|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 25, No 12 (April 20), 2007: pp. 1461-1462 © 2007 American Society of Clinical Oncology. DOI: 10.1200/JCO.2006.10.3648
Cancer Chemoprevention: How Do We Know What Works?Departments of Medicine and Epidemiology, Herbert Irving Comprehensive Cancer Center, Columbia University Medical Center, New York, NY The potential for natural or synthetic substances to reduce the risk of cancer has spurred the biomedical community to identify, isolate, and evaluate such compounds. Recent advances in chemoprevention include the success of tamoxifen and raloxifene in preventing breast cancer,1,2 aspirin and cyclooxygenase-2 inhibitors in preventing colorectal adenomas,3-5 and finasteride in preventing prostate cancer.6 For a substance to be acceptable for prevention, it has to be effective, easy to administer, and have minimal adverse effects. The most rigorous way to determine the efficacy of a potential chemopreventive agent is a randomized controlled trial (RCT), with the measured outcome being the development of cancer, which is similar to the studies on hormones in breast and prostate cancer cited previously. However, because such studies are large, require prolonged follow-up time, and are very costly, they are used for only a select group of potential agents. A second option is to measure an intermediate biomarker as a surrogate for the development of cancer. In this regard, change in mammographic density or proliferative antigen Ki-67 in breast tissue may be used as surrogate end points for breast cancer.7,8 Such studies assume that intermediate biomarkers are reliable predictors of the development of cancer, although the validity of these assumptions may vary. A third approach is an observational study, such as a case-control or cohort study. Such studies have been facilitated by the widespread availability and investigator use of patient databases to explore associations between drug exposure and the subsequent development of cancer.
In this issue of the Journal of Clinical Oncology, Govindarajan et al9 use this latter technique as they report the results of the first large-scale study of thiazolidinediones (TZDs) as chemopreventive agents in humans. This class of oral hypoglycemic agents, which includes rosiglitazone and pioglitazone, is used as monotherapy and in combination with other oral hypoglycemic agents as well as insulin in the management of diabetes mellitus type II. These agents alter insulin resistance by acting on the peroxisome proliferator–activated receptor- This study illustrates the utility of population-based databases in cancer research. We can expect to see a growing number of such studies as the methodology of epidemiological data acquisition and interpretation is further refined and applied. Increasingly, the medical system relies on electronic sources of record keeping for patient care: insurance billing, clinical laboratory systems, tumor registries, and pharmacy dispensing for record keeping. These databases, though not designed for research purposes, are information repositories, which can then be adapted in resourceful and ingenious ways to provide invaluable information relatively cheaply. For cancer, in particular, the existence of population-based tumor registries in virtually every state—spearheaded by the tumor registries of the National Cancer Institute's (Bethesda, MD) Surveillance, Epidemiology, and End-Results Program (SEER)—provide unselected and high-quality tumor information and survival outcomes that can be linked to other systems. To date, the most successful of these databases in the United States has been the link of SEER to Medicare (Baltimore, MD) data, which has been the source of several hundred papers on the quality, efficacy, cost, and consequences of cancer treatments.11,12 The use of database analysis has its strengths as well as limitations. Databases maintained by Veterans Affairs, such as VISN 16, as well as by Medicare and private health maintenance organizations, allow the investigator access to large populations. Due to the population size, databases allow for the analysis of subgroups of patients that are often under-represented in RCTs, such as racial minorities and the elderly. Whereas volunteers for RCTs tend to be younger, healthier, and of a higher socioeconomic status, those included in database-derived studies might reflect the typical patient with multiple comorbidities or less healthy behavior patterns, which might broadly reflect the patients under consideration for treatment.13 The larger numbers of patients also allow the investigator to study rare or uncommon outcomes, such as specific cancers or subsets of cancer, and can address the effect of variations of treatment administration on outcomes. Databases provide an opportunity to observe other late effects, beneficial or harmful, of medications as well. Such effects may not be realized in the limited follow-up of most RCTs. However, studies using databases also have limitations. The information gleaned from databases is observational and, thus, prone to selection and other biases. For example, the patients who were prescribed TZDs in the study by Govindarajan et al had different baseline characteristics compared with those who were not prescribed TZDs, which could have confounded the results. Some of these differences were noted by the investigators, such as race (there were fewer African Americans in the TZD group), average body mass index (which was larger in the TZD group), and glycemic control (which was worse in the TZD group). Baseline differences between groups can be overcome in the statistical analysis with covariate adjustment (as was done for the covariates mentioned previously). However, bias can be introduced with unknown or unmeasured confounders and remain after statistical adjustment. In the study by Govindarajan et al, the authors acknowledge that the status of cigarette smoking was not known. As the main finding was a reduction in lung cancer, smoking status is a crucial piece of information, and its influence on the reported results is unknown. In addition, although prescription medications, such as other oral hypoglycemic agents and insulin, were included in the multivariate analysis, over-the-counter medications, such as aspirin and other nonsteroidal anti-inflammatory drugs—both of which are likely chemoprotective in colon cancer—were not captured in the pharmacy records. This may have obscured a possible effect in the colorectal cancer outcome if the use of nonsteroidal anti-inflammatory drugs was more common in the control arm. Other comorbidities that were not adjusted for include heart failure, renal disease, and hepatic disease—all of which are contraindications to TZD therapy and can be associated with smoking or other confounding factors. Database analysis involves multiple assumptions about both exposure and outcome. In RCTs, subjects are given specific medications at fixed doses for specific durations, and the studies can include mechanisms to monitor patients’ adherence to medications. In contrast, studies that use patient databases, such as the VISN 16, use prescription records and dispensing dates as a surrogate for medication use, both in terms of adherence as well as duration of use. Fortunately, for outcomes such as cancer, the diagnoses tend to be well validated when a tumor registry is used, as tumor registries, such as SEER, record detailed diagnostic information on tumor, histology, stage, and grade. We should greet these database-derived results with a degree of healthy skepticism. As is the case at times with RCTs, conflicting results can emerge. The case of statins as a chemopreventive class offers a cautionary tale. A case-control study that used patient databases in Israel found that the use of statins was associated with a substantial reduction in the risk of the development of colorectal cancer.14 However, a subsequent large cohort study,15 as well as a meta-analysis of RCTs,16 demonstrated no such association. The study by Govindarajan et al illustrates databases’ potential to generate hypotheses regarding chemopreventive agents. As the number of agents worthy of investigation increases, novel study designs will be required to assess the chemopreventive potential of each. As is frequently the case, when large amounts of data permit unrestricted multiple comparisons, algorithms and procedures with appropriate statistical rigor will be mandatory to assure that the findings are biologically and medically meaningful. When we do arrive at the next breakthrough in chemoprevention, we will likely owe the achievement in part to a database-derived study that pointed us in the right direction. As the great philosopher Yogi Berra said, "You can observe a lot by just watching." AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST The authors indicated no potential conflicts of interest. AUTHOR CONTRIBUTIONS Conception and design: Alfred I. Neugut, Benjamin Lebwohl Administrative support: Alfred I. Neugut Manuscript writing: Alfred I. Neugut, Benjamin Lebwohl, Dawn L. Hershman Final approval of manuscript: Alfred I. Neugut, Benjamin Lebwohl, Dawn L. Hershman REFERENCES
1. Fisher B, Costantino JP, Wickerham DL, et al: Tamoxifen for prevention of breast cancer: Report of the National Surgical Adjuvant Breast and Bowel Project P-1 Study. J Natl Cancer Inst 90:1371-1388, 1998 2. Vogel VG, Costantino JP, Wickerham DL, et al: Effects of tamoxifen vs raloxifene on the risk of developing invasive breast cancer and other disease outcomes: The NSABP Study of Tamoxifen and Raloxifene (STAR) P-2 trial. JAMA 295:2727-2741, 2006 3. Baron JA, Cole BF, Sandler RS, et al: A randomized trial of aspirin to prevent colorectal adenomas. N Engl J Med 348:891-899, 2003 4. Bertagnolli MM, Eagle CJ, Zauber AG, et al: Celecoxib for the prevention of sporadic colorectal adenomas. N Engl J Med 355:873-884, 2006 5. Arber N, Eagle CJ, Spicak J, et al: Celecoxib for the prevention of colorectal adenomatous polyps. N Engl J Med 355:885-895, 2006 6. Thompson IM, Goodman PJ, Tangen CM, et al: The influence of finasteride on the development of prostate cancer. N Engl J Med 349:215-224, 2003 7. Pike MC: The role of mammographic density in evaluating changes in breast cancer risk. Gynecol Endocrinol 21:1-5, 2005 (supp 1)[Medline] 8. Fabian CJ, Kimler BF, Anderson J, et al: Breast cancer chemoprevention phase I evaluation of biomarker modulation by arzoxifene, a third generation selective estrogen receptor modulator. Clin Cancer Res 10:5403-5417, 2004 9. Govindarajan R, Ratnasinghe L, Simmons D, et al: Thiazolidinediones and the risk of lung, prostate, and colon cancer in diabetic subjects. J Clin Oncol 25:1476-1481, 2007 10. Kopelovich L, Fay JR, Glazer RI, et al: Peroxisome proliferator-activated receptor modulators as potential chemopreventive agents. Mol Cancer Ther 1:357-363, 2002 11. Potosky AL, Riley GF, Lubitz JD, et al: Potential for cancer related health services research using a linked Medicare-tumor registry database. Med Care 31:732-748, 1993[Medline] 12. Warren JL, Klabunde CN, Schrag D, et al: Overview of the SEER-Medicare data: Content, research applications, and generalizability to the United States elderly population. Med Care 40:IV3-IV18, 2002 (suppl 8) 13. Antman K, Amato D, Wood W, et al: Selection bias in clinical trials. J Clin Oncol 3:1142-1147, 1985 14. Poynter JN, Gruber SB, Higgins PD, et al: Statins and the risk of colorectal cancer. N Engl J Med 352:2184-2192, 2005 15. Jacobs EJ, Rodriguez C, Brady KA, et al: Cholesterol-lowering drugs and colorectal cancer incidence in a large United States cohort. J Natl Cancer Inst 98:69-72, 2006 16. Dale KM, Coleman CI, Henyan NN, et al: Statins and cancer risk: A meta-analysis. JAMA 295:74-80, 2006
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2007 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
nuclear receptor, which plays a role in cell cycle arrest; thus, they may inhibit tumor growth.
