Cancer Prevention

Biomarkers and Surrogate Endpoints in Clinical Trials

Objectives

• Examine the advantages and disadvantages of using biomarkers as surrogate endpoints in cancer prevention trials

• Discuss decision criteria for biomarkers to be considered in clinical trials

• Consider current options substituting for clinical endpoints in chemoprevention strategies

• Address statistical considerations for the validation of biomarkers as surrogate endpoints

- What role does preclinical validation play in establishing surrogate status?

- What approaches should be considered in the evaluation of multiple surrogate endpoints?

- What strategies can be used to interpret and analyze “macroscale” biomarker data obtained from high-throughput technologies (e.g., gene expression patterns)?

- How can statistical modeling incorporating epidemiologic data (including biomarkers) be used in the design of large, randomized clinical trials?

- What resources (e.g., animal models, tissue specimens, sera, genetic and protein data) are needed to ensure the accuracy and precision of novel biomarker assays?

Agenda

Moderators: E. Robert Greenberg, M.D., Dartmouth Medical School W. Ki Hong, M.D., University of Texas M.D. Anderson Cancer Center

Retinoic Acid Receptor Beta as an Intermediate Biomarker in Retinoid Chemoprevention Trials

Theoretical and Practical Considerations in the Use of Surrogate Endpoints in Cancer Prevention Research

Susan S. Ellenberg, Ph.D., Center for Biologics Evaluation and Research, U.S. Food and Drug Administration

abstracts

Risk biomarkers used in the design of chemoprevention trials, and especially in the identification of high‑risk cohorts, can affect study size and duration as well as stratification and analytic considerations. Their value is further enhanced if they are also predictive of response to the intervention. These biomarkers should be of a high sensitivity and specificity; the biomarker assay should be standardized, validated, relatively easy to measure, and should require noninvasive techniques. Since only a fraction of smokers will develop neoplastic lesions, markers of genetic susceptiblity to tobacco carcinogenesis would in theory help identify highest risk smokers for enrollment in chemoprevention trials. A number of low penetrance/high frequency genes are likely to be relevant in the etiology of tobacco‑related cancers. For example, the dose of tobacco carcinogens to which aerodigestive tract tissue is exposed is modulated by genetic polymorphisms in the cytochrome P450 multigene family of enzymes. These enzymes are involved in phase I oxidative processes that may create intermediates more reactive than the parent compounds. The derivative compounds may covalently bind to DNA and form carcinogen‑macromolecular adducts. Phase II metabolic processes generally inactivate these genotoxic compounds through conjugation that promotes cellular excretion. Cancer risk is thus defined by the balance between metabolic activation and detoxification of tobacco carcinogen compounds, as well as by the efficiency of DNA repair. Several phenotypic assays are predictive of risk. These include the host cell reactivation assay that measures DNA repair capability and the mutagen sensitivity assay in which the frequency of in vitro bleomycin‑ or benzo[a]pyrene‑induced breaks in cultured peripheral lymphocytes is quantified as a measure of sensitivity. These assays are not optimal since they require viable lymphocytes and are labor intensive. Nevertheless, our data show that they are predictive of risk and might also be useful to predict patients at higher risk of developing recurrence and/or second primaries. Emphasis should now be placed on studying phenotype/genotype correlations and on comparing markers in surrogate tissue (peripheral lymphocytes) with molecular and cellular changes in the target (lung) tissue. In the near future, microarray technology will enable large‑scale, low‑cost genotyping with the use of automated workstations for extracting DNA and performing DNA amplification, hybridization, and detection. As many as 100 genes at a time can be analyzed for this allele signature analysis. The initial chip being developed will include putative genes for tobacco‑induced cancer susceptibility as well as nicotine addiction. The legal and ethical implications of this emerging technology are immense.

Cancer chemoprevention is the intervention in the multistep process of carcinogenesis by chemical agents (Hong et al. 1995; Hong and Sporn 1997; Lotan 1996). Prevention trials that rely on cancer development as an endpoint are prolonged and require large populations. Therefore, biomarkers are needed to serve as intermediate endpoints (Hong et al. 1995; Hong and Sporn 1997). We investigated the expression of nuclear retinoic acid receptors (RARs) and retinoid x receptors (RXRs) (alpha, beta, and gamma) in surgical specimens collected from normal, premalignant, and malignant head and neck and lung tissues during retinoid chemoprevention trials (Hong et al. 1995; Lotan 1996, 1997) using a nonradioactive in situ hybridization technique (Xu et al. 1994; Xu and Lotan 1998). The level of RARbeta messenger ribonucleic acid (mRNA) was suppressed in 50 to 65 percent of oral premalignant lesions and head and neck squamous cell carcinomas (Xu et al. 1994; Lotan et al. 1995). Treatment of leukoplakia patients with 13‑cis‑retinoic acid caused an increase in RARbeta expression, which was associated with clinical response (Lotan et al. 1995). Similar results were obtained recently with specimens from normal and malignant lung biopsies (Xu et al. 1997). Thus, RARbeta can serve as an intermediate biomarker because its level decreases during the carcinogenic process, it is upregulated by the chemopreventive agent (retinoid), and this upregulation is associated with clinical response. Several ongoing trials will assess in a prospective manner the potential of employing RARbeta as an intermediate biomarker for chemoprevention trials.

Hong WK, Lippman SM, Hittelman WN, Lotan R. Retinoid chemoprevention of aerodigestive cancer: Basic to clinic. Clin Cancer Res 1995;1:677‑686.

Hong WK, Sporn MB. Recent advances in chemoprevention of cancer. Science 1997;278:1073‑1077.

Lotan, R. Retinoids and chemoprevention of aerodigestive tract cancer. Cancer Metastasis Rev 1997;16:349‑356.

Lotan R, Xu X‑C, Lippman SM, Ro JY, Lee JS, Lee JJ, Hong WK. Suppression of retinoic acid receptor beta in oral premalignant lesions and its upregulation by isotretinoin. N Engl J Med 1995;332:1405‑1410.

Xu X, Lotan R (1998). Non‑isotopic in situ hybridization for detection of nuclear retinoid receptor transcripts in tissue sections. In: Methods in Molecular Biology: Retinoids. Redfern C (ed.) Totowa NJ: Humana Press 1998;89:233-246.

Xu X, Sozzi G, Lee JS, Lee JJ, Pastorino U, Pilotti S, Kurie JM, Hong WK, Lotan R. Selective suppression of nuclear retinoic acid receptor beta in non‑small cell lung cancer in vivo: Implications for lung cancer development. J Natl Cancer Inst 1997;89:624‑629.

Xu, X‑C, Clifford JL, Hong WK, Lotan R. Detection of nuclear retinoic acid receptor mRNAs in histological tissue sections using non‑radioactive in situ hybridization histochemistry. Diagn Mol Pathol 1994;3:122‑131.

Xu X‑C, Ro JY, Lee JS, Shin DM, Hong WK, Lotan R. Differential expression of nuclear retinoic acid receptors in normal, premalignant, and malignant head and neck tissues. Cancer Res 1994;54:3580‑3587.

Theoretical and Practical Considerations in the Use of Surrogate Endpoints in Cancer Prevention Research

Because cancer occurs relatively infrequently, prevention trials with cancer endpoints tend to be large, long, and costly. Studies with surrogate cancer endpoints may be smaller, shorter, and cheaper. The key question though, is whether studies with surrogate endpoints give us the right answers about cancer. Using colorectal cancer as an illustration, I will discuss two potential surrogate endpoints: epithelial cell hyperproliferation and adenoma formation. Studies with colorectal mucosal proliferation as an endpoint have been influential (e.g., the calcium chemoprevention story). Although hyperproliferation has been postulated as a relatively early event in carcinogenesis, whether it is a necessary step on the pathway to cancer is uncertain. Alternative pathways to cancer that bypass (and possibly offset) hyperproliferation are plausible, which casts doubt on using hyperproliferation as a surrogate endpoint. The most definitive way to address this uncertainty is to integrate proliferation markers in prevention trials with cancer endpoints, the very studies we were trying to avoid. Less persuasive evidence can be gleaned from observational cohort studies of proliferation versus cancer. Furthermore, even if we establish that hyperproliferation is a good surrogate endpoint in a prevention trial with treatment A (a chemopreventive agent or dietary modulation), there is no guarantee that this marker is a valid surrogate for treatment B. Similar problems are likely to attend the use of other potential molecular and cellular surrogates. In contrast, adenomatous polyp recurrence in the large bowel is widely regarded as a strong surrogate endpoint for intervention studies. Adenoma formation, a relatively late event in carcinogenesis, appears to be a necessary step in the development of most colorectal cancers (the adenoma‑carcinoma sequence). Even for adenomas, however, inferences to cancer may be problematic: (1) A small proportion of cancers may arise from areas of flat dysplasia not readily observable through the colonoscope; (2) preventive interventions may differentially affect the pathways to bad (progressing to cancer) and innocent (not progressing) adenomas, complicating the interpretation of polyp trials; and (3) because recurrent adenomas tend to be small, if an intervention operates primarily in the transition from small to large adenoma, or large adenoma to cancer, then this effect will be largely missed in polyp trials. Cervical intraepithelial neoplasia type 3 (CIN3) may be one of the strongest surrogates for cancer, but this state of severe dysplasia/carcinoma in situ is obviously very close to being invasive cervical cancer. The analogous state for large bowel might be the so‑called advanced adenoma (more than 1 cm in size, villous elements, or high‑grade dysplasia). An adenoma recurrence trial with large or advanced adenomas as endpoints, however, would have to be substantially larger and more expensive than the current generation of polyp trials. For surrogate endpoints in cancer prevention research, study cost and inferential certainty may well be directly related.

Alberts DS, Einspahr J, Ritenbaugh C, Aickin M. Rees‑McGee S, Atwood J, Emerson S, Mason‑Liddil N, Bettinger L, Patel J, Bellapravalu S, Ramanujam PS, Phelps, J, Clark L. The effect of wheat bran fiber and calcium supplementation on rectal mucosal proliferation rates in patients with resected adenomatous colorectal polyps. Cancer Epidemiol Biomarkers Prev 1997;6:161‑169.

Fleming TR, DeMets DL. Surrogate endpoints in clinical trials: Are we being misled? Ann Intern Med 1996;125:605‑613.

Freedman LS, Schatzkin A, Schiffman MH. Statistical validation of intermediate markers of precancer for use as endpoints in chemoprevention trials. J Cell Biochem 1992;16G:27‑32.

McShane LM, Schatzkin A, Wargovich MJ, Woods C, Purewal M, Freedman LS, Corle D, Burt RW, Mateski DJ, Lawson M, Lanza E, Kulldorff M, O'Brien B, Lake W, Moler J. An evaluation of rectal mucosal proliferation measure variability sources in the Polyp Prevention Trial: Can we detect informative differences among individuals' proliferation measures amid the noise? Cancer Epidemiol Biomarkers Prev 1998;7:605‑612.

Prentice RL. Surrogate endpoints in clinical trials: Definition and operational criteria. Stat Med 1989;8(4):431‑440.

Schatzkin A, Freedman LS, Dawsey SM, Lanza E. Interpreting precursor studies: What polyp trials tell us about large bowel cancer. J Natl Cancer Inst 1994;86:1053‑1057.

Schatzkin A, Freedman LS, Dorgan J, McShane L, Schiffman M, Dawsey S. Surrogate endpoints in cancer research: A critique. Cancer Epidemiol Biomarkers Prev 1996;5:947‑953.

Schatzkin A, Freedman LS, Schiffman MH, Dawsey SM. The validation of intermediate endpoints in cancer research. J Natl Cancer Inst 1990;82:1746‑1752.

Schiffman MH. New epidemiology of human papillomavirus infection and cervical neoplasia. J Natl Cancer Inst 1995;87:1345‑1347.

A variety of measurements and outcomes have been proposed as possibly predictive of the development of cancer. The approaches to assessing the predictive value of such variables are complex and increase in complexity when considering multiple variables simultaneously. Nevertheless, it may be the case that combinations of biomarkers or early clinical events (e.g., precancerous changes) will prove to be better predictors of disease than any single variable. Statistical methods for evaluating interventions based on multiple outcomes have been developed; their potential application to the setting of cancer prevention studies will be considered.

Cancer prevention studies evaluate the effectiveness of interventions to reduce cancer risk in population subgroups using the model of randomized, double‑blind clinical trials. Biomarkers are used to enrich the study population with individuals at high risk of breast cancer (i.e., susceptibility markers), measure compliance with the intervention, and serve as surrogate endpoints. A study population at high risk of disease serves both to increase the potential for benefits to study participants and to reduce sample size requirements by increasing the frequency of outcome events. The selection of population groups with germline alterations in major genes such as the breast cancer gene BRCA1 and 2, or common polymorphisms in specific P450 genes may be suitable for targeting specific interventions. Biomarkers of compliance are specific to the intervention. Most critical to trial design is the need for surrogate endpoints that may be identified well in advance of the occurrence of disease. To be useful, these measurable biological events must occur with greater frequency than clinically detectable breast cancer, occur at an earlier point in disease pathogenesis, and be strong predictors of invasive cancer occurrence. Breast parenchymal patterns, depicted on mammograms as variations in radiographic density, illustrate a potential surrogate endpoint. A high percentage relative to a low percentage of density in the breast confers at least a fourfold increased risk of breast cancer. Thus, the correlates of breast density, including hormone replacement therapy, require further evaluation and will be discussed.

Boyd NF, Lockwood G, Byng J, Tritchler D, Yaffe M. Mammographic densities and breast cancer risk. Cancer Epidemiology, Biomarkers Prevention 1998;7:1133‑1144.

Clonality is a fundamental characteristic of human cancer. One transformed cell gives rise to daughter cells that all exhibit the same genetic change that initially provided a growth advantage to the parent cell (Nowell 1976). The faithful transmission of these and other genetic changes in subsequent daughter cells has been well documented in vitro and in vivo (Sidransky et al. 1992). Correlation of these clonal genetic changes with histopathologic progression has led to the development of molecular progression models (Fearon et al. 1990). Thus, genetic markers able to detect the clonal outgrowth of neoplastic cells may be useful in the detection of primary cancers. We have demonstrated that point mutations in critical oncogenes and tumor suppressor genes can be used in polymerase chain reaction detection of rare neoplastic cells in bodily fluid. We have identified ras gene mutations in the stool of patients with colorectal cancer, p53 mutations in the urine of patients with bladder cancer, and both ras and p53 mutations in the sputum of patients with lung cancer (Sidransky 1997). Although widespread microsatellite instability is rare, we found that markers composed of larger repeat sizes (trinucleotides and tetranucleotides) were more likely to display instability in many tumor types (Mao et al. 1994). We have identified more than 90 percent of bladder cancers at initial presentation and followup with a panel of 20 carefully selected markers (Mao et al. 1996; Steiner et al. 1997). Furthermore, we have identified a clonal population of cells several months before cystoscopic detection in at least two patients. Patients without evidence of recurrence (NED) have reverted to normal by molecular analysis. Epigenetic promoter methylation can inactivate tumor suppressor genes and serves as a marker of clonal expansion. Using a panel of differentially methylated genes, including p16, we have shown that 50 percent of lung cancer patients display a methylated DNA pattern in sputum and serum (Ahrendt et al. 1999; Esteller et al. 1999). Novel automated approaches will allow rapid assessment of these tests in prospective trials and rapid integration into the clinical setting. Identification of clonal DNA alterations in clinical samples continues to evolve as a promising method of cancer detection (Ju et al. 1996).

Ahrendt SA, Chow Xu L, Yang SC, Eisenberger CF, Esteller M, Herman JG, Wu L, Decker PA, Jen J, Sidransky D. Molecular detection of tumor cells in bronchoalveolar lavage fluid from patients with early stage lung cancer. J Natl Cancer Inst, in press.

Esteller M, Sanchez‑Cespedes M, Rosell R, Sidransky D, Baylin SB, Herman JG. Detection of aberrant promoter hypermethylation of tumor suppressor genes in serum DNA from non‑smallcell lung cancer patients. Cancer Res 1999 Jan1;59:67‑70.

Fearon ER, Vogelstein B. A genetic model for colorectal tumorigenesis. Cell 1990;61:759‑767.

Ju J, Glazer AN, Mathies RA. Energy transfer primers: A new fluorescene labeling paradigm from DNA sequencing and analysis. Nat Med 1996;2:246‑249.

Mao L, Lee DJ, Tockman MS, Erozan YS, Askin F, Sidransky D. Microsatellite alterations as clonal markers in the detection of human cancer. Proc Natl Acad Sci U S A 1994;91:9871‑9875.

Mao L, Schoenberg MP, Scicchitano M, Erozan YS, Merlo A, Schwab D, Sidransky D. Molecular detection of primary bladder cancer by microsatellite analysis. Science 1996;271:659‑662.

Nowell PC. The clonal evolution of tumor cell populations. Science 1976;94:23‑28.

Sidransky D, Mikkelsan T, Cavenee W, Vogelstein B. Clonal expansion of p53 mutant cells leads to histologic progression in astrocytoma. Nature 1992;335:841‑42.

Sidransky D. Nucleic acid‑based methods for detection of cancer. Science 1997;278:1054‑1058.

Steiner G, Schoenberg MP, Linn JF, Mao L, Sidransky D. Detection of bladder cancer recurrence by microsatellite analysis of urine. Nat Med 1997;3(6):621‑624.