Impact of Managed Care on Cancer Trial Enrollment
http://www.100md.com
《临床肿瘤学》
the Sections of General Internal Medicine
Cardiovascular Medicine, Department of Medicine, Yale-New Haven Hospital Center for Outcomes Research and Evaluation, Section of Health Policy and Administration, Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, CT
ABSTRACT
PURPOSE: To determine the relationship between managed care market activity and cancer trial enrollment.
METHODS: Trial participant data were obtained from the National Cancer Institute. Participants in cooperative group trials of breast, colorectal, lung, or prostate cancer during the years 1996 through 2001 were assigned to counties based on their zip code of residence. Linear regression was used to determine the relationship between county enrollment rate and two measures of county managed care activity (penetration and index of competition [IOC]), adjusting for other county characteristics.
RESULTS: In bivariate analysis, there was a strong inverse correlation between trial enrollment rate and IOC (r = –0.23; P < .001) as well as the proportion of the population uninsured (–0.31; P < .001) and the percentage below poverty (–0.16; P < .001). In the multivariate model, greater county managed care competition (IOC) was inversely related to trial enrollment rate (P < .008 for comparison of each quartile v lowest quartile) after accounting for managed care penetration, proportion uninsured, and other county characteristics. Counties in the lowest quartile of managed care penetration tended to have lower enrollment rates than the remaining counties (r = –0.05; P = .048), while counties in the second, third, and fourth quartiles of penetration all had similar enrollment rates to one another.
CONCLUSION: Cancer trial enrollment rates were suboptimal across all counties, and counties with higher levels of managed care competition had significantly lower enrollment rates. The relationship between managed care penetration and trial enrollment was less consistent. Future efforts to enhance trial participation should address the potential negative influence of market factors.
INTRODUCTION
Recruitment of patients into clinical studies is exceedingly difficult.1,2 A recent Institute of Medicine roundtable has identified insufficient public participation in clinical research as one of two major obstacles that must be overcome in order to "...apply science to better human health in an expeditious fashion."2 A considerable amount of attention has focused on identifying barriers to trial participation. These barriers may be related to patients, physicians, study characteristics, or the health system. Patient- and physician-specific barriers such as sociodemographic characteristics, attitudes, and beliefs have been well documented.3-8
To enhance trial recruitment and maintain the viability of clinical cancer research enterprise, it is also important to further our understanding of the impact of health system factors on trial enrollment. Unfortunately, the clinical environment may be increasingly less likely to support research activities. The increased competition and strained resources associated with managed care could affect trial enrollment at either the patient or physician level. Patients with managed care insurance may be wary of enrolling because of concerns about reimbursement, as private insurers may not cover routine care costs for patients enrolled on trials.9-13
Managed care can also affect physician behavior, as enrolling patients onto trials requires a significant investment of time and money. For each cancer patient enrolled on a trial, clinician-investigators and their staff devote roughly between 150 and 450 hours of time, and expend (nonclinical) costs ranging from approximately $1,300 to $3,900.14 These demands on the resources of prospective clinicians/investigators are occurring during a time of scarce resources, as the cost and reimbursement structure surrounding the care of cancer patients has changed significantly with the growth of managed care organization (MCO) insurance. The decrease in clinical revenues that have been associated with MCO coverage has reportedly led clinicians to be reluctant to devote the additional resources necessary for enrolling patients onto trials.15-17 Anecdotal data suggest that managed care has forced academic researchers to devote more time to clinical activities and less to research.18-22 Other investigators have reported that high regional MCO penetration impedes research productivity of academic medical centers in terms of grants and publications.21,22 To our knowledge, the effect of managed care on trial enrollment in community-based practices has not been explored.
Despite the widespread recognition of inadequate trial participation, little is known about the impact of managed care on trial participation at the population level. We therefore assessed the relationship between participation in cooperative group trials sponsored by the National Cancer Institute (NCI) and two measures of MCO activity—MCO market penetration and index of competition (IOC). The rationale for analyzing IOC arises from concern that the level of competition between MCOs may have a greater impact on a market than penetration.23,24 That is, a market that contains only a single MCO that has a penetration of 80% might be quite different from a market with 80% penetration shared by numerous, competing entities. The IOC is used to quantify this competition, and is defined as 1 minus the sum of the squares of each MCO’s market share.23,24 The IOC applies only to the portion of the market that is enrolled in managed care entities and assumes values ranging from 0 to 1, with more competitive markets having values closer to 1. We hypothesized that clinicians and institutions operating in areas with higher MCO penetration or IOC would face increased pressures to maintain clinical revenues and have fewer resources available to conduct research.
METHODS
Cancer trial enrollment data for 1996 to 2001 were obtained from the NCI Cancer Trials Evaluation Program. We only included participants in NCI-sponsored cooperative group breast, colon, lung, and prostate cancer trials of therapeutic agents who were older than 30 years and had documented racial/ethnic group, age, and sex. Trial participants were assigned to counties according to zip code of their residence; the 10,900 patients (of 86,600) who were missing zip codes were excluded. We also excluded counties that were located in Hawaii or Alaska due to the unique geographic characteristics of these locations. We also excluded counties that had less than one patient enrolled per year during our study period, because we hypothesized a priori that patients residing in those counties had virtually no opportunity to enroll on a trial, regardless of the level of managed care activity, and we wanted to focus on areas in which the opportunity to enroll could have been affected by managed care.
County managed care penetration and IOC estimates for the year 2000 were obtained from the Interstudy County Surveyor Database.25 Additional county characteristics were obtained from the US Census and the Area Resource file.25,26 Geographical Information Systems (GIS) data were used to estimate the distance in (linear) miles between the center of each county and the nearest county that had an NCI research center (defined as either a designated cancer center or a community clinical oncology program [CCOP]). We estimated the proportion of the US population residing in the counties in our study sample by dividing the total populations of the included (sample) counties by the total US population.
The most recently available estimates of cancer incidence according to sex, race/ethnicity, and 5-year age group (1995 through 1999) were obtained from the NCI’s Surveillance, Epidemiology, and End Result (SEER) data.27 Because SEER counties only represent a subset of the US population, cancer incidence rates in each of these groups were then applied to the corresponding population categories and then summed over all categories to obtain cancer incidence estimates for each county.
The trial enrollment rate for each county was defined as the annual number of trial participants divided by the annual number of incident cancer patients. Using the county as the unit of analysis, we performed bivariate analyses comparing county trial enrollment rates with predictor variables of interest using t test, 2, and Spearman’s when appropriate. Continuous variables were converted into categorical variables after reviewing bivariate analyses of each managed care variable in quantiles and collapsing them each into four groups based on distribution of the data. This conversion facilitated interpretation of the resulting model, and was found to have little impact on the model’s performance.
We then performed a multiple linear regression using the natural logarithm of the enrollment rate as the dependent variable, and the two measures of MCO activity (penetration and IOC) as independent variables. Because enrollment rates were highly skewed, we log transformed county rates to approach a normal distribution. Candidate covariates were selected with forward stepwise selection and included the proportion of population without health insurance, proportion below poverty level income, urban/rural influence code, unemployment rate, presence of a teaching hospital, number of hospitals with oncology services, physicians per capita, and distance to the nearest research center. Variables with P < .05 were retained in the final model. In several instances, there was substantial colinearity among the candidate predictor variables. In these instances, we selected the variable that had a stronger correlation with enrollment rate on bivariate analysis for inclusion in the model. To present the results in terms of the relationship between IOC and absolute trial enrollment rate, we estimated the adjusted enrollment rate for counties in each of the IOC groups after holding all other county characteristics at a constant level. Analyses were performed using Stata 6.0 (Stata Corp, College Station, TX).28
RESULTS
County Characteristics
Of the 3,134 counties in our data set, 1,418 had six or more patients enrolled in CGCT during 1996 to 2001 and were included in the analytic sample (Table 1) . The total population of the included counties represented about 87.5% of the US population. As expected, the counties in our sample were significantly more likely to be urban and to have a higher managed care market penetration (median was 14.9% in our sample v 4.9% in the excluded counties; Table 1; P < .001) and IOC (0.588 v 0.287; P < .001). In general, excluded counties were more likely to have lower income levels and fewer teaching hospitals and physicians per capita. Finally, sample counties had far more developed health care and research infrastructure, being more likely to have hospitals with teaching programs and oncology services, to have more physicians per capita, and to be closer to NCI-designated cancer centers or CCOP centers (Table 1).
The median trial enrollment rate for the 1,418 counties included in our sample was 17.1 participants per 1,000 cancer patients (interquartile range [IQR], 10.6 to 26.7). There was a strong inverse correlation relation between index of competition and trial enrollment rate (Table 2; correlation coefficient, –0.23; P < .001). In the 517 counties with the least amount of competition among managed care plans (IOC < 0.5), the median enrollment rate was 20.4 (IQR, 12.3 to 31.6; Table 2). The median enrollment rate decreased to 17.3 in counties with IOC 0.5 to 0.65, and was 15.3 and 12.2 among counties with IOC 0.65 to 0.78, and greater than 0.78, respectively (Table 2). The mean enrollment rates (and 95% CIs) according to IOC are shown in Figure 1.
There was a relationship between managed care penetration and enrollment rates, though this was of borderline significance (Table 2; P = .048). The proportion of people without health insurance was also inversely related to enrollment rate (Table 2; –0.31; P < .001). Counties in the lowest quartile of proportion uninsured (< 11%) had a median enrollment rate of 20.6 (IQR, 13.7 to 29.9); the enrollment rate decreased to 11.4 (IQR, 6.9 to 20.0) in counties with the highest quartile uninsured (> 16.4%).
County socioeconomic status was also related to trial enrollment. Counties with a higher proportion of their population living below poverty level had substantially lower median enrollment rates (correlation, –0.16; P < .0001), decreasing from 19.1 in counties with less than 9.1% below poverty to 14.0 in counties with the highest poverty level.
Geographic location was also related to trial enrollment, as counties with NCI research centers had a median rate of 21.7 (IQR, 13.8 to 32.7), while counties more than 36.5 miles from the nearest center had a median rate of 15.5 (IQR, 9.1 to 24.8; Spearman’s correlation = –0.16; P < .0001). Counties with the highest enrollment fraction were less likely to be have two or more hospitals with oncology services (P < .001), and tended to be nonurban (P < .001). The presence of hospitals with teaching programs was unrelated to trial enrollment (P = .34).
In our final multivariable model, both IOC and managed care penetration were significantly related to county enrollment rate after adjusting for geographic location and population sociodemographic characteristics (Table 2). There was a significant decrease in trial enrollment rate with increasing IOC (Table 3). As the county IOC increased, the adjusted trial enrollment decreased substantially, from 27.8 per 1,000 (95% CI, 22.9 to 33.7) in the lowest IOC group to 14.6 (95% CI, 11.9 to 18.0) in the highest IOC group (Fig 1). Counties in the lowest group of MCO penetration (group 1: < 3.23%) tended to have significantly lower enrollment rates than counties in the remaining groups (P = .005, .009, and .009 for comparison of group 1 v groups 2, 3, and 4, respectively). There was no significant relation between enrollment and MCO penetration for the upper three groups of MCO penetration (ie, group 2 v group 3 v group 4 pairwise comparisons all > .05).
Counties which were urban, at increasing distance from a research center, and with higher proportion of the population uninsured or below poverty also tended to have lower enrollment rates in the multivariate analysis.
DISCUSSION
We found that counties with more competitive managed care markets had significantly lower cancer trial enrollment rates after accounting for managed care penetration, poverty status, and other county-level characteristics. The impact of managed care on trial recruitment may be substantial, as the adjusted enrollment decreased from over 25 per 1,000 patients in the counties with the lowest IOC to about 15 per 1,000 in the counties with the highest IOC. Given that there are more than 600,000 newly diagnosed cases of lung, breast, colorectal, and prostate cancer annually, these relative changes in enrollment that are related to managed care translate into large differences in trial enrollment at the national level.29
The relationship between trial enrollment rate and managed care penetration was less consistent than IOC. We found that enrollment rates in the upper three groups of counties according to managed care penetration did not differ significantly from one another. Although prior efforts have suggested that managed care penetration may hinder research productivity, these studies were performed substantially earlier than our study, when managed care was growing more dynamically.21,22 Another, more recent study found that statewide MCO penetration was not associated with cancer trial enrollment.30 However, there is substantial variation in MCO penetration within states. Hence, our analysis, which accounts for MCO penetration at the county level, provides more rigorous support to the hypothesis that MCO penetration is not a strong barrier to enrollment.
Our findings emphasize the importance of distinguishing between competition between MCO plans and overall penetration. While some recent work may suggest that managed care may not hinder enrollment, our findings suggest otherwise. There are a number of potential mechanisms to explain this phenomenon. Among counties with similar levels of managed care penetration, counties with higher levels of competition among managed care plans may be more likely to have more pressure to control costs. Decreased revenues associated with managed care have affected the research productivity of both individuals and institutions.31 Competitive markets are associated with shortfall of discretionary funds, less institutional support for research, and decreased amount of time spent on research activities.20,31 Prior physician surveys have also found that MCO insurance makes it more difficult for clinicians to provide routine care for their cancer patients or to enroll them on clinical trials.15,32 Researchers in more competitive markets are also more likely to express concerns about access to potential enrollees due to limitations on referrals out of network.33 MCO coverage may also directly affect patients’ ability to participate in trials, as many private insurers will not routinely cover patient care costs for "investigational treatments."11,16 A prior study found that among privately insured patients with HIV, those with MCO insurance were much less likely than those with fee for service to participate in studies.8
Competition among providers may also serve as a barrier to trial enrollment. In our bivariate analysis, counties with proportionately more physicians and/or hospitals with oncology services actually enrolled fewer patients. Although neither of these relations remained significant in the multivariate model, future work should explore whether areas with a relative abundance of oncology services are optimizing opportunities for trial enrollment. Geographic location was also an important factor; counties 36 miles from the nearest NCI research center had significantly lower enrollment rates than counties with such centers. We also found that the presence of a teaching hospital was not associated with increased trial enrollment rates, suggesting that NCI-funded research centers have a more broad affect on trial enrollment than do teaching hospitals. This reinforces the importance of the CCOP network and plans for expansion.34
Some studies have reported that lack of insurance can decrease access to experimental therapies, while others have suggested that uninsured patients are more likely to participate in trials.8,35 The participation of patients without health insurance in trials has raised concerns about undue inducement of participants, who may have few other care options outside of the study setting.36 However, we found that the proportion of the population without health insurance was inversely related to county trial enrollment. This may be attributable to decreased access to care; uninsured patients may be more likely to present with later stage disease and be ineligible for trials on that basis.37-39 Additionally, higher levels of uninsured at the county level may further strain clinical resources at associated institutions, with a grater demand for uncompensated care.
It is important to note that counties with fewer than one trial participant annually were excluded from our analysis, as we had hypothesized a priori that it was unlikely that health system factors would affect enrollment in counties that offered little or no access to research. Although the 1,400 counties in our analytic sample comprised approximately 88% of the US population, the results of our analysis may not be generalizable to the remaining counties. One of the challenges to analyzing trial enrollment rates by county is the lack of available cancer incidence data by county at the national level. To test the validity of our demographic projection approach, we compared estimates of cancer incidence recorded in the SEER registry with those we estimated for the SEER counties; we found the Spearman’s correlation coefficient was 0.97. Additionally, our sample was also restricted to NCI-sponsored cooperative group trials of therapeutic agents; it is unclear whether our results are generalizable to federal or industry-sponsored trials or to prevention studies. Because patients were assigned to counties based on their home address, some patients may have traveled to different counties to enroll in research. We addressed this potential limitation by including distance to the nearest cancer center as a covariate. Finally, it is important to note that the relationship between managed care and trial enrollment was cross-sectional; future work will have to explore whether managed care competition directly results in lower trial enrollment rates.
Trial enrollment rates are currently not optimal for the expedient evaluation of new therapeutic agents. Additionally, patients who could potentially benefit from trial participation, either from access to new experimental treatments or by the important sense of altruism associated with enrolling on studies, face health system issues as one of the important barriers to participation. The NCI has identified expanding the pool of trial participants as a priority and has invested substantially in improving infrastructure and information systems to alleviate "...costs at local clinical trial sites, and for operations, data management, and statistical offices."34 Our findings emphasize the importance of addressing the impact of managed care and lack of health insurance on strained clinical resources, in order to ensure that clinicians have sufficient time to devote to clinical research and that patients are offered the chance to participate. Additionally, it is unclear whether managed care reimbursement policies also pose a barrier, as patients with managed care may be wary of seeking approval and reimbursement. Recent work suggests that clinicians do not receive adequate reimbursement for the costs associated with trial participation.17,40 Adequate cancer trial enrollment depends on more than the knowledge, attitudes, and beliefs of individual patients and physicians. As clinical revenues wane and the health marketplace becomes more competitive, patients, physicians, and investigators may need additional resources to ensure access to trials and timely scientific progress in a challenging clinical environment.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
NOTES
Supported by (C.P.G.) a Cancer Prevention, Control and Population Sciences Career Development Award (1K07CA-90402) and the Claude D. Pepper Older Americans Independence Center at Yale (P30AG21342).
This original work has been neither presented nor published previously.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Office of the Director: The Nation's Investment in Cancer Res: A Budget Proposal for Fiscal Year 2001. Bethesda, MD, National Cancer Institute, 1999
Sung NS, Crowley WF Jr, Genel M, et al: Central challenges facing the national clinical research enterprise. JAMA 289:1278-1287, 2003
Shavers-Hornaday V, Lynch C, Burmeister L, et al: Why are African-Americans under-represented in medical research studies Impediments to participation. Ethn Health 2:31-45, 1997
Foley J, Moertel C: Improving accrual into cancer clinical trials. J Cancer Educ 6:165-173, 1991
Fleming I: Clinical trials for cancer patients: The community practicing physician's perspective. Cancer 65:2388-2390, 1990
Fallowfield L, Ratcliffe D, Souhami R: Clinicians' attitudes to clinical trials of cancer therapy. Eur J Cancer 33:2221-2229, 1997
Swanson D, Ward A: Recruiting minorities into clinical trials: Toward a participant-friendly system. J Natl Cancer Inst 87:1747-1759, 1995
Gifford AL, Cunningham WE, Heslin KC, et al: Participation in research and access to experimental treatments by HIV-infected patients. N Engl J Med 346:1373-1382, 2002
Comis R, Aldige C, Stovall E, et al: A Quantitative Survey of Public Attitudes Towards Cancer Clinical Trials. http://www.cancertrialshelp.org/cnccg_info/news.html, 2000
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
Goodman C: Investigational exclusion, clinical trials, and cancer. NCCN Proceedings 12:37-49, 1998
Vogelzang N, Richards J: Third party reimbursement issues during a phase I trial of interleukin-2. J Nat Canc Inst 81:544-545, 1989
NIH Clinical Trials: Various Factors Affect Patient Participation. Washington, DC, US General Accounting Office, 1999
Emanuel EJ, Schnipper LE, Kamin DY, et al: The costs of conducting clinical research. J Clin Oncol 21:4145-4150, 2003
Wade J, Weinberg P, Bennett C: The effects of managed care on community clinical research. Primary Care Cancer 19:18-25, 1999
Fleming R: Barriers to clinical trials, part I: Reimbursement problems. Cancer 74:2662-2665, 1994
Chaturvedi S: Clinical trials and financial reimbursement. Stroke 29:1256, 1998
Ta KT, Persing JA, Chauncey H Jr, et al: Effects of managed care on teaching, research, and clinical practice in academic plastic surgery. Ann Plast Surg 48:348-354, 2002
Levitt S: Impact of managed care on scholarly activity and patient care: Case study of 12 academic radiology and radiation oncology departments. Radiology 216:618-623, 2000
Campbell EG, Weissman JS, Causino N, et al: Market competition and patient-oriented research: The results of a national survey of medical school faculty. Acad Med 76:1119-1126, 2001
Campbell E, Weissman J, Blumenthal D: Relationship between market competition and activities and attitudes of medical school faculty. JAMA 278:222-226, 1997
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
Wholey D, Christianson J, Engber J, et al: HMO market structure and performance. Health Aff (Millwood) 16:75-84, 1997
Gaskin DJ, Hadley J: The impact of HMO penetration on the rateo of hospital cost inflation. Inquiry 34:205-216, 1997
Publications I: County Surveyor Database 3.0. St Paul, MN, 2003
Quality Resource Systems I: http://www.arfsys.com/
Warren J, Klabunde C, Schrag D, et al: Overview of SEER-Medicare data: Content, research applictions, and generalizability to the United States elderly population. Med Care 40:IV-3-IV-18, 2002 (suppl)
Stata SS: Release: 8.0 Reference Manual. College Station, TX, Stata Press, 2004
Society AC: Cancer facts and figures. 2000
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
Burnett D: Evolving market will change clinical research. Health Aff (Millwood) 15:89-92, 1996
Mortonson L, Leader S, Mallick R: The impact of managed care on oncology practice. Oncology Issues 12:22-27, 1997
Mechanic RE, Dobson A: The impact of managed care on clinical research: A preliminary investigation. Health Aff (Millwood) 15:72-89, 1996
anonymous: The Nation's Investment in research: A plan and budget proposal for fiscal year 2004. Bethesda, MD, National Cancer Institute, 2003
Gorkin L, Schron EB, Handshaw K, et al: Clinical trial enrollers vs. nonenrollers: The Cardiac Arrhythmia Suppression Trial (CAST) Recruitment and Enrollment Assessment in Clinical Trials (REACT) project. Control Clin Trials 17:46-59, 1996
Pace C, Miller FG, Danis M: Enrolling the uninsured in clinical trials: An ethical perspective. Crit Care Med 31:S121-S125, 2003
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
Ferrante JM, Gonzalez EC, Roetzheim RG, et al: Clinical and demographic predictors of late-stage cervical cancer. Arch Fam Med 9:439-445, 2000
Roetzheim RG, Pal N, Tennant C, et al: Effects of health insurance and race on early detection of cancer. J Natl Cancer Inst 91:1409-1415, 1999
Wright J, Levine M: Researching the cost of research. J Clin Oncol 21:4081-4082, 2003(C.P. Gross, H.M. Krumholz)
Cardiovascular Medicine, Department of Medicine, Yale-New Haven Hospital Center for Outcomes Research and Evaluation, Section of Health Policy and Administration, Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, CT
ABSTRACT
PURPOSE: To determine the relationship between managed care market activity and cancer trial enrollment.
METHODS: Trial participant data were obtained from the National Cancer Institute. Participants in cooperative group trials of breast, colorectal, lung, or prostate cancer during the years 1996 through 2001 were assigned to counties based on their zip code of residence. Linear regression was used to determine the relationship between county enrollment rate and two measures of county managed care activity (penetration and index of competition [IOC]), adjusting for other county characteristics.
RESULTS: In bivariate analysis, there was a strong inverse correlation between trial enrollment rate and IOC (r = –0.23; P < .001) as well as the proportion of the population uninsured (–0.31; P < .001) and the percentage below poverty (–0.16; P < .001). In the multivariate model, greater county managed care competition (IOC) was inversely related to trial enrollment rate (P < .008 for comparison of each quartile v lowest quartile) after accounting for managed care penetration, proportion uninsured, and other county characteristics. Counties in the lowest quartile of managed care penetration tended to have lower enrollment rates than the remaining counties (r = –0.05; P = .048), while counties in the second, third, and fourth quartiles of penetration all had similar enrollment rates to one another.
CONCLUSION: Cancer trial enrollment rates were suboptimal across all counties, and counties with higher levels of managed care competition had significantly lower enrollment rates. The relationship between managed care penetration and trial enrollment was less consistent. Future efforts to enhance trial participation should address the potential negative influence of market factors.
INTRODUCTION
Recruitment of patients into clinical studies is exceedingly difficult.1,2 A recent Institute of Medicine roundtable has identified insufficient public participation in clinical research as one of two major obstacles that must be overcome in order to "...apply science to better human health in an expeditious fashion."2 A considerable amount of attention has focused on identifying barriers to trial participation. These barriers may be related to patients, physicians, study characteristics, or the health system. Patient- and physician-specific barriers such as sociodemographic characteristics, attitudes, and beliefs have been well documented.3-8
To enhance trial recruitment and maintain the viability of clinical cancer research enterprise, it is also important to further our understanding of the impact of health system factors on trial enrollment. Unfortunately, the clinical environment may be increasingly less likely to support research activities. The increased competition and strained resources associated with managed care could affect trial enrollment at either the patient or physician level. Patients with managed care insurance may be wary of enrolling because of concerns about reimbursement, as private insurers may not cover routine care costs for patients enrolled on trials.9-13
Managed care can also affect physician behavior, as enrolling patients onto trials requires a significant investment of time and money. For each cancer patient enrolled on a trial, clinician-investigators and their staff devote roughly between 150 and 450 hours of time, and expend (nonclinical) costs ranging from approximately $1,300 to $3,900.14 These demands on the resources of prospective clinicians/investigators are occurring during a time of scarce resources, as the cost and reimbursement structure surrounding the care of cancer patients has changed significantly with the growth of managed care organization (MCO) insurance. The decrease in clinical revenues that have been associated with MCO coverage has reportedly led clinicians to be reluctant to devote the additional resources necessary for enrolling patients onto trials.15-17 Anecdotal data suggest that managed care has forced academic researchers to devote more time to clinical activities and less to research.18-22 Other investigators have reported that high regional MCO penetration impedes research productivity of academic medical centers in terms of grants and publications.21,22 To our knowledge, the effect of managed care on trial enrollment in community-based practices has not been explored.
Despite the widespread recognition of inadequate trial participation, little is known about the impact of managed care on trial participation at the population level. We therefore assessed the relationship between participation in cooperative group trials sponsored by the National Cancer Institute (NCI) and two measures of MCO activity—MCO market penetration and index of competition (IOC). The rationale for analyzing IOC arises from concern that the level of competition between MCOs may have a greater impact on a market than penetration.23,24 That is, a market that contains only a single MCO that has a penetration of 80% might be quite different from a market with 80% penetration shared by numerous, competing entities. The IOC is used to quantify this competition, and is defined as 1 minus the sum of the squares of each MCO’s market share.23,24 The IOC applies only to the portion of the market that is enrolled in managed care entities and assumes values ranging from 0 to 1, with more competitive markets having values closer to 1. We hypothesized that clinicians and institutions operating in areas with higher MCO penetration or IOC would face increased pressures to maintain clinical revenues and have fewer resources available to conduct research.
METHODS
Cancer trial enrollment data for 1996 to 2001 were obtained from the NCI Cancer Trials Evaluation Program. We only included participants in NCI-sponsored cooperative group breast, colon, lung, and prostate cancer trials of therapeutic agents who were older than 30 years and had documented racial/ethnic group, age, and sex. Trial participants were assigned to counties according to zip code of their residence; the 10,900 patients (of 86,600) who were missing zip codes were excluded. We also excluded counties that were located in Hawaii or Alaska due to the unique geographic characteristics of these locations. We also excluded counties that had less than one patient enrolled per year during our study period, because we hypothesized a priori that patients residing in those counties had virtually no opportunity to enroll on a trial, regardless of the level of managed care activity, and we wanted to focus on areas in which the opportunity to enroll could have been affected by managed care.
County managed care penetration and IOC estimates for the year 2000 were obtained from the Interstudy County Surveyor Database.25 Additional county characteristics were obtained from the US Census and the Area Resource file.25,26 Geographical Information Systems (GIS) data were used to estimate the distance in (linear) miles between the center of each county and the nearest county that had an NCI research center (defined as either a designated cancer center or a community clinical oncology program [CCOP]). We estimated the proportion of the US population residing in the counties in our study sample by dividing the total populations of the included (sample) counties by the total US population.
The most recently available estimates of cancer incidence according to sex, race/ethnicity, and 5-year age group (1995 through 1999) were obtained from the NCI’s Surveillance, Epidemiology, and End Result (SEER) data.27 Because SEER counties only represent a subset of the US population, cancer incidence rates in each of these groups were then applied to the corresponding population categories and then summed over all categories to obtain cancer incidence estimates for each county.
The trial enrollment rate for each county was defined as the annual number of trial participants divided by the annual number of incident cancer patients. Using the county as the unit of analysis, we performed bivariate analyses comparing county trial enrollment rates with predictor variables of interest using t test, 2, and Spearman’s when appropriate. Continuous variables were converted into categorical variables after reviewing bivariate analyses of each managed care variable in quantiles and collapsing them each into four groups based on distribution of the data. This conversion facilitated interpretation of the resulting model, and was found to have little impact on the model’s performance.
We then performed a multiple linear regression using the natural logarithm of the enrollment rate as the dependent variable, and the two measures of MCO activity (penetration and IOC) as independent variables. Because enrollment rates were highly skewed, we log transformed county rates to approach a normal distribution. Candidate covariates were selected with forward stepwise selection and included the proportion of population without health insurance, proportion below poverty level income, urban/rural influence code, unemployment rate, presence of a teaching hospital, number of hospitals with oncology services, physicians per capita, and distance to the nearest research center. Variables with P < .05 were retained in the final model. In several instances, there was substantial colinearity among the candidate predictor variables. In these instances, we selected the variable that had a stronger correlation with enrollment rate on bivariate analysis for inclusion in the model. To present the results in terms of the relationship between IOC and absolute trial enrollment rate, we estimated the adjusted enrollment rate for counties in each of the IOC groups after holding all other county characteristics at a constant level. Analyses were performed using Stata 6.0 (Stata Corp, College Station, TX).28
RESULTS
County Characteristics
Of the 3,134 counties in our data set, 1,418 had six or more patients enrolled in CGCT during 1996 to 2001 and were included in the analytic sample (Table 1) . The total population of the included counties represented about 87.5% of the US population. As expected, the counties in our sample were significantly more likely to be urban and to have a higher managed care market penetration (median was 14.9% in our sample v 4.9% in the excluded counties; Table 1; P < .001) and IOC (0.588 v 0.287; P < .001). In general, excluded counties were more likely to have lower income levels and fewer teaching hospitals and physicians per capita. Finally, sample counties had far more developed health care and research infrastructure, being more likely to have hospitals with teaching programs and oncology services, to have more physicians per capita, and to be closer to NCI-designated cancer centers or CCOP centers (Table 1).
The median trial enrollment rate for the 1,418 counties included in our sample was 17.1 participants per 1,000 cancer patients (interquartile range [IQR], 10.6 to 26.7). There was a strong inverse correlation relation between index of competition and trial enrollment rate (Table 2; correlation coefficient, –0.23; P < .001). In the 517 counties with the least amount of competition among managed care plans (IOC < 0.5), the median enrollment rate was 20.4 (IQR, 12.3 to 31.6; Table 2). The median enrollment rate decreased to 17.3 in counties with IOC 0.5 to 0.65, and was 15.3 and 12.2 among counties with IOC 0.65 to 0.78, and greater than 0.78, respectively (Table 2). The mean enrollment rates (and 95% CIs) according to IOC are shown in Figure 1.
There was a relationship between managed care penetration and enrollment rates, though this was of borderline significance (Table 2; P = .048). The proportion of people without health insurance was also inversely related to enrollment rate (Table 2; –0.31; P < .001). Counties in the lowest quartile of proportion uninsured (< 11%) had a median enrollment rate of 20.6 (IQR, 13.7 to 29.9); the enrollment rate decreased to 11.4 (IQR, 6.9 to 20.0) in counties with the highest quartile uninsured (> 16.4%).
County socioeconomic status was also related to trial enrollment. Counties with a higher proportion of their population living below poverty level had substantially lower median enrollment rates (correlation, –0.16; P < .0001), decreasing from 19.1 in counties with less than 9.1% below poverty to 14.0 in counties with the highest poverty level.
Geographic location was also related to trial enrollment, as counties with NCI research centers had a median rate of 21.7 (IQR, 13.8 to 32.7), while counties more than 36.5 miles from the nearest center had a median rate of 15.5 (IQR, 9.1 to 24.8; Spearman’s correlation = –0.16; P < .0001). Counties with the highest enrollment fraction were less likely to be have two or more hospitals with oncology services (P < .001), and tended to be nonurban (P < .001). The presence of hospitals with teaching programs was unrelated to trial enrollment (P = .34).
In our final multivariable model, both IOC and managed care penetration were significantly related to county enrollment rate after adjusting for geographic location and population sociodemographic characteristics (Table 2). There was a significant decrease in trial enrollment rate with increasing IOC (Table 3). As the county IOC increased, the adjusted trial enrollment decreased substantially, from 27.8 per 1,000 (95% CI, 22.9 to 33.7) in the lowest IOC group to 14.6 (95% CI, 11.9 to 18.0) in the highest IOC group (Fig 1). Counties in the lowest group of MCO penetration (group 1: < 3.23%) tended to have significantly lower enrollment rates than counties in the remaining groups (P = .005, .009, and .009 for comparison of group 1 v groups 2, 3, and 4, respectively). There was no significant relation between enrollment and MCO penetration for the upper three groups of MCO penetration (ie, group 2 v group 3 v group 4 pairwise comparisons all > .05).
Counties which were urban, at increasing distance from a research center, and with higher proportion of the population uninsured or below poverty also tended to have lower enrollment rates in the multivariate analysis.
DISCUSSION
We found that counties with more competitive managed care markets had significantly lower cancer trial enrollment rates after accounting for managed care penetration, poverty status, and other county-level characteristics. The impact of managed care on trial recruitment may be substantial, as the adjusted enrollment decreased from over 25 per 1,000 patients in the counties with the lowest IOC to about 15 per 1,000 in the counties with the highest IOC. Given that there are more than 600,000 newly diagnosed cases of lung, breast, colorectal, and prostate cancer annually, these relative changes in enrollment that are related to managed care translate into large differences in trial enrollment at the national level.29
The relationship between trial enrollment rate and managed care penetration was less consistent than IOC. We found that enrollment rates in the upper three groups of counties according to managed care penetration did not differ significantly from one another. Although prior efforts have suggested that managed care penetration may hinder research productivity, these studies were performed substantially earlier than our study, when managed care was growing more dynamically.21,22 Another, more recent study found that statewide MCO penetration was not associated with cancer trial enrollment.30 However, there is substantial variation in MCO penetration within states. Hence, our analysis, which accounts for MCO penetration at the county level, provides more rigorous support to the hypothesis that MCO penetration is not a strong barrier to enrollment.
Our findings emphasize the importance of distinguishing between competition between MCO plans and overall penetration. While some recent work may suggest that managed care may not hinder enrollment, our findings suggest otherwise. There are a number of potential mechanisms to explain this phenomenon. Among counties with similar levels of managed care penetration, counties with higher levels of competition among managed care plans may be more likely to have more pressure to control costs. Decreased revenues associated with managed care have affected the research productivity of both individuals and institutions.31 Competitive markets are associated with shortfall of discretionary funds, less institutional support for research, and decreased amount of time spent on research activities.20,31 Prior physician surveys have also found that MCO insurance makes it more difficult for clinicians to provide routine care for their cancer patients or to enroll them on clinical trials.15,32 Researchers in more competitive markets are also more likely to express concerns about access to potential enrollees due to limitations on referrals out of network.33 MCO coverage may also directly affect patients’ ability to participate in trials, as many private insurers will not routinely cover patient care costs for "investigational treatments."11,16 A prior study found that among privately insured patients with HIV, those with MCO insurance were much less likely than those with fee for service to participate in studies.8
Competition among providers may also serve as a barrier to trial enrollment. In our bivariate analysis, counties with proportionately more physicians and/or hospitals with oncology services actually enrolled fewer patients. Although neither of these relations remained significant in the multivariate model, future work should explore whether areas with a relative abundance of oncology services are optimizing opportunities for trial enrollment. Geographic location was also an important factor; counties 36 miles from the nearest NCI research center had significantly lower enrollment rates than counties with such centers. We also found that the presence of a teaching hospital was not associated with increased trial enrollment rates, suggesting that NCI-funded research centers have a more broad affect on trial enrollment than do teaching hospitals. This reinforces the importance of the CCOP network and plans for expansion.34
Some studies have reported that lack of insurance can decrease access to experimental therapies, while others have suggested that uninsured patients are more likely to participate in trials.8,35 The participation of patients without health insurance in trials has raised concerns about undue inducement of participants, who may have few other care options outside of the study setting.36 However, we found that the proportion of the population without health insurance was inversely related to county trial enrollment. This may be attributable to decreased access to care; uninsured patients may be more likely to present with later stage disease and be ineligible for trials on that basis.37-39 Additionally, higher levels of uninsured at the county level may further strain clinical resources at associated institutions, with a grater demand for uncompensated care.
It is important to note that counties with fewer than one trial participant annually were excluded from our analysis, as we had hypothesized a priori that it was unlikely that health system factors would affect enrollment in counties that offered little or no access to research. Although the 1,400 counties in our analytic sample comprised approximately 88% of the US population, the results of our analysis may not be generalizable to the remaining counties. One of the challenges to analyzing trial enrollment rates by county is the lack of available cancer incidence data by county at the national level. To test the validity of our demographic projection approach, we compared estimates of cancer incidence recorded in the SEER registry with those we estimated for the SEER counties; we found the Spearman’s correlation coefficient was 0.97. Additionally, our sample was also restricted to NCI-sponsored cooperative group trials of therapeutic agents; it is unclear whether our results are generalizable to federal or industry-sponsored trials or to prevention studies. Because patients were assigned to counties based on their home address, some patients may have traveled to different counties to enroll in research. We addressed this potential limitation by including distance to the nearest cancer center as a covariate. Finally, it is important to note that the relationship between managed care and trial enrollment was cross-sectional; future work will have to explore whether managed care competition directly results in lower trial enrollment rates.
Trial enrollment rates are currently not optimal for the expedient evaluation of new therapeutic agents. Additionally, patients who could potentially benefit from trial participation, either from access to new experimental treatments or by the important sense of altruism associated with enrolling on studies, face health system issues as one of the important barriers to participation. The NCI has identified expanding the pool of trial participants as a priority and has invested substantially in improving infrastructure and information systems to alleviate "...costs at local clinical trial sites, and for operations, data management, and statistical offices."34 Our findings emphasize the importance of addressing the impact of managed care and lack of health insurance on strained clinical resources, in order to ensure that clinicians have sufficient time to devote to clinical research and that patients are offered the chance to participate. Additionally, it is unclear whether managed care reimbursement policies also pose a barrier, as patients with managed care may be wary of seeking approval and reimbursement. Recent work suggests that clinicians do not receive adequate reimbursement for the costs associated with trial participation.17,40 Adequate cancer trial enrollment depends on more than the knowledge, attitudes, and beliefs of individual patients and physicians. As clinical revenues wane and the health marketplace becomes more competitive, patients, physicians, and investigators may need additional resources to ensure access to trials and timely scientific progress in a challenging clinical environment.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
NOTES
Supported by (C.P.G.) a Cancer Prevention, Control and Population Sciences Career Development Award (1K07CA-90402) and the Claude D. Pepper Older Americans Independence Center at Yale (P30AG21342).
This original work has been neither presented nor published previously.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Office of the Director: The Nation's Investment in Cancer Res: A Budget Proposal for Fiscal Year 2001. Bethesda, MD, National Cancer Institute, 1999
Sung NS, Crowley WF Jr, Genel M, et al: Central challenges facing the national clinical research enterprise. JAMA 289:1278-1287, 2003
Shavers-Hornaday V, Lynch C, Burmeister L, et al: Why are African-Americans under-represented in medical research studies Impediments to participation. Ethn Health 2:31-45, 1997
Foley J, Moertel C: Improving accrual into cancer clinical trials. J Cancer Educ 6:165-173, 1991
Fleming I: Clinical trials for cancer patients: The community practicing physician's perspective. Cancer 65:2388-2390, 1990
Fallowfield L, Ratcliffe D, Souhami R: Clinicians' attitudes to clinical trials of cancer therapy. Eur J Cancer 33:2221-2229, 1997
Swanson D, Ward A: Recruiting minorities into clinical trials: Toward a participant-friendly system. J Natl Cancer Inst 87:1747-1759, 1995
Gifford AL, Cunningham WE, Heslin KC, et al: Participation in research and access to experimental treatments by HIV-infected patients. N Engl J Med 346:1373-1382, 2002
Comis R, Aldige C, Stovall E, et al: A Quantitative Survey of Public Attitudes Towards Cancer Clinical Trials. http://www.cancertrialshelp.org/cnccg_info/news.html, 2000
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
Goodman C: Investigational exclusion, clinical trials, and cancer. NCCN Proceedings 12:37-49, 1998
Vogelzang N, Richards J: Third party reimbursement issues during a phase I trial of interleukin-2. J Nat Canc Inst 81:544-545, 1989
NIH Clinical Trials: Various Factors Affect Patient Participation. Washington, DC, US General Accounting Office, 1999
Emanuel EJ, Schnipper LE, Kamin DY, et al: The costs of conducting clinical research. J Clin Oncol 21:4145-4150, 2003
Wade J, Weinberg P, Bennett C: The effects of managed care on community clinical research. Primary Care Cancer 19:18-25, 1999
Fleming R: Barriers to clinical trials, part I: Reimbursement problems. Cancer 74:2662-2665, 1994
Chaturvedi S: Clinical trials and financial reimbursement. Stroke 29:1256, 1998
Ta KT, Persing JA, Chauncey H Jr, et al: Effects of managed care on teaching, research, and clinical practice in academic plastic surgery. Ann Plast Surg 48:348-354, 2002
Levitt S: Impact of managed care on scholarly activity and patient care: Case study of 12 academic radiology and radiation oncology departments. Radiology 216:618-623, 2000
Campbell EG, Weissman JS, Causino N, et al: Market competition and patient-oriented research: The results of a national survey of medical school faculty. Acad Med 76:1119-1126, 2001
Campbell E, Weissman J, Blumenthal D: Relationship between market competition and activities and attitudes of medical school faculty. JAMA 278:222-226, 1997
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
Wholey D, Christianson J, Engber J, et al: HMO market structure and performance. Health Aff (Millwood) 16:75-84, 1997
Gaskin DJ, Hadley J: The impact of HMO penetration on the rateo of hospital cost inflation. Inquiry 34:205-216, 1997
Publications I: County Surveyor Database 3.0. St Paul, MN, 2003
Quality Resource Systems I: http://www.arfsys.com/
Warren J, Klabunde C, Schrag D, et al: Overview of SEER-Medicare data: Content, research applictions, and generalizability to the United States elderly population. Med Care 40:IV-3-IV-18, 2002 (suppl)
Stata SS: Release: 8.0 Reference Manual. College Station, TX, Stata Press, 2004
Society AC: Cancer facts and figures. 2000
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
Burnett D: Evolving market will change clinical research. Health Aff (Millwood) 15:89-92, 1996
Mortonson L, Leader S, Mallick R: The impact of managed care on oncology practice. Oncology Issues 12:22-27, 1997
Mechanic RE, Dobson A: The impact of managed care on clinical research: A preliminary investigation. Health Aff (Millwood) 15:72-89, 1996
anonymous: The Nation's Investment in research: A plan and budget proposal for fiscal year 2004. Bethesda, MD, National Cancer Institute, 2003
Gorkin L, Schron EB, Handshaw K, et al: Clinical trial enrollers vs. nonenrollers: The Cardiac Arrhythmia Suppression Trial (CAST) Recruitment and Enrollment Assessment in Clinical Trials (REACT) project. Control Clin Trials 17:46-59, 1996
Pace C, Miller FG, Danis M: Enrolling the uninsured in clinical trials: An ethical perspective. Crit Care Med 31:S121-S125, 2003
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
Ferrante JM, Gonzalez EC, Roetzheim RG, et al: Clinical and demographic predictors of late-stage cervical cancer. Arch Fam Med 9:439-445, 2000
Roetzheim RG, Pal N, Tennant C, et al: Effects of health insurance and race on early detection of cancer. J Natl Cancer Inst 91:1409-1415, 1999
Wright J, Levine M: Researching the cost of research. J Clin Oncol 21:4081-4082, 2003(C.P. Gross, H.M. Krumholz)