Diverticular Inflammation and Complication Assessment-guided prevention of acute diverticulitis: evidence, uncertainties, and the path ahead

Introduction Colonic diverticulosis is among the most common anatomical alterations in Western populations, with prevalence exceeding 60% in individuals over 70 years of age 1,2. While most cases remain clinically silent, a significant subset progresses to symptomatic uncomplicated diverticular disease (SUDD), and a smaller proportion develops acute diverticulitis 3. Although the lifetime risk of acute diverticulitis in individuals with diverticulosis is relatively modest – around 4–5% 4 – its impact on healthcare systems is substantial. In the USA alone, diverticular disease accounts for more than 300 000 hospital admissions annually, with direct costs exceeding 2 billion dollars 4. In Europe, the condition is associated with roughly 13 000 deaths each year 5. Given this burden, preventing the onset or recurrence of acute diverticulitis has long been an important clinical goal. Multiple strategies have been proposed – from high-fiber diets to cyclic poorly absorbed antibiotics, anti-inflammatory agents, and probiotics 6–11. Yet despite decades of research, no approach has consistently demonstrated robust preventive efficacy. One likely reason for these inconclusive results is inadequate patient stratification: individuals with SUDD are often treated as a homogeneous population despite marked heterogeneity in disease severity, extent, and risk. It was in this context that the Diverticular Inflammation and Complication Assessment (DICA) classification was introduced in 2015 12. As the first purely endoscopic scoring system for diverticular disease, DICA assesses extent of disease, number of diverticula, endoscopic signs of inflammation, and presence of complications, stratifying patients into DICA 1 (≤3 points), DICA 2 (4–7 points), and DICA 3 (>7 points) subgroups. This classification proved to be a simple and reproducible endoscopic scoring system for diverticular disease with an overall interobserver agreement ranging from good to very good 13. Subsequent studies suggested that higher DICA categories correlate with increased risk of complications, including acute diverticulitis and need for surgery 14. Its prognostic value was further improved through integration of clinical data into the combined overview on diverticular assessment (CODA) score 15 and in combination with fecal calprotectin 16. Despite these efforts and promising findings, DICA has not been incorporated into major European 17 or American guidelines 18,19, primarily due to limited external validation and lack of evidence that DICA-guided management improves outcomes. However, the recently published Global Guidelines, developed by 32 experts from 14 countries, extensively adopted the DICA classification 20. In this issue of the European Journal of Gastroenterology this difference did not reach statistical significance, likely due to the low number of events 29. The Global International Consensus therefore concluded that rifaximin has no proven role in acute diverticulitis prevention 20. Up to this point, both the existing evidence and the results of Tursi et al. 21 are in agreement. However, the suggestion of harm associated with rifaximin is unexpected and essentially implausible from a biological standpoint. High quality evidence suggests that both SUDD and acute diverticulitis are linked to shifts toward pro-inflammatory taxa 30–33. Rifaximin eradicates SIBO 27 and promotes beneficial genera such as Bifidobacteria, Lactobacilli, and Faecalibacterium prausnitzii 25,34 without disrupting global microbial diversity. A 10-day course of rifaximin has been shown to increase F. prausnitzii abundance in responders with inflammatory conditions including diverticular disease 35. While systemic antibiotics can disrupt the microbiota 36,37 and theoretically increase vulnerability to acute diverticulitis, rifaximin’s pharmacodynamic profile makes such harm mechanistically unlikely. A critical distinction must be made between statistical and clinical significance 38,39. A recent methodological study, involving more than 300 trials, reported discrepancies between statistical and clinical significance in up to 20% of studies 40. Statistical significance indicates only that observed differences are unlikely due to chance; it does not reflect the magnitude, clinical relevance, or biological plausibility of the effect. Study design, analytic approach, sources of bias, and underlying mechanisms must all be considered. Therefore, in this context it is worth conducting a closer re-evaluation of the study by Tursi et al. 21. Study design Although the original cohort was prospectively assembled, the present analysis is a post-hoc, non-randomized comparison of therapeutic strategies. Treatment allocation was discretionary, introducing substantial confounding by indication 41. Rifaximin was more frequently prescribed to symptomatic or moderately severe population (DICA 1–2), whereas milder cases often received no treatment. This imbalance is evident in table 1 of the paper by Tursi et al. 21. Such preexisting differences undermine causal inference. The progressive rise in acute diverticulitis incidence when moving from fiber treatment to mesalazine or rifaximin, and ultimately to their combined use, strongly indicates that treatment selection was driven by the underlying severity of the clinical condition. Statistical approach In this post-hoc analysis, the investigators employed an ‘entropy balancing’ approach 42,43 to mitigate the confounding that inevitably arises in non-randomized treatment settings. Entropy balancing constructs weights for the comparison group so that its covariate distribution matches, as closely as possible, that of the treated group. When applied, the weighted sample more closely resembles the treated population at baseline, allowing the Cox model to estimate hazard ratios under conditions that approximate covariate balance. Despite these apparent advantages, several limitations remain in this post-hoc analysis. Only variables explicitly included in the balancing procedure are adjusted for, leaving room for unmeasured confounding. For example, in the study by Tursi et al. 21, the weighting step included demographics, smoking status, DICA score, and a limited set of symptom scales. Important determinants of treatment selection – such as symptom severity, frequency of pain episodes, prior use of antidiarrheals or antispasmodics, adherence, physician preferences, and local prescribing practices – were not incorporated. The omission of these clinically relevant factors plausibly contributes to the unexpected findings reported. Indeed, when key prognostic variables are excluded, adjusted estimates may still carry substantial residual confounding, even when balance on observed covariates appears adequate. In the context of a Cox model, excluding such variables can also compromise the proportional-hazard assumption, as unmeasured factors can exert time-varying effects on risk. This produces unstable hazard ratio estimates or patterns that do not accurately reflect the underlying disease process. For these reasons, results derived from models that omit essential clinical information should be interpreted cautiously, with the understanding that no statistical adjustment – albeit sophisticated – can fully compensate for the absence of critical covariates. Although the authors emphasize treatment differences within DICA strata, no formal tests of interaction between DICA category and treatment were reported. Without such tests, claims of differential treatment effects across DICA categories remain exploratory at best and statistically unsupported. Furthermore, the analysis involves multiple treatments (n = 5), multiple strata (n = 3), and numerous pairwise hazard ratio comparisons. No adjustment for multiple comparisons was performed, increasing the likelihood of false-positive findings – particularly in subgroups with few events. Protopathic bias Rifaximin is often prescribed intermittently or ‘on demand’. If initiated in patients with prodromal symptoms of an impending acute diverticulitis episode, reverse causation (protopathic bias) 42,43 can occur. No lag-time or exclusion-window analyses were performed to address this concern. Outcome ascertainment Acute diverticulitis was recorded by local investigators without central adjudication or uniform radiologic confirmation. With only 140 acute diverticulitis events among 1945 patients, even modest outcome misclassification could markedly distort hazard ratios. Subgroup findings The harmful association of rifaximin (hazard ratio >1) contradicts established literature showing neutral or beneficial effects and mirrors baseline severity differences across groups. In DICA 2 patients, the apparent protective effect of mesalazine (hazard ratio 0.23) stems from small event numbers, wide confidence intervals, and potential overfitting. These patterns strongly suggest that residual confounding rather than pharmacologic differences explains the findings. Taken together, these limitations indicate that the association between rifaximin and increased acute diverticulitis risk is biologically implausible and methodologically fragile. Indeed, such paradoxical signals are most often methodological artifacts rather than true (and clinically relevant) effects. Conclusions and the path forward The study by Tursi et al. 21 is an interesting methodological exercise, exploring DICA-based risk stratification, but the observed association between rifaximin and acute diverticulitis raises serious questions and suggests: Confounding by indication and baseline disease severity Residual confounding from unmeasured variables Unreported or unstable weights in entropy balancing Lack of interaction testing Multiple comparisons without correction Limited event numbers with wide confidence intervals Protopathic bias The correct interpretation is not that rifaximin increases risk, but that, in real-world practice, it is more often prescribed to patients, who are inherently at higher risk of diverticulitis than those who receive no treatment. The dicothomy between the observed effect and the true expected effect is illustrated in Fig. 1.Fig. 1.: Observed harm most likely reflects treatment selection bias rather than pharmacological effect. HR, hazard ratio.Despite important advances in the management of acute diverticulitis, little progress has been made in its primary or secondary prevention. Existing studies on medical therapies are often of limited quality, and recommendations rely heavily on expert opinion rather than high-quality evidence 18,44–46. A recent systematic review found that RCTs evaluating medical (as well as surgical) interventions for diverticular disease are not robust. Changing a single-outcome event in most studies was sufficient to make a statistically significant study finding not significant 47. Since definitive RCTs are missing, at the present time we are drawing preventive conclusions from the bottom of the evidence pyramid (Fig. 2). Only large, well designed randomized controlled trials with adequate follow-up can provide definitive answers – yet such trials require support from regulatory authorities and industry.Fig. 2.: Hierarchy of evidence for preventive treatments of acute diverticulitis. RCT, randomized controlled trial.Independent studies must validate the prognostic accuracy of DICA in diverse populations. Mechanistic research should investigate the interplay between endoscopic severity, inflammation, and the microbiota in the transition from diverticulosis to diverticulitis. Only through such rigorous work can we determine whether DICA-guided prevention will remain an appealing theoretical construct – or evolve into a reliable tool for personalized care in diverticular disease. Acknowledgements We are indebted to Luigi Gatta, MD, PhD, FACG, AGAF (Oncologic Screening Unit, Azienda USL Toscana Nord-Ovest, ATNO, Livorno, Italy) for the helpful discussions and the valuable insights regarding the statistical analysis of the study results. Conflicts of interest C.S. has served as a speaker, consultant, and advisory board member for Alfasigma, Cosmo Pharmaceuticals and Dicofarm Group, and has in the past received funding from Giuliani Pharmaceuticals and Pfizer. For A.D., there are no conflicts of interest.