A historical review of surgery for peritonitis secondary to acute colonic diverticulitis: from Lockhart-Mummery to evidence-based medicine

The PRISMA flow diagram for systematic reviews is presented in Fig. 1. We identified 2,403 publications using the literature search strategy described above and additional searches through other sources. After excluding 2,186 records following the duplicate removed and the review of the titles and abstracts, 217 abstracts eligible for full-text evaluation remained. After full-text assessment we identified 143 publications that fulfilled the inclusion criteria.

Surgical treatment of acute generalized peritonitis from diverticulitis was described as early as 1910 by Lockhart-Mummery [19] who advocated washing the peritoneum and abdominal drainage, combined, if possible, with suture of the colonic perforation (Tables 4 and 5) [19‐113].

In the same decade Mikulicz described his two-stage technique of intestinal resection and anastomosis in a well-known article entitled “Surgical Experiences with Intestinal Carcinoma” presented for the first time to the Thirty-First Congress of the German Society of Surgery in 1903 [114]. Along with the description of details of his technique for intestinal resection, he also reported his personal experience in 106 patients, 16 of whom underwent a two-stage technique because he considered performing the initial anastomosis to be too hazardous for the treatment of intestinal cancer and so advocating to limiting the procedure in some cases to the resection and a double-barreled colostomy. Mikulicz strongly recommended this two-stage technique for all resections and anastomoses of the large and small bowel when the bowel was obstructed. This technique was then subsequently adopted for the treatment of diverticulitis.

In 1924 an observational study from the Mayo Clinic advocated drainage and suture of the colonic perforation as well as selective use of a diverting colostomy [115]. However, in some cases, fecal fistulas developed and some became chronic. In cases of substantial infection in or close to the colon, or in the presence of a colovesical fistula or fistula with other structures, the mortality decreased substantially with use of a diverting colostomy. However, in some cases the local suture of the perforation had been unsatisfactory because of difficulties visualizing the perforation, as well as difficulty in suturing edematous bowel wall. During the same decade Henri Hartmann proposed his surgical technique consisting of sigmoid resection, burying the rectal stump and performing terminal colostomy for the treatment of rectal cancer, as an alternative to abdomino-perineal resection, commonly called Hartmann’s procedure (HP) [116]. Hartmann had not originally advocated subsequent restoration of intestinal continuity.

In 1930 Rankin and Brown standardized the three-stage procedure developed by Mayo in 1907 [21, 115]. The first stage of the procedure consisted of peritoneal lavage, drainage of any abscess and creation of a proximal colostomy. The second stage was performed after a period of 2 to 4 months and involved resection of the sigmoid colon with end-to-end anastomosis. The third stage consisted of closure of the colostomy a few weeks after the second stage to ensure healing of the anastomosis.

In 1934 Lockhart-Mummery changed his original surgical technique, based only on peritoneal toilette and abdominal drainage by adding the use of a proximal diverting colostomy [23]. The right half of the transverse colon was used to create the colostomy. This approach was subsequently shown to leave the left colon and splenic flexure free from adhesions and favor the ensuing sigmoidectomy [117]. However, drainage and colostomy were associated with a higher mortality and morbidity because further leakage occurred from the site of perforation in spite of the proximal colostomy in some cases. Moreover, the inflamed colon could represent a source of sepsis and leaving this in situ was not seen as an attractive option. For these reasons, many surgeons, including Arhneim and Egger, favored the Mikulicz procedure [24, 118]. However, while most surgeons approved the theoretical advantages of Mikulicz’ exteriorization, they believed that this technique could be performed in very limited cases, because of the surrounding adhesions and edematous bowel [25, 119].

The three-stage procedure was the standard treatment of acute diverticulitis until the late 1940’s. “Palliative” treatment, consisting of formation of a colostomy alone, without subsequent resection, was almost partially abandoned during this period. Interval closure of the colostomy without colon resection led to a high proportion of aggravation of symptoms– over 45 and 70% of patients, in Smithwick’s and Pembertonet al.’s experiences, respectively [27, 28]. During this period, the use of the HP for the treatment of recto-sigmoid cancer decreased in favor of resection and primary anastomosis (PRA). At the same time, HP gained in popularity as a treatment of complications of DD and other emergency conditions while antibiotics were introduced into clinical practice. In 1950 Boyden articulated this approach and proposed a technical variation: a long distal bowel stump was exteriorized through the hypogastrium [29].

During this period, drainage of pus and formation of a proximal colostomy with the aim of controlling severe sepsis was not longer popular: fecal diversion was thought to limit peritonitis and surgeons avoided source control in fear of spreading infection. In the 50’s such ‘palliative’ operations were increasingly considered as unsatisfactory. The resection of the inflamed colonic segment followed by an anastomosis was suggested as a more ‘bold’ alternative. The improvement of anesthetic techniques and antibiotic therapy supported this approach of “eliminating the source of sepsis”, as stated by Crile in 1954 or “source control” as later coined by Marshall [120, 121]. It seems that a direct approach to the source of contamination, with diversion or resection of diseased segments, drainage of abscess, and suction of most of the pus and feces from abdominal cavity, gave the patient the best chance of survival [120]. Also during this period PRA with or without proximal protective colostomy was increasingly reported [32‐35]. The results were good in the presence of a minimum peri-diverticular contamination or intra-mesenteric abscess [34].

The gold standard surgical treatment of complicated diverticulitis was the three-stage procedure with the drainage and colostomy as first stage. Stauton and Smiley supported the diversion of the perforated loop with sigmoid colostomy in the left iliac fossa plus drainage and colostomy [49, 50, 59‐61]. Transverse colostomy was discouraged for fecal peritonitis as it might leave fecal residue proximal to the perforation [122]. Certainly the septic focus was removed from the peritoneal cavity, but the toxins were not removed from the circulation and still exerted their systemic effect [123, 124].

During the 1970s HP gained renewed popularity, because eradicating the source of sepsis had proved its superiority to mere diversion in terms of mortality and complications. Subsequent colostomy closure became an increasingly routine procedure, well standardized and progressively feasible [5, 84‐113].

In the 1980s a landmark systematic review by Krukowski and Matheson was published [125]. The two authors examined the mortality in 36 case series (821 cases of diverticulitis associated with purulent or fecal peritonitis) published between 1957 and 1984 that compared resection versus colostomy without resection. All patients underwent emergency surgery: mortality was 12% in the 316 who underwent resection versus 29% in the 505 who underwent colostomy without resection. Although there was a high risk of selection bias (patients in better health were more likely to undergo resection while patients with poor health status were more likely to receive a colostomy), this report found that, with antibiotics and better fluid resuscitation therapy, a substantial number of patients could undergo resection as an emergency procedure with an acceptable mortality rate. In addition, advocates of resection argued that emergency colectomy avoided the risk of missing colonic cancer (which was estimated to occur in 2–7% of the cases) and decreased morbidity [125].

The 1990s saw the emergence of two-stage resection and HP supported by two RCTs published in Denmark and France [126, 127]. Kronborg, in a single-center study published in 1993, examined 62 patients operated on for diverticulitis complicated with peritonitis, 46 of whom were classified as Hinchey III (purulent peritonitis). Twenty-one were randomized to transverse colostomy, suture and omentoplasty without resection with 100% survival. Six of the 25 patients (24%) randomized to acute resection without primary anastomosis died post-operatively [126]. Kronborg concluded that simple suture of the perforation and omentoplasty with proximal diversion was safer and more effective than acute resection in purulent peritonitis and comparable in effectiveness [126]. Hospital stay was shorter and there were fewer inflammatory relapses after acute resection. Twenty-seven different surgeons operated on the 62 enrolled patients during 14 years. This RCT was suspended early due to slow recruitment (an average of four patients each year) and used subgroup analysis without statistical adjustment, and consequently lacked statistical power [126]. In contrast, the French multicenter prospective RCT included 103 patients with either purulent (Hinchey III) or fecal peritonitis (Hinchey IV) [127]. The primary endpoint was post-operative peritonitis. For the 48 patients who were randomized to colostomy (with perforation closure by suture in the Hinchey IV cases), the postoperative peritonitis rate was high, up to 20%. In contrast, for the other 55 patients randomized to HP emergency resection, the postoperative peritonitis rate was significantly lower, less than 2%. Three studies were published from 1985 to 2000 where the HP was compared to the three-stage technique (Table 6) [126, 128, 129]. Our recent meta-analysis of these three studies analyzed the overall mortality as the primary outcome [130]. A total of 159 patients had colonic resection versus 105 who maintained perforated diseased segment of colon after proximal diversion or suture of the colon perforation. Overall, mortality was 13.6% (20/147) in the colonic resection group versus 24.6% (18/73) after proximal diversion or suture of the colon perforation (with perforated diseased segment of colon maintained). Statistical analysis failed to show a statistically significant lower overall perioperative mortality rate in the colonic resection group compared to the other group (OR 0.53, 95% CI 0.16 to 0.73, P = 0.31) and heterogeneity among the included studies was high (Tau² = 0.71, Chi² = 5.31, I² = 62%) [130]. In 2000, the American Society of Colon and Rectal Surgeons based on expert review of the evidence concluded that segmental colonic resection followed by an end colostomy (i.e., HP) was the most suitable procedure for perforated diverticulitis with peritonitis [131].

During this period a number of systematic reviews and meta-analysis were published comparing PRA versus HP (Table 7) [130, 132‐135].

The first published was Salem’s review in 2004 [132]. This review identified 98 studies (published between 1957 and 2003) on the surgical management of perforated diverticulitis complicated with peritonitis. Perioperative mortality data from patients with diverticular peritonitis undergoing HP (n = 1,051) was calculated for 54 studies. Overall cumulative mortality rate was 19.6% (18.8% for HP and 0.8% for subsequent procedures to restore intestinal continuity). The surgical site infection rate was 29.1% (24.2% for HP and 4.9% for reversal). Stoma complications and anastomotic leaks (after restoring intestinal continuity) occurred in 10.3% and 4.3%, respectively. Of 569 reported cases of PRA from 50 studies, the associated mortality rate was 9.9% (range 0–75%) with an anastomotic leak rate of 13.9% (range, 0–60%) and a surgical site infection rate of 9.6% (range, 0–26). In patients with diverticular peritonitis who underwent PRA the reported mortality and morbidity rates were not higher than those in patients undergoing HP, suggesting that PRA was a safe operative option in this specific population.

However, in 2006, Constantinides et al. published a systematic review of 15 observational studies (13 retrospectives and 2 prospective nonrandomized studies published from 1984 to 2004) comparing PRA with HP for acute diverticulitis in emergency surgery [133]. They found that peri-operative mortality was lower in those patients who underwent PRA compared with those who underwent HP. In addition, there was a trend favoring PRA for surgical complications (surgical site infections, abscesses, and peritonitis). However, it has to be borne in mind that this review was at high risk of selection bias because of the primarily retrospective character of case series Notwithstanding, these data showed that: 1) emergency PRA could be performed in selected cases with a low incidence of anastomotic leak (roughly 6%); 2) PRA and HP had similar operative times; and 3) for the most severe cases (Hinchey IV), PRA and the HP had similar mortality (14.1 vs. 14.4%).

In 2006, the American Society of Colon and Rectal Surgeons updated their guidelines for the treatment of sigmoid diverticulitis dating from 2000 [136]. Emergency sigmoid resection was deemed mandatory for diverticulitis with peritonitis, and the alternatives to the HP consisted of PRA with or without a de-functioning stoma. The role of the PRA (particularly without the use of a de-functioning stoma) remained unclear.

In 2007, Abbas published a systematic review of trials conducted between 1966 and 2003 [134]. Eighteen non-randomized studies reporting on 884 matched patients with complicated diverticulitis were included. There were no significant differences found between PRA and HP in terms of mortality, morbidity, sepsis, surgical site complications, duration of operation or antibiotic therapy. However, again the risk of selection bias was high.

Successively a RCT comparing PRA and HP was published [137]. The study protocol called for a de-functioning ileostomy within the PRA procedure. Ninety patients were randomized to PRA or HP in 14 centres in eight countries during a 9-year period [137]. Thirty-four PRA patients were compared to 56 HP patients. There were no statistically significant differences found in terms of age (P = 0.481), gender (P = 0.190), Hinchey stage III and IV (P = 0.394) and Mannheim Peritonitis Index (P = 0.145). There were no statistically significant differences found in mortality (2.9 vs. 10.7%; P = 0.247) or morbidity (35.3 vs. 46.4%; P = 0.38) after PRA or HP. The rate of restoration of intestinal continuity was similar in both groups (64.7% after PRA and 60% after HP, P = 0.659) after a similar lag time between emergency operation and elective stoma reversal (P = 0.43). The main difference between the two groups was the post-operative complication rate after restoration of intestinal continuity that differed statistically significantly (4.5 vs. 23.5% after PRA and HP, respectively; P = 0.05). However, it is impossible to draw firm conclusions from this RCT because of early termination and enrolment of only 15% of its calculated sample size (600 patients to achieve 90% power to detect 10% difference in mortality).

Another systematic review and meta-analysis on the same topic, published in 2013, compared PRA and HP [130] for the treatment of diverticulitis complicated by peritonitis, including the above mentioned RCT [137]. The overall morbidity rate was 17% (40/235) and 28.37% (84/296) in the PRA and HP groups, respectively (OR 0.46, 95% CI 0.23 to 0.90, P = 0.02). The re-intervention rate after PRA did not differ significantly between the two groups (15.2% versus 24.1% in the PRA and HP, respectively; OR 1.06, 95% CI 0.24 to 4.73, P = 0.94). Successively another RCT was published on this topic; Oberkofler and colleagues randomized 62 patients in four centres with acute perforation of left colon (Hinchey III and IV) to HP (n = 30) or PRA (with de functioning ileostomy, n = 32) with a planned procedure to restore intestinal continuity after 3 months in both groups [138]. Both groups were similar at baseline (Hinchey III: 76% vs. 75% and Hinchey IV: 24% vs. 25%, for HP vs. PRA respectively). The primary outcome was the overall complication rate that was similar in both groups (80% vs. 84%, P = 0.813). The outcomes after the primary colon resection were also similar (mortality 13% vs. 9% and morbidity 67% vs. 75% in HP vs. PRA). This is the first RCT that seems to favour PRA in patients with complicated diverticulitis with peritonitis. However, there is evidence of bias, as highlighted by Panis in his comments, therefore no firm conclusions can be drawn [139]. Succesively Lorusso published a systematic review including 24 studies, in [135] which reported the same our results (Table 8) [137, 138, 140‐162].

Recently laparoscopic peritoneal lavage (LPL) with drainage and antibiotics has been recently introduced into the surgical practice with aim to decrease the rate of HP [163, 164]. In 2009, Toorenvliet’s systematic review identified 231 patients with acute diverticulitis who underwent LPL, drainage and antibiotics therapy [165]. In 95.7% of patients this minimally invasive procedure permitted adequate control of the abdominal and systemic sepsis, with low rates of mortality (1.7%), morbidity (10.4%) and stoma (1.7%). Most patients subsequently had a delayed elective laparoscopic PRA. Patients who did not undergo subsequent resection had a long recurrence free period. The authors concluded that LPL was an effective and safe treatment of peritonitis secondary to perforated diverticulitis [165].

However, the use of peritoneal lavage without primary resection in generalized peritonitis originating from perforated diverticulitis remains controversial. Recently three RCT (DILALA-trial, SCANDIV-trial, LADIES trial) including a total of 343 participants (178 in the lavage group versus 175 in the resection group) have been published on this topic (Table 9) [166‐168].

The DILALA trial included patients with only Hinchey III peritonitis diagnosed by laparoscopy and with 1-year re-operation rate as primary outcome. The preliminary analysis of the short-term results (12 weeks) in 76 patients reported no statistically significant difference regarding morbidity and mortality, statistically significant longer period of abdominal drainage but shorter hospital stay in the LPL group compared to HP group [167].

LADIES was a two-arm trial with 1-year morbidity and mortality as the primary outcome. The LOLA arm compared laparoscopic lavage with sigmoidectomy in 90 patients with Hinchey III diverticulitis [166]. The trial could not prove the superiority of LPL and was terminated due to increased adverse events in this group despite the lack of statistical significance.

In contrast, the SCANDIV trial was able to report on the totality of 199 patients randomized to laparoscopic lavage or to laparoscopic/open resection with or without anastomosis [168]. The primary outcome was 90-day major complications rate according to the Clavien-Dindo classification. The authors reported a non-statistically significant higher incidence of the primary outcome in the LPL group and comparable mortality. However, there were statistically significantly higher rate of abscesses, secondary peritonitis and re-operations and in the LPL group along with missed malignancy in four cases. Despite the shorter operative time in the LPL group (72 vs 149 min), the hospital stay was similar in both groups. However, the study has several limitations such as inclusion of patients with Hinchey I and II and participation of more experienced surgeons in the resection group, which might be a source of significant bias [169].

Slim recently published a letter in which the three RCTs were examined and none showed laparoscopic lavage to be superior. In relation to three meta-analysis of these studies could respond to the question raised by this editorial, but “… came to opposite conclusions…..” [170].

We published the fourth meta-analysis, that failed to demonstrate significant benefits (Table 10) [171‐174]. Overall the quality of evidence was low because of serious concern regarding the risk of bias and imprecisions. A significantly increased rate of intra-abdominal abscess formation (RR = 2.54, 95% CI 1.34 to 4.83) (moderate quality of evidence), was seen with this approach. However, LPL does not appear inferior to traditional surgical resection and may achieve reasonable outcomes (lower rate of post-operative wound infections, R = 0.10, 95% CI 0.02 to 0.51) and less hospital resources (shorter duration of post-operative hospital stay during index admission, WMD =−2.03, 95% CI −2.59 to −1.47).