How to Evaluate a New Treatment for Cancer

William D. DeWys, M.D.
January 5, 2016

In a previous editorial in this journal, Dr. Wittes discussed the initial phase of human studies of a new drug, studies which are aimed at establishing the optimal dose of the drug (Your Patient & Cancer, January 1982, page 24). The next step (Phase II) is aimed at testing the effectiveness of the new drug in a small series of patients. The effectiveness is most often evaluated by monitoring regression of measurable tumor using the criteria discussed by Dr. Carter in this issue (page 53). In general, if one studies 20 patients and observes at least one response, one can be reasonably certain that the actual response rate is probably greater than 15%. Thus one should study at least 20 patients to avoid falsely calling a drug inactive when, in fact, if studied adequately, it might show at least a 15% response rate.

Duration of survival is documented in Phase II trials but cannot be the basis for evaluating a treatment because survival is often influenced more strongly by prognostic factors than by the effects of the treatment (see page 54). For example, patients with a good performance status and no weight loss who are given an ineffective treatment may live longer than patients with a poor performance status and significant weight loss who are given a moderately effective treatment. Duration of survival can be used to compare treatments only when all other factors are equivalent. The research technique used to obtain equivalence of prognostic factors is random assignment to the treatments being compared. This type of study is called a Phase III study. If prognostic factors are well-defined, one may choose to use stratified randomization, which helps to assure that these known factors are evenly balanced between treatments, while the randomization process itself balances undefined prognostic factors.

If a drug has shown acceptable toxicity in Phase I and Phase II, and has shown anti-tumor activity in the Phase II study, it becomes a candidate for a Phase III study. Phase III studies may compare the new drug to the best previous treatment, or may evaluate the addition of the new drug to a previously established treatment.

Physicians conducting these studies must, on the one hand, be sympathetic, caring physicians, but on the other hand must be scrupulously honest scientists. They must not let their desire to see patient improvement bias their judgment of the effectiveness of the treatment.

With these comments as background let us look at the paper in this issue on the use of vitamin C in cancer patients. In the introduction, Drs. Cameron and Pauling present their conclusions as though these conclusions were supported by firm evidence. In fact, they reveal that they “had reached the conclusion in 1971 that vitamin C should be of value in controlling cancer.” Note that they apparently reached this conclusion prior to treating any patients. This is contrary to a sound scientific approach, which requires that one develop a hypothesis (not a conclusion), then design and perform a study, collect and analyze data, test the hypothesis using the data collected, and finally, formulate a conclusion.

The rationale for testing vitamin C was that it might interfere with the infiltrative growth of a tumor and might increase collagen formation round a tumor. With this rationale, the experimental design should emphasize serial measurements of tumor size looking for an arrest of tumor infiltration. How were patients selected for the study group, and how were the comparison patients chosen? The vitamin C patients were cancer patients under the care of Dr. Cameron, while the comparison patients were under the care of other physicians. We are told at patients were assigned “in an essentially random way” to Dr. Cameron or to other physicians, However, this was not randomized assignment to treatment as we understand it in the research setting. One can easily envision a slanted assignment, It example, if patients with the most severe pain, or with evidence of spinal cord compression were assigned directly to a radiation therapist, rather than to Dr. Cameron.

Before the “controls,” who were chosen by this unorthodox method, can be accepted for comparison, we should see an analysis as to comparability for known prognostic factors. We are told that the groups were “matched as to age, sex, and type and location of the primary tumor.” However, we are given no data as to matching by stage, performance status, weight loss, and sites of metastases, all of which are powerful prognostic factors. Thus, we must be skeptical and be suspicious that the ascorbate-treated and control groups are really not comparable prior to treatment.

Evidence of Incomparability

We are given one piece of evidence that demonstrates lack of comparability. Dr. Cameron’s patients were started on vitamin C when they were judged “untreatable,” and their subsequent survival was compared to the survival of the control group from the time they were labeled “untreatable.” However, the criteria for labeling a patient “untreatable” are not given, and it is likely that different physicians use this label differently. One way of analyzing this possibility is to measure the time from initial diagnosis until a patient is labeled “untreatable.” If the definition were applied uniformly to comparable groups of patients, the average time should be the same for the two groups. For the patients reported on by Drs. Cameron and Pauling, we can estimate the time from diagnosis to being labeled “untreatable” by using the data in their Table 1 (see below) and subtracting column D from B for the controls, and Column C from A for the ascorbate-treated.



Taking colon cancer as an example, we see that Dr. Cameron’s cases were followed for only 106 days (458 minus 352) before being called “untreatable,” while the controls were followed for 283 days before being labeled “untreatable.” Thus, at the time they were called “untreatable” Cameron’s patients would be expected to have less advanced disease and a better remaining life expectancy than would the control patients. Thus, their survival would have been expected to be better regardless of what treatment they were given.

Looking at the survival curves (Figure 1 in Cameron and Pauling) reinforces our concern about lack of comparability of the groups from the time of their being labeled “untreatable.” Note that in each control group 20% to 40% of patients died with in a few days of being labeled “untreatable.” These patients truly were “untreatable.” In contrast, in several of Dr. Cameron’s groups there were no early deaths, suggesting his patients had less advanced disease when labeled “untreatable.” For example, in ovarian cancer, all of Cameron’s patients survived more than 90 days.

We noted early in this paper that, based on the proposed rationale, it would be of interest to record serial tumor measurements. Although Drs. Cameron and Pauling state that they observed objective evidence of benefit, it is disappointing that data on tumor measurements, which might support their claim, are not presented.

From this analysis I conclude that the value, if any, of vitamin C in cancer remains unproven. Most, if not all, of the difference between the vitamin C and the comparison group could be due to patient selection and differences in use of the label “untreatable.” Even Drs. Cameron and Pauling seem to concede this when they acknowledge the need for a controlled trial.

Referring back to the strategy discussed at the beginning of this article, one would first like to see a dose-seeking study (Phase I). Although dosage of vitamin C has not been studied systematically, it is likely that the recommended dose in cancer patients would be somewhere near 10 grams a day. Such a dose causes stomach upset, gas formation, and diarrhea in an appreciable number of patients, so that most patients could not tolerate doses much larger than this. One would then like to see a study documenting a tumor- arresting effect. Data on this may exist in the records of Dr.
Cameron, and we would encourage that this be analyzed using the criteria discussed by Dr. Carter (p. 53). However, vitamin C has received so much publicity that further studies may be warranted to clarify the claims made for this agent. A randomized controlled clinical trial is currently in progress at the Mayo Clinic. In this study, patients who have colon cancer with asymptomatic metastases are randomly assigned to vitamin C or placebo. When these patients progress to symptomatic disease they receive chemotherapy. The efficacy will be evaluated by measuring time to progression increase in measurable tumor) and by measurement of survival. Until the results of this trial are
known, vitamin C must be considered to be an unproven agent for the treatment of cancer patients.

This article appeared originally in the May 1982 issue of Your Patient & Cancer, a medical journal that is no longer published. At the time it was written, Dr. DeWys headed the Clinical Investigations Branch of the Cancer Therapy Program at the National Cancer Institute.