Diseases of the Colon & Rectum

, Volume 49, Issue 7, pp 966–981

Primary Resection With Anastomosis vs. Hartmann’s Procedure in Nonelective Surgery for Acute Colonic Diverticulitis: A Systematic Review

Authors

  • Vasilis A. Constantinides
    • Imperial College of Science, Technology and Medicine, Department of Surgical Oncology and TechnologySt. Mary’s Hospital
    • Imperial College of Science, Technology and Medicine, Department of Surgical Oncology and TechnologySt. Mary’s Hospital
    • Imperial College London, Department of Surgical Oncology and TechnologySt. Mary’s Hospital
  • Thanos Athanasiou
    • Imperial College of Science, Technology and Medicine, Department of Surgical Oncology and TechnologySt. Mary’s Hospital
  • Omer Aziz
    • Imperial College of Science, Technology and Medicine, Department of Surgical Oncology and TechnologySt. Mary’s Hospital
  • Sanjay Purkayastha
    • Imperial College of Science, Technology and Medicine, Department of Surgical Oncology and TechnologySt. Mary’s Hospital
  • Feza H. Remzi
    • Department of Colorectal SurgeryThe Cleveland Clinic Foundation
  • Victor W. Fazio
    • Department of Colorectal SurgeryThe Cleveland Clinic Foundation
  • Nail Aydin
    • Department of Colorectal SurgeryThe Cleveland Clinic Foundation
  • Ara Darzi
    • Imperial College of Science, Technology and Medicine, Department of Surgical Oncology and TechnologySt. Mary’s Hospital
  • Asha Senapati
    • Department of SurgeryQueen Alexandra Hospital
Article

DOI: 10.1007/s10350-006-0547-9

Cite this article as:
Constantinides, V.A., Tekkis, P.P., Athanasiou, T. et al. Dis Colon Rectum (2006) 49: 966. doi:10.1007/s10350-006-0547-9
  • 294 Views

Purpose

This study compares primary resection with anastomosis and Hartmann's procedure in an adult population with acute colonic diverticulitis.

Methods

Comparative studies published between 1984 and 2004 of primary resection with anastomosis vs. Hartmann's procedure were included. The primary end point was postoperative mortality. Secondary end points included surgical and medical morbidity, operative time, and length of postoperative hospitalization. Random effects model was used and sensitivity analysis was performed.

Results

Fifteen studies, including 963 patients (57 percent primary resection with anastomoses, 43 percent Hartmann's procedures), were analyzed. Overall mortality was significantly reduced with primary resection and anastomosis (4.9 vs. 15.1 percent; odds ratio = 0.41). Subgroup analysis of trials matched for emergency operations showed significantly decreased mortality with primary resection and anastomosis (7.4 vs. 15.6 percent; odds ratio = 0.44). No significant difference in mortality was observed in trials matched for severity of peritonitis Hinchey > 2 (14.1 vs. 14.4 percent; odds ratio = 0.85). Sensitivity analysis did not reveal significant heterogeneity between the studies for the primary outcome.

Conclusions

Patients selected for primary resection and anastomosis have a lower mortality than those treated by Hartmann's procedure in the emergency setting and comparable mortality under conditions of generalized peritonitis (Hinchey > 2). The retrospective nature of the included studies allows for a considerable degree of selection bias that limits robust and clinically sound conclusions. This analysis highlights the need for high-quality randomized trials comparing the two techniques.

Key words

Systematic reviewPrimary resection and anastomosisHartmann’s procedureDiverticulitisDiverticular disease

The incidence of colonic diverticulosis increases with age and is believed to affect approximately one-half of those older than aged 60 years.1 Inflammation of these diverticula occurs in approximately 20 to 30 percent of these individuals and may require hospitalization. Although the vast majority of cases can be managed medically, a quarter of patients with diverticulitis suffer from perforation of the inflamed segment resulting in generalized peritonitis and/or an intra-abdominal abscess, ultimately requiring surgical intervention.

Emergency surgery for perforated sigmoid diverticula most commonly involves a Hartmann's procedure (HP), in which the diseased segment of bowel is resected, distal rectal stump oversewn, and end colostomy formed.2 The procedure allows staged reanastomosis after complete resolution of pelvic inflammation (often 6 months after emergency surgery), thereby theoretically minimizing the risk of anastomotic leak and related postoperative morbidity. It is important to remember that staged reversal and restoration of bowel continuity has its own morbidity with reported anastomotic leak rates of up to 30 percent, and a reported mortality of up to 14.3 percent.28 Twenty to 50 percent of the patients will never have the colostomy reversed and adapt to living with a permanent stoma.9,10

Primary resection and anastomosis (PRA) of the affected bowel segment has been used in an attempt to minimize the morbidity and mortality associated with the reversal of Hartmann's procedure,3,11 with strategies such as a primary anastomosis and defunctioning stoma (PADS) or intracolonic lavage (PAIL) used to reduce the rate of anastomotic leak.12 The choice of operation is largely influenced by factors, such as the degree of peritoneal contamination, comorbidity, patient age, and mode of presentation. Hartmann's procedure often is performed in patients with several adverse prognostic indicators as perceived by the operating surgeon at the time of operation. PRA often is reserved for patients with more favorable comorbid criteria and a lesser degree of peritoneal contamination.11,13 A systematic review by Salem and Flum14 suggested that mortality and morbidity was not higher after PRA compared with HP; however, a limitation of the study was that it reported on aggregated results of adverse outcomes and did not attempt to perform a quantitative analysis comparing the two treatment modalities.

At present, the difference in outcome between PRA and HP procedures remains undefined, leaving the choice of procedure to the operating surgeon. This study used meta-analytical techniques to compare the two operative strategies with regard to postoperative morbidity, mortality, length of hospital stay, and operative time. It also was designed to identify any heterogeneity and explain any difference in estimates of the outcomes of interest between comparative studies.

Methods

Study Selection

A MEDLINE, Ovid, Embase, and Cochrane database search was performed on all studies published between 1984 and 2004 reporting on primary resection with anastomosis for acute colonic diverticulitis and comparing this to Hartmann's procedure. The following Mesh search headings were used: “comparative studies and diverticular disease/diverticulitis,” “comparative studies and peritonitis,” “diverticular disease/diverticulitis and primary and resection/anastomosis,” “one-stage and diverticular disease/diverticulitis and operation/resection,” “diverticular disease/diverticulitis and Hartmann's and procedure/operation.” The “related articles” function was used to broaden the search, and all abstracts, studies, and citations scanned were reviewed.

Data Extraction

Two reviewers (VC and PT) independently extracted the following data from each study: first author, year of publication, study population characteristics, study design, number of patients operated on with each technique, and end point data (operative time, length of postoperative hospital stay, postoperative complications, and mortality). Postoperative surgical complications included were: anastomotic leak, wound infection, secondary peritonitis/abscess, and stoma complications. Postoperative medical complications included were: septicemia, deep vein thrombosis, upper or lower gastrointestinal bleed, cerebrovascular accident, pulmonary complications, cardiac complications, and urinary complications.

Inclusion Criteria

To enter the analysis, studies had to: 1) report the operative results on the “colon” or use the words “splenic flexure, descending colon, sigmoid colon” to describe the study group; 2) compare PRA and HP operations; 3) report on at least one of the outcome measures mentioned below; 4) contain a previously unreported group (if patient material was reported more than once, we chose the most informative and recent article); and 5) when two studies were reported by the same institution, the analysis included the one of better quality or the most recent publication.

Exclusion Criteria

The following criteria were used to exclude studies: 1) studies in which the outcomes of interest (mentioned below) were not reported on the two techniques or it was impossible to deduce these from the published results; 2) studies that displayed a zero for the outcomes of interest in both PRA and HP groups; 3) studies in which no differentiation was made between right or left colon and type of operation performed; 4) studies in which no differentiation was made between diverticular disease and other pathology and type of operation performed; 5) studies in which no clear distinction was made between individual complications of diverticular disease other than abscess and/or peritonitis (obstruction, fistula, hemorrhage) and type of operation performed; and 6) studies in which the standard deviation of the mean or the range for continuous outcomes of interest (operative time and length of stay) was not reported.

Primary and Secondary End Points of Interest

PRA and HP were compared with regard to mortality (primary end point) and postoperative complications ranging from immediate to late, including wound infection, secondary peritonitis/abscess, stoma complications (comparing PADS cases with HP), and medical complications, including septicemia, deep vein thrombosis, upper and lower gastrointestinal bleed, cerebrovascular accident, pulmonary complications, cardiac complications, and urinary complications (secondary end points). Data were collected on operative time and postoperative length of hospital stay after PRA vs. HP.

Definitions

Mortality was defined as occurring within 30 days of the operation. Anastomotic leak was defined as the presence of contrast medium leakage at the level of the anastomosis during radiologic examination, presence of digestive contents in abdominal drain, or the finding of anastomotic separation during reoperation on the same admission. Wound infection was defined as presence of inflammation and/or purulent discharge and/or positive wound swab for bacterial growth. Secondary peritonitis or abscess was defined according to the clinical findings of abdominal guarding and/or rebound tenderness, pain, fever, diminished bowel sounds and/or intraoperative findings on reoperation. Stoma complications included prolapse, retraction, leakage, necrosis, herniation, or peristomal abscess. Septicemia was defined as the presence of compromised cardiovascular status combined with pyrexia and/or positive blood cultures. Pulmonary complications included atelectasis, pneumonia, adult respiratory distress syndrome, and respiratory failure. Cardiac complications included acute coronary syndrome, myocardial infarction, arrhythmias, heart failure, and cardiogenic shock. Urinary complications included urinary infection, retention, and renal failure. Length of stay was defined as days spent as an inpatient after the operation and excluding readmission for complications or reversal operations. Emergency operations were defined as those operations occurring within 48 hours of admission or if described using the word “emergency” or “urgent” in the study. Severity of peritoneal contamination was graded using the Hinchey classification15 as follows: Grade I = inflammatory phlegmon or abscess confined to the mesentery; Grade II = pelvic abscess; Grade III = purulent peritonitis; Grade IV = fecal peritonitis. The term “perforation” when mentioned in the text refers only to Hinchey Stages III and IV. Finally, operative time, deep vein thrombosis, gastrointestinal bleeding, and cerebrovascular accident were identified as mentioned in each study.

Statistical Analysis

Meta-analysis was performed in line with recommendations from the Cochrane Collaboration and the Quality of Reporting of Meta-analyses (QUORUM) guidelines and the proposal for reporting meta-analysis of observational studies in epidemiology (MOOSE), which was produced in Atlanta.1618 Statistical analysis for categoric variables was performed by using the odds ratio as the summary statistic. This ratio represents the odds of an adverse event occurring in the treatment (PRA) group compared with the reference (HP) group. An odds ratio of <1 favors the treatment group and the point estimate of the odds ratio is considered statistically significant at the P < 0.05 level, if the 95 percent confidence interval does not include the value, one.

Combination of the odds ratios of the outcomes of interest was performed with the Mantel-Haenszel chi-squared method. Yate's correction was used for those studies that contained a zero in one cell for the number of events of interest in one of the two groups.19,20 These “zero cells” create problems with the computation of ratio measure and its standard error of the treatment effect. This was resolved by adding the value 0.5 in each cell of the 2 × 2 table for the study in question, and if there were no events for both PRA and HP groups the study was discarded from the meta-analysis.

In the present study both fixed and random effect models were used. In a fixed effect model, it is assumed that the treatment effect in each study is the same, whereas in a random effect model it is assumed that there is variation between studies and the calculated odds ratio thus has a more conservative value.21,22 In surgical research, meta-analysis using the random effect model is preferable particularly because patients who are operated on in different centers have varying risk profiles and selection criteria for each surgical technique. In the tabulation of the results (Figs. 13), squares indicate point estimates of treatment effect (odds ratio), with 95 percent confidence intervals indicated by horizontal bars. The diamond represents the summary odds ratio from the pooled studies with 95 percent confidence intervals. For continuous variables, such as length of stay and operative time, statistical analysis was performed by using the random effect weighted mean difference as the summary statistic.23

The quality of the nonrandomized studies was assessed by use of the Newcastle-Ottawa Scale with some modifications to match the needs of this study.24 The quality of the studies was evaluated by examining three items: patient selection, comparability of PRA and HP groups, and assessment of outcome. Studies achieving four or more stars were considered as being of higher quality.

Three strategies were used to quantitatively assess heterogeneity. First, data were analyzed by using both random and fixed effect models. Second, graphic exploration with funnel plots was used to evaluate publication bias.22,23 Third, sensitivity analysis was undertaken using subgroups. To do this, the following variables were evaluated: 1) all studies; 2) study size (>50 patients in study); 3) year of publication (inclusive of or after 1990); 4) studies matched for operations performed as emergencies (<48 hours postadmission); 5) studies matched for purulent or feculent peritonitis (Hinchey > II); 6) studies of higher quality (≥4 stars); 7) studies matched for comorbidities; and 8) studies in which simple PRA was used in >50 percent of the total of the primary anastomoses performed.

To translate these results into benefits to clinical outcome, the following parameters were calculated: absolute risk reduction (ARR), which in this case is the difference in the incidence of postoperative complications between PRA and HP groups, and number needed to treat (NNT), which is the number of patients who must be treated (using PRA) to prevent one complication event (NNT = 1/ARR).

Analysis was conducted by use of the statistical software SPSS® version 12.0 for Windows (SPSS Inc., Chicago, IL), Intercooled Stata™ version 8.0 for Windows (Stata Corporation, College Station, TX), and Review Manager® Version 4.2 for Windows (The Cochrane Collaboration, Software Update, Oxford, UK).

Results

Twenty-four studies2548 published between 1984 and 2004 matched the search criteria, comparing PRA vs. HP and reporting the incidence of mortality, postoperative complications, operative time, or length of postoperative hospital stay.25434548 Nine studies were excluded because they contained a patient group undergoing PRA vs. HP that was indistinguishably mixed for cancer as well as diverticulitis.27,4148 Fifteen studies matched our selection criteria and were suitable for meta-analysis.25,26,2840 These included a total of 963 adult patients, of which 547 underwent PRA (57 percent) and 416 HP (43 percent). On review of the data extraction, there was 100 percent agreement between the two reviewers.

The characteristics of these studies are demonstrated in Table 1. The study design was retrospective in 13 studies and prospective, nonrandomized in 2. Nine studies contained at least 50 patients in both the PRA and HP groups and ten studies had a year of publication after 1990. Six studies were matched for the number of emergency operations performed in each group (defined as operation performed <48 hours from the time of admission). Eleven studies were matched for the number of patients presenting with an intra-abdominal abscess or peritonitis in each group (as opposed to other complications of diverticular disease, such as strictures, obstruction, or fistulas). Five studies contained PRA and HP groups that were matched for the severity of peritonitis when Hinchey Stages > II were considered, seven studies were considered to have a quality star rating of ≥4, and six studies contained PRA and HP groups that were matched for patient comorbidities. Finally, only 3 studies contained groups matched for American Society of Anesthesiologists (ASA) grade, and 11studies included a >50 percent proportion of simple PRA from the total of primary anastomoses performed.
Table 1

Characteristics of Included Studies

Study (yr)

Study Type

Cases

Age (yr)

Matching

Pathology

Hinchey Stage

Hinchey ≤2

Hinchey >2

PRA

HP

PRA

HP

PRA

HP

Alanis et al. (1989)25

R

34

26

d

4,5

DD, P, A

31

19

3

7

Blair and Germann (2002)26a

R

33

64

PRA 54, HP 65

3,9

DD, P, A

24

31

9

33

Drumm and Clain (1984)28

R

3

5

d

5

DD, P

0

0

3

5

Gooszen et al. (2001)29

R

32

28

PRA 63,HP 68

1,2,4,5,9,11

DD, P, A

11

9

21

19

Gregg (1984)30b

R

35

25

d

None

DD, O, P, A

d

d

29

17

Hold et al. (1990)31

R

99

76

d

9

DD, P, A

83

45

16

31

Kourtesis et al. (1988)32

R

23

10

d

None

DD, P, A

23

6

0

4

Medina et al. (1991)33

R

3

3

PRA 64, HP 79

5

DD, P

0

0

3

3

Peoples et al. (1990)34

R

11

43

d

None

DD, P, A

0

18

11

25

Regenet et al. (2003)35

PNR

27

33

PRA 65, HP 67

1–11

DD, P

0

0

27

33

Saccomani et al. (1993)36c

R

26

7

d

9

DD, P, A

d

d

d

d

Sarin and Boulos (1991)37

R

19

8

d

None

DD, O, P, A

15

1

4

7

Schilling et al. (2001)38

PNR

13

42

PRA 65, HP 68

1–6,8,9

DD, P

0

0

13

42

Underwood and Marks (1984)39

R

6

15

d

None

DD, P, A

5

6

1

9

Wedell et al. (1997)40

R

183

31

d

2

DD, P, A

169

16

14

15

       

361 (70.5%)

151 (29.5%)

154 (38.1%)

250 (61.9%)

PNR = prospective, nonrandomized; R = retrospective; PRA = primary resection and anastomosis; HP = Hartmann's operation.

Matching: 1 = age; 2 = gender; 3 = American Society of Anesthesiologists; 4 = comorbidity; 5 = Hinchey stage; 6 = Manheim Peritonitis Index; 7 = fever; 8 = inflammatory markers; 9 = emergency/elective surgery; 10 = seniority of surgeon; 11 = specialty of surgeon.

Pathology: P = peritonitis; A = abscess; O = obstruction; DD = diverticular disease.

aStudies in which patients who received bowel preparation were excluded.

bSix PRA and eight HP were performed in indeterminable Hinchey stages and/or in combination with obstruction or fistulas.

cUnable to determine Hinchey staging for each type of operation from data provided in paper.

dNo distinction was made between groups for variable in question.

The outcome measures reported by each of the included studies are shown in Table 2, and Table 3 shows the results from meta-analysis of the studies with regard to postoperative outcomes. This displays the number of studies reporting each outcome, total number of patients in both the PRA and HP groups, summary statistic of odds ratio or weighted mean difference, the 95 percent confidence intervals, and a test of heterogeneity between the studies with the calculated P values. Table 4 displays the type of PRA performed depending on Hinchey stage and operative urgency and shows the outcomes reported.
Table 2

Outcomes Measured

Study (yr)

Outcomes Measured

OT

LOS

MORT

AL

WI

SPA

SC

RO

SEPS

DVT

GIB

CVA

PC

CC

UC

Alanis et al. (1989)25

  

 

   

Blair and Germann (2002)26

 

 

 

       

Drumm and Clain (1984)28

  

            

Gooszen et al. (2001)29

   

 

Gregg (1984)30

 

           

Hold et al. (1990)31

  

 

        

Kourtesis et al. (1988)32

 

 

 

     

  

Medina et al. (1991)33

 

  

      

 

Peoples et al. (1990)34

  

            

Regenet et al. (2003)35

 

 

Saccomani et al. (1993)36

  

     

 

 

Sarin and Boulos (1991)37

  

           

Schilling et al. (2001)38

 

            

Underwood and Marks(1984)39

  

           

Wedell et al. (1997)40

  

           

OT = operative time; LOS = length of stay; MORT = mortality; AL = anastomosis leakage; WI = wound Infection; SPA = secondary peritonitis or abscess; SC = stoma complications; RO = repeat operation; SEPS = septicemia; DVT = deep vein thrombosis; GIB = gastric hemorrhage; CVA = cerebrovascular accident; PC = pulmonary complication; CC = cardiac complication; UC = urinary complications.

Table 3

Summary Statistics for Postoperative Outcomes for PRA vs. HP

 

ORa/WMD

95 Percent CI

No. of Adverse Events/Patients

No. of Studies

H (P Value)b

NNT/NNH

Overall mortality

0.41

0.22–0.77

963

15

0.13

10

Surgical complications

 Wound infection/dehiscence

0.42

0.2–0.9

341

6

0.37

8

 Postoperative abscess/peritonitis

0.43

0.19–0.97

458

6

0.72

21

 Stoma complications

0.51

0.2–1.36

212

4

0.75

NA

Medical complications

 Septicemia

0.48

0.16–1.46

180

3

0.45

NA

 Pulmonary complications

0.35

0.08–1.46

252

6

0.002

NA

 Cardiac complications

0.24

0.06–0.97

126

3

0.81

10

 Urinary complications

0.64

0.23–1.83

213

4

0.51

NA

Operative time

29.7

−3.35–62.74

NA

3

NA

NA

LOS

−2.07

−6.81–2.67

NA

5

NA

NA

OR = odds ratio; WMD = weighted mean difference (used for operative time and LOS); CI = confidence interval; H = between study heterogeneity; NNT = number needed to treat; NNH = number needed to harm; LOS = length of hospital stay; NA = not applicable.

aLess than 1.00 favor treatment with PRA.

bSignificant if < 0.05.

Table 4

Comparison of PRA Types and Studies Discriminating Between PRA Types for Hinchey Stage, Presentation, and Outcomes

Study (yr)

Type of PRA Performed

Hinchey Stage

Operative Urgency

30-Day Mortality

Surgical Complication

Medical Complication

Length of Stay

Operative Time

Total

Simple

PADS

PAIL

≤II

>II

<48 Hours

>48 Hours

Alanis et al. (1989)25a

34

28 (82.4)

5 (14.7)

1 (2.9)

S:28; D:3

S:1; D:2

NS

NS

S:1; D:0

Yes

Yes

S&L: 10.5 D:13.6

No

Blair and Germann (2002)26

33

28 (84.8)

5 (15.2)

0 (0)

NS

NS

NS

NS

No

No

No

No

No

Drumm and Clain (1984)28

3

3 (100)

0 (0)

0 (0)

NA

NA

NA

NA

NA

NA

NA

NA

NA

Gooszen et al. (2001)29

32

0 (0)

32 (100)

0 (0)

NA

NA

NA

NA

NA

NA

NA

NA

NA

Gregg(1984)30

35

23 (65.7)

12 (34.3)

0 (0)

NS

NS

NS

NS

No

No

No

No

No

Hold et al. (1990)31

99

70 (70.7)

29 (29.3)

0 (0)

S:59; D:25

S:11; D:4

NS

NS

S:2; D:2

Yes

No

No

No

Kourtesis et al. (1988)32

23

13 (56.5)

6 (26.1)

4 (17.4)

23

0

NS

NS

No

No

No

S&L: 16.5 D: 35

No

Medina et al. (1991)33

3

3 (100)

0 (0)

0 (0)

NA

NA

NA

NA

NA

NA

NA

NA

NA

Peoples et al. (1990)34b

11

NS

NS

NS

NA

NA

NA

NA

NA

NA

NA

NA

NA

Regenet et al. (2003)35

27

0 (0)

0 (0)

27 (100)

NA

NA

NA

NA

NA

NA

NA

NA

NA

Saccomani et al. (1993)36

26

15 (57.7)

8 (30.8)

3 (11.5)

NS

NS

NS

NS

S&L:0; D:0

No

No

No

No

Sarin and Boulos (1991)37

19

16 (84.2)

3 (15.8)

0 (0)

NS

NS

NS

NS

No

No

No

No

No

Schilling et al. (2001)38

13

13 (100)

0 (0)

0 (0)

0

13

NA

NA

NA

NA

NA

NA

NA

Underwood and Marks(1984)39

6

3 (50)

3 (50)

0 (0)

S:3;D:2

S:0; D:1

NS

NS

No

No

No

No

No

Wedell et al. (1997)40

183

148 (80.9)

35 (19.1)

0 (0)

S:138 D:31

S:10 D:4

NS

NS

S:1; D:1

No

No

No

No

PRA = primary resection and anastomosis; PADS = primary anastomosis and defunctioning stoma; PAIL = primary anastomosis with intracolonic lavage; S = simple; O = defaulting stoma; S&L = combinatio of simple PRA and PAIL; NS = not stated; NA = not applicable; LOS = length of stay; OT = operative times.

Data are numbers with percentages in parentheses unless otherwise indicated.

aSimple PRA and PAIL treated as one group.

bNo distinction made between PRA types.

Postoperative Mortality

All 15 studies reported on mortality, with results from this meta-analysis shown in Figure 1. Two studies showed a statistically significant reduction in mortality in the PRA compared with the HP group.36,40 Meta-analysis of all studies showed a significant difference in mortality of 4.9 percent (27/547) in the PRA group compared with 15.1 percent (63/416) in the HP group (odds ratio (OR) = 0.41; 95 percent confidence interval (CI), 0.22–0.77). Subgroup analysis of studies with at least 50 patients as sample size,25,26,2931,34,35,38,40 showed a significant difference in mortality of 4.5 percent (21/467) in the PRA group compared with 15.5 percent (57/368) in the HP group (OR = 0.36; 95 percent CI, 0.18–0.72). This finding was mirrored in studies with year of publication including and after 199026,29,31,3338,40 in which the mortality rate was 5.2 percent (23/446) in the PRA group compared with 16.4 percent (55/335) in the HP group (OR = 0.38; 95 percent CI, 0.18–0.78). Meta-analysis of all the studies matched for the number of emergency operations performed26,29,31,35,36,38 revealed a significant difference in mortality of 7.4 percent (17/230) in the PRA group compared with 15.6 percent (39/250) in the HP group (OR = 0.44; 95 percent CI, 0.24–0.83).
https://static-content.springer.com/image/art%3A10.1007%2Fs10350-006-0547-9/MediaObjects/NY00000547_FIG_01.gif
Figure 1

PRA = primary resection with anastomosis; OR = odds ratio; CI = confidence interval; n/N = number of deaths/total number of patients in group.

Subgroup analysis for studies matched for the presence of abscess or peritonitis25,26,28,29,31,3335,3840 revealed a similar significant difference in mortality of 5.2 percent (23/444) in the PRA group compared with 15.8 percent (58/366) in the HP group (OR = 0.43; 95 percent CI, 0.21–0.85). Meta-analysis of studies matched for severity of peritonitis Hinchey > 228,29,33,35,38 did not produce a significant difference in outcome with a mortality of 14.1 percent (11/78) in the PRA group compared with 14.4 percent (16/111) in the HP group (OR = 0.85; 95 percent CI, 0.36–2.01). Analysis of all the higher-quality studies26,29,31,35,38 revealed a marginally significant difference in mortality of 7.8 percent (16/204) in the PRA group compared with 14.8 percent (36/243) in the HP group (OR = 0.51; CI, 0.27–0.97).

The results for sensitivity analysis performed for studies comparing PRA vs. HP are shown in Table 5. Analysis of studies matched for patient comorbidities26,29,31,35,36,38 revealed a significant difference in mortality of 7.4 percent (17/230) in the PRA group compared with 15.6 percent (39/250) in the HP group (OR = 0.44; 95 percent CI, 0.24–0.83), whereas analysis of studies with a proportion of simple PRA > 50 percent25,26,28,3033,3638,40 showed a significant difference in mortality of 3.6 percent (17/471) in the PRA group compared with 14.8 percent (44/297) in the HP group (OR = 0.3; 95 percent CI, 0.13–0.69). Finally, for studies matched for ASA grade,26,35,38 analyses did not produce significant difference in mortality with 9.6 percent (7/73) in the PRA group compared with 15.1 percent (21/139) in the HP group. The absolute risk reduction for overall mortality between the PRA and HP groups was calculated to be 0.102, which can be translated in a number of patients needed to treat of 10.
Table 5

Sensitivity Analysis for Postoperative Outcomes

Outcome

Total No. of Patients

Total No. of Studies

ORa

95 Percent CI

P Value

HG P Value

High-quality studies ≥4 stars

 Wound infection/dehiscence

275

4

0.3

0.09–0.98

0.047

0.15

 Postoperative peritonitis/abscess

452

5

0.38

0.16–0.88

0.02

0.88

 Pulmonary complications

180

3

0.3

0.02–3.87

0.35

0.003

 Mortality

567

7

0.45

0.24–0.82

0.009

0.8

 Studies published after 1990

 Wound infection/dehiscence

248

4

0.5

0.23–1.08

0.08

0.43

 Postoperative peritonitis/abscess

398

5

0.47

0.2–1.09

0.08

0.68

 Pulmonary complications

159

4

0.77

0.27–2.2

0.62

0.77

 Mortality

781

10

0.38

0.18–0.78

0.009

0.09

Studies with sample size >50 cases

 Wound infection/dehiscence

275

4

0.3

0.09–0.98

0.047

0.15

 Postoperative peritonitis/abscess

452

5

0.38

0.16–0.88

0.02

0.88

 Pulmonary complications

180

3

0.3

0.02–3.87

0.35

0.003

 Mortality

835

9

0.36

0.18–0.72

0.004

0.14

Studies matched for operative urgency

 Wound infection/dehiscence

248

4

0.5

0.23–1.08

0.08

0.43

 Postoperative peritonitis/abscess

392

4

0.41

0.17–0.98

0.05

0.85

 Pulmonary complications

153

3

0.85

0.28–2.58

0.78

0.71

 Mortality

480

6

0.44

0.24–0.83

0.01

0.48

Studies matched for Hinchey >II

 Wound infection/dehiscence

120

2

0.21

0.06–0.77

0.02

0.95

 Postoperative peritonitis/abscess

126

3

0.68

0.2–2.33

0.54

0.54

 Pulmonary complications

126

3

0.7

0.22–2.28

0.55

0.6

 Mortality

189

5

0.85

0.36–2.01

0.71

0.66

Studies matched for comorbidities

 Wound infection/dehiscence

248

4

0.5

0.23–1.08

0.08

0.43

 Postoperative peritonitis/abscess

392

4

0.41

0.17–0.98

0.05

0.85

 Pulmonary complications

153

3

0.85

0.28–2.58

0.78

0.71

 Mortality

480

6

0.44

0.24–0.83

0.01

0.48

Proportion of simple PRA > 50%

 Wound infection/dehiscence

221

4

0.53

0.2–1.46

0.22

0.3

 Postoperative peritonitis/abscess

338

4

0.37

0.13–1.07

0.07

0.46

 Pulmonary complications

132

4

0.2

0.02–1.25

0.09

0.05

 Mortality

768

11

0.3

0.13–0.69

0.005

0.1

Studies matched for ASA

 Mortality

212

3

0.59

0.23–1.5

0.26

0.7

OR = odds ratio; WMD = weighted mean difference (used for continuous variables); CI = confidence interval; HG = between study heterogeneity; ASA = American Society of Anesthesiologists; NA = not applicable.

aLess than 1.00 favor treatment with PRA.

A “funnel plot” of the 15 studies used in the meta-analysis for postoperative mortality is shown in Figure 2. This is scatter plot of the treatment effects estimated from individual studies on the horizontal axis (OR), against a measure of mortality on the vertical axis (standard error of the logarithm of the odds ratio (SE[logOR])). One of the studies lay outside the limits of the 95 percent confidence intervals but exerted no significant publication bias on the overall effect, as proved by Begg's and Egger's tests (P = 0.98). When tested, the heterogeneity between studies was not statistically significant in any of the subgroups as well as in the overall group (P value for heterogeneity 0.13; Table 4). Meta-regression analysis was performed to establish the relative importance and significance of any covariates on postoperative mortality. Presence of peritonitis (P = 0.98), severity (Hinchey ≤2, P = 0.063; and Hinchey > 2, P = 0.094) and emergency surgery (P = 0.377) were not found to exert significant influence on outcome. To analyze the possible heterogeneity arising from the different types of PRA (simple, PADS, and PAIL), further meta-regression analysis was performed and did not reveal a significant influence of PRA type on mortality (Simple PRA, P = 0.775; PADS, P = 0.861; PAIL, P = 0.934).
https://static-content.springer.com/image/art%3A10.1007%2Fs10350-006-0547-9/MediaObjects/NY00000547_FIG_02.gif
Figure 2

SE[logOR] = standard error of the logarithm of the odds ratio; OR = odds ratio.

Power Calculation for the Primary Outcome of Interest

The overall incidence of postoperative mortality between studies in the HP group was 63 of 416 (approximately 15 percent). To rule out a 30 percent relative risk reduction (from 15 to 10.5 percent) with a 5 percent significance level and 80 percent power, a traditional randomized, controlled trial would require 906 patients in each arm.

Postoperative Surgical Complications

Results for meta-analysis of surgical complications are represented as a forest plot (Fig. 3). The incidence of wound infection and postoperative abscess/peritonitis was found to be significantly less in the PRA groups (OR = 0.42 and 0.43, respectively) whereas there was no difference in the incidence of stoma complications (OR = 0.51; 95 percent CI, 0.2–1.36).
https://static-content.springer.com/image/art%3A10.1007%2Fs10350-006-0547-9/MediaObjects/NY00000547_FIG_03.gif
Figure 3

PRA = primary resection with anastomosis; OR = odds ratio; CI = confidence interval; n/N= number of complications/total number of complications in group.

Sensitivity analysis (Table 5) revealed a significantly lower wound infection rate with PRA when studies of high-quality, sample size > 50, and those matched for severity of peritonitis were considered. There was, however, no significant difference when studies with a proportion of simple PRA > 50 percent, or those matched for emergency operations and patient comorbidities were considered. In the case of postoperative peritonitis, considering studies of high quality, sample size > 50, matched for emergency operations, and patient comorbidities significantly favored the PRA group. This was not the case when studied with proportion of simple PRA > 50 percent and those matched for Hinchey > II were considered.

Postoperative Medical Complications

Results for meta-analysis of medical complications are represented as a forest plot (Fig. 4). Only cardiac complications were found have marginal significance in favor of the PRA group (OR = 0.24; 95 percent CI, 0.06–0.97). No difference was found between the groups in the incidence of septicemia, pulmonary, and urinary complications.
https://static-content.springer.com/image/art%3A10.1007%2Fs10350-006-0547-9/MediaObjects/NY00000547_FIG_04.gif
Figure 4

PRA = primary resection with anastomosis; OR = odds ratio; CI = confidence interval. n/N = number of complications/total number of complications in group in group.

Operative Time and Length of Stay

Analysis of the three studies29,35,38 reporting on operative time and five studies25,26,29,32,33 reporting on length of stay did not reveal a significant difference between the two groups.

Discussion

The results of this study suggest that patients selected for PRA have a significantly reduced postoperative mortality compared with HP in the emergency setting. On consideration of PRA and HP performed under conditions of purulent or fecal peritoneal contamination (Hinchey > 2), no significant difference in mortality was found. This may be the reflection of the careful selection of patients undergoing PRA that may form a group of particularly good prognostic outlook and are hence able to withstand the increased risk of anastomotic leakage and other surgical complications. Whether an anastomosis per se can reduce mortality is unclear and needs to be the subject of further research. In the case of PRA, we were unable to identify any beneficial effect of intracolonic lavage or fecal diversion on postoperative mortality. Significant patient selection bias because of the retrospective nature of the included studies limits extrapolation of the results to the clinical setting, because, traditionally, patients with favorable physiology tend to undergo PRA, with HP reserved for patients more likely to suffer a postoperative complication.

The finding that patients selected for PRA had a lower incidence of wound infection compared with HP is subject to significant patient selection bias. It also may be related to the fact that in the latter, cross-contamination from the colostomy is a definite risk. In the case of PADS, a similar cross-contamination risk exists. Interestingly, the overall reduction in wound infection rates after PRA seemed to diminish as the proportion of simple PRA procedures increased, suggesting that PADS and PAIL may have some effect on reducing wound infection after PRA. The same applies to postoperative peritonitis in whichthe overall comparison favored PRA, but this beneficial effect was not seen in selected subgroups in the sensitivity analysis. When considering postoperative complications, such as medical adverse events, operative time, and length of hospital stay, it is important to remember the small number of patients reporting each complication and its effect on the results.

The PRA procedure has the risk of anastomotic leakage compared with HP, which was reported at 5.5 percent (30/543) in this study. This could not be compared with peristomal leaks or leaks from the rectal stump after HP because the outcome was not reported by the included studies. Factors that have been thought to predispose to breakdown of colonic anastomoses include a loaded colon at the time of anastomosis45,49 and the presence of bacterial peritonitis.46 In an attempt to reduce the incidence of anastomotic leak, PAIL32,35,36 and PADS25,26,2932,36,37,39,40 procedures were used by several groups. The studies did not, however, separate the outcome from patients in whom these techniques were used compared with those in whom simple PRA was performed (Table 4). It was, therefore, impossible to conclude whether intraoperative bowel irrigation or fecal diversion have an impact on anastomotic leakage, which needs to be the subject of future randomized, controlled research.

The finding that operative time was not significantly different between PRA and HP suggests that the latter may not be justified only on the basis of performing a “quicker operation” in the emergency setting. It is important to note that no distinction was made between PADS or PAIL and simple PRA. The former may considerably increase the operative time taken to perform PRA. This is because in PADS a proximal defunctioning stoma has to be formed, and in the case of PAIL, a Foley catheter has to be inserted through the appendix and the colon irrigated with copious amounts of saline until the effluent is clear. The results of the present study, therefore, may not represent the “true” difference between PRA and HP procedures.

Similarly the results of this study suggest that HP cannot be justified over PRA solely on the basis of length of postoperative stay, which were not significantly different when comparing the two groups. It also is important to note that in the case of HP, restoring bowel continuity requires subsequent readmission and reoperation, both of which are associated with further time and cost as well as the risk of postoperative adverse events.

Statistically quantitative comparisons, such as those included in this study, have several limitations that must be taken into account when considering its results. Despite the efforts for standardization, outcome measures were less well-defined, which limited the proportion of studies for which outcomes were comparable. It was not possible to match all patient groups for patient age, ASA grade, comorbid conditions, operative urgency, and Hinchey stage, all of which are factors known to affect outcome for patients undergoing emergency surgery for diverticulitis. Neither the allocation of treatment nor the assessment of outcome was blinded, and it is important to consider publication bias, particularly in meta-analytic research based on published studies as well as variation in inclusion criteria, treatment protocols, operative technique, and outcome assessment between studies. Finally, 13 of 15 studies were retrospective in nature, and thereby allocation of treatment was based on the operating surgeon's judgment once access to the abdominal cavity was gained. It is obvious that for cases in which peritoneal contamination was Hinchey stage > 2, and patients in whom perioperative physiologic variables were unfavorable,11,13 a Hartmann's procedure was almost always performed. This source of selection bias is unfortunately inherent to the studies and cannot be reliably controlled for with statistical methods.

This study could not account for the experience of the operating surgeon and its effect on postoperative outcome, which is an important consideration, because it has been suggested that PRA is a more technically demanding procedure. The effect of specialization on emergency surgery for diverticulitis also is important, because a colorectal surgeon, for example, may be more likely to perform PRA compared with HP.50

Conclusions

This article attempts to highlight the individual merits and weaknesses of PRA compared with HP as the primary emergency surgical treatment for acute diverticulitis. Although the results of this study are marred by patient selection bias, patients selected for PRA have been shown to have comparable mortality to those treated with HP in the setting of peritonitis stages Hinchey > 2. Furthermore, patients selected for PRA have a lower mortality than those treated with HP in the emergency setting. Clinical extrapolation of the results should, therefore, be done with extreme caution until high-quality, randomized studies become available to clarify the issue. The use of statistical tools, such as sensitivity analysis and meta-regression, have enabled us to control for the effect of important factors on mortality: namely, the presence and severity of peritonitis, mode of operation, and the type of PRA performed, all of which had no direct impact on the mortality difference reported between the two groups. However, we have not been able to control for the patients intraoperative physiologic status, which is clearly an important factor that would determine the choice of procedure. For example, a patient suffering intraoperative cardiac ischemia would be more likely to undergo an HP by the operating surgeon, even if the initial intent would be to perform a PRA. Previously mentioned limitations, including a small sample size, mean that a randomized, controlled trial comparing PRA and HP in patient groups who are matched for age, gender, ASA grade, and degree of peritonitis and intraoperative physiologic status, taking into account the impact of proximal diversion and intraoperative colonic lavage, is eagerly awaited.

Copyright information

© The American Society of Colon and Rectal Surgeons 2006