors. In a preliminary model that adjusted only for cancer type, we found that 5.7% of the variation in the probability that a particular patient was elderly was explained by the protocol onto which they were enrolled, 2.0% of the variation by was explained by the hospital (recruitment center) level, and 55.2% of the variation by was explained by the individual patient level.

In the full multivariate model, black, Hispanic, and Asian patients were significantly (P < .001) less likely to be elderly than were white patients (OR, 0.51 and 95% CI, 0.44 to 0.58; OR, 0.49 and 95% CI, 0.40 to 0.59; and OR, 0.58 and 95% CI, 0.44 to 0.77, respectively; Table 2). Among colorectal cancer patients, there was no significant association between sex and age, whereas men with lung cancer were significantly more likely to be older than were women with lung cancer (P = .047; comparison not shown in Table 2). Patients residing closer to the enrollment center were significantly more likely to be elderly (OR, 1.31; 95% CI, 1.24 to 1.38), as were patients in areas with increased poverty (OR for intermediate and high poverty v low poverty, 1.16 and 95% CI, 1.04 to 1.30; OR, 1.25 and 95% CI, 1.13 to 1.38, respectively). Participants with colorectal, lung, and prostate cancer were each more likely to be older than were breast cancer patients (P < .05 for all comparisons).

Two county characteristics were associated with age. Patients residing in counties with a higher proportion of the population 65 years of age were more likely to be older (OR, 1.26; 95% CI, 1.14 to 1.39). Conversely, residents of counties with more physicians per capita were less likely to be elderly (OR, 0.88; 95% CI, 0.81 to 0.95).

Variation Across Recruitment Centers
Figure 2A shows the distribution of all 910 recruitment centers according to the proportion of patients who were elderly. The median proportion of trial participants at an individual center who were elderly was 29.2% (IQR, 12.0 to 50.0). Figure 2B shows the distribution of recruitment centers according to elderly enrollment after adjusting for patient, protocol, recruitment center, and county characteristics, as well as for the specific protocols recruiting at each center. The median adjusted proportion of patients who were elderly was 24.9% (IQR, 24.0% to 26.1%). Centers that were outliers in this distribution are noted with dark bars on the figure. There was a significantly higher number of outlier centers than would be expected by a normal distribution (68 in our sample v six that would be expected in a normal distribution, respectively; Shapiro-Wilk test, P < .0001 for normality). We found that 31 centers had an adjusted elderly enrollment rate 20.8% (which was 1.5 x IQR below the first quartile of adjusted rates), and 47 centers had an adjusted elderly enrollment rate 29.3% (1.5 x IQR above the third quartile).

View larger version (23K): [in this window] [in a new window]   Fig 2. (A) Distribution of all 910 recruitment centers according to the proportion of patients who were elderly. (B) Distribution of recruitment centers according to elderly enrollment after adjusting for patient, protocol, recruitment center, and county characteristics, as well as for the specific protocols recruiting at each center.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors’ Disclosures of... REFERENCES

Despite the prominence of patient and protocol factors in the enrollment of older persons, we also found that substantially more recruitment centers than expected had an adjusted percentage of elderly that were considered outliers. Although half of the 910 centers in our sample had an adjusted percentage of elderly between 24% and 26%, there was still substantial variability across centers, even after accounting for the different protocols for which each center was recruiting patients. It is unclear why some centers were outliers. Attitudes of investigators and recruitment personnel may vary across centers, and resources to facilitate the enrollment of older persons may also vary across centers. Future work should identify investigator and recruitment center characteristics that are associated with the enrollment of older persons.

The geographic location of trial participants was important in two ways. Both travel distance from the trial recruitment center and the underlying age distribution of the population were related to recruitment of older persons. Our finding that participants who lived within 7 miles of a recruitment center were significantly more likely to be elderly is consistent with other work suggesting that travel time is a frequently cited reason for declining to enroll.³³ Enrolling onto a study sometimes entails undergoing treatment and clinical assessment in facilities that are further from patients’ homes than they would otherwise use. Older persons may have more barriers to travel because of impairments in mobility or cognition. Prior work has suggested that patients who travel farther to enroll onto a study may be more likely to survive.³⁴ Considered with our findings, it is possible that not only are older persons less likely to travel longer distances to enroll, but those who do may be a particularly healthy subset of the elderly. Hence, facilitating travel by addressing logistic barriers for older persons may improve not only the number of older persons but also the generalizability of the results by enabling patients with a broad spectrum of health status to enroll.

We also found marked variation in enrollment of older persons across cancers, even after accounting for the underlying age distribution of patients with each type of cancer. That is, even after accounting for the fact that prostate cancer patients are more likely to be elderly than are breast cancer patients, we still found that the adjusted odds of prostate cancer participants being elderly was more than five times the odds for breast cancer patients. This suggests that the approach to designing trials and recruiting elderly patients varies substantially across cancer types. Future work should explore whether approaches that have been successful in some disease entities (ie, prostate cancer) may also help to promote the design and conduct of trials that are applicable to older persons with other cancer types (ie, breast).

As expected, trials with specific exclusion criteria according to age, performance status, and advanced-stage disease were less likely to enroll older persons. This finding suggests strongly that efforts to enhance the enrollment of older persons should focus on designing trials that are accessible by—and applicable to—elderly persons. However, suggesting relaxation of exclusion criteria as a remedy would be an oversimplification. In many cases, strict exclusion criteria may reflect concerns about potential toxicity and risks involved with the therapy under investigation. That is, strict exclusion criteria may be a proxy for the degree of perceived risk involved with the new therapy, and the desire of investigators to test the proof of concept of new therapies in the subset of patients who are most likely to benefit. To enhance the body of evidence that is applicable to elderly patents, it will be necessary to consider the full portfolio of publicly funded studies and ensure that some trials focus specifically on clinical questions that are germane to older cancer patients. For instance, an ongoing trial, Comparison of Combination Chemotherapy Regimens in Treating Older Women Who Have Undergone Surgery for Breast Cancer, is limited to patients 65 years of age. Rather than simply excluding patients with renal insufficiency, the protocol calls for dose medications on the basis of creatinine clearance.³⁵ The results of this ongoing study will help to address the knowledge gap that resulted from prior exclusion of elderly women from trials of adjuvant therapy for breast cancer.^35,36 Promoting more studies such as this will have a great impact on ensuring the applicability of trials to the general population.

Older patients and their clinicians may also believe that increased age presents increased risk involved with trial enrollment in comparison with younger patients. Thus, patients and their clinicians may be less willing to participate, regardless of structural barriers. Hence, distance traveled may reflect more than a simple logistic hurdle—it may reflect the fact that older people may be less willing to expend the additional time and energy required to travel and participate in trials, even if they were able to travel. It is also important to note that there must be a balance between removing barriers and undue enticements: clinicians, researchers, and policy makers should promote and facilitate enrollment of elderly patients, but should not be coercive.

Our statistical approach was particularly robust in that we were able to analyze our data at the individual patient level, yet still account for clustering of patients within specific trials and recruitment centers.⁴ Because our analysis was limited to NCI-sponsored cooperative group trials, it is unclear whether our findings can be extrapolated to other cancer types or to industry-sponsored studies. We also limited our sample to trials in which at least 5% of patients were elderly; hence, our findings may not apply to trials that enrolled fewer older patients.

In summary, we found that among participants in NCI-sponsored cooperative group trials, patient race, distance from enrollment center, cancer type, and protocol exclusion criteria were significantly associated with the enrollment of older persons. We also noted that more centers than expected had an adjusted elderly enrollment rate that was considered an outlier. Our results suggest that several interventions may enhance the enrollment of older persons: NCI cooperative groups should expand into areas with more elderly and more minorities, and provide support for patients in need of travel assistance as well as for clinician-investigators operating in areas with highly competitive health markets. Although it is clear that the research question and eligibility criteria are important factors, future efforts should also identify and evaluate specific strategies that could be applied at the patient or institutional level to facilitate the participation of older persons in cancer research.

	Authors’ Disclosures of Potential Conflicts of Interest

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors’ Disclosures of... REFERENCES

 
The authors indicated no potential conflicts of interest.

	NOTES

 
Supported by a Cancer Prevention, Control and Population Sciences Career Development Award (1K07CA-90402), a Beeson Career Development Award (1 K08 AG24842), and the Claude D. Pepper Older Americans Independence Center at Yale (P30AG21342).

Authors’ disclosures of potential conflicts of interest are found at the end of this article.

	REFERENCES

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors’ Disclosures of... REFERENCES

 

1. Murthy VH, Krumholz HM, Gross CP: Participation in cancer clinical trials: Race-, sex-, and age-based disparities. JAMA 291:2720-2726, 2004[Abstract/Free Full Text]
2. Trimble E, Carter C, Cain D, et al: Representation of older patients in cancer treatment trials. Cancer 74:2208-2214, 1994[CrossRef][Medline]
3. Hutchins L, Unger J, Crowley J, et al: Underrepresentation of patients 65 years of age or older in cancer-treatment trials. N Engl J Med 341:2061-2067, 1999[Abstract/Free Full Text]
4. Lewis JH, Kilgore ML, Goldman DP, et al: Participation of patients 65 years of age or older in cancer clinical trials. J Clin Oncol 21:1383-1389, 2003[Abstract/Free Full Text]
5. Gross CP, Filardo G, Mayne ST, et al: The impact of socioeconomic status and race on trial participation for older women with breast cancer. Cancer 103:483-491, 2005[CrossRef][Medline]
6. Simon MS, Du W, Flaherty L, et al: Factors associated with breast cancer clinical trials participation and enrollment at a large academic medical center. J Clin Oncol 22:2046-2052, 2004[Abstract/Free Full Text]
7. Extermann M: Measurement and impact of comorbidity in older cancer patients. Crit Rev Oncol Hematol 35:181-200, 2000[Medline]
8. Mandelblatt J, Andrews H, Kerner J, et al: Determinants of late stage diagnosis of breast and cervical cancer: The impact of age, race, social class, and hospital type. Am J Public Health 81:646-649, 1991[Abstract/Free Full Text]
9. Schwartz KL, Crossley-May H, Vigneau FD, et al: Race, socioeconomic status and stage at diagnosis for five common malignancies. Cancer Causes Control 14:761-766, 2003[CrossRef][Medline]
10. Daugherty C, Ratain MJ, Grochowski E, et al: Perceptions of cancer patients and their physicians involved in phase I trials. J Clin Oncol 13:1062-1072, 1995[Abstract]
11. Joffe S, Weeks JC: Views of American oncologists about the purposes of clinical trials. J Natl Cancer Inst 94:1847-1853, 2002[Abstract/Free Full Text]
12. Panageas KS, Schrag D, Riedel E, et al: The effect of clustering of outcomes on the association of procedure volume and surgical outcomes. Ann Intern Med 139:658-665, 2003[Abstract/Free Full Text]
13. Ross S, Grant A, Counsell C, et al: Barriers to participation in randomised controlled trials: A systematic review. J Clin Epidemiol 52:1143-1156, 1999[CrossRef][Medline]
14. Schwartz CE, Fox BH: Who says yes? Identifying selection biases in a psychosocial intervention study of multiple sclerosis. Soc Sci Med 40:359-370, 1995
15. Yeomans-Kinney A, Vernon SW, Frankowski RF, et al: Factors related to enrollment in the breast cancer prevention trial at a comprehensive cancer center during the first year of recruitment. Cancer 76:46-56, 1995[CrossRef][Medline]
16. 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 20:2109-2117, 2002[Abstract/Free Full Text]
17. Vogt TM, Ireland CC, Black D, et al: Recruitment of elderly volunteers for a multicenter clinical trial: The SHEP pilot study. Control Clin Trials 7:118-133, 1986[CrossRef][Medline]
18. Fichtenberg C, Glantz S: Association of the California tobacco control program with declines in cigarette consumption and mortality from heart disease. N Engl J Med 343:1772-1777, 2000[Abstract/Free Full Text]
19. US Census Bureau: 2000 Census. http://www.census.gov/support/2000/SF3/
20. Ingram D, Parker J, Schenker N, et al: United States Census 2000 population with bridged race categories. Washington, DC, National Center for Health Statistics, 2003
21. National Cancer Institute: Search for clinical trials: Advanced, 2004. http://www.nci.nih.gov/search/clinical_trials/search_clinicaltrialsadvanced.aspx/
22. Schag CC, Heinrich RL, Ganz PA: Karnofsky performance status revisited: Reliability, validity, and guidelines. J Clin Oncol 2:187-193, 1984[Abstract]
23. IMS Publications: County Surveyor Database 3.0. St Paul, MN, 2003
24. Campbell E, Weissman J, Blumenthal D: Relationship between market competition and activities and attitudes of medical school faculty. JAMA 278:222-226, 1997[Abstract]
25. Moy E, Mazzaschi A, Levin R, et al: Relationship between National Institutes of Health research awards to US medical schools and managed care market penetration. JAMA 278:217-221, 1997[Abstract]
26. Wholey D, Christianson J, Engber J, et al: HMO market structure and performance. Health Aff (Millwood) 16:75-84, 1997[Abstract]
27. Gaskin DJ, Hadley J: The impact of HMO penetration on the rate of hospital cost inflation. Inquiry 34:205-216, 1997[Medline]
28. Quality Resource Systems I: Area Resource File, 2005. http://www.arfsys.com/
29. Donald A, Donner A: Adjustments to the Mantel-Haenszel chi-square statistic and odds ratio variance estimator when the data are clustered. Stat Med 6:491-499, 1987[Medline]
30. Browne WJ, Goldstein H, Rasbash J: Multiple membership multiple classification (MMMC) models. Stat Modelling 103-124, 2001
31. Hanley J, McNeil BJ: The measure and use of the area under the receiver operating characteristic (ROC) curve. Radiology 143:29-36, 1982[Abstract]
32. Stata SS: Release 8.0 Reference Manual. College Station, TX, Stata Press, 2004
33. Chang BH, Hendricks AM, Slawsky MT, et al: Patient recruitment to a randomized clinical trial of behavioral therapy for chronic heart failure. BMC Med Res Methodol 4:8, 2004[CrossRef][Medline]
34. Lamont EB, Hayreh D, Pickett KE, et al: Is patient travel distance associated with survival on phase II clinical trials in oncology? J Natl Cancer Inst 95:1370-1375, 2003[Abstract/Free Full Text]
35. National Cancer Institute: Clinical Trials PDQ, 2004. http://www.cancer.gov/search/ViewClinicalTrials.aspx?cdrid=68891&version=HealthProfessional&protocolsearchid=1121902
36. Early Breast Cancer Trialists’ Collaborative Group: Polychemotherapy for early breast cancer: An overview of the randomised trials. Lancet 352:930-942, 1998[CrossRef][Medline]

This article has been cited by other articles:

				J. H. Stewart IV, A. G. Bertoni, J. L. Staten, E. A. Levine, and C. P. Gross Participation in Surgical Oncology Clinical Trials: Gender-, Race/Ethnicity-, and Age-based Disparities Ann. Surg. Oncol., December 1, 2007; 14(12): 3328 - 3334. [Abstract] [Full Text] [PDF]

				F. Keime-Guibert, O. Chinot, L. Taillandier, S. Cartalat-Carel, M. Frenay, G. Kantor, J.-S. Guillamo, E. Jadaud, P. Colin, P.-Y. Bondiau, et al. Radiotherapy for Glioblastoma in the Elderly N. Engl. J. Med., April 12, 2007; 356(15): 1527 - 1535. [Abstract] [Full Text] [PDF]

				K. R. Yabroff, W. W. Davis, E. B. Lamont, A. Fahey, M. Topor, M. L. Brown, and J. L. Warren Patient Time Costs Associated With Cancer Care J Natl Cancer Inst, January 3, 2007; 99(1): 14 - 23. [Abstract] [Full Text] [PDF]

				J O'Loughlin, E Dugas, K Maximova, and N Kishchuk Reporting of ethnicity in research on chronic disease: update Postgrad. Med. J., November 1, 2006; 82(973): 737 - 742. [Abstract] [Full Text] [PDF]

				D. Crivellari, D. Lombardi, G. Corona, C. Massacesi, R. Talamini, R. Sorio, M. D. Magri, C. Lestuzzi, A. Lucenti, A. Veronesi, et al. Innovative schedule of oral idarubicin in elderly patients with metastatic breast cancer: comprehensive results of a phase II multi-institutional study with pharmacokinetic drug monitoring Ann. Onc., May 1, 2006; 17(5): 807 - 812. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Save to my personal folders Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Gross, C. P. Articles by Krumholz, H. M. Search for Related Content PubMed PubMed Citation Articles by Gross, C. P. Articles by Krumholz, H. M.