Right Approach in ChemoPrevention Studies?

Are We Taking the Right Approach in Planning Chemoprevention Studies?

Gad Rennert

Evidence gained in randomized controlled trials (RCTs) is considered superior to that gained in association epidemiological studies, such as cohort or case-control studies. Association studies are prone to methodological bias, which is assumed to be overcome by proper randomization of the study population.

The decision to embark on an RCT usually relies on preliminary data from human association studies supported by biological data from other types of studies and is not made lightly. RCTs require years of intervention and follow-up and are hence costly.

RCTs have failed to validate several epidemiological hypotheses for cancer prevention. For example, -carotene, shown in many epidemiological studies to be associated with reduced incidence of many cancers, among them lung cancer,[1] was found in RCTs to increase the probability of lung cancer in smokers.[2,3]

Fiber supplementation in the diet, shown in association studies to reduce the risk of colorectal cancer,[4] failed to reduce colorectal adenoma formation.[5] In contrast to a recent study showing an inverse association between use of statins and colorectal cancer,[6] historical RCTs of effects of statins in populations at risk for cardiovascular disease failed to confirm such a cancer-preventive effect.[7]

This seemingly repeated failure of RCTs to confirm associations identified by epidemiology datasets and biological inferences might be explained by deficiencies in the design and conduct of the observational association studies, as well as deficiencies in the design of the randomized chemoprevention trials.

Reasons for the failure of observational studies include selection bias and recall bias, but also include difficulty in representing lifestyle behaviors (such as diet) over a lifetime, which could be considered a measurement bias.

Further problems arise when considering that the timing might be as important as the type of exposure. Critical susceptibility periods are often unknown. If a critical exposure time falls during a period in the remote past (childhood, early adulthood), it cannot be represented properly in classical case-control studies of cancer.

Multiple behaviors and crosstalk between various exposures, combined with the unique nature of each participant's genotype, complicate any analysis (or dataset).

RCTs might fail or lack reproducibility because we are unable to control for genetic differences in susceptibility, individual lifestyle choices or other, as yet unknown, risk factors. In trials of pharmaceutical or nutrition-based interventions, a wrong choice of intervention (such as a synthetic product expected to mimic naturally occurring dietary components), dose, schedule, or timing over lifespan can result in failure of the study. Experimental subjects could undermine therapeutic trials by poor adherence to intervention schedules or to endpoint diagnostic procedures.

These problems compound the classical methodological pitfalls of study power, analysis strategy and control group contamination, which are especially typical with interventions (such as common dietary components or commonly used medications) involving common behaviors.

The failure of the -carotene trials as interventions for the prevention of lung cancer in smokers could, for example, result from a false hypothesis that -carotene is the only or the major bioactive carotenoid in yellow-green leafy vegetables, when in fact the active ingredient could have been another relevant carotenoid isomer, another nutritional component or a combination of nutrients found in whole foods.[3]

Differences in susceptibility to cancer between current and past smokers have also been suggested as reasons for the failure of the trials,[1] and an interaction of smoking and -carotene in enhancing rather than inhibiting carcinogenesis has been demonstrated.[8]

We need to think of remedies we can apply. For population-based case-control studies, strict methodological standards should ensure high compliance of participants and optimal sampling of population-based controls. Data within the non-population-based studies, commonly found in the literature, are difficult to interpret and might be susceptible to bias.

An effort should be made to estimate the likelihood of important exposures or markers of disease risk (such as weight) in different time periods over the lifetime of the participant, to correct for lifetime variability and identify critical time points.

Self-reported information (such as use of medications) should be externally validated. The degree of internal coherence of the findings and differences in exposure effects between demographic groups (e.g. gender, age and ethnicity) should be assessed. Outstanding findings need in-depth investigation.

Optimally, gene-environment interaction analysis should be routinely included. Such a strategy requires the definition of a minimum set of genetic markers that regulate or influence the activation, disposition or intracellular effects of carcinogens or procarcinogens.

This strategy might also require the definition of genetic markers for a specific exposure (such as MTHFR status for folate consumption) or ethnic group (such as BRCA1 and BRCA2 mutations in the study of risk factors for breast cancer in Ashkenazi Jews).

Such a set or sets of markers will need to be updated regularly as new mechanisms and new markers are discovered. Including evaluation of gene-environment interactions in classical population-based association studies requires larger study cohorts, which should have an impact on the policies of the funding agencies.

For randomized prevention trials, strategies should be tailored to the intervention and the expected associated bias. For studies employing an intervention that is not commonly part of daily life (e.g. an uncommon prescription drug or unique nutritional derivative), fear of contamination is not of major concern.

The study population usually includes well-defined high (genetic) risk populations. In studies involving an intervention that is easily accessible to populations or that modulates human behavior (e.g. diet modification), the target population will be closer to the general (and highly heterogeneous) population.

In these studies fear of contamination (i.e. use of the intervention by the placebo group) leading to reduced power is high. Such population heterogeneity (including a significant genetic heterogeneity) can obscure an effect when such actually exists. Major confounders need identification at the design phase to allow stratification of participants. Randomization should then take place in the identified strata.

Before embarking upon an RCT of purified compounds thought to delay or reverse carcinogenesis ('chemoprevention'), it is important to define an appropriate 'preventive index'. Preliminary datasets should provide evidence that healthy populations will tolerate long-term administration of the test agent, and these datasets might include definition of the minimal effective dose.

Lower doses might actually offer better effectiveness than higher doses. For example, low-dose tamoxifen was shown to exert mainly antiestrogenic effects while full-dose tamoxifen has an estrogenic effect,[9] low-dose aspirin might have a better preventive effect against colorectal cancer than high-dose aspirin,[10] and -carotene might have been given in too high a dose for smokers in the lung cancer prevention trials.[2]

Finally, finding a 'pristine' population that might permit rigorous trials of a commonly available intervention (e.g. aspirin or statin treatment) could prove difficult and expensive. In today's wired, worldwide village, such trials could be an experience of the past.


Omenn GS (1998) Chemoprevention of lung cancer: the rise and demise of beta carotene. Annu Rev Public Health 19: 73-99 -Tocopherol, Beta-Carotene Cancer Prevention Study Group (1994) The effect of vitamin E and -carotene on the incidence of lung cancer and other cancers in male smokers. N Engl J Med 330: 1029-1035 Omenn GS et al. (1996) Effects of a combination of -carotene and vitamin A on lung cancer and cardiovascular disease. N Engl J Med 334: 1150-1155 Park Y et al. (2005) Dietary fiber intake and risk of colorectal cancer: a pooled analysis of prospective cohort studies. JAMA 294: 2849-2857 Rennert G (2002) Dietary intervention studies and cancer prevention. Eur J Cancer Prev 11: 419-425 Poynter JN et al. (2005) Statins and the risk of colorectal cancer. N Engl J Med 352: 2184-2192 Dale KM et al. (2006) Statins and cancer risk: a meta-analysis. JAMA 295: 74-80 Wang XD et al. (1999) Retinoid signaling and activator protein-1 expression in ferrets given -carotene supplements and exposed to tobacco smoke. J Natl Cancer Inst 91: 60-66 Guerrieri-Gonzaga A et al. (2006) Preliminary results on safety and activity of a randomized, double-blind, 2 ~ 2 trial of low-dose tamoxifen and fenretinide for breast cancer prevention in premenopausal women. J Clin Oncol 24: 129-135 Baron JA et al. (2003) A randomized trial of aspirin to prevent colorectal adenomas. N Engl J Med 348: 891-899

Reprint Address

Gad Rennert, CHS National Cancer Control Center, Carmel Medical Center, 7 Michal St, Haifa 34362, Israel. Email: rennert@tx.technion.ac.il

Nat Clin Pract Oncol. 2006;3(9):464-465. Thanks to MEDSCAPE

References: Chemo Prevention Studies

Nat Clin Pract Oncol. 2006;3(9):464-465. Thanks to MEDSCAPE

Remember we are NOT Doctors and have NO medical training.

This site is like an Encyclopedia - there are many pages, many links on many topics.

Support our work with any size DONATION - see left side of any page - for how to donate. You can help raise awareness of CAM.