Two RCTs have been completed comparing outcomes between Hartmann’s procedure and primary anastomosis in patients undergoing surgery for acute diverticulitis [15, 16]. These studies, however, fare poorly on the Cochrane Collaboration’s tool for assessing risk of bias [39]. Additionally, both studies were terminated prematurely due to lack of accrual of patients.

Oberkofler et al. conducted a multicenter RCT in Switzerland to compare Hartmann’s and primary anastomosis with loop ileostomy in patients with left-sided diverticulitis [15]. Their analytic approach considered the initial operation together with the subsequent ostomy reversal. Their power analysis included a very liberal estimate of expected differences in complication rates (40 % for primary anastomosis, 80 % for Hartmann’s), and estimated that 68 patients should be enrolled. During the 3 years that the study was conducted, the researchers were only able to recruit a total of 62 patients (30 in Hartmann’s and 32 in primary anastomosis + ileostomy group). In addition, 52 potential study patients were not assessed for eligibility because of the surgeons’ choice not to enroll patients resulting in the potential for significant selection bias [15]. Their analysis revealed differences in several endpoints in favor of primary anastomosis with loop ileostomy. Only 15 of 26 (58 %) end colostomies (after Hartmann’s procedure) were eventually reversed, whereas the stoma reversal rate after ileostomy was significantly higher at 90 % (26/29, P < 0.012). Diverting ileostomies were reversed much earlier than the end colostomies after Hartmann’s procedure (median 3 months vs. 6 months, respectively). The rate of severe complications (20 % vs. 0 %, P = 0.046), as well as the total number of complications per patient (median 1 vs. median 0, P < 0.001), was significantly higher after reversal of Hartmann’s procedure (colostomy) compared to ileostomy reversal. Anastomotic dehiscence, sepsis, and bleeding occurred only after reversal of the end colostomy. Furthermore, the duration of the operation (183 min vs. 73 min, P < 0.001) as well as the hospital stay (9 days vs. 6 days, P = 0.016) was significantly longer after reversal of Hartmann’s procedure. Of note, all the advantages of primary anastomosis with diverting ileostomy relate to the reversal operation.

Binda et al. from Norway conducted a multicenter RCT, but terminated it prematurely as they could recruit only 15 % of the target sample size (300 patients in each group) in 9 years [16]. No conclusions could be drawn from this study.

Two meta-analyses have been performed that examined evidence regarding outcomes in patients undergoing Hartmann’s procedure vs. primary anastomosis. The first of these, conducted by Constantinides et al. in 2006 included a total of 15 studies; 10 of these studies were published between 1984 and 1995 and 5 after 1995 – these 5 studies are a part of our review [36]. Results from this meta-analysis show lower mortality with primary anastomosis than with Hartmann’s operation, (4.9 % vs. 15.1 %). Another meta-analysis of 14 studies was performed by Cirocchi et al. in 2013, and also found lower mortality rates with primary anastomosis than Hartmann’s procedure (9.8 % vs. 22.0 %) in the treatment of acute diverticulitis. The authors, however, found that the heterogeneity of the included studies was very high and recommended that their findings be interpreted with caution [35].

Despite the intuitive appeal of relying on meta-analyses as a quantitative synthesis of existing evidence, there is good reason to discount the findings from these two studies. First, the technique of meta-analysis does not apply well to small, non-randomized studies with heterogenous populations/interventions. This limitation was articulated nicely in the study performed by Cirocchi [35]. Second, these studies are ambiguous as to whether they are estimating the clinical burden of the initial operation or the initial operation plus any subsequent operations (to restore intestinal continuity).

Three studies have been conducted using secondary databases in order to compare outcomes of primary anastomosis vs. Hartmann’s procedure for acute diverticulitis [32–34].

In 2012, Gawlick et al. published a study using patient data from the NSQIP database in 2005–2009 to analyze 2018 patients undergoing surgery for acute diverticulitis [34]. This study used wound classification (contaminated and dirty) as a surrogate marker for severity in patients who underwent emergent surgery with a diagnosis code of diverticulosis or diverticulitis. The study found no significant difference in the risk of infectious complications, return to the operating room, prolonged ventilator use, death, or hospital length of stay between Hartmann’s procedure and primary anastomosis with diversion. In examining the subgroup of patients where the operation was classified as dirty/infected, however, the adjusted mortality rate was twice as high when primary anastomosis with diversion was performed compared to the Hartmann’s procedure.

Also in 2012, Masoomi et al. published a study using discharge data from the NIS between 2002 and 2007 to analyze 99,259 patients undergoing primary anastomosis with diversion vs. Hartmann’s procedure for acute diverticulitis [33]. This study found a lower complication rate in the primary anastomosis (plus diversion) group compared with the Hartmann’s group (primary anastomosis: 39.06 % vs. Hartmann’s: 40.84 %; p = 0.04). Mortality was lower in the primary anastomosis group (3.99 % vs. 4.82 %, p = 0.03). However, patients in the Hartmann’s group had a shorter mean length of stay (12.5 vs. 14.4 days, p < 0.001) and lower mean hospital costs (USD 65,037 vs. USD 73,440, p < 0.01) compared with the primary anastomosis group. This study, while based on a very large cohort of patients, may suffer from issues regarding the granularity and accuracy of administrative coding. The International Classification of Disease (ICD) coding scheme is not a perfect system in terms of describing the type of operation performed, and there is the potential that many of the patients in this study were mischaracterized in terms of the type of surgical care they received.

In 2013, Tadlock et al. published a study using patient data from the NSQIP database in 2005–2008 to analyze 1313 patients undergoing surgery for acute diverticulitis [32]. Three operative approaches were analyzed: Hartmann’s procedure, primary anastomosis without diversion, and primary anastomosis with diversion. In this study, the 30-day mortality was 7.3 %, 4.6 %, and 1.6 %, respectively (P = 0.163), while surgical site infections occurred in 19.7 %, 17.9 %, and 13.2 % of patients (p = 0.59). In addition, the three groups did not have significant differences in surgical infectious complications, acute kidney injury, cardiovascular incidents, or venous thromboembolism after surgery. The authors of this study concluded that primary anastomosis in the acute setting is a safe alternative to a Hartmann’s procedure, with no significant difference in mortality or postoperative surgical site infections.

As with meta-analyses, the results from large database studies should be interpreted with caution. Statistical differences in outcomes may not always be clinically significant due to the large sample sizes. This is illustrated by the small difference in complication rate between the primary anastomosis group (39.06 %) compared with the Hartmann’s procedure group (40.84 %) in the NIS study above which was statistically significant (p = 0.04). More importantly, the translation of clinical phenomena into accurate representation in codes (ICD or otherwise) may lead to inaccuracy, bias, and confounding.