The Author Replies

The setting out of the statistical headings on tables and figures. Table 2 shows the multivariate regression analysis results. Figures 1 to 3 are the univariate results. I should have spotted the error in the Table 2 heading and corrected this at the time of the proofs.

The quality of the surgery on outcome as reflected by the frequency of splenectomy. As stated in the article, we matched each patient based on surgical center, sex, similar age (±5 years), pathologic stage and site (colon or rectum) of disease, preoperative ASA (±1), and operation, in part to try and control for quality of surgery. We did not match for surgeon, although we did consider it. However, this proved too difficult. We had to look over a ten-year period in the five major hospitals (>10,000 colorectal resections) to find 55 patients who had an incidental splenectomy (approximately 0.05 percent) during left-sided (sigmoid and rectal) resections. In a recent study from the United States1 with a total of 41,999 non-Stage 4 colorectal cancer resections, inadvertent splenectomy was undertaken in 241, i.e., 0.57 percent of all resections with a marked increase in odds ratio for left-sided resections (sigmoid 2.689 (95 percent CI, 1.494–4.839); rectosigmoid 2.56 (95 percent CI, 1.305–5.023); and rectum 1.389 (95 percent CI, 0.697–2.77)). Therefore, our study would seem to have a very low level of inadvertent splenectomy.

The effect of blood loss on outcome. We assessed blood loss by the number of units transfused. The reason for this was twofold: 1) We know that transfusion and not blood loss per se has significant immunologic effects as demonstrated in the transplant situation2; and 2) It is easy to measure, although estimations of intraoperative blood loss can be difficult (especially retrospectively).

The effect of low-volume surgeons on outcome. I agree that low-volume surgeons have repeatedly been shown to have poorer outcome than high volume; however, outcome is a not a simple relationship3,4 and this is not the issue that we were studying.

Dr. Fujita makes the point that the influence of splenectomy on outcome has been most widely studied in regard to upper gastrointestinal cancers. In noncolorectal surgery, such as surgery for pancreatic, esophageal, or gastric cancer, splenectomy has been suggested to impair survival.5‐8 In a Japanese randomized trial on splenectomy and gastrectomy, the possibility was raised that the immunosuppressive acidic protein (IAP) level may influence the outcome of splenectomy.5 In patients with high IAP levels (>580 μg/ml), splenectomy improved the prognosis. Conversely in patients with lower IAP values, the preservation of the spleen demonstrated a significant benefit to survival.

A study from China reported on 449 patients operated on for gastric cancer that had undergone total gastrectomy with curative intent between 1991 and 1995.6 They found that the recurrence rate in the splenectomy group was 48.1 percent compared with 22.6 percent in the spleen-preserved group (P = 0.01). Among transfused patients it was 40.7 percent compared with 26.5 percent among nontransfused patients (P = 0.086). There was no significant difference in the mean survival between the splenectomy group and the spleen-preserved group in a subgroup analysis by stage. Multivariate analysis identified splenectomy as an independent risk factor for recurrence.

A study from Germany on the influence of splenectomy on survival after surgery for esophageal carcinoma in 404 patients (operated on from September 1985 to July 2003) found that 34 (8.4 percent) patients had an incidental splenectomy and that this was associated with an increase in intraoperative blood loss and the need for blood products (P < 0.0001).7 However, there were no significant differences in pulmonary, general, or surgical complications between patients with and without (P > 0.05) splenectomy. Although the median survival of 13.9 months recorded in patients without splenectomy was longer compared with a survival rate of 8.9 months for patients after splenectomy, the difference did not reach statistical significance (P = 0.315). An analysis of survival time (log-rank) did not yield any differences for patients with and without concomitant splenectomy (P > 0.05).

In a study from New York from a database of 555 pancreatic resections for adenocarcinoma at a single institution, splenectomy had been undertaken in 55 (9.9 percent) and was considered on Cox multivariate analysis as an independent significant variable (P = 0.001) for poor survival.8 An earlier study by the same group of 332 patients undergoing pancreatectomy, in which 326 having confirmed local or regional disease only formed the study cohort and only 37 underwent concomitant splenectomy (11.4 percent). It was found that splenectomy did not affect outcome.