A 1999 investigation1 found that industry sponsorship of cost-effectiveness analyses (CEAs) for oncology drugs was associated with lower likelihood of reporting unfavorable conclusions relative to CEAs with other sponsorship. Over the past 15 years, the CEA literature for oncology drugs has expanded dramatically, and oncology now accounts for the largest single pharmaceutical sales area worldwide.2 We sought to determine whether the association between industry sponsorship and CEA results has persisted.
Methods
We examined CEAs for breast cancer, the most common target diagnosis among oncology CEAs (36% of studies).3 We obtained data on all such CEAs published between 1991 and 2012 from the Tufts Cost-Effectiveness Analysis Registry, which was created by searching MEDLINE for all English language CEAs using the key words “QALYs” (quality-adjusted life-years), “cost-utility analysis,” and “breast cancer.” From the registry we extracted study characteristics, results (cost per QALY), and registry-assigned quality ratings (which ranged from 2 to 6).4 We considered a study industry-sponsored if a pharmaceutical company provided funding or if 1 or more study authors was a company employee. Study authors provided sponsorship information for 13 studies with unclear funding information in the publication.
We converted each study’s results to 2013 US dollars using purchasing power parity conversion factors, categorized study results based on 3 thresholds ($50 000, $100 000, and $150 000/QALY), and classified studies as “cost-effective” if all results were equal to or more favorable than the chosen threshold, “not cost-effective” if none were, or “mixed” otherwise (note: each study could contain multiple analyses, with varying assumptions). Using JMP Pro statistical software (version 11.0.0, SAS Institute), we tested bivariate associations between industry sponsorship and study characteristics. We then fitted logistic regressions to estimate independent associations between industry sponsorship and study results, adjusting for drug class, cancer stage targeted, and study quality score.
Results
Of 105 CEA studies, 65 were industry funded (Table 1). Study quality ratings were nonsignificantly higher among industry-sponsored studies (mean rating, 4.8 vs 4.4 among studies with other sponsorship; P = .09).
Table 1.
Characteristics of Published Cost-effectiveness Studies of Drugs for Breast Cancer Treatment With Industry and Other Sponsorshipa
| Characteristic | No. (%) | P Valueb | |
|---|---|---|---|
| Pharmaceutical Company-Sponsored (n = 65) | Other Sponsorship (n = 40) | ||
| Drug classc | |||
| Hormonal therapy | 30 (46.2) | 12 (30.0) | |
| Primary prevention | 1 (3.3) | 8 (66.7) | <.001 |
| First line | 6 (20.0) | 1 (8.3) | |
| Adjuvant | 21 (70.0) | 3 (25.0) | |
| Second line | 2 (6.7) | 0 | |
| Chemotherapy | 20 (30.8) | 17 (42.5) | |
| First line | 3 (15.0) | 1 (5.9) | .02 |
| Adjuvant | 9 (45.0) | 15 (88.2) | |
| Second line | 8 (40.0) | 1 (5.9) | |
| Bisphosphonated | 8 (12.3) | 1 (2.5) | |
| Hematopoietic growth factor | 6 (9.2) | 2 (5.0) | |
| Targeted therapye | 6 (9.2) | 10 (25.0) | |
| Cancer stage targeted | |||
| Preventive | 1 (1.5) | 8 (20.0) | <.001 |
| Early (stage IIIA or below) | 37 (56.9) | 26 (65.0) | |
| Advanced | 26 (40.0) | 5 (12.5) | |
| Both early and advanced | 1 (1.5) | 1 (2.5) | |
| Average quality rating of paper, mean (SD)f | 4.8 (0.8) | 4.4 (1.0) | .09 |
P values for categorical variables (drug class, cancer stage) are from fisher exact tests; P value for the continuous variable (average quality rating) is from a Wilcoxon rank sum test.
P = .04 for difference in drug class evaluated (hormonal, chemotherapy, bisphosphonate, hematopoietic, targeted) between industry-sponsored and other sponsored studies.
Includes denosumab.
Includes trastuzumab and lapatinib.
Tufts registry reviewers assigned each study a quality score from 1 (lowest quality) to 7 (highest quality) using the following criteria: whether the authors correctly computed the incremental cost-effectiveness ratios, performed a sensitivity analysis, correctly used and specified the health economic assumptions used in the study, and appropriately and explicitly estimated the utility weights.
Industry-sponsored studies were statistically significantly more likely than other-sponsored studies to report favorable cost-effectiveness results: 75.4% vs 40.0% at $50 000/ QALY (P = .004), 80.0% vs 57.5% at $100 000/QALY (P = .03), and 87.7% vs 67.5% at $150 000/QALY (P =.04) (Table 2). Among the subset of CEAs with high quality ratings (≥4.5), industry-sponsored studies were more likely to report favorable findings (75.5% vs 45.5%, P = .04, at the $50 000/QALY threshold).
Table 2.
Relationship Between Results and Sponsorship of Cost-effectiveness Studies of Drugs for Breast Cancera
| Cost-effectiveness Threshold, QALY, $ | Sponsored Studies Reporting Cost-effective Results, No. (%) | Reporting Cost-effective Results, Adjusted Odds Ratio (95% CI) | P Value | C Statistic | |
|---|---|---|---|---|---|
| Industry | Other | ||||
| 50 000 | 49/65 (75.4) | 16/40 (40.0) | 4.01 (1.55–10.92) | .004 | 0.72 |
| 100 000 | 52/65 (80.0) | 23/40 (57.5) | 3.14 (1.14–9.08) | .03 | 0.69 |
| 150 000 | 57/65 (87.7) | 27/40 (67.5) | 3.27 (1.05–11.08) | .04 | 0.68 |
Abbreviation: QALYs, quality-adjusted life-years.
Adjusted odds ratios, 95% CIs, and P values are from logistic regressions predicting cost-effectiveness results as a function of industry sponsorship, with adjustment for the variables presented in Table 1: drug class (hormonal, chemotherapy, or other), cancer stage, and study quality score.
Discussion
Our analysis of breast cancer CEAs suggests that pharmaceutical industry–sponsored studies continue to be more likelytoreportfavorableestimatesthanstudieswithotherspon-sorship. These findings have multiple possible explanations.
First, most CEAs have retrospective designs, which can allow investigators to identify and then conduct, based on early looks at clinical and resource profiles, those trials most likely to yield positive outcomes. Second, potential conflicts of interest exist. Pharmaceutical companies can exert influence through grants, educational funds, or manuscript review requirements. Investigators set the values assigned to quality of life, determine the price and duration of interventions, and make other methodological choices that can affect study findings. Making these choices transparently and before results are known could enhance the credibility of CEAs.5
Our study has limitations. We examined drugs for breast cancer only. Financial relationships between pharmaceutical companies and researchers were considered, but other less readily detectable factors influencing study findings may exist.
Additional studies are needed to determine whether similar associations between industry sponsorship and results exist for CEAs of drugs treating other cancers. Registering CEAs and their methods at inception could help address the conflicts of interest that might underlie these associations.
Acknowledgments
Funding/Support: This study was supported in part by the National Cancer Institute (1R01CA165609-01A1), the South Carolina SmartState Program, and the Doris Levkoff Meddin Medication Safety Center.
Role of the Funder/Sponsor: These sponsors did not have a role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.
Footnotes
Conflict of Interest Disclosures: None reported.
Additional Contributions: We thank Peter J. Neumann, ScD (Tufts Medical Center) and the Center for the Evaluation of Value and Risk in Health at Tufts Medical Center for establishing, maintaining, and making available the Cost-Effectiveness Analysis Registry. Dr Neumann did not receive compensation for his contribution.
Author Contributions: Dr Lane had full access to all of the data in the study and takes responsibility for the integrity of the data and accuracy of data analysis.
Study concept and design: All authors.
Acquisition, analysis, or interpretation of data: Lane, Friedberg.
Drafting of the manuscript: Lane.
Critical revision of the manuscript for important intellectual content: Friedberg, Bennett.
Statistical analysis: Lane, Friedberg.
Administrative, technical, or material support: Bennett.
Supervision: Friedberg, Bennett.
References
- 1.Friedberg M, Saffran B, Stinson TJ, Nelson W, Bennett CL. Evaluation of conflict of interest in economic analyses of new drugs used in oncology. JAMA. 1999;282(15):1453–1457. doi: 10.1001/jama.282.15.1453. [DOI] [PubMed] [Google Scholar]
- 2.Innovation in cancer care and implications for health systems. Global Oncology Trend Report. IMS Institute for Healthcare Informatics; [Accessed Nov 4, 2015]. website http://www.imshealth.com/deployedfiles/imshealth/Global/Content/Corporate/IMS%20Health%20Institute/Reports/Secure/IMSH_Oncology_Trend_Report.pdf. [Google Scholar]
- 3.Greenberg D, Earle C, Fang CH, Eldar-Lissai A, Neumann PJ. When is cancer care cost-effective? a systematic overview of cost-utility analyses in oncology. J Natl Cancer Inst. 2010;102(2):82–88. doi: 10.1093/jnci/djp472. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Center for the Evaluation of Value and Risk in Health (CEVR) [Accessed July 27, 2014];The Cost-Effectiveness Analysis Registry website. https://research.tufts-nemc.org/cear4/
- 5.Rennie D, Luft HS. Pharmacoeconomic analyses: making them transparent, making them credible. JAMA. 2000;283(16):2158–2160. doi: 10.1001/jama.283.16.2158. [DOI] [PubMed] [Google Scholar]
- 6.Thorat T, Changelosi M, Neuman PJ. Skills of the trade: the tufts cost-effectiveness analysis (CEA) registry. J Benefit-Cost Analysis. 2012;3:1–9. [Google Scholar]