Article Text
PDF
Statistics from Altmetric.com
- Acute severe colitis
- clinical decision-making
- colectomy
- IBD clinical
- IBD surgery
- management
- prediction
- ulcerative colitis
Acute severe ulcerative colitis is a potentially life-threatening condition. It is most readily identified by the criteria of Truelove and Witts,1 which have been adopted by international bodies: a patient who has a bloody stool frequency of six or more per day and tachycardia (>90 bpm, or temperature greater than 37.8°C, or anaemia (haemoglobin <10.5 g/dl), or an elevated erythrocyte sedimentation rate (ESR) (>30 mm/h).1–4 However, the response to intensive treatment with corticosteroids has remained unchanged for 50 years. In a systematic review of 32 trials of steroid therapy for acute severe colitis involving 1991 patients from 1974 to 2006, the overall response to intravenous hydrocortisone, methylprednisolone, or betamethasone was 67% (95% CI 65% to 69%), and 29% (95% CI 28% to 31%) came to colectomy.5 The only major difference from 1955 was that mortality was 1% (95% CI 0.7% to 1.6%) and not 7%. The colectomy rate did not change between 1974 and 2006.
The UK national inflammatory bowel disease audit in 2008 identified a mortality of 2.8% among 317 patients.6 Old age dominated death (11/16 who died were aged 80–89years), but the morbidity in young people has yet to be quantified. A nationwide linkage analysis in Scotland also strongly implicates age as the critical determinant of mortality in patients hospitalised for ulcerative colitis—the 3-year mortality in 245 patients aged 65 years and over was 39%, whereas none of 212 patients aged 30 years or less died (p<0.001).7 Co-morbidity was independently associated with mortality, but it remains unclear whether older people die from co-morbidity, complications of immunosuppression, delayed surgery, or surgery itself. The prime concern in managing acute severe colitis must be to reduce both morbidity and mortality.
We address two principal clinical dilemmas in managing acute severe colitis: how to identify at an early stage those who are likely to fail intensive treatment, and when to start rescue medical therapy in time so that surgery, if it becomes necessary, is not inappropriately delayed as outcome may deteriorate (see page 130).8 9
Answers to these questions are not mutually exclusive and demand the most taxing clinical judgement. For those who fail intravenous steroids, currently available ‘rescue’ medical therapy consists of ciclosporin or infliximab. Tacrolimus also appears to work,10 but not visilizumab.11 Patients naturally want to avoid colectomy. Physicians frequently prevaricate, faced with making either an irrevocable decision for colectomy, or prolonging therapy with potentially toxic immunomodulators, which may in turn compromise outcome after surgery. Surgeons also want to avoid delayed surgery, mindful of the increase in postoperative complications.12 They are also concerned about the long-term outcome of patients who escape colectomy on one admission, but will need colectomy later on. It may help patients to know that whereas, overall, 12% of patients with ulcerative colitis come to colectomy, the rate is 40% in those who have one or more admissions with acute severe colitis.13 14 The likelihood of needing re-admission (assuming that colectomy has been avoided on the first admission, as it will be in up to 80%) is 36%. If prospective data from 15/49 patients who had an incomplete response to medical therapy can be extrapolated (judged by a stool frequency greater than three per day or visible blood in their stools on day 7), 40% came to colectomy within 6 months and 85% had a colectomy over the next decade, compared with 15% in those responding completely to medical therapy.15 Ciclosporin and infliximab may reduce the risk of colectomy in the short term without compromising safety,16 although care should be taken to distinguish between complications related to emergency colectomy and elective colectomy, but neither appear able to prevent colectomy in the long term. After ciclosporin, 88% of 142 patients17 came to colectomy over 7 years, whereas 3 years after infliximab 12/24 (50%) came to colectomy, compared with 16/21 (76%) who received placebo, although more in the infliximab group received continuation treatment with azathioprine.18
Factors that may predict the need for colectomy can be broadly divided into clinical, laboratory, radiological, endoscopic and genetic. Many have focused on steroid failure, but some prospective studies also included patients treated with ciclosporin or infliximab.19 20 Other techniques may prove useful in the future, including proteomics, microbiomics or metabolomics, because they are linked more closely to the pathogenesis than consequences of inflammation. Most predictive indices are composites, combining two or more factors. To be clinically useful, an index must be easy to remember, simple to apply and reliable. Validation of indices for acute severe colitis is needed to avoid surgical decisions based on clinical impression, rather than objective assessment. On the other hand, there is a danger that indices become self-fulfilling, resulting in colectomy because criteria are met, when it might otherwise have been avoided. It should be remembered that current indices relate to the failure of steroid therapy and do not necessarily translate to failure of other therapies, or infection complicating colitis.
Clinical criteria associated with the need for colectomy
Stool frequency
Clinical markers generally depend on the objective measures of stool frequency, pulse, or temperature. A stool frequency greater than 12 per day on day 2 was associated with 55% colectomy in a retrospective study of 189 admissions in 166 patients,21 whereas a frequency greater than eight per day on day 3 of intensive treatment predicted colectomy in 85% (p<0.001) on that admission in a prospective analysis of 51 admissions.19 Stool frequency has been validated in 128 children: in the only prospective study yet to compare different indices, the number of daily stools on day 3 (closely followed by the amount of blood in the stool) was the main factor identified in multivariate analysis associated with lack of response to intravenous steroids.20 The value of this paediatric ulcerative colitis activity index (PUCAI)22 to facilitate the decision about ‘rescue’ therapy has been demonstrated.23 The change in stool frequency may also have predictive value. In a prospective study of 67 patients, lack of response to steroids was predicted by a less than 40% reduction in stool frequency within 5 days.24 The question of what counts as a ‘stool’ is not trivial. There is no standardisation and it is generally regarded as an evacuation (be it blood, liquid, or solid stool) that is counted as a bowel movement by the nursing staff. This is inevitably imprecise. Other clinical factors have been considered, including disease extent, duration, previous therapy (steroids, thiopurines), number of previous attacks and even gender (table 1), but all arise from retrospective analysis of admission outcome and none are in widespread use. Disease activity is independent of disease duration and age in prospective studies (table 2), even if old age is associated with mortality.
Retrospective studies identifying criteria to predict outcome in acute severe colitis
Prospective studies identifying criteria to predict outcome in acute severe colitis
Truelove and Witts' criteria on admission
Clinical criteria on admission also help predict outcome, rather than after 3 days' treatment. Recent data suggest that the number of Truelove and Witts' criteria1 on admission are associated with colectomy.14 The more of these criteria in addition to a bloody stool frequency of six or more per day, the more severe the systemic inflammatory response, and it is not surprising that the biological severity of an attack of colitis predicts colectomy. In a retrospective study of 294 episodes of colitis in 186 patients, the risk of colectomy was 9% (11/129) if patients had one additional criterion, compared with 31% (29/94) if two additional criteria were present and 48% (34/71) if three or more additional criteria were present (p=1.4×10−5; OR 4.35, 95% CI 2.20 to 8.56 one criterion vs two or more).14 Therefore, simply counting the number of additional criteria on admission helps identify those patients at higher risk of colectomy.
Laboratory criteria associated with the need for colectomy
Biochemical markers include C-reactive protein (CRP) and albumin, among others (table 1). Both CRP and albumin are measures of the inflammatory response. Although the ESR is one of the original Truelove and Witts' criteria, it has not been shown to be of predictive value in recent prospective studies (table 2).
C-reactive protein
The CRP crops up in several indices after prospective study and in the only validated index (PUCAI). It is the key biochemical marker in both the Oxford and the Swedish indices,34 in conjunction with stool frequency. On the third day of treatment, when the CRP was 45 mg/l or greater and the stool frequency three to eight, then 85% patients came to colectomy (p<0.00625, corrected for multiple comparisons).19 The combination is being used to decide when to initiate rescue therapy with either ciclosporin or infliximab in a prospective trial (CONSTRUCT, ISRCTN22663589). In the prospective Outcome of Steroid Therapy in Colitis Individuals (OSCI) study20 to evaluate short-term corticosteroid response rates in 128 children hospitalised with acute severe colitis, the significant predictors at day 3 were: nocturnal diarrhoea (OR 3.4, 95% CI 1.9 to 6.1), number of daily stools (OR 2.7, 95% CI 1.7 to 4.3), amount of blood in stool (OR 4.2, 95% CI 2.0 to 8.9) and CRP (OR 1.3, 95% CI 1.1 to 1.6). The faecal calprotectin differed (p=0.039), but was not significant by OR.20 Confirmation in different patient groups provides reassurance that the CRP is a useful, objective marker of predictive value for steroid failure. This resonates with clinical practice, because CRP is commonly measured and stool frequency can simply be monitored in patients with acute severe colitis. There is some evidence that the rate of change in CRP during intensive treatment predicts response,49 but this has not yet been quantified in a clinically useful tool.
Albumin
Albumin synthesis is suppressed by pro-inflammatory cytokines. A low albumin in acute severe colitis should cause concern and has been associated with colectomy in retrospective case series: 42% with an albumin less than 30 g/l at the end of the first day came to colectomy in the early series of 189 admissions from St Mark's,21 and in a large series from Edinburgh.41 Sensitivity was increased by combining albumin with clinical (stool frequency) and radiological data. However, multivariate analysis in prospective case series has not identified albumin as an independent marker of colectomy.19 20 24 Nevertheless, response to ciclosporin was predicted by an Edinburgh index greater than 5, which includes albumin and has been validated at another institution.50
Faecal calprotectin
Calprotectin is a neutrophil-derived protein that correlates with the severity of intestinal inflammation. The discriminant ability of faecal calprotectin to predict colectomy, corticosteroid, or infliximab non-response was investigated in 90 inpatients with acute severe colitis.44 Twenty-one received infliximab and 31 came to colectomy, including 11/21 who had received infliximab. Faecal calprotectin was strikingly elevated (median 1020 μg/g) and significantly higher in patients coming to colectomy (1200 vs 887, p=0.04), with a 97% specificity for a predictive threshold of 1922 μg/g. The sensitivity was low (24% at this threshold), even if a high proportion (87%) of such patients came to colectomy in the next year. Calprotectin only showed a trend towards significance for predicting corticosteroid or infliximab response. This is similar to data from the OSCI study.20 Although calprotectin predicted non-response to steroids (72% at a threshold >6150 μg/g), it did not predict response to infliximab and was significantly less effective than the combination of stool frequency and CRP (in the PUCAI and Oxford indices) at predicting colectomy. At this stage faecal calprotectin has potential for predicting response to treatment, but more studies are needed. Care must be taken to distinguish predictive factors for acute severe colitis from lesser degrees of activity.
Other markers
Other laboratory markers have been used, from fibrinogen to cholinesterase activity (table 1), but none have been validated. An Italian group have suggested that a pH greater than 7.4 predicts colectomy,27 perhaps because an alkaline pH occurs with hyperventilation associated with abdominal distress, but the acid-base buffering capacity does not make this likely to be a sensitive or reproducible tool. A prospective study of four faecal markers found that although all four were substantially elevated in 101 children with acute severe colitis (calprotectin 4215 μg/g (2297–8808); lactoferrin 212 μg/g (114–328); M2-pyruvate kinase 363 U/g (119–3104); and S100A12 469 μg/g (193–1112)), none were responsive to change and the clinical PUCAI was better at predicting poor response to intravenous corticosteroids.48
Radiological criteria associated with the need for colectomy
Radiological criteria include the presence of colonic dilatation greater than 5.5 cm (associated with a 75% need for colectomy), or mucosal islands on a plain abdominal radiograph (75% colectomy),21 or (most recently) single stage colonic MRI. A plain radiograph in acute severe colitis is an audit measure of care (http://ibdaudit.rcplondon.ac.uk/2008/). A radiograph not only excludes colonic dilatation, but also estimates the distribution of colitis, because the distal distribution of faeces approximates to the proximal distribution of colitis and colonic mucosal oedema is usually visible. Mucosal islands are remnants of mucosa surrounded by ulceration, visible as circular opacities en face, which provides collateral evidence of the severity of inflammation. As with other predictive markers, they are often overlooked or their significance is not recognised. The presence of an ileus (indicated by three or more small bowel loops of gas) was associated with colectomy in 73% in a retrospective study,46 but only 50% in a prospective study from the same institution.19 The depth of colonic ulceration after gentle air insufflation identified 42/49 patients with deep ulcers that were associated with the need for colectomy,33 but this is not widely used in practice.
Onto the predictive stage has come colonic MRI. In a prospective study, 22 patients with acute severe colitis underwent unprepared MRI (coronal breath hold T2-weighted HASTE images) within 48 h of admission.51 Six colonic segments were consensus-scored for inflammation by haustral loss, mesenteric oedema and mural thickness. Admission total colonic inflammatory score, but not admission CRP or stool frequency, was associated with extended inpatient stay greater than 7 days, or colectomy (positive predictive value 87%). This approach has yet to be validated and the outcome that generally matters is colectomy, rather than an extended inpatient stay. Access to MRI is likely to be a limiting factor in many hospitals and it needs to be compared with the simpler day 3 CRP and stool frequency criteria, but quantifying risk by imaging has a direct appeal. There is no evidence, however, that the more accessible techniques of CT or ultrasound have predictive value, even though they may assist decision-making by evaluating the extent of colitis, colonic diameter, or complications such as intramural gas.
Endoscopic criteria associated with the need for colectomy
The presence of deep or large ulcers has been associated with colectomy in retrospective studies. Deep ulceration must precede colonic perforation and the timing of surgery should prevent perforation occurring, so this is expected.2 52 Endoscopy is an important component of care in acute severe colitis, to confirm the diagnosis and exclude confounding factors such as cytomegalovirus infection, although not usually repeated before initiating rescue therapy. Extensive deep colonic ulceration in 46/85 patients with acute severe colitis was associated with colectomy in 93%, compared with 26% without such lesions.53 Ulceration reached the circular muscle layer or deeper in 42/43 of the colectomy specimens. This has been independently confirmed,40 and deep ulceration at the index examination has also been associated with a higher long-term risk of surgery.35 43 Because of risks of perforation, flexible sigmoidoscopy is preferable to colonoscopy.2
In retrospective studies, a circular argument again remains possible, because patients with deep ulcers are more likely to have surgery when the clinician is convinced of their prognostic importance. In a placebo-controlled study in acute severe colitis,45 the severity of endoscopic lesions predicted neither response to infliximab, nor colectomy. The results of CySIF (comparing infliximab to ciclosporin as a rescue therapy, NCT00542152), in which endoscopy is performed but not used for decision-making, are awaited. Interobserver variation in endoscopic assessment also needs to be recognised. Just 76% agreement on the endoscopic categorisation of severe colitis was found among 10 inflammatory bowel disease specialists, with even lower agreement for lower degrees of severity.54
Genetic markers associated with colectomy
Genome-wide scanning has identified over 50 genetic determinants associated with susceptibility to ulcerative colitis (www.ibdgenetics.org). While providing new insights into disease pathogenesis, the potential clinical application of these gene discoveries remains uncertain. Detailed prospective studies are needed to address key issues relating genotype to phenotype—including natural history, progression to severe disease, response to therapy, for the known determinants—either individually, or combined with clinical or other markers. The few data available were described in candidate gene studies of cohorts identified retrospectively after a severe attack. Therefore, 1–3% of the general population express HLA DRB1*103, but 16% of those who come to colectomy express this haplotype.55 A polymorphism in the gene for the multidrug resistance (MDR-1) efflux pump is also associated with steroid resistance and therefore colectomy in ulcerative colitis.56 Although genetic polymorphisms have the potential to predict the outcome of disease in an individual from the time of diagnosis, they cannot be used for decision-making when colectomy threatens. The potential for pharmacogenetic profiling—to predict response or intolerance to corticosteroids, ciclosporin, or biological therapies, and target therapy appropriately—is being realised in other fields,47 but will rely on adequately designed prospective studies.
Composite indices predicting the need for colectomy
Combining clinical and laboratory data intuitively makes sense when developing a predictive index for complex clinical cases. The combination of CRP greater than 45 mg/l and a stool frequency of three to eight per day, or a stool frequency greater than eight per day on day 3 (‘Oxford index’) is probably simplest. Eighty-five per cent of those who met these criteria came to colectomy on that admission.19 The Swedish (‘fulminant colitis’) index was derived from retrospective data on 97 patients 34: the index is calculated on day 3 (stool frequency/day+0.14×CRP mg/l) and has been prospectively evaluated in a clinical trial, in which the positive predictive value of a score of 8 or greater for colectomy within 90 days was 69%.45 The Seo index is more complex, combining weighted components of stool frequency, pulse rate, albumin and haemoglobin to give a positive predictive value and negative predictive value to predict short-term colectomy of 52% and 97%, respectively, when the score was 180 or greater.37 57 The Edinburgh predictive index includes stool frequency, the presence of colonic dilatation and hypoalbuminaemia, derived from a retrospective study of 167 admissions, with a colectomy rate of 40%.41 Stool frequency and hypoalbuminaemia (<30 g/l) were weighted, to create a simple score: 85% of those who scored 4 or more came to colectomy. Colonic dilatation gave a score of 4, which trumped other components, as it should in clinical practice. It is not entirely clear whether the score is best applied on day 1 (an advantage) or day 3, but the simplicity appeals.
Some of these indices have been compared, albeit in children, although fortunately there are more similarities than differences in both presentation and outcome between adults and children with acute severe colitis.2 20 In a prospective study the PUCAI, followed closely by the Oxford index, strongly predicted response when compared with other measures (Swedish and Seo indices, CRP, or faecal calprotectin), p<0.001.20 Sensitivity matters most at an early stage and on day 3, a PUCAI greater than 45 had a negative predictive value of 94% for patients likely to fail intravenous corticosteroids. By day 5, specificity for needing rescue therapy matters most and a PUCAI score greater than 65 had a positive predictive value of 100%.
Conclusions
We can and should introduce metrics into clinical decision-making. A condition with a high morbidity and definable mortality such as acute severe colitis is a prime example. There is no preferred index, although the PUCAI and Oxford indices appear to work better than others in prospective validation in children. Apart from a rough estimate gained by counting the number of additional Truelove and Witts' criteria,14 we cannot yet predict outcome reliably on the day of admission. The importance is to use one of the indices to bring objectivity to decision-making, to allow time for the patient and family to come to terms with potential outcomes (figure 1). Numbers in an objective index provide a threshold that should trigger contingency planning. When, for instance, set criteria on day 3 (eg, CRP >45 mg/l and stool frequency three to eight, or stool frequency greater than eight) are met, then therapeutic options beyond steroids should be discussed with the patient, consultation with a colorectal surgeon arranged, stomatherapist introduced, and a decision made that day about ‘rescue’ therapy or surgery. However, numbers cannot replace clinical judgement, so the experienced clinician also considers the age of the patient, previous pattern of colitis and therapy, radiographic appearance, endoscopic evidence of deep ulceration and the patient's views. Contingency planning allows time for rescue therapy to work within the bounds of safety. The ‘worst’ outcome is not surgery, but the mortality associated with acute severe colitis.58
Algorithm of selected factors for predicting outcome in acute severe colitis. All percentages indicate risk of colectomy on that admission, except where otherwise indicated. See text for references and details. AXR, plain abdominal radiograph; CRP, C-reactive protein; PUCAI, paediatric ulcerative colitis activity index; T&W, Truelove and Witts'.
Acknowledgments
It is with a great sense of loss that the authors pay tribute to Marc Lémann, who died unexpectedly on 26 August 2010, after this paper was submitted. He was an outstanding clinical investigator with wit and wisdom, and also a personal friend.
References
Footnotes
Competing interests None.
Provenance and peer review Commissioned; externally peer reviewed.
Linked Articles
- Recent advances in clinical practice