Ulcerative colitis (UC) is a chronic inflammatory bowel disease (IBD) that is characterized by frequent flares, hospitalizations, surgery and psychological morbidity [1,2]. Acute severe ulcerative colitis (ASUC) is a severe exacerbation of UC that warrants intravenous corticosteroids and may require emergency colectomy [3]. Corticosteroid failure necessitates salvage therapy with infliximab or ciclosporin which have demonstrated good short term efficacy [4-7]. However longer term prognosis after medical salvage remains guarded, where up to 50% of patients still proceed to colectomy in the absence of maintenance biologic therapy [8-10]. Conversely, a proportion of patients may have an initial period of significant symptoms, but develop only mild symptoms or even remission in the subsequent 10 years of their disease course [11].

To date, the ability to distinguish individuals who require maintenance infliximab following salvage therapy from those who could be continued on conventional therapy, such as immunomodulators, has not been established. The ability to accurately predict response to infliximab salvage and the subsequent risk of colectomy is crucial in order to decide the most appropriate maintenance therapy, which is an important concept in modern precision medicine [12,13]. Moreover, the optimal timing and dosing regimen of infliximab salvage is subject to ongoing research [14], therefore a predictive score for infliximab failure early in the presentation might facilitate timely decisions for potential dose escalation, additional line medical salvage, or early colectomy [15,16].

Biomarkers are increasingly recognized to play a crucial role in precision medicine, and useful biomarkers to aid in clinical prediction are needed [17,18]. We hypothesize that an improved understanding of the dynamics of biomarkers during infliximab salvage is crucial to identify candidate predictors of response, as well as to determine ideal cutoffs, which may change over time. The C-reactive protein (CRP)-albumin ratio (CAR) after infliximab salvage has been identified as a potential predictor of colectomy [15], yet the optimal timing of measurements and the utility of serial measurements require further evaluation. Recently, novel biomarkers such as the neutrophil-lymphocyte ratio (NLR) and platelet-lymphocyte ratio (PLR) have been shown to predict mucosal healing in UC after infliximab [19,20] but their utility in ASUC has not been studied. Additionally, the CRP-lymphocyte ratio (CLR) has been studied in other clinical settings, but not in UC [21].

The primary aim of this study was to evaluate and compare the dynamic response of serum biomarkers in responders compared with nonresponders to infliximab salvage during steroid-refractory ASUC, and to determine their utility in predicting the need for colectomy at 12 months following the initial presentation. The secondary aim was to determine the utility of these biomarkers in predicting early colectomy by 1 month.

Biomarkers have the potential to predict outcomes and tailor therapeutic options to the individual, which are core components of modern IBD management algorithms and precision medicine [17,26]. Our study has highlighted the difference in biomarker dynamics between patients who required and avoided colectomy and demonstrated the potential utility of CAR and CLR early in the treatment course (day 3 post salvage) in predicting infliximab failure. With the expansion of efficacious treatments such as newer biologic therapies and small molecule inhibitors [27-30], the ability to predict disease trajectory in ASUC should inform treatment choice—for example, a patient who is likely to require longer term colectomy despite infliximab salvage may warrant more proactive optimization of maintenance therapy with a biologic agent (a top-down approach), while patients with a lower risk of colectomy may only require conventional therapy [31-33]. Although the ability to continue biologic therapy in all patients may appear attractive, a careful risk-benefit assessment should be made given such a strategy introduces an additional risk of infections and may not be cost-effective [34,35]. Our study further stresses the importance of discovering an optimal management algorithm after ASUC, as the risk of colectomy in our study appeared to be dependent on the type of maintenance therapy received.

This study has demonstrated that the utility of biomarker ratios is predicated by the early dynamics of the constituent biomarkers after infliximab salvage. CRP falls in most patients following infliximab, but the fall is attenuated in patients who proceed to colectomy. Even if such patients avoid an early colectomy, there is still a higher likelihood of longer-term colectomy, so clinicians need to remain vigilant in such patients and have a low threshold to continue aggressive medical therapy. Lymphocyte count rises in most patients shortly after salvage therapy yet appears to remain static in patients who fail infliximab. The association between elevated CRP and both steroid and infliximab failure has been demonstrated previously and is expected [15,23]. An association between lymphocyte count is much less understood, although our results support a recent study which showed that lymphocytosis occurs after anti-tumor necrosis factor (anti-TNF) therapy and may be associated with treatment response [36]. In other disease entities, anti-TNF therapy purportedly impairs marginalization of lymphocytes to the site of inflammation as part of its mechanism of action, thereby causing a transient increase in circulating lymphocytes [37]. Although subject to replication in other IBD cohorts, the lack of a transient lymphocytosis after infliximab may thus suggest ongoing lymphocyte marginalization to the inflamed gut and hence portend infliximab failure.

As expected, serum albumin remained lower across all time points in patients who failed infliximab. The association with hypoalbuminemia and infliximab failure has been attributed to increased colonic infliximab clearance with the former [38,39]. Indeed, a higher baseline clearance and lower serum concentrations of infliximab have been shown to predict treatment failure and colectomy [16,40]. A suppressed albumin is also reflective of the systemic inflammatory burden and is a known poor prognostic marker in multiple disease entities. Surprisingly, although a higher platelet count has been associated with steroid failure in ASUC [23], there was no association with infliximab failure in our study. Although we have shown that platelet count and neutrophil count were not useful as acute phase reactants, their suppression later in the treatment course might be useful to indicate a response to therapy and predict mucosal healing [19].

These data emphasize that the timing of biomarker measurement is crucial. Not only did biomarkers perform better when calculated on day 3 than on day 1, biomarkers and their ratios were shown to change over time, so cutoffs derived from optimizing sensitivity and specificity will change depending on which day the tests are performed. A previous study that showed an optimal CAR cutoff for 0.37 unfortunately did not specify the timing of the measurements [15]. The optimal cutoffs identified in our cohort were CAR > 0.47 mg/g and CLR > 6.0 mg/10⁹, which both applied to just under a third of patients in this study. Using either of the ratios, the probability of requiring colectomy by 12 months in patients identified as high risk was 50%–57% (or 29%–37% in those who avoided index colectomy), and in over two-thirds of patients identified as low risk, the probability of requiring colectomy was only 4%–6% (or 3%–5% correspondingly). CLR seemed to provide the best risk stratification: using day 3 CLR, low-risk patients not on biologic maintenance only had a 3% probability of requiring colectomy, compared to high-risk patients on biologic maintenance (20%) and high-risk patients not on biologic (44%). CLR therefore had substantial utility in identifying patients likely to fail infliximab, independent of other variables associated with colectomy such as age, baseline steroid requirement and type of maintenance therapy used following ASUC as shown by our multiple regression analysis.

In addition to guiding the choice of maintenance therapy, predictive scores may also guide optimal salvage therapy in ASUC, which remains an area of ongoing research. Observational studies have suggested a possible improved outcome with accelerated infliximab dosing [41], although this remains contentious and has yet to be confirmed in a randomized trial [14,42]. The use of a predictive score calculation on day 3 post infliximab salvage may therefore be useful in deciding whether the patient is likely to respond to therapy, require further (potentially intensified) medical therapy or require surgical intervention. The ability to quantify the risk of having a poor response can also better facilitate shared decision making between patients and clinicians [43]. However, before these scores can be applied in clinical care, prospective controlled validation is required, specifically by demonstrating an improved outcome with the incorporation of biomarker-guided decision making in ASUC management algorithms.

This study has also validated and compared existing indices in predicting colectomy during index admission for ASUC after infliximab salvage. The Lindgren score and Ho index were originally derived to predict steroid failure, while the Mayo-related scores reflect general disease severity, but it is unsurprising that they also predict short term outcomes after infliximab given they incorporate clinical findings and reflect the immediate clinical response. However, it is notable that, with the exception of the Lindgren score, objective biomarkers such as CAR and CLR are superior in predicting longer term infliximab failure. Additionally, the CAR and CLR were superior to the initial endoscopic Mayo subscore and subsequent Mayo scores, which are normally considered important indicators of disease severity [44]. The objectivity, and therefore reliability of biomarkers makes them a more appealing candidate as predictive scores. Yet the continued relatively strong performance of the Lindgren score and the correlation between stool count and colectomy suggests the inclusion of clinical markers in prediction scores may still be useful.

The predominant limitation of this study is the retrospective design and therefore it is subject to information bias due to misclassification. A small proportion of data were missing however the risk of bias was mitigated by the use of multiple imputation for biomarker data. Also, though most patients received standard infliximab dosing (5 mg/kg at weeks 0, 2, and 6), this was not consistent and was at the discretion of the treating clinician. The strong association with CRP and early colectomy is confounded by auto-correlation, where the early CRP trend is used as part of the subjective decision-making process by surgeons for index colectomy. However, early CRP is not used for decisions regarding later colectomy (the focus of this study) beyond index admission at our institution. Additionally, a small proportion of patients had mild disease on endoscopy which may suggest an alternative cause may have been contributing to their ongoing symptoms. Fecal calprotectin was not routinely performed at our centers and therefore could not be analyzed. Endoscopy was performed on admission prior to infliximab and not routinely following salvage treatment, therefore we could not investigate the utility of endoscopic findings immediately post infliximab. The lack of long term clinical, biomarker and endoscopic outcome data represents a limitation to this study, where ideally these alternate endpoints would be studied in addition to 12-month colectomy. Further, biomarker calculations later during the disease course (such as at week 6 post infliximab) may provide additional predictive value, however these were not uniformly tested in all patients and therefore could not be analyzed in this study. The study may not have been powered to identify additional independent predictors of colectomy in multivariable analyses, however this was not a primary focus. The study was performed at a single health service which may limit its external validity. Further, this study was observational and therefore the utility of the predictive scores evaluated is only speculative, and confirmatory controlled studies are required to prove their usefulness in ASUC management algorithms.

In conclusion, in this large series of patients presenting with steroid-refractory ASUC, simple and routinely performed biochemistry in the biomarker ratios CAR and CLR calculated on day 3 post infliximab salvage show significant promise in predicting infliximab failure and need for longer term colectomy by 12 months. Thus, biomarkers have the potential to aid in decision making for the optimal maintenance strategy after infliximab salvage. These biomarkers are advantageous because they are simple to calculate, widely available, objective and exhibit dynamics that align well with current mechanistic understandings of anti-TNF therapy and the rapidly changing nature of ASUC. Furthermore, CAR and CLR had superior predictive performance for 12-month colectomy compared with the Mayo score, partial Mayo score, Ho index and Travis score, while the Lindgren score performed similarly. Prospective controlled studies are required to confirm the utility of incorporating predictive scores in current ASUC management algorithms.