A news article published online in Clinical Oncology News in February suggests that oral paclitaxel preceded by an inhibitor of p-glycoprotein is an effective and attractive regimen for metastatic breast cancer (https://bit.ly/2KypPot). The article—based on the results of a study presented at the 2019 San Antonio Breast Cancer Symposium (SABCS)—reported higher response rates, longer survival, less neuropathy, and less hair loss with oral compared with intravenous paclitaxel (abstract GS6-01).

At least 4 reasons exist for clinicians and federal agencies not to embrace these results: 

1. The description of the treatment regimen in the SABCS oral presentation was, at best, incomplete and misleading.
2. The design of the treatment protocol was biased in favor of the experimental treatment, since inclusion in the primary analysis (called modified intention to treat) required only 1 day of IV paclitaxel but 16 days of oral paclitaxel. Study participants who do poorly in the first 2 weeks of the experimental arm would not be included in the primary analysis, making the experimental arm look better than it actually was, since those on IV paclitaxel were not similarly excluded! This was a potentially major fault, but probably did not affect the results of this study. 
3. An inordinately large number of patients were lost to follow-up and censored on the analyses of overall survival and progression-free survival (PFS). The validity of conclusions drawn from these analyses is questionable.
4. The study was substantially—if not predominantly—conducted in Central America and the Caribbean. This region is poor and underdeveloped, and without a substantial history of conducting clinical cancer research. An underfunded pharmaceutical start-up might choose such an area to save money, but it is concerning because manipulation of data or data analysis is more likely to go unchallenged in such a setting.

Aside from the clear-cut deliberate failure to accurately describe the study regimen, I know of no evidence of deception or fraud in the study presentation. That does not mean I am not worried about it.

How Was Oral Paclitaxel Actually Given?
Tipped off by a reporter that the SABCS presentation had inadequately described the drug administration schedule, during the question and answer (Q&A) period I asked the presenter, Gerardo Umanzor, MD, a medical oncologist at Hospital del Valle, in San Pedro Sula, Honduras, to tell attendees the details of drug administration and fasting used in the study. He described only 1 hour of fasting between encequidar and paclitaxel. He also said drug administration was a minor issue.

At the time, standing at the microphone, I had no clear-cut information that this was incorrect or incomplete, but I was able to review the study protocol with one of the investigators about a week after SABCS. Oral paclitaxel was given 3 days in a row every week, without a break. For analysis purposes, a cycle of therapy was defined as 3 weeks. In effect, the treatment was continuous without breaks except for toxicity or refusal of therapy. Each dose of paclitaxel—an average of eleven 30-mg capsules—was preceded by a 7-hour fast. One hour before the planned oral paclitaxel, the patient swallowed a 15-mg tablet of encequidar (the p-glycoprotein inhibitor that prevents the intestinal lining from pumping the paclitaxel back into the intestinal cavity and keeping it from entering the body). The patient was told to fast for 4 hours after paclitaxel ingestion. Thus, the protocol stipulated 11 hours of fasting 3 days per week for the duration of therapy.

Obviously, the small pharma start-up called Athenex behind encequidar knows that an 11-hour fast 3 days per week, every week, makes the encequidar-paclitaxel regimen unattractive. Investigators have since told me that the company is trying to validate an abbreviated fasting period in a small pharmacokinetic study. The price of Athenex stock moved up and down vigorously around the time of the San Antonio presentation. I believe that Athenex had a lot at stake in the SABCS presentation and the unmentioned 11-hour fast.

Subsequent correspondence with Dr Umanzor confirmed the fasting and drug administration schedule I have described. Dr Umanzor attributes the lack of completeness in his reply to my question about fasting and drug schedule to limited time. It takes 14 seconds to say, “The protocol required fasting for 6 hours before encequidar, 1 hour after encequidar and before oral paclitaxel, and then 4 hours after the oral paclitaxel.“ It would have taken another 10 seconds to say, “Many patients finished the entire process at home at 10 AM, and had little difficulty with the drug administration.” The latter is a paraphrase from Dr Umanzor’s email to me on January 14, 2020.

The sin here is not the 11-hour fast. As a clinical oncologist, I have been giving IV paclitaxel for 22 years. Despite the 11-hour fast, if oral paclitaxel with a p-glycoprotein antagonist produces less neuropathy, fewer allergic reactions and less hair loss, and requires no steroid premedication, I would offer it to my patients, especially if it works better, and even if it requires prophylactic antiemetics and antidiarrheals. 

The sin here is the duplicity. If the presenter and the drug company were obscuring the schedule of drug administration, can we trust them to have accurately reported response and toxicity?

How Important Is the Modification of the Intention-to-Treat Analysis?
As pointed out during the Q&A period by Alan Coates, MD, a statistician and visiting scientist at the Garvan Institute of Medical Research, in Sydney, Australia, and a retired medical oncologist, the 2-week period of required drug compliance in the oral paclitaxel group introduces “guarantee time bias” in time to progression and survival analyses; it also increases apparent response rate and decreases toxicity by allowing exclusion of patients removed early because they are not doing well. No such removal occurred in the IV paclitaxel arm, for which 1 dose of drug was sufficient to be counted in the study analysis.

A careful analysis of the presentation slides suggests that this modification probably had little effect on the results reported. It seems that 10.9% of the patients given IV paclitaxel and 11.3% of those given oral paclitaxel were excluded in the modified intention-to-treat analysis. If many patients assigned to oral paclitaxel were excluded for non-compliance in the first 3 weeks, the rates of exclusion from the modified intention-to-treat analysis would have been much higher for this group. The rate of exclusion for never starting treatment was 1.45% for the 137 patients given IV therapy and 0.37% for the 265 patients given oral therapy. The remaining reasons for exclusion from the modified analysis were the absence of disease evaluable for response on CT scans (both groups) and failure to complete 15 days of treatment (oral paclitaxel group only). If the investigators tell us how many patients from each group were excluded for lack of evaluable disease on CT scan, then we can calculate the number experimental group patients excluded for early noncompliance. If it is only 1 or 2 patients, then it is unlikely to have had a major effect on the results reported.

High Rates of Censoring in PFS and Overall Survival
Based on the data presented at SABCS, PFS did not appear to be significantly prolonged by oral paclitaxel; the median went from 8.3 to 9.3 months. However, 40.4% of the patients given oral paclitaxel and 36.8% of those given IV paclitaxel were censored because they were lost to follow-up while still responding. The reasons for this were not discussed in the presentation. In his later correspondence, Dr Umanzor noted that the investigators did not plan to study duration of response but that responses to oral paclitaxel were longer than anticipated.

It may be that many patients refused further therapy and returned to inaccessible villages. In a study in which the primary end point was response by independent radiological review, there may have been no funding for follow-up CT scans, or even examinations and evaluations to determine dates of relapse. Regardless, the high rates of loss to follow-up make the PFS analysis suspect.

The analysis of overall survival also is somewhat suspect as presented. The modified intention-to-treat analysis shows oral paclitaxel improved median overall survival from 16.9 to 27.9 months (hazard ratio [HR], 0.680; P=0.035 by log-rank test). The medians were similar when the entire intention-to-treat groups were analyzed, but the HR was only 0.762, and the result was not significant (P=0.114 by log-rank test). At the time of analysis, 31.1% of the oral group and 41.6 % of the IV group had died, about 18% of each group had been lost to follow-up, and 51% of the oral group and 40% of the IV group were still alive. 
In such a relatively small study, it should not be difficult to obtain dates of death for all or nearly all patients. Before the FDA and other reviewing agencies allow any improved survival claim for oral paclitaxel, and before peer-reviewed publication, the percentage of patients lost to follow-up should be reduced to the low single digits. This is important, since survival advantages are hard to come by in metastatic breast cancer, and the existence of one would justify many inconveniences for oral paclitaxel.

Should We Trust Data Coming Predominantly From Central America?
Dr Umanzor in his correspondence to me stated that most sites “have participated in many international trials“ and that the involved oncologists had “a lot of experience in clinical research.” Dr Umanzor, who was not the study chairman, noted that he did not know how the sites and investigators were chosen. He informed me that an independent US company (Intrinsic Imaging) analyzed CT scans for response.

I doubt it will prove productive to further investigate the choice of study sites and participating oncologists. Instead, I suggest an audit of local data from major contributing sites compared with study data with regard to response, toxicity, and, if the censoring for loss to follow-up is reduced, overall survival. Such audits are routine in cooperative groups in the United States. The cost of the audit will be trivial compared with the likely drug costs involved in treating just a few patients with oral paclitaxel and encequidar. Whether the FDA can pay for the audit, or whether it should require Athenex to pay for the audit, is a matter for negotiation. If Athenex balks at paying for the audit, the FDA should point out that Athenex should have been complete and honest in its presentation of the study protocol in San Antonio, and that failure to accurately report the study protocol has consequences.

Dr Vogl reported no relevant disclosures.