The role of colony stimulating factors in small cell lung cancer: Why the question is still unsolved

Abstract

The use of Colony Stimulating Factors (CSF) to reduce the incidence and consequences of chemotherapy-induced febrile neutropenia was approved more than 10 years ago on the basis of two randomised trials [1,2] evaluating Granulocyte Colony Stimulating Factors (G-CSF) in small cell lung cancer (SCLC). The American Society of Clinical Oncology then recommended the prophylactic use of either G-CSF or Granulocyte Macrophage Colony Stimulating Factor (GM-CSF) in addition to any chemotherapy regimen associated with a more than 40% risk of febrile neutropenia [3]. Since then, various attempts have been made to test their ability to improve the therapeutic effects of cytotoxic chemotherapy in SCLC, either by increasing the dose per course, or by reducing the interval between the courses, or by using the recruitment and re-infusion of haematopoietic stem cells. Overall, the conclusions drawn from the potential impact of these strategies on treatment results are controversial. Furthermore, although the pivotal studies using G-CSF were randomised and placebo controlled, and despite the demonstration of long term response and survival equivalence of both arms in one of them [4], some studies have raised the question of a possible deleterious effect of GM-CSF [5,6] and generated uncertainties in the scientific community. In such a context, the publication of the paper by Berghmans et al. in the present issue of the ‘Lung Cancer’ journal [7] is particularly appropriate. The authors present the results of a meta-analysis based on the data from the literature on the potential impact of G-CSF and GM-CSF on anti-cancer treatment outcomes in SCLC. They have pooled 12 randomised studies and applied to the 12 corresponding publications two previously published quality scales. The scores proved to be satisfying to this group who has wellknown expertise in the field [8] and led the authors to consider these publications as adequate to perform the meta-analysis. Because of the above-mentioned heterogeneity in the design and objectives of the different studies, they have artificially chosen to split them into three subgroups based on arbitrarily applied endpoint criteria, namely ‘maintenance of planned dose-intensity’, ‘accelerated chemotherapy’ and ‘concentrated chemotherapy’. Since they find a detrimental effect on response in the first group and a reduced survival in the third (which only contains one study, the negative trial by Pujol et al. [6]) they come to the conclusion that ‘the published data do not support the routine use of haematological CSF in the treatment of SCLC’. Although the weakness of meta-analyses based on published data only is well-known, the provocative paper by Berghmans et al. has the considerable interest to offer a useful selection of the appropriate trials to be included in a meta-analysis on individual data, a much more valid method to answer such a difficult question. In 1996, the meta-analysis group in cancer, which conducted and published several meta-analyses based on individual patient data in the past 10 years [9 /11] decided to perform a meta-analysis on survival in SCLC by pooling the seven randomised studies on GCSF available at that time, most of them in abstract forms. These clinical trials had been promoted either by an institutional group of investigators [12 /15] or by the pharmaceutical industry [1,2,16]. The principal investigators of these studies gave us consent to collaborate. The Chugai corporation gave us full access to their data base. Unfortunately, despite multiple contacts with the Amgen corporation, including specific requests from the trial investigators, we never obtained the authorization to use the data from the Amgen G-CSF trials [1,2] and had to give up the project. Finally, the paper by Berghmans et al. raises the issue of the accessibility of data collected in clinical trials. We * Corresponding author Lung Cancer 37 (2002) 125 /126