Predictors for colectomy in patients with acute severe ulcerative colitis: a systematic review and meta-analysis

Abstract

Objectives

Acute severe ulcerative colitis (ASUC) poses challenges to patient management owing to its high surgical rate. This study aimed to identify predictors of colectomy in patients with ASUC.

Design

This is a systematic review and meta-analysis.

Data sources

PubMed and Web of Science were searched up to April 2024.

Eligibility criteria

Studies on the predictors of colectomy in adult patients with ASUC were eligible.

Data extraction and synthesis

Two reviewers independently extracted the data using a prespecified data collection sheet. A qualitative synthesis was performed in tabular form. Random-effect meta-analyses were conducted using OR and 95% CI.

Results

Forty-two studies were included in the systematic review. The reported variables can be categorised into biomarkers, auxiliary examination findings, demographic and clinical characteristics, and drug factors. Through meta-analysis, albumin (OR 0.39 (95% CI 0.26 to 0.59) per 1 g/dL increment, I²=0.0%), high C reactive protein level (2.63 (1.53 to 4.52), I²=29.6%), high erythrocyte sedimentation rate level (2.92 (1.39 to 6.14), I²=0.0%), low haemoglobin level (2.08 (1.07 to 4.07), I²=56.4%), fulfilling the Oxford criteria (4.42 (2.85 to 6.84), I²=0.0%), extensive colitis (1.85 (1.24 to 2.78), I²=47.5%), previous steroids (1.75 (1.23 to 2.50), I²=17.7%) or azathioprine (2.25 (1.28 to 3.96), I²=0.0%) use, and sarcopenia (1.90 (1.04 to 3.45), I²=0.0%) were identified as valuable predictors for colectomy within 1 year. The ulcerative colitis endoscopic index of severity (OR 2.41 (95% CI 1.72 to 3.39), I²=1.5%) was the only predictor found to predict colectomy over 1 year.

Conclusion

Identification of these predictors may facilitate risk stratification of patients with ASUC, drive personalised treatment and reduce the need for colectomy.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Acute severe ulcerative colitis (ASUC) presents challenges in patient management with its high incidence and surgical rate.

• Despite various predictors of colectomy, there is a lack of systematic review of their predictive efficacy, which is important for improving clinical decision-making.

WHAT THIS STUDY ADDS

• Meta-analysis revealed that albumin levels, elevated C reactive protein, increased erythrocyte sedimentation rate, decreased haemoglobin, fulfilment of the Oxford criteria, extensive colitis, prior use of steroids or azathioprine, and sarcopenia are significant predictors of colectomy within 1 year.

• The ulcerative colitis endoscopic index of severity emerged as the sole predictor for colectomy beyond 1 year.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Identification of significant predictors for colectomy may facilitate risk stratification of patients with ASUC and drive personalised treatment, thereby reducing the need for colectomy.

Introduction

Acute severe ulcerative colitis (ASUC), characterised by bloody stools and systemic symptoms such as fever, tachycardia and anaemia, is a potentially life-threatening condition of ulcerative colitis (UC).¹ Approximately 25% of patients with UC will experience such severe attacks at least once during their lifetime,^{2 3} with a projected mortality of approximately 1% and colectomy rates of approximately 30% during the first hospitalisation.^{2 4}

Intravenous corticosteroids have been considered the cornerstone of therapy for hospitalised individuals with ASUC,¹ while medical rescue therapy, including infliximab (IFX) and ciclosporin, is regarded as the first line that can be implemented to avoid emergency colectomy as much as possible for patients with steroid-refractory ASUC.^{1 5} However, colectomy as a definitive treatment for ASUC should be considered seriously from the beginning of admission. The increased risk of postoperative complications and mortality is associated with delayed or emergency surgery.⁶ Therefore, early identification of patients at high risk for colectomy is beneficial for clinicians to adopt targeted and more active medical treatment to prevent disease progression. Additionally, in the case of unavoidable surgery, it enables proactive decision-making regarding the right timing and method of surgery, with adequate preoperative preparation, rather than performing emergency surgery.

Previous studies have proposed several variables that can predict the outcomes of colectomy in patients with ASUC, including clinical symptoms, biomarkers, endoscopic or imaging parameters, and drug factors. Among these, C reactive protein (CRP), albumin, erythrocyte sedimentation rate (ESR) and faecal calprotectin are the most frequently studied variables with prognostic potential.^{6 7} However, the predictive power of some risk factors for colectomy has been mixed across studies, and the lack of systematic studies to address these controversies has hindered the use of these prognostic factors in clinical practice.⁷

Therefore, this study aimed to systematically review the predictors of colectomy in patients with ASUC identified in the existing literature and to further estimate the associations between these predictors and colectomy risk by meta-analysis.

Methods

Registration and reporting

This systematic review and meta-analysis is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses 2020 statement (online supplemental table S1).⁸ The study protocol was registered in the International Prospective Register of Systematic Reviews (registration number CRD42022361905).

Search strategy

We systematically searched PubMed and Web of Science, and the final search was conducted in April 2024. Additionally, the reference lists of the enrolled studies and related reviews were manually searched as supplements. The search terms included UC, colectomy, prediction, prognosis and their corresponding synonyms. The search was restricted to articles published in English. Document types other than original articles, such as review papers, conference abstracts, letters and editorial materials, were filtered. Online supplemental table S2 provides comprehensive details of the search strategies used.

Study selection

Literature pertaining to the predictors of colectomy in adult patients with ASUC was deemed eligible, including both cohort and case-control studies with full texts. A diagnosis of ASUC was made based on the Truelove and Witts criteria, which were defined as six or more bloody stools per day and at least one manifestation of systemic inflammation, including fever, tachycardia, anaemia or increased inflammatory markers,⁹ or physician assessment in cases where patients with UC exhibited severe clinical symptoms. We did not impose any restrictions on the baseline settings of the included studies, and any pharmaceutical intervention subsequent to baseline was allowed. The primary outcome of our study was the occurrence of colectomy within a 1-year timeframe after baseline, and the secondary outcomes included colectomy that occurred within 90 days and during follow-up of more than 1 year.

To determine whether the retrieved articles met the pre-established eligibility criteria, two reviewers (ZF and RC) independently screened the titles and abstracts and then conducted comprehensive reviews of the full texts of the selected articles. Discrepancies were resolved through consultation to reach a consensus.

Data extraction

Two reviewers (ZF and JZ) independently extracted the data from the enrolled studies using a prespecified data collection sheet. The data recorded included the principal author, publication year, study design and data source, country of the study, sample size, participant characteristics, definition of ASUC, outcomes and follow-up period, risk factors studied, concomitant medications and measures of effect such as OR and its corresponding 95% CI. Discrepancies in extracted information were resolved by a third reviewer (RC) through adjudication.

Study quality assessment

The Newcastle-Ottawa Scale was used to evaluate the quality of the studies, which comprised three dimensions as follows: selection of the population (four items), comparability of the study groups (one item) and the outcome assessment (three items). Each item was assigned a maximum score of 1, except for comparability, which was assigned a maximum score of 2. Studies with total scores ranging from 4 to 6 and 7 to 9 were considered of moderate and high quality, respectively, and were included in subsequent analyses. The assessments were independently completed by two reviewers (ZF and DW). Disagreements were resolved through discussion.

Data synthesis and statistical analysis

To evaluate the overall predictive effects of the candidate predictors proposed in the included studies on colectomy in patients with ASUC, data synthesis was conducted separately for different predictors and outcomes. We categorised outcomes into colectomies occurring within 1 year, within 90 days and after more than 1 year while disregarding the potential variances within these respective periods. For predictors with available ORs and corresponding 95% CIs in two or more studies, we performed a random-effect meta-analysis to calculate the overall ORs and 95% CIs. The ORs for continuous and categorical variables were pooled separately, whereas categorical variables with slightly different cut-off values were considered the same variable. Cochran’s Q test and the I² statistic were used to assess the presence and extent of statistical heterogeneity among the studies, respectively. An I² statistic >50% or p≤0.10 indicated substantial heterogeneity.¹⁰ For risk factors mentioned in only one study or lacking sufficient data for OR calculations, the area under the receiver operating characteristic curve, sensitivity, specificity or the mean/median differences of the risk factors between two patient groups (those undergoing colectomy vs those not) were summarised via tabulation and descriptive analysis.

Given that Truelove and Witt’s criteria have long been the gold standard for diagnosing ASUC, we performed a sensitivity analysis of predictors from studies employing Truelove and Witt’s criteria for diagnosis to test the robustness and reliability of the results of the primary outcome.

All analyses were conducted using Stata MP V.16.0.

Results

Study selection and characteristics

We retrieved 10 753 records from the databases (5944 from PubMed and 4809 from the Web of Science), with 8175 records remaining after the removal of duplicates. Through title and abstract screening, 80 articles were deemed eligible and subjected to full-text evaluation. Ultimately, 42 articles were included in this review, of which 35 had sufficient data to be included in the meta-analysis¹¹,⁴⁵ (online supplemental figure S1).

As shown in table 1, four of the eligible studies had prospective designs,^{24 28 30 46} and the rest were retrospective analyses. Colectomy occurred within 1 year in 27 studies,¹¹,²³³⁴ and during a follow-up period of more than 1 year in 10 studies.²⁵²⁷,32 38 50 51 Five studies^{24 26 33 43 52} reported both 1-year and long-term outcomes. Sample sizes ranged from 18 to 489 patients. Thirty studies used the Truelove and Witts criteria to define ASUC,¹¹,^{1517 20} whereas 12 studies diagnosed ASUC based on severe acute clinical attacks.^{13 15 16 18 19 23 25 28 33 34 38 40 42 50} The detailed study characteristics are provided in online supplemental table S3. Given the random effect analysis, the pooled rates of colectomy within 90 days, 90 days to 1 year, and more than 1 year of the included studies were 0.21 (95% CI 0.18 to 0.24), 0.29 (95% CI 0.25 to 0.34) and 0.38 (95% CI 0.32 to 0.45), respectively (online supplemental figure S2).

Table 1. Summary of eligible studies.

Characteristics	Included studies (n=42)
Region
America	8 (19.0)
Asia	8 (19.0)
Europe	17 (40.5)
Others	9 (21.4)
Study design
Retrospective	38 (90.5)
Prospective	4 (9.5)
Case-control	4 (9.5)
Cohort	38 (90.5)
Sample size	112.5 (18–656)
Definition of acute severe ulcerative colitis
Truelove and Witts criteria	30 (71.4)
Other criteria	12 (28.6)
Follow-up period
≤12 months	14 (33.3)
>12 months	28 (66.7)
Concomitant medications
Corticosteroids	38 (90.5)
Infliximab	26 (61.9)
Ciclosporin	23 (54.8)
Others	15 (35.7)

Data were presented as counts N (percentage) or median (range).

Study quality assessment

Online supplemental table S4 shows the results of the quality assessment using the Newcastle-Ottawa Scale for each study. All included studies had a total score between six and nine, with one and 41 studies study rated as moderate³⁸ and high quality, respectively.

Predictors for colectomy available for meta-analysis

Predictors for colectomy within 1 year

In 29 studies, multiple candidate predictors for colectomy within 1 year among patients with ASUC were identified, including biomarkers, past or current drug factors, demographic and clinical characteristics, and other auxiliary examination results.

Figure 1A illustrates the overall predictive power of biomarkers, including albumin, CRP, ESR and haemoglobin, for the risk of colectomy within 1 year. Four studies^{14 18 23 39} reported albumin level as a notable predictor for colectomy within 1 year, with each 1 g/dL increase in albumin level decreasing the probability of colectomy by approximately 61% (pooled OR (95% CI): 0.39 (0.26 to 0.59), I²=0.0%, p=0.454). Moreover, when compared with a high albumin level, the pooled OR of a low albumin level (<30 g/L) in five studies¹¹¹⁵,^{17 21} was 2.34 (95% CI 1.61 to 3.39, I²=0.0%, p=0.931). High levels of CRP (≥25 mg/L) have been reported in five studies.¹⁵,^{1721 40} An overall OR of 2.63 (95% CI 1.53 to 4.52, I²=29.6%, p=0.224) suggested that it was an important risk factor for colectomy within 1 year. Four studies³⁴,³⁷ demonstrated the predictive value of the Oxford criteria, which were defined as CRP levels exceeding 45 mg/L in conjunction with 3–8 bowel movements in 24 hours, or independently, more than 8 bowel movements per day on days 3 of intravenous corticosteroids. Compared with patients failing to satisfy the Oxford criteria, those who met the Oxford criteria exhibited a statistically significant elevation in risk of colectomy, with an overall OR of 4.42 (95% CI 2.85 to 6.84, I²=0.0%, p=0.701). However, no significant predictive effect of CRP as a continuous variable was found despite heterogeneity^{13 14 18 39} (OR (95% CI): 1.00 (0.98 to 1.02), I²=60.4%, p=0.056). Similarly, patients with high levels of ESR (>30 mm/hour) were shown to have a nearly three times higher risk of colectomy than patients with low ESR levels, with a pooled OR of 2.92 (95% CI 1.39 to 6.14, I²=0.0%, p=0.792) from two studies.^{17 21} Four other studies^{11 16 17 21} have reported on the performance of low haemoglobin levels (≤12 g/dL) in predicting colectomy in 1 year. Although some heterogeneity was observed among studies (I²=56.4%, p=0.076), the result indicated that a low haemoglobin level tended to increase the risk of colectomy within 1 year (OR (95% CI): 2.08 (1.07 to 4.07)).

Figure 1. Meta-analysis on predictors for colectomy within 1 year. (A) Biomarkers. (B) Drug factors. (C) Demographic and clinical characteristics. (D) Auxiliary examination results.

Prior corticosteroid exposure was found to significantly influence the risk of colectomy within 1 year, with a pooled OR from six studies^{15 17 20 22 39 40} of 1.75 (95% CI 1.23 to 2.50, I²=17.7%, p=0.299) (figure 1B). Similarly, patients with previous azathioprine use were more than twice as likely to undergo colectomy than those without prior exposure (figure 1B). The pooled OR of two studies^{33 39} was 2.25 (95% CI 1.28 to 3.96, I²=0.0%, p=0.927). Prior immunomodulator use,^{20 21} baseline aminosalicylate use,^{12 17 39} baseline immunomodulator use^{12 13 17} and high-dose IFX induction (10 mg/kg)^{12 14 19 44} were not significantly associated with future colectomy, although heterogeneity existed among studies of the former two factors (both I²>50%) (figure 1B).

No demographic factors exhibited predictive value for colectomy within 1 year, including age, newly diagnosed UC, sex, body mass index, disease duration and smoking status (figure 1C). However, there was considerable heterogeneity between studies on age, both as a continuous (I²=83.4%, p=0.000)^{12 13 18 42} and dichotomous variable (I²=86.5%, p=0.000).^{15 21 41 45} Besides, some heterogeneity observed among the three studies^{16 17 44} on newly diagnosed UC (I²=64.4%, p=0.060), 11 studies^{1112 15},18 20 21 33 39 44 on sex (I²=48.7%, p=0.041), and two studies^{12 21} on disease duration (I²=57.0%, p=0.127).

Disease extent was the most frequently investigated variable among the auxiliary examination findings, as reported in nine prognostic studies of ASUC.¹¹¹⁵,18 20 23 33 39 Compared with patients with localised colitis, those with extensive colitis were more likely to undergo colectomy within 1 year, with an overall OR of 1.85 (95% CI 1.24 to 2.78, I²=47.5%, p=0.055) (figure 1D). Similarly, sarcopenia, a decrease in skeletal muscle mass and strength measured using CT at the third lumbar vertebra level, was a notable predictor of colectomy within 1 year. The pooled OR of two studies^{11 18} was 1.90 (95% CI 1.04 to 3.45, I²=0.0%, p=0.750) (figure 1D). No significant association was found between cytomegalovirus infection status and the occurrence of colectomy, as well as between Mayo endoscopic subscore and colectomy (figure 1D).

Predictors for colectomy within 90 days

Eighteen variables in 18 studies^{1114 17},23 33 34 36 37 39 were included in the meta-analyses for colectomy within 90 days. Among them, each 1 g/dL increase in albumin level and baseline use of aminosalicylate reduced the risk of patients undergoing colectomy within 90 days, with pooled OR values of 0.37 (95% CI 0.24 to 0.57, I²=0.0%, p=0.709) and 0.45 (95% CI 0.27 to 0.75, I²=0.0%, p=0.476), respectively (figure 2A). Five variables were identified as risk factors for colectomy within 90 days, including fulfilling the Oxford criteria (OR (95% CI): 4.69 (2.99 to 7.34), I²=0.0%, p=0.950), high ESR levels (OR (95% CI): 2.92 (1.39 to 6.14), I²=0.0%, p=0.792), previous corticosteroids exposure (OR (95% CI): 1.91 (1.19 to 3.07), I²=37.4%, p=0.172), previous azathioprine use (OR (95% CI): 2.25 (1.28 to 3.96), I²=0.0%, p=0.927) and sarcopenia (OR (95% CI): 2.60 (1.23 to 5.49), I²=0.0%, p=0.476) (figure 2A). Furthermore, the pooled results indicated that current smokers and cytomegalovirus infection were not significantly associated with the occurrence of colectomy within 90 days. The remaining variables did not demonstrate a predictive effect on colectomy, though significant heterogeneity among the studies was observed (figure 2B).

Figure 2. Meta-analysis on predictors with predictive value for colectomy within 90 days. (A) Variables with predictive values. (B) Variables without predictive values.

Predictors for colectomy during a follow-up period of more than 1 year

Six variables—UC endoscopic index of severity (UCEIS), disease extent, sex, stool frequency, age as a continuous variable and age >35 years—were proposed in several different studies to explore whether they could predict the occurrence of colectomy during a follow-up period longer than 1 year. Among them, UCEIS was the only proven satisfactory risk factor for colectomy, with a pooled OR of three studies^{26 27 32} of 2.41 (95% CI 1.72 to 3.39, I²=1.5%, p=0.362) (figure 3). Disease extent and sex, with seven^{2425 29},³³ and nine^{2425 27},^{29 31} studies, respectively, showed no significant effect on the probability of long-term colectomy, although certain heterogeneity was observed among the studies.

Figure 3. Meta-analysis on predictors for colectomy during a follow-up period of more than 1 year.

Sensitivity analyses

Among the variables for colectomy within 1 year, 13 variables were reported in studies with different diagnostic criteria, of which 10 variables had a sufficient number of studies defining ASUC using the Truelove and Witts criteria. These were ultimately included in the sensitivity analysis to test the consistency of the results.¹¹,^{1517 20} Similar associations were found between 1-year colectomy and all variables, except for albumin level and low haemoglobin level, which failed to maintain the statistically significant pooled results in sensitivity analysis despite the presence of heterogeneity (online supplemental figure S3).

Predictors for colectomy not available for meta-analysis

We systematically summarised the variables for the prediction of colectomy (regardless of positive or negative results) that did not qualify for the meta-analysis (online supplemental table S5). A total of 70 candidate predictors for colectomy within 1 year in patients with ASUC were discussed in 26 studies,¹¹,1720 21 23 24 26 33 35 while 52 predictors for colectomy during a longer follow-up period were discussed in 12 studies.^{2425 27},29 31 32 38 43 50 In addition to biomarkers studied in the meta-analysis, faecal calprotectin was another promising predictor of colectomy within 1 year, demonstrating good predictive capacity across multiple studies. Different auxiliary examination findings, such as endoscopic factors, radiological findings and the detection of Clostridium difficile infection, have also been discussed in various studies. Additionally, the predictive power of several indexes used to predict intravenous steroid failure, such as the Ho index, Lindgren score, and Travis index, for colectomy has also been raised. However, while research on concomitant medication on admission was relatively abundant, there was limited research on the prediction of colectomy by previous drug exposure. The only two studies on prior biological exposure presented contradictory results, with previous use of anti-tumour necrosis factor (TNF) agents reported as a protective factor for colectomy, while prior use of IFX was deemed to have no impact on colectomy within 1 year.

Discussion

Predicting the risk of colectomy in patients with ASUC, so as to implement more aggressive treatment decisions and refined care for high-risk individuals, is integral to avoiding emergency surgery and is crucial for ASUC management. In this study, we systematically reviewed 42 articles on the potential predictors of colectomy in patients with ASUC and summarised the predictive effects of multiple variables through meta-analysis and descriptive synthesis. To the best of our knowledge, this is the first systematic review and meta-analysis focusing on the risk factors for colectomy in patients with ASUC. The variables identified with definite predictive values in this study can not only be used in the clinical context as independent predictors of colectomy but are also thought to be more significant for inclusion in the development of future predictive models.

In the current study, 10 variables were identified as effective predictors of colectomy within 1 year through meta-analysis. Increased CRP level, ESR level, bowel movement frequency and disease extent, indicating a higher degree of inflammation, were identified as risk factors for colectomy, along with the previous use of steroids or azathioprine. Conversely, improvements in nutritional and health status indicators, such as albumin level, haemoglobin level and skeletal muscle mass and strength, may contribute to lower odds of surgery. Fulfilling the Oxford criteria was identified as being associated with a higher risk of colectomy within 1 year, serving as the most effective tool currently available for predicting this outcome. Similar findings were established in predictors for 90-day colectomy. In addition, an increase in the UCEIS score was the only factor found to increase the likelihood of undergoing colectomy more than a year later. A recently published systematic review predicting outcomes in patients with ASUC stated that the majority of scoring systems for predicting response to intravenous corticosteroids, as well as short-term and long-term colectomy, incorporated clinical or laboratory parameters, and endoscopic assessment using UCEIS was also applied for outcome prediction.⁵³ Variables selected in these scores, including albumin, CRP, stool frequency, UCEIS and previous anti-TNF or thiopurines, overlapped with the predictors found in our study. Another study by Adams et al developed and validated a prediction model composed of CRP, albumin and UCEIS to predict steroid non-response in patients with ASUC.⁵⁴ While the clinical focus may differ, these findings underscore the universality of these factors across various clinical contexts of ASUC and further support their association with colectomies. Additionally, our study identified some predictors with predictive values that are less commonly considered in current scoring systems, such as haemoglobin levels, disease extent and concurrent sarcopenia, which will be a driving force for developing and refining prediction models specific to the outcome of colectomy. Furthermore, the results revealed that for certain predictors such as CRP level, a categorical form may be more significant in predicting the outcome of colectomy than a continuous form, suggesting that the monotonic or linear associations between these predictors and the outcome may require further scrutiny, and greater emphasis should be placed on the determination of optimal cut-off values in future prognostic studies.

Fifteen, ten and five variables were found to be unrelated to colectomy within 1 year, within 90 days and over a longer period, respectively. Notably, the predictive effects of 17 of these variables were significantly heterogeneous across the included studies. Restricted by the limited number of studies, we did not further explore the specific origins of this interstudy heterogeneity. However, this remained understandable because our eligibility criteria did not impose strict specifications on study types and designs, sources of patient data, baseline clinical settings, interventions during follow-up or specific timing of outcome occurrence. Therefore, we had limited confidence in guaranteeing that these variables could not predict the risk of colectomy in patients with ASUC. Further research on the predictive capacity of these variables for colectomy should be conducted, preferably on a large scale, with a prospective design in a specific clinical setting, and with satisfactory generalisability.

In our systematic review, numerous variables were not included in the quantitative synthesis due to an insufficient number of studies or the absence of ORs for analysis, and they were comprehensively summarised in tabular form, which resulted in our inability to derive an overall effect of these variables in predicting colectomy. For instance, the only two studies discussing prior exposure to IFX and anti-TNF agents have yielded conflicting results, meaning that current evidence remains inadequate to definitively ascertain their effects on colectomy, though the meta-analysis showed that prior azathioprine exposure was a risk factor for colectomy. These controversial results may be related to whether the previous medication was administered as monotherapy or combination therapy. Due to the lack of additional studies to distinguish and compare these approaches, we are unable to draw specific conclusions. However, the qualitative results could still provide valuable information for clinicians and researchers, as all potential factors investigated in the existing studies were clearly presented. A notable topic for subsequent investigations is to validate the predictive power of these variables further to gain greater insight into their validity and robustness in predicting outcomes related to colectomies. In the meantime, additionally exploring the potential of other variable types to predict the outcome of colectomy is also of great importance. For instance, incorporating radiological findings may improve the accuracy of prediction of colectomy; however, current evidence is relatively insufficient. The potential value of omics data, such as genomics and metabolomics, for ASUC prognosis has also been previously proposed,⁵⁵ but further progress is still needed. In addition, the diversity of reported results in the available studies highlights the need to follow uniform reporting guidelines in prognostic studies, which is essential for improving transparency, reproducibility and the comparison and synthesis of evidence.

IFX, ciclosporin and tacrolimus have been recognised as salvage therapeutic options for patients with steroid-refractory ASUC, acting as alternatives to surgical intervention.⁵⁶ However, not all patients could respond to salvage therapy and maintain colectomy-free survival.^{57 58} In cases where salvage therapy with a particular drug fails, the adoption of sequential therapy with another agent is controversial, as secondary treatment failure may result in delaying surgery, potentially increasing mortality or the risk of postoperative complications.⁵⁹ Furthermore, existing evidence showed that even if the treatment is effective, a significant proportion of patients would experience relapse or even require colectomy during follow-up.^{57 59} Thus, early prediction of the probability of colectomy and risk stratification of patients could aid in timely surgical decision-making for high-risk patients, allowing adequate preparation for colectomy at the most appropriate time. This study contributes to the existing literature on this topic. The remaining challenges include the lack of prospective assessments of the optimal timing of surgery, the controversy of the advantages and disadvantages of sequential therapy and the benefits of personalised dosing strategies, such as prolonging administration periods, increasing dosages and accelerating dosing.^{1 60} Therefore, further normative studies are needed to address these issues.

This study has some limitations. First, colectomy among patients did not occur at a specific time node but occurred within a prespecified time range in our study, which may have contributed to some variability. However, our results showed that for most predictors, there was no significant heterogeneity among the studies in the calculation of the overall effect sizes. Second, we used unadjusted OR as the effect size for pooling without considering the influence of potential confounding factors, which may have affected the results. Variations in populations, study designs and outcomes across studies could have led to OR value without adjustment for confounders that are more likely to deviate from the true value, which could have contributed to the observed heterogeneity. Third, the number of articles that could be used for the meta-analysis was relatively limited for some predictors, which may have affected the persuasiveness and robustness of our results. Furthermore, the lack of information in the original texts to calculate the pooled OR values prevented some studies from being included in the meta-analysis, which may have introduced bias. Therefore, we conducted a tabulation summary of other parameters reflecting the predictive power of these factors as a supplement to the meta-analysis, hoping that it could provide some reference for clinical practice or future research.

In conclusion, this systematic review and meta-analysis of the predictors of colectomy in patients with ASUC identified several significant factors such as albumin, CRP, ESR, haemoglobin, Oxford criteria, UCEIS, disease extent, previous use of steroids or azathioprine and concurrent sarcopenia, as predictors for colectomy risk. In future studies and clinical practice, a comprehensive consideration of different factors, including biomarkers, endoscopic and CT findings, demographic and clinical characteristics, and drug factors, will facilitate risk stratification and personalised treatment, thereby reducing the occurrence of colectomy and improving patient prognosis.

Data availability statement

Data are available upon reasonable request.