Zubrod had his own schemes to keep leukemia research out of trouble. As new drugs, combinations, and trials proliferated, Zubrod worried that institutions would be caught at cross-purposes, squabbling over patients and protocols when they should really be battling cancer. Burchenal in New York, Farber in Boston, James Holland at Roswell Park, and the two Emils at the NCI were all chomping at the bit to launch clinical trials. And since ALL was a rare disease, every patient was a precious resource for a leukemia trial. To avert conflicts317, Zubrod proposed that a “consortium” of researchers be created to share patients, trials, data, and knowledge.
The proposal changed the field. “Zubrod’s cooperative group model galvanized cancer medicine,” Robert Mayer (who would later become the chair of one of these groups) recalls. “For the first time,318 an academic oncologist felt as if he had a community. The cancer doctor was not the outcast anymore, not the man who prescribed poisons from some underground chamber in the hospital.” The first group meeting, chaired by Farber, was a resounding success. The researchers agreed to proceed with a series of common trials, called protocols, as soon as possible.
Zubrod next set about organizing the process by which trials could be run. Cancer trials, he argued, had thus far been embarrassingly chaotic and disorganized. Oncologists needed to emulate the best trials in medicine. And to learn how to run objective, unbiased, state-of-the-art clinical trials, they would need to study the history of the development of antibiotics.
In the 1940s, as new antibiotics had begun to appear on the horizon, physicians had encountered an important quandary: how might one objectively test the efficacy of any novel drug? At the Medical Research Council in Britain, the question had taken on a particularly urgent and rancorous note. The discovery of streptomycin, a new antimicrobial drug in the early forties, had set off a flurry of optimism that tuberculosis could be cured. Streptomycin killed tuberculosis-causing mycobacteria in petri dishes, but its efficacy in humans was unknown. The drug was in critically short supply, with doctors parrying to use even a few milligrams of it to treat a variety of other infections. To ration streptomycin, an objective experiment to determine its efficacy in human tuberculosis was needed.
But what sort of experiment? An English statistician named Bradford Hill (a former victim of TB himself) proposed an extraordinary solution. Hill began by recognizing that doctors, of all people, could not be entrusted to perform such an experiment without inherent biases. Every biological experiment requires a “control” arm—untreated subjects against whom the efficacy of a treatment can be judged. But left to their own devices, doctors were inevitably likely (even if unconsciously so) to select certain types of patients upfront, then judge the effects of a drug on this highly skewed population using subjective criteria, piling bias on top of bias.
Hill’s319 proposed solution was to remove such biases by randomly assigning patients to treatment with streptomycin versus a placebo. By “randomizing” patients to each arm, any doctors’ biases in patient assignment would be dispelled. Neutrality would be enforced—and thus a hypothesis could be strictly tested.
Hill’s randomized trial was a success. The streptomycin arm of the trial clearly showed an improved response over the placebo arm, enshrining the antibiotic as a new anti-TB drug. But perhaps more important, it was Hill’s methodological invention that was permanently enshrined. For medical scientists, the randomized trial became the most stringent means to evaluate the efficacy of any intervention in the most unbiased manner.
Zubrod was inspired by these early antimicrobial trials. He had used these principles in the late 1940s to test antimalarials, and he proposed using them to lay down the principles by which the NCI would test its new protocols. The NCI’s trials would be systematic: every trial would test a crucial piece of logic or hypothesis and produce yes and no answers. The trials would be sequential: the lessons of one trial would lead to the next and so forth—a relentless march of progress until leukemia had been cured. The trials would be objective, randomized if possible, with clear, unbiased criteria to assign patients and measure responses.
Trial methodology was not the only powerful lesson that Zubrod, Frei, and Freireich learned from the antimicrobial world. “The analogy of drug resistance320 to antibiotics was given deep thought,” Freireich remembered. As Farber and Burchenal had discovered to their chagrin in Boston and New York, leukemia treated with a single drug would inevitably grow resistant to the drug, resulting in the flickering, transient responses followed by the devastating relapses.
The situation was reminiscent of TB. Like cancer cells, mycobacteria—the germs that cause tuberculosis—also became resistant to antibiotics if the drugs were used singly. Bacteria that survived a single-drug regimen divided, mutated, and acquired drug resistance, thus making that original drug useless. To thwart this resistance, doctors treating TB had used a blitzkrieg of antibiotics—two or three used together like a dense pharmaceutical blanket meant to smother all cell division and stave off bacterial resistance, thus extinguishing the infection as definitively as possible.
But could two or three drugs be tested simultaneously against cancer—or would the toxicities be so forbidding that they would instantly kill patients? As Freireich, Frei, and Zubrod studied the growing list of antileukemia drugs, the notion of combining drugs emerged with growing clarity: toxicities notwithstanding, annihilating leukemia might involve using a combination of two or more drugs.
The first protocol was launched321 to test different doses of Farber’s methotrexate combined with Burchenal’s 6-MP, the two most active antileukemia drugs. Three hospitals agreed to join: the NCI, Roswell Park, and the Children’s Hospital in Buffalo, New York. The aims of the trial were kept intentionally simple. One group would be treated with intensive methotrexate dosing, while the other group would be treated with milder and less intensive dosing. Eighty-four patients enrolled. On arrival day, parents of the children were handed white envelopes with the randomized assignment sealed inside.
Despite the multiple centers and the many egos involved, the trial ran surprisingly smoothly. Toxicities multiplied; the two-drug regimen was barely tolerable. But the intensive group fared better, with longer and more durable responses. The regimen, though, was far from a cure: even the intensively treated children soon relapsed and died by the end of one year.
Protocol I set an important precedent. Zubrod’s and Farber’s cherished model of a cancer cooperative group was finally in action. Dozens of doctors, nurses, and patients in three independent hospitals had yoked themselves to follow a single formula to treat a group of patients—and each one, suspending its own idiosyncrasies, had followed the instructions perfectly. “This work is one of the first comparative studies322 in the chemotherapy of malignant neoplastic disease,” Frei noted. In a world of ad hoc, often desperate strategies, conformity had finally come to cancer.
In the winter of 1957, the leukemia group launched yet another modification to the first experiment. This time, one group received a combined regimen, while the other two groups were given one drug each. And with the question even more starkly demarcated, the pattern of responses was even clearer. Given alone, either of the drugs performed poorly, with a response rate between 15 and 20 percent. But when methotrexate and 6-MP were administered together, the remission rate jumped to 45 percent.
The next chemotherapy protocol, launched just two years later in 1959, ventured into even riskier territory. Patients were treated with two drugs to send them into complete remission. Then half the group received several months of additional drugs, while the other group was given a placebo. Once again, the pattern was consistent. The more aggressively treated group had longer and more durable