Article Text
PDF
Abstract
Objectives We aimed to determine whether changes in acute severe colitis (ASC) management have translated to improved outcomes and to develop a simple model predicting steroid non-response on admission.
Design Outcomes of 131 adult ASC admissions (117 patients) in Oxford, UK between 2015 and 2019 were compared with data from 1992 to 1993. All patients received standard treatment with intravenous corticosteroids and endoscopic disease activity scoring (Ulcerative Colitis Endoscopic Index of Severity (UCEIS)). Steroid non-response was defined as receiving medical rescue therapy or surgery. A predictive model developed in the Oxford cohort was validated in Australia and India (Gold Coast University Hospital 2015–2020, n=110; All India Institute of Medical Sciences, New Delhi 2018–2020, n=62).
Results In the 2015–2019 Oxford cohort, 15% required colectomy during admission vs 29% in 1992–1993 (p=0.033), while 71 (54%) patients received medical rescue therapy (27% ciclosporin, 27% anti-tumour necrosis factor, compared with 27% ciclosporin in 1992–1993 (p=0.0015). Admission C reactive protein (CRP) (false discovery rate, p=0.00066), albumin (0.0066) and UCEIS scores (0.015) predicted steroid non-response. A four-point model was developed involving CRP of ≥100 mg/L (one point), albumin of ≤25 g/L (one point), and UCEIS score of ≥4 (1 point) or ≥7 (2 points). Patients scoring 0, 1, 2, 3 and 4 in the validation cohorts had steroid response rates of 100, 75.0%, 54.9%, 18.2% and 0%, respectively. Scoring of ≥3 was 84% (95% CI 0.70 to 0.98) predictive of steroid failure (OR 11.9, 95% CI 10.8 to 13.0). Colectomy rates in the validation cohorts were were 8%–11%.
Conclusions Emergency colectomy rates for ASC have halved in 25 years to 8%–15% worldwide. Patients who will not respond to corticosteroids are readily identified on admission and may be prioritised for early intensification of therapy.
- ulcerative colitis
- clinical decision making
Data availability statement
Data are available upon reasonable request. Please contact the corresponding authors for access to anonymised data.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Acute severe colitis (ASC) is an important cause of morbidity and mortality in UC, requiring hospitalisation and, frequently, colectomy.
Biological agents, initially anti-tumour necrosis factor therapies and later vedolizumab and ustekinumab, are now routinely used in maintenance strategies.
For over 20 years, the accepted management of ASC has involved 3 days’ treatment with intravenous corticosteroids, followed by use of accepted protocols to assess response.
WHAT THIS STUDY ADDS
By performing a historical comparison within the same unit, we provide insight into how outcomes of ASC have changed in the age of biologics.
We identify and validate threshold values of C reactive protein, albumin and a validated endoscopic activity index on admission as predictors of steroid non-response in ASC in the biologics’ era.
We show that a simple clinical score based on these parameters—the Admission Model for Intensification of Therapy in Acute Severe Colitis—accurately identifies patients who are unlikely to benefit from initial management with corticosteroid monotherapy.
Two further cohorts from Australia and India enabled validation of these findings, making this the first replicated prognostic score for ASC which does not rely on waiting to assess steroid response after 3 days. It is the only score derived and validated in patients exposed to biologics.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
We demonstrate that the outcome of ASC has altered dramatically. Innovations in management, notably the introduction of protocol-driven management algorithms, have halved rates of emergency surgery.
On admission, prediction of the effectiveness of initial management with parenteral corticosteroids is now possible using a simple clinical scoring system based on routine blood tests and endoscopic appearances.
This offers an opportunity for early escalation of treatment: introducing second-line therapy or surgery, potentially avoiding futile steroid exposure and reducing the duration of hospitalisation.
We confirm that patients with a suboptimal response to steroids remain at increased risk of readmission and colectomy in the year following discharge, providing potential for further improvement in outcomes by identifying patients for early escalation of medical therapy.
Introduction
UC is a common, chronic, immune-mediated illness affecting approximately 1 in 200 of the UK population and increasing in prevalence worldwide.1 2 Approximately 25% of patients will develop an episode of acute severe colitis3 (ASC), defined as frequent bloody diarrhoea with evidence of significant systemic disturbance (fever, tachycardia, anaemia and elevated inflammatory biomarkers): this complication necessitates hospitalisation for intensive medical management or surgery.4 Landmark trials in the 20th century demonstrated the efficacy of corticosteroids as first-line medical therapy.4 The subsequent success of ciclosporin or infliximab as rescue therapy in a proportion of patients with no response to corticosteroids5–8 led to protocol-driven algorithms of management with reduced mortality and colectomy rates.9–14 Nevertheless, consistent data suggest that approximately 40% of ASC episodes fail to respond to intravenous corticosteroids.9 10 14 Predictive models to guide use of medical rescue therapy based on early response to corticosteroids, typically by day three, are widely used.9 10 It remains unclear to what extent this has improved outcomes, or whether earlier identification of steroid non-responders is possible.
We examined outcomes of patients admitted to the Translational Gastroenterology Unit in Oxford, UK, between 2015 and 2019 and compared these with prospective historical data available from the same centre in the pre-biologics era (1992–1993)10 and with outcomes in specialist units in India and Australia.
We aimed to develop and validate a model to predict steroid response at admission so that rescue therapy or surgery could be expedited. We define and validate in the Oxford population a simple index for risk stratification on the day of admission, involving serum albumin and C reactive protein (CRP) concentrations, with endoscopic severity, as assessed by the Ulcerative Colitis Endoscopic Index of Severity (UCEIS).15–17 This model, the Admission Model for the Intensification of Therapy in Acute Severe Colitis (ADMIT-ASC) index, was then successfully validated in independent cohorts in Queensland, Australia and New Delhi, India.
Methods
Study design
All patients (≥16 years) admitted with ASC in Oxford between 1 May 2015 and 31 October 2019 were identified from a prospectively maintained endoscopy database (Endobase, Olympus) at Oxford University Hospitals NHS Foundation Trust, serving a population of 700 000. Per local protocol, all patients admitted with suspected ASC underwent flexible sigmoidoscopy on admission, with UCEIS15 16 recorded in all patients. This was cross-referenced against the Oxford BRC IBD Biobank database (InfoFlex) to assess completeness of capture. Patient case records were reviewed by two IBD clinicians (VG and WM) to ensure consistent recording of diagnosis and management. Sample sizes were targeted to exceed those used in previous studies, for example, Travis et al (n=51), on which the current day three criteria are based.10 As a retrospective audit of routine clinical examinations and outcomes, patient and public involvement was not included in the design and conduct of the research.
Definitions
The diagnosis of UC was confirmed for each patient using established criteria.18 ASC was defined according to modified Truelove and Witts’ criteria as ≥6 bloody stools/day with one or more markers of systemic disturbance (tachycardia >90 beats/min, temperature >37.8°C, haemoglobin <105 g/L or CRP >30 mg/L).12 Steroid non-response was defined as administration of medical rescue therapy (ciclosporin or anti-tumour necrosis factor (TNF)) or colectomy during the admission. The 1996 Oxford criteria were used to define need for escalation of medical therapy (stool frequency of >8 daily, or stool frequency of three to eight daily and CRP of >45 mg/L on the third day after initiating intravenous hydrocortisone).10
Management
Inpatient management in all three centres followed the same established Oxford protocol and international guidelines and was overseen in specialist units with endoscopic assessment on admission, and day 3 assessment of need for escalation of therapy.10 Inpatient management consisted of intravenous hydrocortisone (100 mg four times per day), rectal hydrocortisone (100 mg two times per day), intravenous fluids, and thromboprophylaxis with low molecular weight heparin.11 19 Response was assessed on the third day using the Oxford criteria.10 Medical rescue therapy protocol was ciclosporin 2 mg/kg intravenously until response, then 5 mg/kg orally for 3 months; or infliximab 5 mg/kg at 0, 2 and 6 weeks, rarely with a higher dose, based on previous therapy and the judgement of the treating consultant gastroenterologist (Oxford n=3, Gold Coast n=5, Delhi, n=2), if necessary after multidisciplinary discussion.
Outcomes
The primary outcome was steroid non-response during admission with ASC. We separately assessed need for biological therapy, ciclosporin or colectomy during admission. Readmission rates with ASC and colectomy rates at 12 months and within the total follow-up period were also assessed. All outcome measures were prespecified.
Data collection
In all centres, data were collected on demographic details (gender and age), UC history (disease duration, medication history and extraintestinal manifestations), admission clinical parameters (stool frequency, heart rate, temperature, haemoglobin, platelet count, white cell count and CRP), admission radiographic findings (presence of toxic megacolon on abdominal radiograph defined by a transverse colon diameter >5.5 cm) and endoscopic findings (UCEIS score). Duration of intravenous steroids, medical rescue therapy or colectomy during admission, readmission with ASC, advanced therapy post discharge (biologics or tofacitinib) and colectomy, both at 12 months and over the total follow-up period, were also collected. Readmissions with ASC were treated as separate events.
Predictive index development
A predictive index was created using the Oxford cohort, with only the final model validated in the Australian and Indian cohorts. We aimed to create a simple and memorable index which could be used at the bedside without reference tables or software. Initial modelling was performed using multiple classification techniques to identify a performance baseline against which the simple index could be compared, ensuring that the simplification did not overly effect performance.
Models included up to 10 untransformed parameters available on admission (age, albumin, CRP, current biological treatment, disease duration, haemoglobin, platelets, sex, stool frequency and UCEIS). One value for CRP and four values for albumin were missing; only patients with complete data were used; no imputation was undertaken; and outcome data were complete for all patients.
Models including a subset of the 10 parameters above were tested, with one or more thresholds at memorable values for each parameter, and with both even and weighted valuation of each parameter passing its threshold. Models were ranked by Akaike information criterion to identify the top 10 with the final index chosen from these, taking into account simplicity and ease of implementation (eg, weighing practical difficulties of adding accurate stool frequency measurement against a marginal increase in performance in the discovery cohort).
Predictive index validation
For validation of the index outcomes in two independent cohorts were analysed.
All direct adult admissions of UC patients meeting Truelove and Witts’ criteria were retrospectively identified in Gold Coast University Hospital, Queensland, Australia (January 2015–April 2020), and the All India Institute of Medical Sciences, New Delhi, India (August 2018–May 2020). In both these cohorts, the index was applied and correlated with outcomes, thereby including 110/128 Australian patients who had a UCEIS score available (all with complete albumin and CRP data), and 62 Indian patients (all with complete albumin, CRP and UCEIS data).
Statistical analysis
Analysis was performed in R 3.6. Univariable analysis used Fisher’s exact test for binary variables, logistic regression for all other variables, and for multivariable logistic regression. P values are given as corrected false discovery rates (FDRs).20 ORs for univariable and multivariable logistic regression show CIs calculated using the profile likelihood method,21 and binomial proportion CIs are given for summary statistics of the final model. Preliminary statistical modelling of steroid non-response in the discovery cohort was performed using caret22 to perform 100 repeats of 10-fold cross-validation. Blinding during analysis was not possible, but the definition of steroid response was based on prespecified criteria and determined prior to modelling, and selection of a final predictive index in the discovery cohort was completed before data from either validation cohort were accessed.
Results
Discovery cohort
The discovery cohort (tables 1 and 2) comprised 131 admissions, including 38 index presentations (42% male, median age 41.2). These admissions represent 117 patients admitted once and 14 patients admitted twice (online supplemental table 1).
Supplemental material
Admission demographics in the current Oxford cohort
Day 1 admission clinical, laboratory and endoscopic data in 131 admissions for Oxford cohort
Response to corticosteroids was seen in 52/131 ASC episodes (40%). No demographic parameters were associated with steroid response (table 1) or ASC recurrence. Median UCEIS score was 6 (IQR 5–7) overall, 6 (IQR 5–7) for non-responders and 5.5 (IQR 5–6) for responders. Baseline CRP differed between non-responders and responders (median 43 mg/L vs 101 mg/L, p=2.7×10−6), but albumin, haemoglobin, stool frequency and platelet count did not (table 2).
In total, 71 ASC episodes (54%) involved medical rescue therapy, with 35 (49%) given ciclosporin and 36 (51%) anti-TNF therapy (adalimumab in two cases due to longer-term intravenous access). After follow-up, 60/117 (51%) of patients were undergoing advanced therapy (24 infliximab, 9 adalimumab, 21 vedolizumab, 4 tofacitinib and 2 in clinical trials).
Overall, 19 (15%) patients required colectomy during admission, 8 undergoing rescue therapy (2 patient preference, 1 readmission a month following a previous ASC episode, 1 toxic megacolon and 4 patients admitted on vedolizumab therapy). Over a median follow-up of 22 (1–49) months, 14 (12%) patients required readmission with a further episode of ASC.
In 26/131 events (20%), the patient had prior exposure to anti-TNF or vedolizumab (26/93 (28%), excluding index presentations), with 17/131 (13%) events occurring despite current advanced therapy (3 infliximab, 3 adalimumab and 11 vedolizumab). One infliximab-treated patient responded to steroids and continued treatment; two had accelerated anti-TNF rescue therapy, neither requiring surgery. Two of three adalimumab patients responded to intravenous steroids (subsequently discharged on vedolizumab and tofacitinib, respectively); the other had surgery without rescue therapy. Of the 11 patients admitted on vedolizumab, 4 responded to intravenous steroids (1 continued vedolizumab; 2 switched to tofacitinib); 3 received rescue therapy (1 ciclosporin and 2 infliximab), and 5 five had surgery during the admission (one after unsuccessful infliximab rescue therapy).
Comparison of outcomes with prebiological cohort in Oxford
In comparison to 1992–199310 (table 3), the proportion of episodes treated with ciclosporin was unchanged (27%), but overall use of medical rescue therapy doubled to 54% due to the use of infliximab in 27% of episodes in the later time period. This coincides with colectomy rates during admission reducing from 29% to 15% (p=0.033), and a reduction in readmission with ASC from 35% to 12% (p=0.0017), despite significantly longer follow-up (12 vs 22 months, p<0.007).
Historical comparison of outcomes in Oxford
The rate of response to intravenous steroids was similar (41% vs 40%, p=0.87). In both cohorts, colectomy rates in the year following admission were eightfold lower in those who responded to steroid treatment than in those receiving medical rescue therapy (5% vs 40% p=0.013, and 2 vs 17% p=0.010). The overall rate of colectomy within a year fell from 43% (21/49) to 21% (24/117) (p=0.0044).
Predictive analyses of outcome in the discovery cohort
On univariable analysis, baseline CRP (FDR corrected p=0.00066, OR 1.02), albumin (p=0.0066, OR 0.89), UCEIS (p=0.015, OR 1.62) and number of Truelove and Witts’ criteria (p=0.0066, OR 2.43) were significantly associated with steroid non-response (online supplemental figure 1). Of the Truelove and Witts’ criteria, only CRP was independently significant.
Of the components of the UCEIS score, the ‘erosion and ulceration’ component was independently significant (p=0.0066, OR 2.68), and stronger than the overall score. Details of univariable regressions for steroid non-response and rescue therapy are given in table 4 and online supplemental table 2).
Results of univariable analysis in the discovery cohort
In multivariable logistic regression (online supplemental table 3), only CRP was independently significant for steroid non-response (p=0.0076, OR 1.02). Any biological treatment on admission was associated positively with infliximab rescue therapy (p=0.0049, OR 5.65) and negatively with ciclosporin (p=0.016, OR 0.13).
Predictive models of steroid non-response in the discovery cohort
Steroid non-response was modelled in the discovery cohort with multiple classification approaches (online supplemental table 4 and online supplemental figure 2), revealing CRP to be the most informative variable, followed by UCEIS, in particular the erosion and ulceration component (online supplemental figure 3). Average prediction accuracy in left out samples from cross-validation ranged from 57% (95% CI 47% to 66%) for the K-nearest neighbour (KNN) to 68% (95% CI 59% to 77%) for multivariate adaptive regression splines (MARS), though across all models, accuracy was high for those predicted to be at high or low risk of steroid non-response and worse for those of intermediate risk. receiver operating characteristic (ROC) curves for these predictions are shown in online supplemental figure 4, with mean area under the ROC curve values ranging from 0.59 for KNN and AdaBoost to 0.75 for MARS, naive Bayes and sparse partial least squares models. This modelling demonstrates that prediction of steroid response and non-response is possible, but that identifying extremes of risk is more feasible than accurately predicting risk in all patients.
Selection of a predictive index
Indices were tested in the discovery cohort with a range of thresholds and with both evenly and variably weighted criteria. The final index consists of four points, one each for albumin of ≤25 g/L (19% of patients) and CRP of ≥100 mg/L (34%), with UCEIS of ≥4 (96%) scoring 1 point or 2 points for UCEIS score of ≥7 (29%). These selected features were all significant in univariable regression (aforementioned), and correlations between them are shown in online supplemental table 5. Of potential fifth components (in decreasing order of utility: stool frequency ≥14, haemoglobin ≤90 g/L, current biologics, platelets ≥500×109/L and male sex) only stool frequency made a nominal improvement, but this was not selected for the final model as the added complexity and difficulty of accurately measuring stool frequency outweighed the minor difference in performance (7% improvement in sensitivity at a cost of 4% loss of specificity). The best performing model where UCEIS on admission endoscopy is not available comprised CRP ≥100, albumin ≤25 and haemoglobin ≤120, and is detailed in online supplemental table 6.
Scores of 3 (14% of patients) and 4 (8%) identified patients with rates of non-response to steroids of 83% and 100% (table 5). Use of a score threshold of ≥3 to identify high-risk patients therefore yields a positive predictive value (PPV) of 0.89 (95% CI 0.78 to 1.00), with an OR for steroid non-response of 7.6 (6.3 to 8.8) (table 6). In comparison, PPVs for steroid non-response using single-component models of CRP of ≥100 and albumin of ≤25 are 0.46 and 0.53, respectively. No significant association exists between current biological treatment on admission and score (p=0.82), and no correlation was observed between score and non-response to medical rescue therapy.
Breakdown of steroid response in discovery and validation cohorts by ADMIT-ASC score involving C reactive protein of ≥100 mg/L (1 point), albumin of ≤25 g/L (1 point), and UCEIS score of ≥7 (2 points) or ≥4 (1 point)
Summary statistics for final model performance predicting steroid non-response using ADMIT-ASC scores of 2–4 as thresholds in the discovery, validation, and combined validation cohorts
Validation cohorts
At Gold Coast University Hospital, a total of 128 patients (50% male, median age 35 years) presented with ASC between 1 January 2015 and 30 April 2020. Demographics and clinical parameters are given in online supplemental tables 9 and 10. At presentation, 12 (9%) patients were on biological therapy (8 anti-TNF and 4 vedolizumab), and for 41 patients (32%), the admission was their index presentation of UC.
Medical rescue therapy was given to 51 (40%) patients (4 ciclosporin and 47 infliximab). Ten (8%) patients had a colectomy during admission, with 1 having direct colectomy. All features selected for the final index were significant in univariable regression (online supplemental table 11). Validation of the index was performed in the 110 (86%) patients with a UCEIS score recorded (all with CRP and albumin results). There were no significant differences between this subset and the complete Australian cohort (online supplemental table 12).
At the All India Institute of Medical Science, 62 patients (40% male, median age 35.5 years) presented with ASC between August 2018 and May 2020. Demographics and clinical parameters are given in online supplemental tables 13 and 14). Medical rescue therapy was given to 17 (27%) patients (2 ciclosporin and 15 anti-TNF), and 7 (11%) patients had a colectomy during admission.
When the final model was applied to the Australian cohort with available UCEIS scores (n=110), a similar proportion of patients scored ≥3 (18% vs 21%, p=0.63); in the Indian cohort, this proportion was smaller (8%, p=4.6×10−7). However steroid response rates for each score were similar across all three cohorts (table 5), apart from patients with the intermediate scores of 1 or 2 who were more likely to undergo rescue therapy or surgery in the UK cohort (p=0.10 and 0.002, respectively). Scores of 0 and 4 were 100% predictive of steroid response and non-response, respectively, across all three cohorts, although this represents <10% of patients.
In the Indian cohort, median CRP levels were significantly lower than both UK and Australian cohorts. In comparison to the Australian cohort, the Indian cohort therefore achieved lower scores, but the observed steroid response rate for any given score remained remarkably similar (table 5).
Proposed implementation from pooled results
Although scoring 4 was 100% predictive of steroid non-response, only 13 patients (4.3%) scored 4. Applying a threshold of 3 identified 17.5% of the combined validation cohorts and was highly specific (0.96, 95% CI 0.93 to 1.00), with a PPV of 0.84 (0.70 to 0.98) and OR of 11.9 (10.8 to 13.0). Use of this threshold to advance treatment would prevent one delayed treatment for every 8.2 (95% CI 5.8 to 13.7) patients assessed. Table 6 shows results for thresholds of 2–4 in the discovery cohort and both validation cohorts separately and combined.
Discussion
The key findings from the present study are the development and international validation of a simple clinical index to predict likelihood of steroid response on admission with severe UC; and the evidence of a significant improvement in the outcome of severe colitis, with a halving in the need for acute surgery in the modern era.
ADMIT-ASC index
ADMIT-ASC is a simple clinical index that can be applied on the day of admission to predict steroid non-response in ASC, using only two routine blood tests (CRP and albumin) and a validated index of endoscopic severity (UCEIS) with high interinvestigator and intrainvestigator reliability.16 Our data are novel in demonstrating independent replication notwithstanding geographical and ethnic differences between three populations.
With a threshold of ≥3 the index identifies a subset of approximately one in five patients with highly active disease, who are unlikely to respond to parenteral corticosteroid therapy11 and may benefit from consideration of escalation of therapy or surgery on the day of admission. These data provide a rationale to refine the paradigm for management, which is currently predicated on treating all patients with parenteral steroids and assessing steroid response on day 3. The choice of threshold for escalation needs to balance the perceived implications and costs of avoidable rescue therapy with the risk of delayed treatment and the same threshold may not be considered appropriate in all settings internationally.
A combination of CRP of ≥100 mg/L, albumin of ≤25 g/L and a UCEIS score of 7–8 (2 points) on day 1 results in a score of 4 and high likelihood of steroid non-response (100% in both discovery and two validation cohorts), regardless of whether the patient is already established on biologics. A score of 3–4 is associated with an 84% chance of non-response to steroids, and we propose a score of 3 or greater in a biological-naïve patient is an indication for early medical rescue therapy. In a patient admitted on biological therapy a score of 3 mandates surgical discussion early in admission. As a counterpoint, those patients with a score of 2 or less may be managed by conventional approach, with reassessment of steroid response and need for escalation at day 3. A score of 0 may be particularly reassuring since we found complete concordance with steroid response across the cohorts.
The current study confirms previous observations from ourselves and others that the biological severity on admission, as defined by Truelove and Witts, predicts resistance to therapy.9 10 14 17 23–28 Our data suggest that CRP is the key parameter associated with predicting non-response to steroids.
The critical difference between this score and others is that reproducibility has been demonstrated across multiple cohorts. The only recent comparable index using data on admission is the ACE ("Albumin, CRP and Endoscopy") index, developed using data generated in the period before biologicals were widely used in UC.14 In contrast to the present study, this score was calculated and tested in the same cohort without cross-validation or a random split into learning and testing sets, meaning that the model may overestimate expected performance in unseen cases; additionally, patients not meeting strict criteria for ASC were included. Notwithstanding these differences, the components of the index are broadly comparable, though the Physicians Global Assessment was used rather than UCEIS.
It is relevant that other studies of predictive markers for ASC outcomes have identified the level of inflammation on admission as a critical factor in predicting clinical response to steroids, with some early data even predating the introduction of the day 3 criteria and modern rescue therapy.17 Previous studies show correlations between steroid response on day 3 and albumin, CRP and endoscopic severity, as well as stool frequency, haemoglobin, heart rate and calprotectin, among others.9 10 14 17 29–39 In situations where early endoscopy is not available, an alternative panel could be used such as the one we demonstrated combining CRP, albumin and haemoglobin.
Defining risk on the day of admission carries advantages in decision-making, contingency planning and patient counselling, especially in those at high-risk of steroid non-response who are accurately identified by the index. The index is simple and implementable without other changes in practice; all the parameters are readily and reproducibly defined, without the awkward need for assessment of stool frequency.
There are, of course, limitations. The study is retrospective in design and did not allow reliable assessment of preadmission therapy in the discovery cohort. Nonetheless, patient identification appeared complete, and painstaking searches of available databases and paper case notes were performed to avoid ascertainment bias.
We did not formally capture disease morbidity or complications of treatment, other than in the context of colectomy rates or readmission. These are subject to regular audit and there is no signal of complication rates associated either with drug therapy or colectomy.40 We also recognise that although the proposed index accurately identifies patients at a very high or low likelihood of response to steroid treatment, there remains a significant proportion who score 1 or 2 in whom steroid responsiveness is difficult to predict, and while the predictive value of a result indicating steroid non-response is high, the sensitivity to detect all non-responders is not. The study was performed in three specialist centres where management was the responsibility of IBD-focused clinicians who followed protocols to optimise outcomes developed in Oxford and adopted internationally. Nevertheless, most (80%) of the patients came from the local population, as in any secondary care hospital.
The replication in two cohorts from different genetic backgrounds, previous treatment, environmental exposures, age and disease duration is a particular strength of our study. It is interesting to note that median CRP levels were significantly lower in the Indian cohort. Variations in both baseline and inflammation-associated CRP levels have been explained by ethnicity, sociodemographic and environmental factors, and genetic polymorphisms.41–47 Body mass index (BMI) has been implicated as the strongest correlate accounting for 15% overall variation in CRP41; and although BMI was not routinely recorded in our cohorts, the WHO Global Health Observatory 2016 estimates the mean age-standardised BMI of Indian adults (21.8) is lower than British and Australian adults (both 27.1).
Historical perspective
This study allows comparison of outcomes within a single specialist unit between the 1990s and 2010s, over which time significant changes in management of ASC have been introduced. We first demonstrate that urgent colectomy rates have halved from 29% to 15% in 2015–2019, in comparison with the 1992–1993 series from Oxford,10 consistent with previous studies within our unit showing a consistent rate of 25%–30% until the mid-1990s and a subsequent fall.10 23 39 48–51 Elsewhere, there are recent studies with similar acute colectomy rates from the UK (14.9%–21.4%),25 52 53 Ireland (18%),54 USA (12.4%),55 Australia (13.3%)56 and Portugal (13.4%).57 However, overall rates vary widely from 2.8% (6/217) in a Finnish study58 to 75.9% (41/54) in a French study,59 and within countries, there can be substantial variability, such as 7%–76% in France.59 60
A key difference between the 1992 and 1993 study from Oxford and the present study is the doubling of the proportion undergoing medical rescue therapy. In Australia, where approximately 40% of patients underwent rescue therapy with ciclosporin or infliximab, we report an acute colectomy rate of <10%, and in India, the colectomy rate was 11%, with 27% given rescue therapy. This provides compelling evidence that recent alterations in treatment strategy, including intervention with biological therapies, have improved clinical outcome in ASC. Standardised protocols for assessing response, guiding care and emphasis on multidisciplinary involvement, appear important to set standards for care, inform patient discussion and as a basis for further audit.
Furthermore, readmission and colectomy rates after discharge have fallen by two-thirds and a half, respectively. The risks of colectomy in those without a clear steroid response remain eightfold elevated in the year following discharge.10 24 The early identification of this group with initiation and maintenance of longer-term effective treatment therefore remain a priority beyond management of the acute episode.
Conclusions
Two major findings from this study have clinical importance. First, the reduction in need for surgery and improved outcomes in the modern era are highly relevant to daily practice for clinicians involved in the care of patients with ASC. Second, we propose that this predictive model, which has been validated across three centres, may be considered for translation to practice as a guide to early stage decision-making.
Data availability statement
Data are available upon reasonable request. Please contact the corresponding authors for access to anonymised data.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and was approved by the National Health Service Health Research Authority (REC 16/YH/0247 and 09/H1204/30), in Gold Coast by Health Service Human Research Ethics Committee (Ref: LNR/2020/QGC/67173), and in Delhi All India Institute of Medical Sciences by the ethics committee (Ref: IEC-261/04.05.2018). This is a retrospective study involving examination of routine clinical examinations and clinical outcomes.
Acknowledgments
The authors acknowledge the clinical and academic contributions made to this work by Professor Satish Keshav, who sadly passed away before the study was completed. We acknowledge the contribution of the Oxford IBD cohort study BRC, which is supported by the NIHR Oxford Biomedical Research Centre, University of Oxford. The authors thank Sister and staff of the specialist wards, IBD specialist nurses and administrative staff, as well as our patients. The authors are also grateful to the other Oxford IBD Cohort Investigators: Carolina V Arancibia-Cárcamo, Adam Bailey, Ellie Barnes, Elizabeth Bird-Lieberman, Barbara Braden, Jane Collier, James East, Alessandra Geremia, Lucy Howarth, Satish Keshav, Paul Klenerman, Simon Leedham, Fiona Powrie, Astor Rodrigues, Peter Sullivan and Holm Uhlig.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
AA, VG and WM are joint first authors.
Twitter @ColleenMcGreg15
Contributors JS proposed the study and is guarantor. WM, VG, DS, PKR, SKu and SKe collected the clinical data. AA and PKR performed the analysis. AA, JS, VG, WM and TPC contributed to the study design and prepared the first draft, and all authors interpreted the results, and reviewed and contributed to the final manuscript.
Funding This work is funded by the Oxford NIHR Biomedical Research Centre, Gastroenterology and Mucosal Immunity Theme.
Disclaimer The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.
Competing interests AW reported personal fees outside the submitted work from Ferring Pharmaceuticals, Janssen and Takeda. SPLT reported outside the submitted work receipt of grants/research support from AbbVie, Buhlmann, Celgene, IOIBD, Janssen, Lilly, Pfizer, Takeda, UCB, Vifor and Norman Collisson Foundation; consulting fees from AbbVie, Allergan, Amgen, Arena, Asahi, Astellas, Biocare, Biogen, Boehringer Ingelheim, Bristol-Myers Squibb, Buhlmann, Celgene, Chemocentryx, Cosmo, Enterome, Ferring, Giuliani SpA, GSK, Genentech, Immunocore, Immunometabolism, Indigo, Janssen, Lexicon, Lilly, Merck, MSD, Neovacs, Novartis, NovoNordisk, NPS Pharmaceuticals, Pfizer, Proximagen, Receptos, Roche, Sensyne, Shire, Sigmoid Pharma, SynDermix, Takeda, Theravance, Tillotts, Topivert, UCB, VHsquared, Vifor and Zeria; speaker fees from AbbVie, Amgen, Biogen, Ferring, Janssen, Lilly, Pfizer, Shire and Takeda; and no stocks or share options. JS received lecture fees from Takeda and from the Falk Foundation.
Patient and public involvement Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.