Member Spotlight: Alexander Melamed, MD, MPH

A San Francisco native, Dr. Alexander Melamed is quite enjoying his second official move to New York far more than his first. At age 23, fresh from college, he moved to a small town in upstate New York to work in a whisky distillery. 

“When I realized that wasn’t for me, I moved to New York City and studied for the MCAT,” he says with a laugh. “I lived in a hovel in Brooklyn, and I worked part-time as a paralegal. That only lasted a few months before I headed back to California. Now, after some 15 years, it is really great being back here in New York, and under better circumstances.”

Dr. Melamed, a member of the Cancer Population Science program at the Herbert Irving Comprehensive Cancer Center (HICCC), joined Columbia in July of 2019 as assistant professor of obstetrics and gynecology at Columbia’s Vagelos College of Physicians and Surgeons. A gynecologic oncologist and clinical researcher, Dr. Melamed focuses on using quasi-experimental study designs and sensitivity analyses to improve causal inference in oncology outcomes research, with an emphasis in ovarian and cervical cancers. His interests lie in generating credible evidence from observational data.

“What’s particular about my work isn’t the scale of the data we study, but more so how we  apply different kinds of analytic methods that have not been traditionally applied in epidemiological research in the past,” he says. “I am interested in study designs better known in fields such as economics and social sciences, and seeing how those can be useful in figuring out causal relationships in healthcare data.”

Dr. Melamed has recently received a KL2 Mentored Career Development award. The award, administered by Columbia’s Irving Institute for Clinical and Translational Research, gives junior faculty exposure to training and mentoring, and supports their trajectory in patient oriented multidisciplinary research.

What is your research focus?

Broadly, what I do is called health services and comparative effectiveness research. I’m interested in figuring out what kind of cancer therapies are effective for which type of patients in the real world. A lot of the way we figure that out is with clinical trials, but there are more questions than there are budgets for clinical trials. How the effectiveness of therapies in the real world is not always the same as in randomized trial settings. Some questions cannot be addressed in trials for ethical or feasibility reasons. In my work, we try to use observational data to answer those kinds of questions. We are interested in making casual inferences from observational data in the context of cancer care.

A conventional epidemiologic view is that observational studies are for hypothesis generation, but that it is not possible to determine cause and effect relationships outside of a randomized clinical trial. I think that is untrue. Humans figure out cause and effect relationships without  randomized trials all the time: I know that if I skip coffee, I am going to have a headache. To my knowledge this has never been tested in a randomized trial. There is a human ability to deduce how certain phenomena are interrelated, how x  causes y. In a medical setting, it’s  complicated because causes are often probabilistic, and there are complex webs of causes and effects. Nonetheless, there are methodological approaches which can be used to apply our innate ability to reason causally to observational data that, I believe, can help to answer important questions in cancer care delivery. 

What are you currently working on?

There is a persistent debate among gynecologic oncologists about whether and when to give patients chemotherapy prior to conducting surgery—called neoadjuvant chemotherapy—in advanced ovarian cancer. What’s interesting to me in this controversy is that even though several randomized trials have shown that neoadjuvant chemotherapy results in similar long term outcomes in ovarian cancer patients, and with less toxicity, the dominant discourse is that primary surgery should be the standard of treatment for anyone who can tolerate the approach and is likely to have a successful resection. One of the reasons that many oncologists seem to disbelieve the randomized evidence is that multiple observational studies have found that patients who receive neoadjuvant chemotherapy tend to have shorter survival then those who receive surgery first. However, since gynecologic oncologists reserve neoadjuvant chemotherapy for women who have the highest burden of cancer, and those who are too frail to tolerate a high-risk operation, the better survival observed after upfront surgery is probably the result of patient selection, rather than a beneficial effect of the upfront surgery.

I’ve been working on how we can use quasi-experimental study designs to evaluate this hypothesis. One of the insights is that sometimes it makes more sense to look at group-level data, like at hospital or regions, rather than focusing exclusively on patient-level outcomes. So, for example, if upfront surgery causes patients to live longer, one might expect to see that patients treated in hospitals that perform more upfront surgery have better outcomes than those treated at hospitals that use the procedure less often. You can also look at change over time as well: how did outcomes change in hospitals that adopted the use of neoadjuvant chemotherapy enthusiastically in recent years compared with those that remained lukewarm about the approach? These kinds of “ecological” analyses have traditionally been discounted in epidemiology, but I believe that studying a practice at a variety of scales can help to tease out cause-and-effect relationship.

What has grown out of this research, are question about the role of surgery for metastatic cancer in general. Gynecologic oncologists operate on people who have advanced stage ovarian cancer, and we believe that surgery has an important role in care for these patients. But for many cancer types, surgery is rarely performed in the advanced setting. We are looking across cancer sites to investigate how surgery is used in the upfront setting. What we have found in preliminary data is that there is a tremendous amount  of variation not only across cancer types, but also geographically. I’m really interested in how this heterogeneity of treatment arose. What generalization can we make by looking across cancer site? Is there evidence that surgery should be performed more often in the metastatic setting in cancer types where it currently performed infrequently? And conversely, is there evidence that there are patients receiving surgery that may not be beneficial?

When and how did you get interested in cancer care and research?

I went into gynecologic oncology because I liked the patient population. Women who tend to be diagnosed with gynecologic malignancies are often older, they have raised families, and lived lives, and this patient population just really drew me in. I felt a huge amount of compassion for them. I knew I couldn’t do anything else. I also like gynecologic oncology surgery – it’s interesting and complex. I really enjoy combining my two areas of research interests: public health and oncology.

Where do you see your work having an impact?

At its best, my work can improve the way cancer care is delivered. In 2018, we had a paper published in the New England Journal of Medicine that helped to define the role of minimally invasive surgery in cervical cancer patients. Using epidemiological approaches, our study suggested that minimally invasive surgery was associated with a higher risk of death than open surgery among women who underwent radical hysterectomy for early-stage cervical cancer. This went against the grain at the time, and as a result of this study and a randomized trial that demonstrated the same findings, the National Comprehensive Cancer Network and the European Society for Gynecologic Oncology have changed their guideline to make open surgery the standard therapy for their early-stage cervical cancer patients.

Until the 19th century, bloodletting--the practice of removing blood for a patient to treat a disease—was thought to be beneficial for almost every illness. Physicians and barber-surgeons, who recommended and performed bloodletting, were learned and well intentioned, we think that they rarely helped patients with this approach. We all would like to think that we aren’t doing anything today in oncology that is like bloodletting, and that all our current treatments are beneficial to patients. But I think experience shows that is naïve. As physicians and researchers, we must consistently subject what we’re doing to rigorous scrutiny in order to get closer and closer to our goal of providing more help to our patients and doing less harm. A lot of what my work hopes to do is to separate what clinical practices are beneficial and when.

What interested you the most about coming to Columbia and joining the HICCC?

Jason Wright, MD, the chief of gynecologic oncology at Columbia does some of the highest caliber observational research in gynecologic oncology. I got to know him through the Society of Gynecologic Oncology, and I was able  to collaborate with him while I was still a trainee in Boston. What I loved about his work is that his approach is unbiased by the dogma of the field, and he is straightforward and objective. The Herbert Irving Comprehensive Cancer Center has an extremely vibrant Cancer Population Science program. I am working closely with the Center for Cancer Outcomes Research and Evaluation on several ovarian cancer care projects.

What motivates your research?

The most important thing to me is to improve the health and the lives – longevity and quality - of my patients. I am also interested in in the puzzle-solving part of cancer research. It turns out it’s really hard to make convincing argument about causal relationships based on observational data. When I approach a research project, I am always interested in how the findings may impact patients and future research, but the process itself, being able to look at data in ways that are innovative and forming interpretations that are persuasive, is also a lot of fun.