In the January issue of the Archives, Bennett et al1 described a prospective analysis of 1194 patients undergoing laparoscopic-assisted colorectal surgery.1 The rate of postoperative complications with operations performed by high-volume (>40 procedures) surgeons was significantly lower than with operations performed by low-volume (<40 procedures) surgeons (9.9% vs 18.6%, P<.001). The authors assert that these findings support the presence of a "learning curve" for laparoscopic-assisted colectomy.
However, a close examination of the data suggests otherwise. The authors used volume as a dichotomous variable (high vs low), and their cutoff point (40 procedures) was defined retrospectively. Had the authors selected a different volume threshold, their conclusions would have been different. Furthermore, if the analysis were performed using volume as a continuous variable, a learning curve would not have been discernible at all. In fact, if the numbers reported in the article's Table are used to plot volume vs