Protein C and Survival from Sepsis
Abstract & Commentary
Synopsis: Coagulopathy in sepsis is a marker of severe disease. Mortality was directly related to the reduction in level of protein C at admission and at 44 hours in a subgroup analysis of a prospective study of patients with severe sepsis. Problems with selection bias, industry authorship, and data presentation are present in this report.
Source: Yan SB, et al. Chest. 2001;120(3):915-922.
The role of coagulopathy in sepsis is an ongoing area of investigation. Therapeutic interventions directed toward interfering with the coagulation cascade have generally not been effective in modifying the course of the disease or preventing death. Recently, the recognition that, at different times during the septic process, coagulation may be enhanced or inhibited, has led to renewed interest in manipulating components that regulate coagulation. Specifically, data demonstrating a beneficial effect of administration of the coagulation inhibitor, activated protein C, in reducing mortality from sepsis, have recently been published.1
Yan and colleagues retrospectively examined the relationship of the level of protein C early in the course of severe sepsis in a highly selected subgroup of patients who were prospectively studied for another intervention. Patients were identified who had appropriate coagulation variables drawn during a prospective, randomized double-blind, placebo-controlled trial of ibuprofen.2 This study originally consisted of 455 patients entered from many institutions over a period of more than 5 years, from October of 1989 through March of 1995, and failed to demonstrate any significant survival benefits from ibuprofen administration.
Seventy patients from this original group were studied; 10 in each of the following groups based on platelet level, prothrombin time (PT), and partial thromboplastin time (PTT): 3 groups with only 1 abnormality; 3 groups with 2 abnormalities, 1 group with no abnormalities. The number of patients with all 3 abnormalities consisted of only 6 individuals, and Yan et al elected to eliminate this group from the analysis. Patients were randomly selected from the pool of patients meeting these criteria, although the selection process was not detailed in the paper. All patients entered into this analysis had D-dimer, fibrinogen, and protein C levels determined at entry and again at 44 hours. The selected patients were pooled and the relationships of platelet counts, D-dimer concentrations, and fibrinogen and protein C levels to mortality, ventilator-free days, and ICU-free days were determined. As it turned out, 83% of the patients selected for the present study were from the placebo arm of the ibuprofen study.
A total of 90% of the patients studied had low serum protein C levels at entry, and the reduced level of protein C showed a trend toward higher mortality (P = 0.19). A low level at 44 hours predicted higher mortality (P = 0.04). ICU-free and ventilator-free days correlated linearly with levels of protein C levels. The dramatic relationship of protein C level and 30-day mortality is shown in the graph, which was adapted from that in the original article (see Figure 1, below).
Blood Sugar Levels of Protein C and 30-day Morality in Severe Sepsis
Comment by Charles G. Durbin, Jr., MD
Yan et al’s interpretation of this study’s main results—that protein C level is a marker of the severity of disease—may well be correct. However, there are a number of worrisome features about the design of the study and the presentation of the data that raise the issue of bias. Retrospective examination of data collected for another purpose is a research technique fraught with pitfalls, and subgroup analysis (with incomplete description of the methods used) is particularly concerning. On the bright side, careful reading of this paper makes these concerns apparent and should induce the reader to be cautious in accepting its conclusions.
The most serious problem with this study is the patient selection process. Of the original 450 eligible patients, only 70 were selected. Little information is presented on the selection process: how the subgroups were determined was described, but why were equal numbers from each group studied? Why not study a random selection of all patients? The most important question to answer is: were there important differences between the patients selected for this analysis and those not selected? This is not addressed in the paper. Why were the 6 patients with abnormalities in all coagulation parameters excluded?
The obvious conclusion about the selection process is that the patients selected were chosen because they demonstrated the effect seen, which could be used in support of administration of an expensive new drug owned by the company that sponsored the study and whose employees are among the authors of the paper. The potential conflicts of interest posed by author affiliation with the sponsoring company are stated in the paper, as required by the ethical standards of scientific publishing, but they would appear to be substantial. With these concerns, interpretation of the implications of the data presented in this paper is exceedingly difficult.
Data presentation is another area where careful examination by the reader may be necessary to avoid jumping to unwarranted conclusions. Figure 1 shows how truncating an axis can change the visual impression of the magnitude of an effect. With a "normalized axis" the effect appears much less pronounced (see Figure 2, below).
The Same Data as in Figure 1, Presented with a "Normalized" Vertical Axis
References
1. Bernard GR, et al. N Engl J Med. 2001;344(10): 699-709.
2. Bernard GR, et al. N Engl J Med. 1997;336(13): 912-918.
You have reached your article limit for the month. Subscribe now to access this article plus other member-only content.
- Award-winning Medical Content
- Latest Advances & Development in Medicine
- Unbiased Content