CIBEREHD is funded by the Instituto de Salud Carlos III from the Spanish Ministry of Health.
Decision tree for early introduction of rescue therapy in active ulcerative colitis treated with steroids†
Article first published online: 6 APR 2011
Copyright © 2010 Crohn's & Colitis Foundation of America, Inc.
Inflammatory Bowel Diseases
Volume 17, Issue 12, pages 2497–2502, December 2011
How to Cite
Mañosa, M., Cabré, E., Garcia-Planella, E., Bernal, I., Gordillo, J., Esteve, M., Zabana, Y., Gassull, M. A. and Domènech, E. (2011), Decision tree for early introduction of rescue therapy in active ulcerative colitis treated with steroids. Inflamm Bowel Dis, 17: 2497–2502. doi: 10.1002/ibd.21634
- Issue published online: 9 NOV 2011
- Article first published online: 6 APR 2011
- Manuscript Accepted: 10 DEC 2010
- Manuscript Received: 12 OCT 2010
- inflammatory bowel disease;
- ulcerative colitis;
- predictors of response;
Corticosteroids are the treatment of choice for moderate-to-severe active ulcerative colitis (UC) but up to 30%–40% of patients fail to respond. It has been reported that early clinical-biological parameters may identify those patients at high risk of colectomy. The aim was to identify predictors of rapid response to systemic steroids in moderate-to-severe attacks of UC.
Consecutive patients treated with prednisone 1 mg/kg/day for moderate-to-severe attacks of UC were prospectively included. Clinical and biological parameters at 3 and 7 days after starting steroids were recorded. Response was defined as mild or inactive UC activity at day 7 (as assessed by the Montreal Classification of severity) together with no need for rescue therapies (cyclosporin, infliximab, or colectomy). A logistic regression analysis was performed to identify those independent predictors of response. In addition, a decision-tree analysis was also performed.
Sixty-eight percent of patients (64 out of 94) responded to steroids. In the univariate analysis the number of bowel movements, rectal bleeding, platelet count, and C-reactive protein (CRP) levels at day 3 were associated with response at day 7, but only rectal bleeding was found to be an independent predictor in the logistic regression analysis. Conversely, the classification and regression tree (CART) model included these four variables. The decision-tree model showed a higher sensitivity in predicting a rapid response to steroids than the logistic regression one.
Rapid response to steroids in active UC attacks can be predicted after 3 days of treatment by simple clinical and biological parameters. A decision-tree model for early introduction of rescue therapies is provided. (Inflamm Bowel Dis 2011)
Ulcerative colitis (UC) is a chronic inflammatory condition that typically produces recurrent flares of mucosal inflammation. Systemic steroids are still the first-line treatment for moderate-to-severe acute UC attacks.1 However, 30%–40% of patients with severe flares do not respond to intensive intravenous steroid therapy and need rescue medical therapies or even colectomy.2, 3 Moreover, up to 45%–50% of patients treated either with oral or intravenous steroids will not achieve a complete remission within 30 days.4, 5 Historically, there has been a lack of consensus in defining steroid-refractoriness in active UC. This clinical situation has been defined by very heterogeneous criteria, such as lack of clinical response to 7–10 days of intravenous (IV) prednisolone at 1 mg/kg/day6 or persistence of active disease despite prednisolone up to 0.75 mg/kg/day over a period of 4 weeks.1
A rapid clinical response is desirable in cases of moderate-to-severe active UC in order to avoid patient clinical deterioration and nutritional worsening. On the other hand, the use of steroids is associated with a high rate of side effects, and great concerns about steroid-related risk of infections have recently been raised.7–10 In this era, when the usefulness of immunomodulators and biologics is widely recognized, long-term steroid therapy has to be deemed unacceptable. Nevertheless, if appropriately used steroids still play a role in the management of acute UC due to their proven efficacy and their low cost.
Until the mid-1990s, proctocolectomy was the only therapeutic alternative in patients unresponsive to steroids. Nowadays, cyclosporine and infliximab have become medical alternatives to proctocolectomy in this clinical setting. Both drugs have shown their efficacy in randomized controlled trials (RCTs), avoiding urgent colectomy in 60%–80% of steroid-refractory UC patients.11, 12 Severity of the flare has been identified as the only prognostic factor for lack of response to both cyclosporine13, 14 and infliximab.12 The availability of predictive factors of response to steroid therapy should allow the earlier indication of rescue therapies. This might improve their efficacy while decreasing the rate of complications associated with prolonged steroid use and protracted severe disease activity (such as infections or malnutrition). Available data suggest that a simple and early evaluation by means of clinical, biological, and/or radiological markers could easily predict the outcome of severe attacks of UC treated with steroids, but most of them arise from retrospective studies.3, 15–23
The aim of the present study was to identify simple clinical and biological predictors of rapid response (within 7 days) to steroid treatment in patients with moderate to severe UC attacks and to elaborate a decision algorithm that could be easily used in clinical practice.
PATIENTS AND METHODS
All consecutive UC patients treated with oral or IV steroids at the physician discretion for moderate-to-severe attacks in three tertiary Spanish centers from September 2004 to February 2009 were prospectively included. Diagnosis of UC was based on the Lennard-Jones criteria24 and disease severity was defined according to the Montreal Classification.25 Briefly, this classification follows the classical Truelove criteria and scores UC severity as mild (S1), moderate (S2), or severe (S3) depending on the number of bowel movements (≤4 stools/day, 5–6 stools/day, or >6 stools/day) and the presence or absence of signs of toxicity (tachycardia, fever, anemia, or increased erythrocyte sedimentation rate [ESR]). We chose the Montreal Classification as a simple method to group patients by the severity of the flare at the time steroids were started and because it is easy to obtain in clinical practice, whereas other clinical indexes require biological markers (ESR, albumin, hemoglobin, etc.) or even endoscopic parameters that are not always immediately available. Patients were included before or within the first 24 hours of steroid therapy. Those patients with disease limited to the rectum and those who had received steroids in the preceding 3 months were not included. Oral steroids were given at a dosage of 1 mg/kg/day of prednisone (or the equivalent dose of 6-methyl-prednisolone in case of IV administration). For the study purposes, response was defined as a disease Montreal severity score of S0 or S1 (<4 stools/day without blood and no systemic symptoms) at day 7, together with no need for rescue therapies. In these cases the steroid tapering schedule as well as the maintenance therapy was decided by the physician in charge. In case of clinical worsening or lack of response, patients received rescue therapy (cyclosporine, infliximab, or colectomy) according to medical decision.
All patients (both in- and outpatients) were visited at baseline, day 3, and day 7 after starting steroid therapy. Demographic (age, gender) and epidemiological data (smoking habits, family history of inflammatory bowel disease, previous appendectomy), as well as data related to UC history (age at diagnosis, time from diagnosis, disease extent, number of acute attacks, previous drug therapies, duration of the current attack) were recorded. Vital signs and clinical observations (pulse, temperature, number of bowel movements, passage of blood or mucus, tenesmus, urgency, rectal irritation, abdominal pain or tenderness, extraintestinal manifestations, Montreal Classification of severity), acute phase reactants (white blood cell count, ESR, C-reactive protein [CRP], fibrinogenemia, platelet count) and other biological parameters that might be altered as a consequence of disease activity (hemoglobin, serum albumin, potassium) were assessed 3 and 7 days after starting steroid therapy. UC extent was defined as the most extensive colonic involvement evaluated by endoscopy or abdominal scintigraphy (when complete colonoscopy could not be performed) prior to the inclusion in the study.
The study was evaluated and approved by the Ethics Committee of the steering center (Hospital Universitari Germans Trias i Pujol), and patient consent was obtained in each case.
Comparison of quantitative and qualitative variables between responders and nonresponders was made with the Student's t-test and the χ2 test, respectively. Those variables reaching a P-value lower than 0.05 in the univariate analysis were included in a stepwise logistic regression analysis to identify independent predictors for response after 7 days of steroids. Multivariately adjusted P-values to enter and remove terms to the model were <0.1 and >0.15, respectively.
In addition, data were also analyzed by means of decision tree analysis (DTA). Whereas logistic regression enables the identification of independent variables associated with the defined target, DTA allows adding a weight to the variables. DTA splits data into binary branches (classification and regression tree [CART] option) according to the values of variables and continues splitting branches in an iterative process that leads to the target value. Each split depends on the value of only one variable. DTA rejects those variables that play no role in arriving at the target.
Receiver operating curves (ROC) were built to compare the prognostic yield of both analyses.26 Statistical analyses were performed using the SPSS 15 package for Windows (Chicago, IL) and RV v. 2.10.1 (R. Ihaka and R. Gentleman, University of Auckland, New Zealand).
A total of 94 UC patients were included in the study with a median age of 37.5 years (range, 18–81) and a median time from UC diagnosis of 37 months (range, 0–324). Of interest, almost two-thirds of patients had extensive UC, 57% of patients were included at disease onset, and 78% of the whole series had a severe flare as assessed by the Montreal Classification (Table 1). All inpatients received subcutaneous LMW heparin to prevent thromboembolic events while admitted.
|Whole Series (n = 94)||Response (n = 64)||Failure (n = 30)||P|
|Male gender (%)||55 (58)||35 (55)||20 (66)||0.38|
|Age (years)||37 (17–81)||39 (17–81)||35 (18–69)||0.26|
|Familial history of IBD||6 (6)||3 (5)||3 (10)||0.38|
|Smokers||12 (13)||8 (12.5)||4 (13)||0.99|
|Distal||34 (36)||26 (41)||8 (27)|
|Extensive||60 (64)||38 (59)||22 (73)|
|Time from UC diagnosis (mo)||30 (0–324)||40 (0–324)||32 (0–240)||0.97|
|New onset UC||54 (57)||34 (53)||20 (67)||0.31|
|Treatment prior to steroids||0.87|
|None||26 (28)||18 (28)||8 (27)|
|Mesalazine||52 (55)||36 (56)||16 (53)|
|Thiopurines||16 (17)||10 (16)||6 (20)|
|Previous episodes of steroid-refractoriness||8 (8.5)||4 (13)||4 (6)||0.26|
|UC severity at inclusion||0.026|
|Moderate||21 (22)||19 (30)||2 (7)|
|Severe||73 (78)||45 (70)||28 (93)|
At day 7, 64 patients (68%) achieved clinical response as previously defined. No differences at baseline were observed between responders and nonresponders at day 7, except for disease severity (P = 0.026) and the proportion of patients with >4 bowel movements (68% and 93%, respectively, P = 0.018).
Among those parameters measured at day 3, severity of the flare, >4 bowel movements/day, rectal bleeding, high CRP, and high platelet count were associated with nonresponse in the univariate analysis (Table 2). Stepwise logistic regression analysis identified rectal bleeding (odds ratio [OR] = 20.3, 95% confidence interval [CI] = 2.3–174.2, P = 0.006) at day 3 as the only independent predictor of response to steroids at day 7.
|Response (n = 64)||Nonresponse (n = 30)||P|
|Severity (Montreal 0-1/2/3) (%)||53/30/17||7/13/80||<0.01|
|<4 bowel movements (%)||44||13||<0.01|
|Rectal bleeding (%)||45||97||<0.01|
|Platelet count (×109/L)||324||415||<0.01|
The same input variables were used in the DTA (Fig. 1). For the whole dataset, rectal bleeding was the first splitting variable; in the absence of rectal bleeding at day 3, 97% of patients could be correctly classified as steroid response. CRP was identified as the second splitting variable in this node and a best cutoff level of 43 mg/L was identified; lower levels allowed an appropriate classification of responders in 64% of cases. Platelet count was the following splitting variable; in case of platelet count lower than 318 × 109/L, 100% of patients were responders. Finally, the number of bowel movements became the last splitting variable, with 63% of the remaining patients correctly classified as responders if bowel movements were less than four.
The CART and logistic regression models were similar in terms of specificity and diagnostic yield, but the former showed a trend to a higher sensitivity (Fig. 2).
Steroids have been the first-line treatment of moderate-to-severe acute flares of UC since the studies by Truelove and Witts in the 1950s.27 No other drug demonstrated a similar efficacy until the appearance of cyclosporine in the 1990s28 and, more recently, infliximab.29 Steroids remain the gold standard therapy in this clinical setting mainly because their rapid therapeutic onset (within hours or days) and their low cost and ease of monitoring. However, it has been reported that 20% of patients fail to improve with steroids and a further 30% will improve only partially within 30 days of treatment.4, 5 In these cases, prolonged inflammatory activity usually worsens the patient's nutritional status and clinical condition. Long-term use of steroids is associated with an unacceptable rate of adverse events. Moreover, great concerns have been recently raised since several studies showed that steroid use is one of the most important risk factors for infectious complications among inflammatory bowel disease patients, especially when combined with other immunosuppressant drugs.7–10, 30
To date, only a few prospective studies trying to identify predictors of response to steroids in UC have been reported. Travis et al5 evaluated 51 consecutive episodes of severe UC flares and found that the most simple rule to predict treatment failure was the persistence of >8 stools/day or 3–8 stools/day together with CRP concentration >45 mg/L after 3 days of steroid therapy. Benazzato et al,16 in another prospective study, evaluated 67 episodes of severe UC, found ESR and body temperature at baseline and the change in the number of bowel movements from baseline to the day 5 as the only independent predictors of steroid failure. Some retrospective studies supported that both persistence of a high number of bowel movements15, 17, 22, 23 and hypoalbuminemia17, 19 were associated with a poorer outcome. All these studies were designed to identify predictors of medical treatment failure as defined by the need for colectomy in the short term, but not by the requirement of other rescue therapies such as cyclosporine or infliximab.5, 16, 17
In clinical practice, a proportion of patients present only a partial or slow response to steroids that is not so bad as to lead to colectomy in the short term. However, Travis et al5 found that patients with a severe UC flare with a partial response to steroids who avoided immediate colectomy had shorter remission periods and higher requirements of both immunosuppressive therapy and colectomy within the following year. Finally, despite the use of sequential therapies in recent years, the rate of medical therapy failure remains around 20%–30%.5, 16, 17, 22, 27
In our view, the relevant decision concerns the early switching to rescue therapies, not merely avoiding colectomy, and also to reduce the rate of steroid-related adverse events by shortening their use in those patients who will not draw any therapeutic benefit from them. In fact, time-bound strategies for evaluation of therapeutic success have been recently recommended in order to improve patient outcomes and reduce side effects.31 For this reason we chose to define treatment failure as the persistence of symptoms after 7 days of steroid therapy instead of the need for colectomy. Our results agree with previous studies in confirming that an early decrease in CRP levels as well as early clinical improvement (assessed by rectal bleeding and bowel movements) are good predictors of response to steroids, and that this can be predicted soon after starting therapy. Moreover, our results indicate that these simple parameters do not predict the risk of colectomy but the likelihood of needing rescue therapies or prolonged steroid use. To our knowledge, this is the first study providing a decision-tree algorithm in UC for time-bound therapeutic escalation. The CART model allows deciding whether to introduce early rescue therapies or to persist on steroids with an accurate prognostic yield and an acceptable sensitivity. In this sense, the absence of rectal bleeding after 3 days on steroid therapy is associated with a high probability of response at 1 week; conversely, in case of persistence of rectal bleeding together with CRP levels higher than 43 mg/L rescue therapies should be considered. In contrast to previous studies, the number or the variation of bowel movements had a lesser relative weight in the CART model; in our opinion, the absolute number of bowel movements may be subjective as far as those patients with more rectal symptoms (that may not correspond to those with more severe flares) should overestimate the number of stools per day.
Radiological and endoscopic parameters were not assessed in this study, although they have been associated with poorer outcomes. However, it must be pointed out that these tests are not always available at a precise timepoint in clinical practice, and findings at endoscopy or plain abdominal radiograph may incorporate some degree of subjectivity. In addition, although the presence of large ulcers at rectosigmoidoscopy,16, 20 and intestinal gas, mucosal islands, or colonic dilatation on a plain abdominal radiograph,17, 21, 22 have been associated with a worse clinical outcome, only the latter was independently associated with colectomy in a unique retrospective study.17
Finally, most published studies were restricted to patients with severe UC admitted for IV therapy. We intentionally included both severely ill inpatients and outpatients with moderate disease treated with oral steroids. Two different epidemiological studies found that a proportion of patients with partial response to a first course of steroids may develop steroid dependence or require colectomy within the following year.4 Moreover, hospital admission for intravenous treatment after an unsuccessful course of oral steroids has been associated with a higher risk of steroid resistance.15 Thus, predicting clinical response to steroids may be clinically relevant even in patients with moderate disease activity treated by the oral route. In clinical practice, the choice of the IV route is not only based on the severity of the flare but also on other circumstances.32
In summary, although steroids should be considered as first-line therapy in moderate-to-severe active UC because of their efficacy and low costs, their prolonged use is unacceptable and early introduction of rescue therapies is warranted. Our results confirm that complete response to steroid therapy can be easily predicted as soon as 3 days after starting treatment by means of simple clinical and biological parameters. Whether early introduction (after 3 days of steroids) of cyclosporine or infliximab would increase their efficacy and/or reduce morbidity should be evaluated appropriately in future studies.
The authors thank Abbott Laboratories for providing statistical advice.
- 8Corticosteroids pose an increased risk of serious infections: an interim safety analysis of the ENCORE registry. Gastroenterology. 2008; 134( suppl 1): A–140., , , et al.
- 9Infliximab use in patients with inflammatory bowel disease and the risk of serious bacterial infections. Gastroenterology. 2008; 134( suppl 1): A–140., , , et al.
- 14Predictive factors of response to cyclosporine in steroid-refractory ulcerative colitis. Am J Gastroenterol. 2008; 103: 637–642., , , et al.Direct Link:
- 25Toward an integrated clinical, molecular and serological classification of inflammatory bowel disease: report of a Working Party of the 2005 Montreal World Congress of Gastroenterology. Can J Gastroenterol. 2005; 19( suppl A): 5A–36A., , , et al.
- 32Long-term colectomy rate after intensive intravenous corticosteroid therapy for ulcerative colitis prior to the immunosuppressive treatment era. Am J Gastroenterol. 2007; 102: 2513–2519., , , et al.Direct Link: