Although median survival is a reasonable endpoint to highlight, it is only 1 of many relevant outcome factors to discuss, and, most important, this mathematical figure simply does not define the survival of any particular patient.
Maurie Markman, MD
In an elegant commentary discussing patients and their courageous decisions to fight cancer, Lawrence H. Einhorn, MD, a pioneer of modern oncology, highlights the differences between clinical oncologists’ textbook learning about outcomes and their actual real-world experience.1
Particularly moving in Einhorn’s opinion was the story of a young woman, 32, discovered to have stage III inoperable lung cancer. She experienced major treatment-related morbidity and required management for metastatic disease, as well as a second primary cancer (renal cell).
Despite these daunting challenges, she was able to survive for 15 years and had the truly wonderful opportunity to watch her young children grow. Her clinical course and ultimate survival did not follow what is written in textbooks or what would likely be considered the “correct answer” on a medical oncology board examination. Einhorn suggests that “her love of life and family” may have had something to do with her atypical longevity.The personal experience of the late Stephen Jay Gould, PhD, an internationally acclaimed evolutionary biologist, must be included in any discussion of the critical difference between claims of prognostic certainty and the objective reality that prognostic data may not have direct relevance in individual cases of cancer.
Following his own diagnosis of peritoneal mesothelioma, Gould was dismayed to discover in Harvard’s Countway Library of Medicine that he likely had just 8 months to live. However, this scientist critically examined the data and concluded that although it was correct about the median survival for a population of individuals with this devastating malignancy, there was no reason his own outcome would not ultimately be found at the end of the long tail of the peritoneal mesothelioma survival curve.
In a landmark and truly inspiring essay titled “The Median Isn’t the Message,” Gould highlighted the very serious error associated with concluding that an individual patient with cancer would experience a survival outcome defined by the mathematical term median simply because “50% of the entire population lives a longer and 50% lives a shorter period of time.”2
Further, the experiences noted above illustrate the very real danger of attaching too much weight to statistical averages. Another example concerns the news reports in the lay press regarding the introduction of novel antineoplastic strategies approved by the FDA. These articles frequently cite the reported median survivals of the investigative versus control patient populations as the maximum magnitude of benefit a single individual might anticipate.
In fact, although median survival is a reasonable endpoint to highlight, it is only 1 of many relevant outcome factors to discuss, and, most important, this mathematical figure simply does not define the survival of any particular patient.A recent publication in the gynecologic cancer literature that explored the survival outcomes for women who experienced initial recurrence of epithelial ovarian cancer clearly emphasizes the point.3 The study involved an examination of Surveillance, Epidemiology, and End Results—Medicare database records of 2369 women in this most difficult clinical setting who were >65 years of age, who underwent primary surgical cytoreduction and platinum-based chemotherapy, and whose disease unfortunately recurred >3 months after diagnosis.
The reported median overall survivals from the time of documented disease progression <6 months, 7 to 12 months, and >12 months following the completion of chemotherapy were 13 months, 18 months, and 27 months, respectively.3
Therefore, 1 interpretation of these data might be that patients whose ovarian cancer recurred >6 months following the completion of primary therapy will survive for approximately another year. Those whose disease recurs between 7 to 12 months will survive about 1.5 years; if recurrence develops more than a year after the last platinum treatment, the survival will be about 2 years.
But is this the only way to describe the reported survival? Although the median figures just cited are accurate (based on the previously noted report) and would likely be disturbing for most patients and families in this clinical setting, a closer examination of the survival curves tells a different story.
As asked by Gould, why should individual patients be required to focus on the medians? What is wrong with looking at the tails of the reported survival curves of real-world patients who are just like them? Why would the outcome associated with a tail not aptly describe the survival of an individual patient?
In fact, the data in this report suggest that of the patients with ovarian cancer who experience disease progression <12 months after completing primary therapy (surgery and platinum-based chemotherapy), approximately 10% survive at least 4 years and 5%, at least 5 to 6 years. Further, for patients whose disease recurs >12 months after initial treatment completion, approximately 10% will survive at least 6 to 7 more years, and at least 5% will survive an additional 10 years. These figures are far superior to the median survivals previously quoted.
Gould’s question remains as relevant today as it was when he first posed it in 1985—17 years before his death: “Why can’t I be in the 5% to 10% of the population at the tail end of that clearly distressing curve?”
I am not suggesting that we ignore sober facts or that it is ever appropriate to offer unrealistic prognostic information. However, if objective data indicate that a more favorable outcome is conceivable at a particular point in the natural history of a malignancy, why would it be wrong for a patient (and her or his family) to entertain that hope?