Abstract
An important health policy issue is the low rate of patient enrollment into clinical trials, which may slow down the process of clinical trials and discourage their supply, leading to delays in innovative life-saving drug treatments reaching the general population. In the US, patients’ cost of participating in a clinical trial is considered to be a major barrier to patient enrollment. In order to reduce this barrier, some states in the US have implemented policies requiring health insurers to cover routine care costs for patients enrolled in clinical trials. This paper evaluates empirically how effective these policies were in increasing the supply of clinical trials and speeding up their completion, using data on cancer clinical trials initiated in the US between 2001 and 2007. Our analysis indicates that the policies did not lead to an increased supply in the number of clinical trials conducted in mandate states compared to non-mandate states. However, we find some evidence that once clinical trials are initiated, they are more likely to finish their patient recruitment in a timely manner in mandate states than in non-mandate states. As a result, the overall length to completion was significantly shorter in mandate states than in non-mandate states for cancer clinical trials in certain phases. The findings hint at the possibility that these policies might encourage drug innovation in the long run.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
The National Cancer Institute [23] states “Lack of third-party reimbursement for clinical trials may be one of the most critical barriers to patient participation.” The American Cancer Society [1] reported that 60% of patients do not take part in clinical trials due to fears of having their insurance denied. Also, Gross et al. [13] states “In addition to these important patient and physician factors, payer reimbursement policies are a frequently cited barrier to recruiting patients into clinical studies.”
Under the policies, routine care costs are paid for by health insurance providers, not clinical trial sponsors. Hence, from the perspective of clinical trial sponsors, states that mandate reimbursement of routine care costs should be more attractive than non-mandate states, all else being equal, as long as the mandates indeed increase clinical trial participation.
Since the mandates require that the financial burden of routine care costs be borne by health insurance providers, not by sponsors, the mandates have no direct impact on the sponsors’ costs of conducting trials.
Suppose that the mandates result in an increase in the number of clinical trial participants but do not affect the provision or length of clinical trials. Then, while the mandates would have implications for those patients whose financial burden has been eased, the mandates would not have any implications for innovation of drugs and therapies.
To our knowledge, financial compensation to patients for trial participation is limited in its scope as there are ethical concerns that substantial compensation could “ unduly influence participation and thus obscure risks, impair judgment or encourage misrepresentation” [12]. According to Grady [12], most clinical research studies offer considerably less than US $500 for participation, mainly for parking, meals, inconvenience and travel. Thus, it seems that such financial incentives are a relatively minor component in patients’ decision on whether to participate in a clinical trial.
Routine care costs are significant. According to Wagner et al. [27], cancer patients’ average 1-year routine care costs are US $24,660.
The term life-threatening disease includes AIDS-related conditions and a number of rare diseases such as Parkinson’s disease.
According to ClinicalTrials.gov, clinical trials typically proceed in four phases. Phase 1 is the first stage of testing in a small group of human subjects and its goal is to evaluate the initial safety of a new drug or procedure. Once initial safety is confirmed in phase 1 trials, phase 2 trials are conducted on a larger group of people to generate more information about its safety and benefits. Phase 3 trials are randomized experiments during which the performance of the new drug or procedure will be compared to the performance of a current standard treatment. After these phases, the developer of the drug or procedure applies for FDA approval. Once approved, phase 4 trials are conducted to continue evaluation. Due to the different nature of phase 4 trials, we focus on trials in phases 1–3 in this paper.
Medicare applies almost purely to elderly patients (Medicare also covers some people who cannot work due to disabilities such as Lou Gehrig’s disease); in our analysis, we do not use trials that focus exclusively on elderly patients in order to minimize the impact of this Medicare policy on our results. Overall, roughly 25–30% of participants in cancer clinical trials are people of 65 or older [16]. Since we exclude trials exclusively focusing on elderly patients, the fraction should be even lower in our sample. Our analysis can be viewed as examining whether the state mandates’ impact on non-elderly patients is significant enough to affect provision and completion time of clinical trials.
They used a Poisson model (the same methodology we use in one of our analyses) where the dependent variable is the number of clinical trial participants.
They used a logistic regression in which the dependent variable was a dummy for being older than 65 years.
Zarin et al. [28] provide a nice discussion of data from ClinicalTrials.gov.
ClinicalTrials.gov started its operation in November 1999, so 1999–2000 is in the early phase of operation for ClinicalTrials.gov. As a result, we were concerned about accuracy of the data during that initial phase and decided to focus on post-2001 data.
Publication is a very good marketing tool to get doctors to adopt new therapies if the results are good. Publication is also valuable for increasing the profile of the physician or scientist who is conducting the trial. Hill et al. [15] reviewed internal documents by Merck and found that the marketing department was actively designing trials for their marketing value, while Loscalzo [18] reports the importance of publication for academic institutes.
One might worry that the new registration policy could have different impacts on mandate states vs non-mandate states. Suppose that the number of clinical trial starts does not change over time in both mandate states and non-mandate states but, after the registration policy, trials run in mandate states are simply more likely to register due to their greater willingness to publish. In that case, we would observe an increase in the number of registered trials in mandate states but not in non-mandate states, although the mandates in fact have no impact whatsoever (we thank an anonymous referee for raising this possibility). We attempt to check this possibility empirically by including an interaction between the mandate variable and post-2005 dummy in our Poisson regression (in Section “Impact of mandates on provision of clinical trials”) where the dependent variable is the number of initiated clinical trials. If it is indeed the case that the registration policy (which was implemented in 2005) affected mandate states and non-mandate states differentially in the direction discussed above, we would expect a positive and significant coefficient on the interaction between the post-2005 dummy and mandate variable. However, we find that the coefficient on the interaction term is sometimes positive and sometimes negative and always insignificant. Thus we find no empirical evidence that the registration policy affected mandate states and non-mandate states differentially.
We were able to obtain data for 2003–2007, and we assume that the elderly population in a given state was the same for years 2001, 2002 and 2003.
This can be seen by the fact that those who are the most vocal advocates of the state mandates and have exerted significant lobbying efforts to implement the legislation were patient advocates such as the American Cancer Society. For instance, an article by the American Society of Clinical Oncology mentions “As a volunteer for the American Cancer Society, Mangialardi was instrumental in getting legislation passed in her home state of Illinois that prohibits health insurance companies from dropping a covered individual because he or she enrolled in a clinical trial.” (http://www.jopasco.org/content/2/6/298.full). Since legislation was passed as a response to requests from patient advocates, they are unlikely to be correlated with, say, the expected trend of clinical trial supply.
We also tried a negative binomial, which is more flexible. The results are almost identical to those from Poisson distribution, because the dependent variable is not over-dispersed. Hence, we report results from Poisson regression only.
In many cases, a given clinical trial is conducted in multiple states. As long as a trial is conducted in state s, we include that trial in defining Y st, regardless of whether state s is the sole location of the trial or not. Hence, a trial that is conducted in state a and state b will be included in both Y at and Y bt.
When we aggregate our data to the quarterly level, M st will be adjusted accordingly as follows. If a reimbursement policy is introduced in state s in January 2002, M st for the first quarter of 2002 will be 1. If the policy is introduced in state s in February 2002, M st for the first quarter of 2002 will be 0.66. If the policy is introduced in state s in March 2002, M st for the first quarter of 2002 will be 0.33. Interpretation of our results will be not affected by this adjustment.
The incidence and mortality measures are summed over the 16 cancer types for each state and year. The ratio of mortality to new cases is computed using the summed measures.
State dummies will capture time-invariant differences across states such as income difference (since income difference across states is likely to be stable over time).
These and other unreported results are available from the authors upon request.
Note that this ambiguity is not present when we use the number of trials as a dependent variable. No firm will conduct more trials due to low enrollment in each trial, because different clinical trials are for different experimental therapies. On the contrary, multiple sites for a given trial administer the exactly same therapy.
Successfully completed trials do not necessarily mean positive outcomes for drug efficacy. Although it is likely that a trial would not be completed if early results from the trial are strongly negative (e.g., the first few patients treated with the drug die), for less clear-cut cases sponsors might decide to complete the trial in order to have a large enough sample size to more precisely determine drug efficacy.
It might seem to make sense to tailor mandate variables for each trial, by checking the presence of mandates during the life of that particular trial. We decided not to take this approach. Since trials that last longer are more likely to experience adoption of the mandates during their lives, there is a selection and endogeneity issue introduced by trial-specific mandate variables.
For these variables, we used the weighted average across states, where a state’s weight is given by the proportion of trial sites located in that particular state for trial i.
We define a single measure of the mandate variable by calculating the fraction of time when the mandate is in place among states where the trial is located. For instance, if a trial has three locations in state A where the mandate was in place throughout the sample period, four locations in state B where the mandate was available for 20% of the sample period, and five locations in state C where the mandate was never implemented, the measure of mandate for this trial would be \(\frac{3\times1+4\times0.2+5\times0}{3+4+5}=0.32\). Clearly, this measure takes a higher value when the trial is more concentrated in always mandate states than in never mandate states.
Censoring dates for ongoing and recruiting trials were the end of the sample period, while censoring dates for terminated, suspended and withdrawn ones were the dates of termination, suspension and withdrawal, respectively. Note that since ongoing trials have not yet ended, it is not possible to measure duration to the end for these trials. Thus, we treat ongoing trials as right-censored in our duration framework. Treating withdrawn and suspended trials as censored is justified as long as we are willing to assume that the hazards of exit from the sample due to withdrawal or suspension is independent of the hazards of exit from the sample due to successful completion, conditional on observables [19].
Results from a more flexible specification, Cox proportional hazard model, are very similar. Thus, we report only results from the exponential duration model.
We thank an anonymous referee for making this point.
References
American Cancer Society: Clinical trials: State Laws Regarding Insurance coverage. http://www.cancer.org/docroot/ETO/content/ETO_6_2x_State_Laws_Regarding_Clinical_Trials.asp(2008)
Angrist, J., Krueger, A.: The effect of age at school entry on educational attainment: an application of instrumental variables with moments from two samples. J. Am. Stat. Assoc. 87, 328–336 (1999)
Baquet, C., Mishra, S., Weinberg, A.: A descriptive analysis of state legislation and policy addressing clinical trials participation. J. Health Poor Underserved 20(2S), 24–39 (2009)
Bertrand, M., Duflo, E., Mullainathan, S.: How much should we trust differences-in-differences estimates?. Q. J. Econ. 119(1), 249–275 (2004)
Chirikos, T., Ruckdeschel, J., Krischer, J.: Impact of clinical trials on the cost of cancer care. Med. Care 39(4), 373–383 (2001)
Comis, R., Aldige, C., Stovall, E., Krebs, L., Risher P., Taylor, H.: A quantitative survey of public attitudes towards cancer clinical trials. http://www.cancertrialshelp.org/CTHpdf/308-9.pdf (2000)
Conner K., Vachani, C.: Insurance coverage for clinical trial costs: beyond the recent New Jersey agreement. OncoLink (2004)
Dimasi, J., Hansen, R., Grabowski, H.: The price of innovation: new estimates of drug development costs. J. Health Econ. 22, 151–185 (2003)
Finkelstein, A.: Static and dynamic effects of health policy: evidence from the vaccine industry. Q. J. Econ. 119(2), 527–564 (2004)
General Accounting Office: NIH clinical trials: various factors affect patient participation. http://www.gao.gov/archive/1999/he99182.pdf (1999)
Goldman, D., Berry, S., McCabe, M., Kilgore, M., Potosky, A., Schoenbaum, M., Schonlau, M., Weeks, J., Kaplan, R., Escarce, J.: Incremental treatment costs in National Cancer Institute-sponsored clinical trials. J. Am. Med. Assoc. 289(22), 2970–2977 (2003)
Grady, C.: Payment of clinical research subjects. J. Clin. Investig. 115(7), 1681–1687 (2005)
Gross, C., Murthy, V., Li, Y., Kaluzny, A., Krumholz, H.: Cancer trial enrollment after state-mandated reimbursement. J. Natl. Cancer Inst. 96(14), 1063–1069 (2004)
Gross, C., Wong, N., Dubin, J., Mayne, S., Krumholz, H.: Enrollment of older persons in cancer trials after the medicare reimbursement policy change. Arch. Intern. Med. 165, 1514–1520 (2005)
Hill, K., Ross, J., Egilman, D., Krumholz, H.: The advantage seeding trial: a review of internal documents. Ann. Intern. Med. 149(4), 251–258 (2008)
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)
Janson, S., Alioto, M., Boushey, H.: Attrition and retention of ethnically diverse subjects in a multicenter randomized controlled research trial. Control. Clin. Trials 22, S236–S243 (2001)
Loscalzo, J.: Clinical trials in cardiovascular medicine: benefit, bias, hyperbole. Circulation 112, 3026–3029 (2005)
Lunn, M., McNeil, D.: Applying cox regression to competing risks. Biometrics 51(2), 524–532 (1995)
Meyer, B.: Natural and Quasi-experiments in economics. J. Bus. Econ. Stat. 13, 151–161 (1995)
Murthy, V., Krumholz, H., Gross, C.: Participation in cancer clinical trials: race-, sex-, and age-based disparities. J. Am. Med. Assoc. 291(22), 2720–2726 (2004)
National Cancer Institute: Doctors, patients face different barriers to clinical trials. http://www.cancer.gov/clinicaltrials/developments/doctors-barriers0401(2001)
National Cancer Institute: report of the National Cancer Institute clinical trials program review group. http://deainfo.nci.nih.gov/advisory/bsa/bsa_program/bsactprgmin.pdf (1997)
Sica, D.: Premature termination of clinical trials: lessons learned. J. Clin. Hypertens. 4(3), 219–225 (2002)
Stitzenberg, K., Sigurdson, E., Egleston, B., Starkey, R., Meropol, N.: Centralization of cancer surgery: implications for patient access to optimal care. J. Clin. Oncol. 27(28), 4671–4678 (2009)
Taylor, H., Leitman, R.: Misconceptions and lack of awareness greatly reduce recruitment for cancer clinical trial. HarrisInteractive 1(3), 1–3 (2001)
Wagner, J., Alberts, S., Sloan, J., Cha, S., Killian, J., O’Connell, M., Van Grevenhof, P., Lindman, J., Chute, C.: Incremental costs of enrolling cancer patients in clinical trials: a population-based study. J. Natl. Cancer Inst. 91, 847–53 (1999)
Zarin, D., Ide, N., Tse, T., Harlan, W., West, J., Lindberg, D.: Issues in the registration of clinical trials. J. Am. Med. Assoc. 297(19), 2112–2120 (2007)
Author information
Authors and Affiliations
Corresponding author
Additional information
We thank two anonymous referees and the editor for their insightful comments.
Appendix
Appendix
ClinicalTrials.gov
The main data set was obtained from ClinicalTrials.gov for 16 different cancer categories. Trials were obtained by using the general search function to search for trials associated with each cancer category. Data was then gathered from each web page associated with the cancer trials on start date, locations, and other aspects of the trials. Sponsor types were obtained using ClinicalTrials.gov categories, but splitting and recombining different categories in order to obtain federal, industry, university and non-profit organization.
There are several observations in regards to the data that are of importance for the validity of the results. The first is that approximately 4,120 trials out of 17,000 trials did not report location information. Conditional on reporting start date of the trials, almost 20% of trials between 2000 and 2004 do not have location data compared to less than 5% in 2005–2007. Secondly, a number of trials combine several phases under one page and this accounted for approximately 10% of all trials. Trials with combined phases were counted once for each phase. Third, about 6% of trials were limited to youth below the age of 18 and about 1% of all trials were restricted to people 65 and older. These were ultimately tossed out with the main focus concentrated on trials with few age restrictions. Fourth, all trials that were denoted as accepting healthy patients were dropped for phase 1, but not phase 2 and above since it is assumed that these patients are generally used as a comparison for unhealthy patients for phase 2 and above. Fifth, we limit attention to trials that started in 2007 or before since there may be some lag in reporting locations that are recruiting. Sixth, we did not eliminate gender specific trials since there are many cancers that are gender specific, for example breast and prostate cancers. Seventh, many trials span multiple conditions with about 70% of all trials covering between 2 and 4 of the 16 cancer conditions listed, and 20% associated with only one cancer condition. Eighth, we do not include phase 4 trials in our analysis since they are trials that are conducted after drug approval, and therefore have a very different nature compared to trials in phases 1–3.
Rights and permissions
About this article
Cite this article
Chun, N., Park, M. The impact of health insurance mandates on drug innovation: evidence from the United States. Eur J Health Econ 14, 323–344 (2013). https://doi.org/10.1007/s10198-012-0379-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10198-012-0379-6