Ethical issues involved in the conduct of clinical trials have been a source of concern and sometimes frustration for cancer treatment investigators, particularly when it comes to deciding between randomization and historical controls.
As was expected, the confrontation between the two most outspoken investigators on opposite sides of that issue provided plenty of grist for that argument recently at the Second International Conference on the Adjuvant Therapy of Cancer in Tucson.
Charles Moertel, director of the Mayo Comprehensive Cancer Center, believes that not only is randomization ethical in most cases but it is also the only way that reliable comparisons can be made in many phase 3 and 4 studies.
Emil (Jay) Freireich, chief of Developmental Therapeutics at M.D. Anderson Hospital, believes that randomization “borders on the unethical’,’ and that historical controls can be at least’ as reliable as randomization, if not more so.
Moertel opened the debate in his presentation on adjuvant therapy of gastrointestinal tumors. Improvements in surgery and pathology make it difficult or impossible to use historical controls in colorectal carcinoma studies, Moertel contended.
“In C Dukes lesions, those with nodal involvement, cure rates are generally quoted in the 20-30% range,” Moertel said. “It has also been assumed that the results of surgical treatment have essentially been at a plateau over the past quarter century. As we consider surgical adjuvant therapy, it is important that we scrutinize such statistics and assumptions to be sure that they really apply to large bowel cancer today.
“National end result statistics would make it appear that there has, in fact, been a steady improvement in five year survival over a 30-year period-from 53 to 71% in localized disease, and from 27 to 44% in regional disease…. The contention that improved surgical technique has produced these apparently improved results may be true, but this is providing surgical pathology technique has not changed over this same period of time. If, however, the pathologist has been progressively more meticulous in his staging, a lot of tumors that were called B Dukes yesterday.are today going to be moved into the C Dukes column with a resultant apparent improvement in survivorship. As the quality of either surgery or surgical pathology or both improves in our university and community cancer centers, we must anticipate that stage for stage the reported results of surgical treatment are going to be better today than they were yesterday.
“All of this is background for the conclusion that if we apply some type of surgical adjuvant intervention to patients today, and compare results to those we achieved yesterday, we are going to have a positive study whether the surgical adjuvant treatment is effective or not. Just a cursory review of the literature will quickly show that historically controlled surgical adjuvant studies are always positive-this for literally any type of cancer.”
Freireich had his chance to respond during a panel discussion which he chaired with Moertel.
“I heard two statements about historical controls that knocked me off my chair,” Freireich said. “Knowing Dr. Moertel, I understand his bias. There were glaring deficiencies in his interpretation of historical data. Despite what Dr. Moertel said, there is no quick fix. He pointed out, in a clear, moment, that randomized trials can give incorrect results. We won’t do ourselves or science any good by flat out saying that historical controls always are worse (than treated patients in a study). That’s not only silly, it’s false.”
Freireich suggested that two of “Freireich’s
Laws” should be invoked in developing experimental
“One. You cannot replace the human frontal lobe with a checkoff sheet. There is no replacement for careful, thoughtful analysis of data. There will be and are situations where historically controlled studies are clearly superior to randomization. And there may be times when randomizing is better.
“The randomized trial, even when conducted by the best of us, including Dr. Moertel, borders on the unethical. Ed Gehan (head of the Dept. of Biomathematics at M.D. Anderson) has said that the best criteria for determining what is ethical is to find out which plan would be chosen by statisticians when they have cancer.” Freireich said that he has known five statisticians who had cancer. “They asked who was the best doctor available to treat their disease. None asked which randomized trial they should select.
“We cannot allow ourselves to become so marginally ethical that the patients reject us,” Freireich continued.
“Freireich’s Law No. 2,” he said, is that “all knowledge is historical, and for those who feel the best controls have relevance, they are the best controls.”
Moertel responded by saying, “I think you and I are a lot closer on how to manage clinical trials than you may think.” He referred to an article Freireich wrote for the New England Journal of Medicine in which he described criteria and justification for historical controls. “I found little to disagree with,” Moertel said. “The problem is that, although you write this so beautifully, you don’t practice it yourself.”
An example for the proper use of historical controls would be when there is substantial evidence for a big difference between treated and untreated patients. “I agree. But you take this same principle and use it where it is not indicated,” Moertel said. “Also, you point out that it is necessary for historical controls to be tightly aligned to the treatment group… but you don’t practice it…. When you get into 5-FU, methyl CCNU, with and without BCG, you can’t turn the clock back to neanderthal days. Those studies are not proper for historical controls.”
Moertel said he agreed that it was “nonsense” to suggest that phase 1 trials be randomized, as demanded (so far to no avail) by R.S.K. Young, FDA group leader for oncology.
Freireich commented, in comparing applause received by Moertel to his own from the approximately 1,000 physicians at the conference that “you have won round one. Round two is yet to come.”
Emil (Tom) Frei, director of the Sidney Farber Cancer Institute, also a member of the panel, said, “No patient was ever cured by statistics. The problem with this discussion is that it revolves around statistics. It is more important to do the study right.” Frei noted that many of the papers presented at the conference called for new anticancer agents. “I want to emphasize that I’m not sure we’re doing right with the agents we have, particularly 5-FU. This is an active agent. One can get steep dose response rates. Most of the doses (used in trials) are relatively trivial, intervals are long, and 5-FU can be compromised by nitrosureas. In osteosarcoma, the difference between successful and unsuccessful studies probably is due to dose rates.
“The direction to go,” Frei continued, “is to increase doses up front. You may have a problem with patients, but that probably can be counterbalanced by shortening the treatment period. The problem is to eradicate micro-metastases early. Short term, intensive treatment is the way to go. It is likely that 5-FU will work if used properly.”
Frei said he wanted to emphasize : The best surgery will be performed by skilled, experienced surgeons. “The same goes for radiotherapy. And if chemotherapy is to be done effectively, it needs to be done by someone with expertise in medical oncology.”
Bernard Fisher, chairman of the National Surgical Adjuvant Breast Project and another panel member, said he disagreed with “Freireich’s Law No. 1, that you can’t replace the human frontal lobe with a checkoff list. In some cases I know about, it would be more desirable to have a checkoff list.”
Fisher said that “as a lab investigator, I would never think of doing ail experiment without a concomitant control. In clinical trials, because of the heterogeneity of humans, I wouldn’t do less than with animals. A clinical trial is a scientific problem solving exercise. To compare, you have to have comparability. If you are going to use historical controls, the burden is on you to show proof of comparability.
“As it is with randomized trials,” Freireich said.
“Exactly. Exactly,” Fisher agreed.
John Ultmann, director of the Univ. of Chicago Cancer Research Center, presented an overview of the conference. On the issue of historical vs. randomized controls, Ultmann said that most of the conference participants seemed to agree current controls “are justified when you aren’t sure that benefits outweigh the risks, especially when you want to detect small differences.”