| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Eur J Cardiothorac Surg 2006;29:855-856
© 2006 Elsevier Science NL
Letter to the Editor |
Hemodynamics and Cardiovascular Radiology Service, Policlinico San Donato, San Donato Milanese, Italy
Received 16 January 2006; accepted 7 February 2006.
* Corresponding author. Address: Servizio di Emodinamica e Radiologia Cardiovascolare, Centro Cardiovascolare "E. Malan", Policlinico San Donato, Via Morandi 30, 20097 San Donato Milanese (MI), Italy. Tel.: +39 3408626829; fax: +39 0252774585. (Email: gbiondizoccai{at}gmail.com).
Key Words: Coronary artery disease • Coronary artery bypass graft surgery • Clinical outcome research
We read with interest the article by Seccareccia et al. [1] apprasing short-term outcomes of patients undergoing coronary artery bypass graft surgery in Italy, as well as the editorials by Nashef [2] and Menicanti [3]. Specifically, we found intriguing the potential implications for health policy and clinical practice, i.e. the comparison between hospitals according to adjusted outcome rates and focusing possible policy interventions on underperforming surgical centers [4]. However, such interpretation of the results from Seccareccia et al. is limited by a number of methodological drawbacks.
First, the period of enrolment varied largely from center to center. This may confound results by means of selection bias, i.e. small volume centers providing data concerning a specific time frame when complications might have been very few or none at all.
Second, it is difficult to compare high-volume to low-volume centers. Several data suggest that only institutions performing an adequate number of procedures per year can minimize complications. Thus, it appears appalling that low volume centers could outperform high-volume centers. The play of chance in small samples is the most likely explanation, as the occurrence of few events can enormously change the overall event rate of the best-performing hospital, given its very small caseload (146 per year).
Third, the authors do not mention the rate of missing or inaccurate data. Investigators familiar with multivariable techniques know well that multivariable analysis can only be carried out for patients with complete datasets. The implications of wrong or inaccurate data are otherwise self-evident.
Fourth, validation of statistical models is a critical issue as well. The potential fallacy of even highly sophisticated multivariable models, given the risk of overfitting, incomplete discrimination and calibration is well known [5]. While the splitting procedure performed by the authors to validate their model is a good starting point, this is probably not enough. Bootstrapping, an established internal validation technique based on resampling, should have been performed as well [5]. Moreover, while the multivariable model from Seccareccia et al. may perform reasonably well in the overall population, it is uncertain whether this also holds true in the outlying hospitals. The only means to address this would be to appraise accuracy, discrimination ability, and calibration in each subsample.
In addition, no mention of testing for interactions is provided. No complex statistical model can be envisaged without such thorough appraisal, and this fact is based not only on statistical but also on clinical grounds (e.g. the risk of a patient with concomitant chronic obstructive pulmonary disease (COPD) and acute pulmonary edema (APE) is likely greater than the simple sum of the individual risks of COPD and APE separately). Finally, a propensity analysis could also have benefited the robustness of the analysis, providing more support to the authors’ conclusions.
In conclusion, while the potentials of clinical outcome research studies in cardiovascular surgery in general, and of the work by Seccareccia et al. in particular, are several-fold [4], given the underlying assumptions and inherent limitations, this paper should be mainly regarded as hypothesis-generating until further external validation is available.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
