| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters to the Editor |
Department of Clinical Methodology, Epidemiology, and Statistics, Civico e Benfratelli Hospital National Trust, Piazza Nicola Leotta 2, 90127 Palermo, Italy. e-mail: s.corrao@tiscali.it
Editor:
I have read with great interest the work of Dr Schillinger and colleagues (1) in the June 2004 issue of Radiology. They studied the effect of smoking after lower-limb endovascular intervention and surprisingly concluded that smoking 10 or more cigarettes daily was associated with a reduced rate of intermediate-term restenosis. Before reaching this kind of conclusion, however, some considerations were needed about both study methodology and statistical analysis.
First, this study was a prospective study based merely on an existing database. Therefore, it had not been designed for the study purpose. This methodologic matter is of great importance. Indeed, 62 (9%) of 712 patients were soon excluded because of incomplete baseline and follow-up data. Thus, 650 patients were considered. However, two observers found discrepancies in the smoking histories of 71 (11%) of 650 patients. Moreover, follow-up data regarding smoking habits were available for only 332 (51%) of the 650 patients.
Therefore, in summary, the authors aimed to study 712 patients, but only 332 patients had complete data. Moreover, the authors stated that smokers were younger than nonsmokers. Table 1 shows a significant difference in patient sex, also. Thus, assumptions for comparison were lacking. In spite of all of that, however, the authors used a Cox proportional hazards analysis with adjustments for possible confounding effects of a large number of variables. For example, I want to remind the authors that Hasdai et al (2) found, by means of bivariate methods, a positive association between successful percutaneous revascularization and smoking. Obviously, however, performance of sensible multivariate analysis led to avoidance of absurd conclusions. I would also like to remind the authors that confounders are not the only problem one can face when managing data. Indeed, suppressors and interaction variables can invalidate the results of a study. In this case, the misleading study design and possibly also an indiscriminate use of many variables in the multivariable model (confounders plus suppressors) or some interactions (not considered) very likely led to conclusions without biologic plausibility.
In conclusion, I think and hope that no physician will advise patients to smoke or inhale carbon monoxide to avoid restenosis after lower-limb endovascular intervention.
REFERENCES
Department of Internal Medicine II, Division of Angiology, University of Vienna Medical School, Währinger Gürtel 18–20, Vienna A-1090, Austria. e-mail: martin.schillinger@meduniwien.ac.at
We read with interest the letter of Dr Corrao concerning our article (1). We appreciate the critical comments and agree on most of the mentioned study limitations. However, some issues may be discussed further.
First, being aware of the potential study limitations, we concluded only that there was an association between smoking and restenosis; we did not refer to a causal relationship in the conclusion. Second, Dr Corrao obviously misinterpreted the number of patients included in the study: In fact, 650 patients were followed up for restenosis, and all of these patients were included in the final analysis, since baseline data of smoking habits were available in these patients. Smoking habits during follow-up were available in only 332 patients, but these data were not included in any formal statistical analysis. We therefore analyzed 650 of 712 patients—as stated in the abstract and results section.
Third, to address the issue of potential confounders: We agree that in a nonrandomized setting, imbalances between groups remain a major statistical problem. With reference to the observed imbalances of patient ages, however, we carefully assessed possible confounding effects. Knowing the limitations of this approach, we adjusted the final model for patient age, and we tested for interactions between smoking and all possible confounding variables included in the model—as stated in the materials and methods section (in the description of statistical methods). Interactions were indeed considered and were excluded by means of multiplicative interaction terms and log likelihood ratio tests. Furthermore, to our knowledge, patient age has not been described as a clinically relevant risk factor for restenosis after femoropopliteal interventions.
In conclusion, we agree with Dr Corrao that smoking should not be recommended for anyone, particularly not for our patients with vascular conditions. Published animal data on CO inhalation showed promising results, however, and CO as a potent antiinflammatory molecule may be worth further investigation.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
