Proliferation Happens

The use of surrogate end points in trials of colonic cancer chemoprevention has great appeal. Potential agents could be tested more quickly using far fewer patients than would be needed for trials using the development of colonic cancer or adenomas as end points. In this issue of THE JOURNAL, Holt and colleagues¹ report that increased dietary intake of low-fat dairy foods can alter putative intermediate biomarkers of colonic cancer risk. This study raises 2 important questions about the use of surrogate end points for colonic cancer: (1) are there any surrogate end points, short of adenoma formation, that reliably reflect colonic cancer risk and (2) is there currently a defined role for surrogate end point studies in decision making about which agents to test in colonic cancer chemoprevention trials? The answer to both of these questions probably is no.

An ideal surrogate end point biomarker for colonic cancer would be causally linked to the process of colonic carcinogenesis, would be easily and reliably assayed, would be modulated by chemopreventive agents, and its modulation would correlate with a decreased risk of cancer or adenoma.² Colonic epithelial proliferative indices have been extensively studied and are the most widely used of the putative surrogate end point biomarkers of colonic cancer risk. The normal-appearing colonic mucosa of patients with colonic neoplasia appears to have an abnormal proliferative pattern, characterized by an expansion of the proliferative compartment with or without an increase in the proliferative rate.³ However, the fundamental relationship of proliferation to carcinogenesis remains unclear,⁴ the discriminant value of proliferative measurements for identifying high-risk subjects appears to be low,⁵ and the reliability of proliferative measurements in the colon is uncertain.^{5 - 10} All of the commonly used proliferative assays are technically demanding, and they do not seem to correlate highly with each other.^{8 - 10} A recent large study found so little interindividual variation in proliferation and so many other sources of variance (ie, clinical center, month of biopsy specimen collection, intraindividual variability over time, interbiopsy specimen variability, and assay variability) that the investigators concluded it would be unlikely that the measures of proliferation could ever provide meaningful information about individual risk for colonic cancer.⁸

The operational difficulties with the measurement of colonic proliferative markers in clinical trials are illustrated in the article by Holt et al.¹ The proliferative parameters measured were not stable over time, so that many of the statistically significant differences between the 2 groups were due as much to changes in proliferative parameters in the control group as they were to changes in the treatment group. A similar drift in proliferation in the control population was seen in the only other large human study, which suggested that calcium supplements down-regulate colonocyte proliferation.¹¹ The timing of proliferative changes also were not consistent during the study by Holt et al.¹ A significant difference in the whole crypt labeling index was found at 12 months but not at the 6-month time point, whereas the opposite pattern was seen for the labeling index at the top of the crypt. These types of variability suggest that if there is a treatment effect from low-fat dairy food, it is either too small or too variable to be consistently detected above background variations of colonocyte proliferation.

Colonic carcinogenesis is thought to be a multi-step process driven by sequential rounds of mutation followed by expansion of mutant clones that have acquired a growth advantage.¹² One or more of the mutational events in the process of colonic carcinogenesis could cause hyperproliferation in individual or clonal clusters of crypts. Only if such clonal events occurred frequently would they be detectable in random biopsy specimens of the normal-appearing colonic mucosa. The most popular mechanistic hypothesis linking hyperproliferation to colonic cancer is that sustained hyperproliferation of the colonic mucosa creates a fertile field for the mutagenic events that drive carcinogenesis. Although pathologic states such as inflammation and cancer are associated with increased proliferation, epithelial proliferation in the colon is a normal, physiologically regulated process. Dietary modulation of proliferation typically affects the proliferative rate, whereas age can affect both the rate and the size of the proliferative compartment.^{13 - 14} Small changes in proliferation within normal mucosa, therefore, cannot necessarily be considered pathological events. In short, "proliferation happens."

What is the role of surrogate end point trials in decision making for proceeding with chemoprevention trials using adenomas or cancer as end points? Baron and colleagues¹⁵ assessed proliferation as an intermediate end point in their colonic adenoma chemoprevention trial of calcium supplementation and found no effect of calcium supplements on colonic epithelial proliferation. Had the investigators used that surrogate end point to decide about conducting their trial, they would have not proceeded with the adenoma prevention study.¹⁶ However, they reasoned that calcium could affect cancer risk through pathways unrelated to proliferation or that it could alter proliferation in ways unrelated to the measures they used. They completed their trial and found that calcium supplementation indeed reduced the rate of new adenoma formation.¹⁶ Similarly, if investigators in the National Polyp Prevention Trial had used intermediate proliferative markers for clinical trial decision making, they would have not continued their study in light of the findings of their proliferative analyses.⁸ A study of intermediate markers with negative results therefore seems not to be helpful in decision making about proceeding with a trial otherwise justifiable using adenomas or cancer as end points. A positive finding is only a little more helpful in that it provides encouragement to proceed to a larger study with adenomas or cancer as end points, but it does not preclude the need for the further study. Intermediate end point biomarker studies, in and of themselves, currently have little value in decision making about testing chemopreventive agents in larger trials, as they are not predictive enough to preclude further study and any positive findings still would need to be validated against end points of adenomas or cancers. Formal validation studies in such trials are needed to advance intermediate endpoint biomarker analysis to the level of utility and clinical trials decision making.

The important practical question is how proliferation might predict colorectal cancer risk. If the small effects on proliferation observed in the study by Holt et al¹ are assumed to be real, there are 3 possible implications for colorectal cancer: (1) hyper-proliferation occurs early enough in the causal chain to set an early trajectory that results in a larger effect on colorectal cancer risk, (2) the ultimate effect on colorectal cancer is of about the same magnitude as the effect on proliferation, or (3) the ultimate effect on colorectal cancer is even smaller than the proliferation effect because of dilution by other factors closer to the causal pathway to colorectal cancer. Epidemiologic research suggests the third possibility is most likely. As mentioned by Holt et al,¹ the first few studies of the relationship between dietary calcium and colorectal neoplasia, conducted in the mid-1980s, supported the hypothesis that dietary calcium reduces colorectal cancer risk.¹⁷ However, later studies, using better prospective methods, did not support the initial findings.^{17 - 18} Among the 5 prospective studies published in the past 10 years, there is a hint of an inverse association between dietary calcium and colorectal cancer, but at best the findings indicate only a 5% to 25% reduction in colorectal cancer risk among those persons in the highest categories of calcium intake.^{17 - 18} In contrast to the weak associations observed between dietary calcium and colorectal neoplasia, epidemiologic studies have shown stronger and more consistent patterns of associations with other modifiable behavioral factors, including dietary fat, fruits and vegetables, smoking, aspirin, and physical activity.^{19 - 20}

There are many reasons to include ample amounts of calcium in the diet, especially for women at risk for osteoporosis. However, there may be a downside to increased calcium intake for older men, as high-calcium diets can suppress the levels of circulating vitamin D and perhaps thereby increase the risk for prostatic cancer.²¹ Therefore, the risks and benefits of any dietary recommendation must be fully considered, even for foods usually viewed as wholesome. Epidemiologic evidence, although imperfect because of uncertainties of dietary assessment and confounding by choice, are at present more informative about what we should eat to prevent colorectal cancer than are short-term studies of physiologic biomarkers of unknown or unproven importance.