|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 25, No 23 (August 10), 2007: pp. 3555-3557 © 2007 American Society of Clinical Oncology. DOI: 10.1200/JCO.2007.12.6599
In ReplyDepartments of Pediatrics, Anesthesiology and Critical Care, and Health Care Systems, The University of Pennsylvania School of Medicine and The Wharton School; Center for Outcomes Research, The Children's Hospital of Philadelphia, Philadelphia, PA
Department of Statistics, The Wharton School, The University of Pennsylvania, Philadelphia, PA We are grateful for the comments of the five letters and the editorial by Dr Cannistra1 that accompanied the publication of our article.2 Most of the variation in medical practice reflects the variation in the health of patients. The existence of two different specialties treating the same disease—specifically, gynecologic oncologists and medical oncologists treating ovarian cancer in the Medicare population—provides a rare opportunity to study the effects of variations in medical practice that are, to a considerable extent, not determined by variation in the health of patients. In our grant proposal to the National Cancer Institute, we anticipated, correctly as it turned out, that medical oncologists would use chemotherapy more intensively than gynecologic oncologists. We did not know whether more intensive treatment would lengthen survival, although we hypothesized that it would. Obviously, more intensive treatment has the potential to cause adverse effects that are far from pleasant for patients, so one hopes to see that more intensive treatment confers some tangible benefit. We did not think of our study as a contest between medical oncologists and gynecologic oncologists as professional groups; rather it was conceived of as a potential comparison of different approaches to the treatment of ovarian cancer, with these differences driven by different backgrounds in training and emphasis. The editorial and the letters do not dispute the facts as we presented them. By and large, the editorial and letters agree with the four main findings of our study. The matching compared gynecologic oncologists' and medical oncologists' patients who were similar in terms of clinical stage, tumor grade, age, surgeon type, comorbidities, Surveillance, Epidemiology, and End Results (SEER) site, year of diagnosis—in all a total of 36 baseline characteristics. Similar patients who received their postoperative chemotherapy from gynecologic oncologists or medical oncologists had virtually identical survival. The medical oncologist–treated patients received more intensive therapy. Consistent with this more intensive therapy was a higher rate of common adverse effects often associated with chemotherapy. Where we sense disagreement concerns the interpretation of these findings. Drs Markman, McGuire, and Cannistra believe that our findings showing identical survival are not surprising because survival is driven by the biology of the disease. This, of course, raises an obvious question. If the more intensive chemotherapy provided by medical oncologists does not affect survival, because survival is determined by biology and not by treatment, then given the adverse effects of intensive treatment, are medical oncologists justified in treating more intensively? In this context, Dr Cannistra suggests that medical oncologists may be better at offering hope than gynecologic oncologists, and Drs Hoffman and Iqbal suggests that medical oncologists are better trained in communication skills than their gynecologic oncologist counterparts. Are these hopes realistic? Would patients elect the more intensive approach if they knew it conferred no benefit in terms of survival and was accompanied by increased adverse events? Is this choice accurately communicated to patients? Alas, the SEER-Medicare data do not speak to these issues, and we are not inclined to speculate in the absence of evidence. We observed that medical oncologist–treated patients had more adverse events as defined by nausea, vomiting, diarrhea, dehydration, mucositis, anemia, thrombocytopenia, neutropenia, and peripheral neuropathy. Drs Markman, McGuire, Hoffman and Iqbal, and Cannistra correctly observed that there are other aspects of quality of life that we were unable to investigate using merged SEER-Medicare data. We agree; for instance, we did not have clinical data on pain, certainly an important consideration. Dr Cannistra suggests that quality of life may be improved even when chemotherapy associated adverse events are increased. This may be the case, but given the almost identical survival in both groups, and worse chemotherapy-related toxicity in the medical oncologist group, this postulated improvement in quality of life would need to have occurred in a manner just enough to alleviate tumor toxicity, but not enough to have any impact on survival. Dr Cannistra presumes that what is not measured favors the care provided by the medical oncologist group, but of course what was not measured was not measured. The SEER-Medicare data do not allow us to determine if the quality of life in the medical oncologist group was better or worse than the gynecologic oncologist group with respect to such things as pain. Drs Hoffman and Iqbal suggest we provide CIs for Table 4, not just P values, in order to be reassured that there really was more treatment and toxicity in the medical oncologist group. The 95% CI for the typical difference between the number of weeks on chemotherapy in the medical oncologist–gynecologic oncologist group, based on Wilcoxon's signed rank test for all years, was 1.5 to 5 weeks, and for adverse event weeks, the medical oncologist–gynecologic oncologist typical difference CI was 3 to 7 weeks. The typical difference is only part of the picture. To better see the differences in treatment and toxicity between the medical oncologist and gynecologic oncologist groups, we provide Figure 1.
Compare Figure 1A (the medical oncology patients) with Figure 1C (the gynecology oncology patients). For each of the 344 matched pairs of patients, we plot the 1-year relationship between weeks of chemotherapy and weeks with toxicity. Remember, again, that these gynecologic oncologists' and medical oncologists' patients appeared similar at baseline and had virtually identical survival. Note that the gynecologic oncologists' distributions are contained within a corner of the medical oncologists' distribution. Many gynecologic oncologists and medical oncologists treat patients in a similar fashion, but quite a few medical oncologists treat much more intensively with many more adverse events. The same pattern is found for the all years data in Figures 1B and 1D. Given the positive correlation between treatment and toxicity, and given the very different distributions by specialty, there seems little doubt that treatments and toxicity differed by specialty group. Furthermore, our analysis tends to understate the differences between medical oncologists and gynecologic oncologists because it is, in the language of clinical trials, an intent-to-treat analysis, in which patients remain in the groups to which they were initially assigned, even if they drop out or switch groups. Once we assigned patients to the medical oncologist or gynecologic oncologist group at month 3, we retained these assignments even if, for example, a gynecology oncology patient switched to a medical oncologist later in their course, for instance in response to cancer at another site. The resulting possibly more intensive chemotherapeutic approach by the medical oncologist would still be assigned to the initial gynecologic oncologists for our analysis. Hence, as with intent-to-treat analyses in clinical trials, our analysis tends to understate the difference between the two groups, medical oncologist and gynecologic oncologist. Dr Markman misunderstands the structure of our data. He incorrectly suggests that we could only adjust for baseline variables from SEER, and that we lacked important information about comorbidities. However, precisely for the reasons mentioned by Dr Markman, our study utilized a newly available data set that merged SEER data to inpatient and outpatient Medicare claims. We therefore adjusted for many patient comorbidities not part of the SEER database, that were in Medicare claims based on hospital or physician office bills, not patient self reports. The matching included adjustments for the following comorbidities: arrhythmia, angina, chronic heart failure, hypertension, peripheral vascular disease, chronic obstructive pulmonary disease, asthma, liver dysfunction, diabetes, anemia, coagulopathy, hyperthyroidism, rheumatoid arthritis, and fluid and electrolyte abnormality. For each of these comorbidities, we performed a look-back procedure2 to obtain comorbidities that were present within 90 days of being admitted for initial surgery. Drs Markman, McGuire, Hoffman and Iqbal, and Cannistra, suggest that medical oncologists and gynecologic oncologists should work together as a team. Apparently, when reviewing the SEER-Medicare data, this teamwork does not always exist—as was pointed out by Dr Berchuck. The fact that we observe these differences in treatment style suggests that the excellent cooperation experienced in Dr Cannistra's practice may not be present in many regions and medical centers throughout the United States. In summary, all authors appear to agree on the basic facts reported in our article.2 The controversy seems to concern whether greater intensity of treatment by medical oncologists is justified given that no difference in survival was observed, and measurable adverse effects appear to be more frequent in the medical oncologist-treated group than in the gynecologic oncologist–treated group. AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST The author(s) indicated no potential conflicts of interest.
REFERENCES
1. Cannistra SA: Gynecologic oncology or medical oncology: What's in a name? J Clin Oncol 25:1157-1159, 2007 2. Silber JH, Rosenbaum PR, Polsky D, et al: Does ovarian cancer treatment and survival differ by the specialty providing chemotherapy? J Clin Oncol 25:1169-1175, 2007
Related Correspondence
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2007 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
