Michael Marco reports back from FDA-land, where there is always a story to tell, and usually more than one way to tell it. He dedicates this his latest reportage to the memory of fellow treatment warrior Jesse Dobson, who died last fall, on September 23.
When a fellow treatment activist, one I've admired and respected since day one, said to me on his death bed, "Get Doxil approved," he probably never imagined what an arduous task that would be. However, on February 14th, yes, St. Valentine's Day, the FDA's Oncologic Drug Advisory Committee (ODAC) recommended that Liposome Technology Inc.'s DOX-SL (liposome encapsulated doxorubicin, a.k.a. Adriamycin) be granted accelerated approval by a vote of 10-to-0, with one abstention.
How this vote was decided warrants an explanation, and for some involved, a therapist's davenport. In retrospect, I feel fortunate to have witnessed this hearing first-hand. Officially, I was serving on the ODAC panel as a guest/ad hoc expert/consumer representative. That in itself was somewhat of an historical occasion because never before had the FDA's Division of Oncologic Drugs allowed a treatment activist to be a member of the ODAC panel. In the morning, I was allowed to sit "at the table" and in the afternoon, when Zeneca's breast cancer NDA was to be heard, a treatment activist from the breast cancer community took my seat. We, however, were not allowed a vote. Hopefully this will change, and a treatment activist/patient representative seat "at the table" will continue with all future ODAC hearings.
It must first be mentioned that the DOX-SL NDA was problematic from the get-go. The application contained data on 77 patients from DOX-SL trial #30-12, a compassionate use study for AIDS patients with KS who had failed or were intolerant to standard combination chemotherapy. #30-12 was also the salvage trial for patients who had failed in the control arms of #30-10 (DOX-SL vs. ABV) or #30-11 (DOX-SL vs. BV). The FDA reviewer had hoped that the patients in the NDA would be those failures from the other two controlled studies, however, only 3 of the 77 really were. Thus, the overwhelming majority of these patients were essentially on an open-label protocol. The median CD4
count for these patients was 13, and 78% of them were deemed as "poor risk" KS patients according to ACTG criteria. All of the patients had cutaneous lesions; 40% had oral lesions; 26% had pulmonary involvement.
Eligibility for Evaluation
The FDA's medical reviewer, Anthony Murgo, MD, did an intense job at analyzing the sponsor's data. Immediately, he decided that only 63 out of the 77 patients were truly "refractory" to prior chemotherapy. This flew in the face of an earlier assessment by the investigator team of Susan Krown, Ronald Mitsuyasu and Donald Northfelt who, on behalf of LTI, had unanimously decided that all 77 met the strictest of refractory/intolerant definitions.
Murgo then threw out 35 additional patients (45%) because they did not meet all the protocol entry criteria. This was both fair and unfair and has led to a continuing controversy. 22 of the 77 (29%) patients did not meet the 4 week wash-out period and were granted exemptions to get DOX-SL sooner. There are, of course, two angles here. The principal investigators and LTI were being kind enough to let these rapidly progressing KS patients on drug earlier; however, the protocol said no prior cytotoxic therapy for 28 days (it's stated in black and white), and it was clearly not followed word for word. So, here you have a situation where compassion and good science are pitted against protocol violations and messy data. This is unfortunately a rule industry must adhere to: if you write it in your protocol, you'd better follow it!
There was also confusion over how a patient's response to therapy would be assessed. Originally, LTI was using a ranged counting system as its primary endpoint. This method of counting the number of lesions by ranges (less than 10, 10-25, etc.) was put forward by Krown et al. of the ACTG in their 1989 Journal of Clinical Oncology article. Later, LTI exchanged their primary endpoint with their secondary endpoint: which uses 5 marked lesions as representative of the entire body. This was done because they were unaware that the Krown criteria had been revamped in 1992 and because the FDA told them that the range counting approach was not "specific" enough.
Susan Krown, in a smart-yet-powerful black wool suit with a Julie Christie (a la "Shampoo") hairdo explained to ODAC how we used to evaluate lesions for response in the past versus how we do it today, adding that there is little if no data on refractory KS patients. But the ODAC panel had apparently already taken in its critical mass of explanations for the day and, sporting dreary gazes as if they'd just reviewed a transcript of the Yalta conference, were clearly too baffled to absorb any more.
If this sounds confusing, you should have seen the look on the faces of the ODAC members who were struggling to digest all of this during the presentations. The confusion between what patients would be considered evaluable or whether an intent-to-treat analysis should be used and compounding that with deciding on whether to use an out-dated ACTG criteria system or indicator lesions as a marker for response was nothing but ugly.
Sustained Responses... LTI screwed up initially by not stating in its protocol that patients had to sustain their complete or partial response (complete response or partial response) for a minimum of four weeks. This is the basic of basics, and all oncologists learn this in "day one" of medical school. So, the FDA (before the hearing) asked LTI to re-analyze its data and only select those patients who had held their response for at least 21 days. Thus, for indicator/marker lesions, the data went from 68% partial response to 34% partial response; and with using the old ACTG criteria, the partial responses went from 56% to 43%.
For assessing the indicator/marker lesions with an intent-to-treat analysis, 26% achieved a partial response; and when only looking at those patients deemed evaluable for response, 47.6% achieved a partial response. When using the old ACTG criteria, the FDA found that 11.7% achieved a partial response on an intent-to-treat analysis, and 26.5% of only the evaluable patients achieved a partial response.
There is not a great deal of discrepancy, considering the margin of error, in what the FDA found vis-a-vis what LTI found after selecting out those patients who held their response for a minimum of 21 days. Where the discrepancy lay was in the two party's assessment of clinical benefits, or so-called "soft endpoints."
These clinical benefits were based on four areas: reduction in KS-related pain, reduction in lesion-associated edema, changes in lesion color, and complete flattening of previously raised lesions. In all these four categories combined, LTI said that 60 to 70% of all patients improved. This is not what Murgo found. He said: "In a small minority of patients (8%), treatment with DOX-SL resulted in what appeared to be clinically meaningful patient benefit that was substantial and reasonably well-documented. In the remainder of the patients, there was little or no documented evidence of a clinically meaningful benefit."
This discrepancy was never fully resolved. How could LTI have such inflated numbers, but so how could the FDA's assessment be so low? The only thing that saved LTI and DOX-SL was the fact that three patients with KS who had been on DOX-SL testified to the fact that the drug had seemingly worked for them and that they were living, and walking, proof of the "meaningful patient benefit." A few ODAC members remembered these patientsÕ early morning testimony and said that they so empathized with these patients that they were inclined to accept their word over Murgo's assessment. Neverthe-less, many came away from the hearing saluting Murgo. He took LTI's Cadillac data and turned it into a second-hand Volkswagon. It still ran, and it'd get you from Chelsea to Hackensack, if you really wanted to go there, that is.
In Kaposi's sarcoma as in many other illnesses, so-called clinical benefit is sometimes as important as if a patient achieves a complete or partial response. Some fifteen years into the epidemic, the quantification of clinical benefit remains something that is difficult to monitor and record. Much of this is the fault of present KS clinical trial design and the fact that pain and edema are subjective (yet so very real), and cannot readily be recorded on a case report form which is then analyzed by computer.
How the Voting Turned
One of the biggest debates centered around the first question to be voted on which asked if this data set was from an adequate well-controlled trial. Donald Abrams, the voting KS expert/guest said that in no way was this an adequate or well-controlled study and in fact that the data was so sloppy he could hardly consider it a real trial. (He did not tell the panel, nor did I bring it up, that he just happened to be the second author on Paul Volberding's 1985 sloppy vinblastine study, (Ann Intern Med 103:335-38, 1985) which never even specified the criteria used for assessing the "significant" changes of lesion nodularity and pigmentation that would actually warrant a partial response. Nor, did he explain to the ODAC members (who readily admitted they knew little if anything about KS and AIDS) how difficult it was to treat KS patients in the realm of exacerbated HIV disease.
It was, surprisingly, the righteous and dogmatically conservative Robert Temple (the smartest of the bunch) who made a case that this data was, in fact, unwittingly well-controlled because these patients had failed everything else! They could use getting-worse-due-to-no-therapy as their own historical control.
The next two votes, ("Is their some hint of efficacy?" and "Is it relatively safe?") passed as "yes." The fourth vote on full approval was the most difficult and awkward. All of the discussion was heading downhill toward defeat. Don Abrams redeemed himself here with a moment's sanity when he asked, "If we vote 'no' on full approval, can we vote again on accelerated approval?"
It was charming to see many of the ODAC members look relieved when it dawned upon them that they could feel good about voting against full approval but still save this drug from extinction by giving a green light to accelerated approval. Ironically, most of the ODAC members didn't even understand what accelerated approval was. Nor had they used it and benefitted from it in the way Abrams had in the past. So after the vote was 0-to-10 not to sanction full approval, ODAC voted 10-to-0 in favor of accelerated approval.
The FDA and LTI will now lock horns attempting to draw up, and agree upon, plans for post-marketing trials. The real trick will be to see how they can equate "meaningful clinical benefit" with efficacy. Unexpected complications are not all that difficult to imagine, as I goaded Temple just before my exit, "You better get on top of LTI quickly, because if you let them dawdle for a few years before they have to come back, there might not be an FDA."