Anesth Analg 2003;96:1843-1844
© 2003 International Anesthesia Research Society
LETTERS TO THE EDITOR
Preoperative Risk Factors of Intraoperative Hypothermia in Major Surgery Under General Anesthesia
Karel G.M. Moons,
Wilton van Klei, and
Cornelis J. Kalkman
Department of Anesthesiology, Division of Perioperative Care, Emergency Medicine, and Julius Centre for Health Sciences and Primary Care, University Medical Centre, Utrecht, Utrecht, The Netherlands
To the Editor:
Kasai et al. recently derived and validated a multivariate preoperative model to predict the absolute probability of intraoperative hypothermia (1). We acknowledge the relevance and potential utility of such a prediction model. However, we believe that the presented model cannot yet be used in practice due to a design error.
The authors used a case-control design to derive their model. However, only data from cohort studies allow derivation of (models to estimate) absolute probabilities. In a case-control design investigators are free to choose the number of cases and controls and can therefore "manipulate" beforehand the absolute probabilities (2,3). We illustrate this with a simple example using the (dichotomous) predictor sex and data from their own study. The authors selected 200 patients with (cases) and 200 without (controls) hypothermia from a cohort. The sex ratio among controls was 119 (60%) males versus 81 (40%) females, and among cases 138 (69%) versus 62 (31%) (see Table 1, numbers without parentheses). The mean probability of intraoperative hypothermia was apparently 50% (200/400), the absolute probability for males 138/257 = 54% and for females 62/143 = 43%. If the investigators had selected, e.g., 400 controls (see Table 1, numbers within parentheses), the mean probability would become 33%, the probability for males 37% and for females 28%. In fact, each different ratio of patients with and without hypothermia will give different overall and sex-specific probabilities. This applies to continuous predictors (like age and heart rate) similarly, as well as to combinations of predictors in a multivariate model. Nevertheless, if the original cohort size and number of patients with and without hypothermia is known, an easy method to derive a correct prediction model from case-control studies is available (4). As this is the case in the study of Kasai et al., this analysis can easily be carried out and we look forward to see the results.
View this table:
[in this window]
[in a new window]
|
Table 1. Association of sex with intraoperative hypothermia for two different ratios of cases and controls: 1:1 and 1:2
|
|
References
- Kasai T, Hirose M, Yaegashi K, et al. Preoperative risk factors of intraoperative hypothermia in major surgery under general anesthesia. Anesth Analg 2002; 95: 1381–3.[Abstract/Free Full Text]
- Hennekens CE, Buring JE. Epidemiology in medicine. Boston: Little, Brown & Co, 1987.
- Rothman KJ, Greenland S. Modern epidemiology. 2nd ed. Philadelphia: Lippincott-Raven Publishers, 1998.
- Moons KG, Bots ML, Salonen JT, et al. Prediction of stroke in the general population in Europe (EUROSTROKE): is there a role for fibrinogen and electrocardiography? J Epidemiol Community Health 2002 (suppl 1); 56: i30–6.[Abstract/Free Full Text]
