How badly should a drug perform to not get approved?

The advantage with the addition of bevacizumab to radiotherapy and temozolomide for newly diagnosed glioblastoma was studied in two phase 3 trials in 2014 - the AVAglio study and RTOG0825 study - both of which showed that bevacizumab didn't improve survival. Now, in a new phase 3 trial, addition of bevacizumab to lomustine versus lomustine alone was compared for recurrent glioblastoma and demonstrated, again, no OS advantage.

The study authors conclude in the abstract: "Despite somewhat prolonged progression-free survival, treatment with lomustine plus bevacizumab did not confer a survival advantage over treatment with lomustine alone in patients with progressive glioblastoma". A very reasonable conclusion for an industry funded study. Bevacizumab had received an accelerated approval in this setting based on improved response rates in a prior phase 2 single arm trial. Now this properly powered phase 3 randomized trial showed no survival advantage with adding bevacizumab-the HR for OS is only 0.95 (95% CI 0.74 to 1.21).

This trial is thus an example of why we need proper phase 3 RCTs to make final approval decisions. Anyway, you'd assume that the earlier accelerated approval based on response rates for bevacizumab in this setting should be revoked, right? Here's the surprise of the year: FDA Grants Bevacizumab Full Approval for Glioblastoma. Earlier, FDA approved sunitinib for adjuvant treatment of resected high risk renal cell cancer despite one trial showing it didn't improve either DFS or OS and another showing it improved only DFS but not OS. So my question is: how badly should a drug perform to actually not get approved by this FDA? I reckon if you can show that your drug is non-inferior to placebo with the non-inferiority margin of your choice, you can still get your drug approved.

ADJUVANT trial

In a trial of gefitinib versus chemotherapy in the adjuvant treatment of resected stage II-IIIA (N1-N2) EGFR positive non small cell lung cancer, gefitinib significantly improved disease-free survival (DFS). The overall survival data are immature. However, a look at the DFS curves reveals something interesting: both the curves touch the X-axis by 48 months. In other words, all patients relapse at year 4. The purpose of adjuvant treatment is to ensure cure rates are higher but in this trial no patient got cured. The trial's name is ADJUVANT, but the question is: was that actually an adjuvant treatment?

No SOLACE with SOLE

Postmenopausal women with hormone positive breast cancer who had completed 4-6 years of adjuvant endocrine therapy were randomized in the SOLE trial to either continuous or intermittent letrozole for an additional 5 years.  Intermittent letrozole didn't decrease adverse events compared with continuous, but seems probably safe if patients desire treatment holidays.

In another trial among HER2 positive breast cancer patients, the no-chemotherapy regimen of pertuzumab plus TDM1 achieved significantly lower rates of pathological complete response when used as neoadjuvant therapy compared with chemotherapy plus trastuzumab plus pertuzumab, which remains the standard for this setting.

ESMO 2017 studies

Durvalumab and osimertinib both showed substantially improved PFS versus placebo and the standard EGFR TKIs respectively in the PACIFIC and FLAURA trials respectively-both presented at ESMO earlier and published in NEJM in November. However, OS data are unavailable for PACIFIC and immature for FLAURA. I have already discussed the issues with the interpretation of these trials in two earlier videos.

The RANGE trial comparing ramucirumab plus docetaxel versus placebo plus docetaxel for metastatic urothelial cancer after progression on platinum-based therapy was also presented at ESMO 2017 and is now published in the Lancet. The summary is that adding ramucirumab increases PFS (not OS) by a median of 1.3 months, increases toxicities and costs a lot. With the newer options of immunotherapies available, I doubt ramucirumab is of any value in this setting. I have discussed this in an earlier video.

One forest plot is worth a hundred essays

You must have already read hundreds of essays on the pitfalls of subgroup comparisons. Still, conclusions based on subgroup analyses remain rampant in the medical literature. But this one forest plot makes the pitfalls with subgroup analyses poignantly clear: patients treated with abiraterone in the STAMPEDE trial didn't receive benefit if they were diagnosed on a Monday versus the other days. Now imagine, how many conclusions have we based on such subgroup comparisons?

Let me take a selfie

Should we need an RCT for every intervention in the era of precision oncology? This is a question that frequently comes up in recent discussions. Undermining RCT evidence in lieu of speed for drug approvals is risky, while Real World evidence indeed remains important to expand on current knowledge and drive the design of subsequent clinical trials on the basis of findings from observational studies.  We try to address these issues in our latest commentary titled "Real-World Evidence and Randomized Studies in the Precision Oncology Era: The Right Balance".

Bishal Gyawali, MD, PhD completed his training in medical oncology at the Graduate School of Medicine, Nagoya University, Japan, and obtained a PhD as a Japanese government scholar. He works as a medical consultant at the Anticancer Fund, a not-for-profit organization based in Belgium as well as holding an affiliation at the Institute of Cancer Policy, UK. His areas of clinical and research interests include evidence-based oncology practice, global oncology, cancer policy, cancer management in resource-limited settings, financial toxicities of cancer treatment, clinical trial methods and supportive treatment of cancer. He has no conflicts of interest to declare. Dr Gyawali tweets at @oncology_bg. Read his previous blog posts here.