Findings in Relation to the Decisional Dilemma(s)

Most of the clinical questions posed by this systematic review (SR) about nonsurgical management of patients with acute colonic diverticulitis and medical and surgical interventions to prevent recurrence remain unanswered. Much of the evidence base is sparse and many of the studies, though of at least fair methodological quality, did not address the most pertinent clinical questions or were underpowered to effectively do so.

Strengths and Limitations

With few exceptions, the evidence base examined in this SR is sparse or of low SoE. As noted, many important clinical questions have not been addressed by sufficient numbers of studies that meet basic criteria (for most questions, comparative studies with appropriate adjustment for inherent differences between compared groups). Evidence is particularly sparse for questions related to the benefits and harms of CT scanning for acute diverticulitis, the appropriateness of outpatient management of uncomplicated or mildly complicated diverticulitis, interventional radiology for nonsurgical complicated diverticulitis, and various interventions for prevention of recurrent diverticulitis. In addition, there is very limited evidence regarding which patients might benefit most from (or be most harmed by) the various interventions. The lack of evidence about heterogeneity of treatment effects (which patients would most benefit), arguably, is most important for elective surgery because, despite the strong evidence of an important clinical benefit to surgery, clearly elective surgery cannot, and probably should not, be recommended for all patients with a history of acute diverticulitis. It is of paramount importance to determine criteria to establish who would most benefit.

Only for patients undergoing colonoscopy have studies systematically addressed which patients are at highest risk of outcomes. However, while the studies have found that older patients and those with recent complicated diverticulitis are at particularly high risk of CRC and advance colonic neoplasia, the studies comparing patients with diverticulitis to the general population have not evaluated whether younger patients or those with recent uncomplicated diverticulitis, specifically, are at higher risk of CRC than patients in the general population. Also, importantly, the studies have not adequately addressed whether patients who undergo colonoscopy after diverticulitis are at decreased risk of dying from CRC compared to patients who forgo colonoscopy. Ultimately, this is the primary unanswered clinical question pertaining to colonoscopy.

From a methodological perspective, it was common that studies were underpowered (too small) to address the most important clinical outcomes, failed to address the clinically important outcomes, or were inadequately analyzed. For many of the questions pertaining to treatment dilemmas, the randomized controlled trials (RCTs) tended to be too small (thus, underpowered) to detect differences between treatments in important, but relatively rare, clinical outcomes (such as treatment failure, unplanned emergency surgery, and death). The RCTs mostly evaluated less clinically important outcomes. Many of the nonrandomized comparative studies (NRCSs) were designed to be large enough to address at least some of the clinically important outcomes, but did not, or did not adequately, control for the inherent differences between groups. Thus, the findings of these NRCSs may have been biased toward findings that more intensive interventions are associated with worse outcomes (because the more intensive interventions were mostly used in the sicker patients who, by definition, are at highest risk of poor outcomes). Several of the colonoscopy and elective surgery studies were based on registries or administrative databases. However, these data sources are unlikely to be accurate or sufficiently granular about differences in disease severity across patients and other clinical factors such as patient comorbidities, not to mention patient preferences and life goals, which can influence the threshold for intervention (e.g., whether to undergo colonoscopy or to have elective sigmoidectomy).

We believe that our literature search was complete and did not systematically miss studies. We did not reject any study due to language restrictions or study setting (including country). It appears that the large majority of studies that were unavailable to us were conference abstracts, so we might have missed some cutting-edge studies. We restricted the evidence base to the past 30 years, based on changing diagnostic criteria for acute diverticulitis in the 1990s. We might have, thus, missed some important older studies that might still be pertinent. However, none of the stakeholders we collaborated with knew of such studies or were concerned by the choice of dates. While we restricted some study designs based on sample sizes, we do not think the smaller studies would have altered conclusions. Additional studies of the harms of elective surgery might have made our estimates more precise but are unlikely to have changed our overall conclusions that surgical complications are uncommon. Smaller comparative studies are highly unlikely to have been adequately analyzed. Our protocol did not cover all management decisions for the care of patients with acute diverticulitis or history of diverticulitis; for example, we did not address questions related to dietary restrictions during episodes of acute diverticulitis.

We were fairly liberal about decisions to perform meta-analyses. However, where one might have reasonably chosen not to meta-analyze studies (because of clinical heterogeneity of included studies or post hoc decision making), we explicitly point this out. We chose to use meta-analysis mostly as an indicator of possible effect (or of likelihood of an outcome or finding) rather than to provide precise estimates. In particular, for meta-analyses of colonoscopy findings (rates of findings) and elective surgical harms, we conducted meta-analyses to provide an indication of how common (or rare) outcomes are. For evaluations of elective surgery complications, we acknowledge that we did not adequately account for the differences across studies of surgery or patient characteristics. However, no clear patterns were seen across studies to explain the statistically large differences in surgical complication rates.

Applicability

The evidence base, even where insufficient to make conclusions about intervention effect, appeared to be generally applicable to patients with either suspicion of acute diverticulitis, diagnosed acute diverticulitis, or history of diverticulitis (depending on the evaluated intervention). Most studies (at least for nonsurgical interventions) described their eligibility criteria sufficiently to determine that the included participants are those for whom the intervention is potentially appropriate. However, many studies did not provide sufficient detail to understand the detailed level of severity of disease or of potential risk factors for poor outcomes. Arguably, more importantly, as described above, studies rarely evaluated subgroups (except for studies of colonoscopy) and failed to address heterogeneity of treatment effect. Such analyses could allow a better understanding of whom the findings are most applicable to. Many of the single group studies of elective surgery (often from registries or other large databases) did not clearly describe their included patients.

The one caveat about applicability in regard to patient or disease characteristics is that the large majority of studies were conducted in “western” countries, where left-sided diverticulitis is predominant. Only four studies were from East Asia (specifically South Korea and Japan), where right-sided diverticulitis is predominant.

Implications for Clinical Practice, Education, Research, or Health Policy

This review was nominated by the American College of Physicians to summarize the evidence base for a planned new clinical practice guideline on management of patients with diverticulitis. This goal informed the scope of the review to primarily address the needs of nonsurgical decision makers and patients. Unfortunately, many of the important questions about which interventions should be used for which patients remain either unanswered or answered with only low SoE. It is likely that many specific recommendations for management will be weak suggestions based largely on expert opinion. These include important questions related to benefits and harms of CT imaging, appropriateness of outpatient management of mild acute diverticulitis, interventional radiology for complicated diverticulitis, who needs antibiotic treatment and choice of antibiotics, whether colonoscopy is needed for patients under age 50 (particularly those with uncomplicated diverticulitis), what nonsurgical interventions are effective to reduce the risk of recurrence (and who would most benefit), and which patients should be referred for possible elective surgery to prevent recurrent diverticulitis.

Future Research

There is a clear need for high-quality research to address all these issues. Ideally, large-scale, multicenter RCTs should be conducted in unrestricted populations (i.e., without eligibility restrictions that may reduce applicability) with appropriate subgroup analyses. RCTs should be large enough to evaluate potential clinically important differences in rates of the most important outcomes to patients (e.g., death, treatment failure, emergency surgery, and time to recurrence) and important harms, adverse events, and complications (e.g., risk of C. difficile infection from antibiotics, which can be devastating for patients who already have diverticulitis; postoperative death; and permanent stomas).

Alternatively, large databases should be adequately analyzed to compare interventions. It is our strong belief that no (or rare) future studies should be considered that compare groups of patients who are inherently different without adequate adjustment for these differences. Unadjusted comparisons of, for example, hospitalized versus discharged patients or those who undergo or do not undergo percutaneous drainage of abscesses, can generally only conclude that sicker patients (who are, for example, more likely to be hospitalized or to undergo percutaneous drainage) fare worse. Ideally, propensity score analysis (or similar techniques) should be used. These analyses estimate the likelihood that each patient had one or the other intervention and control for this likelihood. They generally require relatively large numbers of patients for whom there is granular data about their risk factors for outcomes.

Furthermore, future studies should emphasize evaluations of heterogeneity of treatment effect to better understand which patients may most benefit from (or may be most harmed by) a given intervention. This can be done relatively simply with subgroup analyses, but more sophisticated evaluations may be appropriate. As for the NRCSs, it is important that the subgroup comparisons be adequately adjusted. For example, in a given set of patients, those with complicated diverticulitis may be fundamentally different from those with uncomplicated disease (beyond the presence or absence of abscesses).

Conclusions

Many questions remain inadequately answered regarding the best management of patients with acute diverticulitis or to prevent future recurrences. Prior reviews have demonstrated that CT imaging accurately diagnoses acute diverticulitis. For selected patients, outpatient management may be as effective an inpatient care. For patients with acute uncomplicated diverticulitis, it may be safe and appropriate to forgo antibiotics. The evidence base is inconclusive, though, about choice of antibiotic regimen for patients with complicated diverticulitis. The evidence is insufficient to assess the clinical value of percutaneous drainage. Patients with recent episodes of diverticulitis are at risk of having undiagnosed CRC or advanced colonic neoplasia, particularly if they are at least 50 years of age or have had complicated diverticulitis. The use of 5-ASA does not reduce (and may increase) the risk of recurrence of diverticulitis but is not more harmful than placebo. Patients with a history of complicated diverticulitis or who have smoldering or frequently recurring diverticulitis who undergo elective surgery are at greatly reduced risk of recurrent diverticulitis; serious surgery-related adverse events are uncommon. However, for elective surgery in particular, and for all other evaluated interventions, the evidence does not adequately address which patients would benefit most from a given intervention. There is a compelling need for future, well-conducted studies that address both effectiveness (and harms) of interventions and heterogeneity of treatment effect.