SEARCH

SEARCH BY CITATION

Keywords:

  • cyclosporine;
  • ulcerative colitis;
  • steroid resistant;
  • predictive factors

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Background: One-third of patients with steroid-refractory ulcerative colitis (UC) do not respond to cyclosporine and require colectomy. Since alternative pharmacological treatments for this condition are available, it is pertinent to identify factors that predict response. The objective of this study was to determine predictive factors of response prior to cyclosporine administration, with validation in an independent cohort.

Methods: The 2 cohorts of patients were identified from prospectively established databases. All patients had received 1 mg/kg/day prednisolone or equivalent for at least 5 days before cyclosporine. The efficacy measure was need of early surgery (within 3 months).

Results: From 1998 to 2005, 34 patients were treated in 1 institution (derivation cohort) and 38 patients in the second institution (validation cohort). Eleven patients in the derivation cohort and 9 patients in the validation cohort underwent early colectomy. Univariate analysis in the derivation cohort demonstrated a significant association of colectomy with C-reactive protein (P = 0.012) and the Ho index before initiation of cyclosporine (P = 0.013). Regression analysis showed that only the Ho index (P = 0.011) had an independent predictive value. The Ho index predicted need of colectomy, with an area under the characteristic receiver operating curve of 0.79 (95% confidence interval [CI], 0.59–0.99) in the derivation cohort and 0.74 (95% CI, 0.53–0.96) in the validation cohort. The cutoff point with the best sensitivity and specificity ratio was ≥5.

Conclusions: The Ho-based predictive score is a good predictor of response to cyclosporine and avoidance of colectomy, and may aid in the indication of this treatment for management of steroid-resistant UC.

(Inflamm Bowel Dis 2007)

Ulcerative colitis (UC) is a chronic inflammatory bowel disease with a heterogeneous natural history. Approximately 15% of patients with UC will suffer a severe attack requiring hospitalization.1 Corticosteroids are used as first-line therapy in moderate and severe flares of UC, achieving remission rates of 54%–86.9%.2, 3 Cyclosporine has shown efficacy for treatment of steroid-refractory UC. Lichtiger et al4 observed that a greater percentage of steroid-refractory patients treated with cyclosporine had a response in comparison with those receiving placebo (82% versus 0%). In subsequent control and uncontrolled series of patients response rates ranging from 56%–93% have been reported.5–8 In addition, it has been shown that low-dose intravenous cyclosporine (2 mg/kg/d) has similar efficacy to high-dose cyclosporine (4 mg/kg/d) and may be associated with fewer adverse effects.9 Although oral cyclosporine has not been directly compared with intravenous cyclosporine, open-label studies suggest also a high efficacy in the treatment of steroid-refractory UC.10–13 However, overall the available evidence indicates than one-third of patients with steroid-refractory UC will not respond to cyclosporine and require colectomy.

Recently, the efficacy of infliximab for induction and maintenance of remission in UC has been evidenced in 2 large controlled trials (ACT I and ACT II),14 and the superiority of infliximab over placebo in the treatment of steroid refractory UC has been proven in an open-label study.15 Until a direct comparison of infliximab and cyclosporine in the treatment of steroid-refractory UC is available, the clinician faces a difficult decision between these 2 treatment options in UC patients refractory to steroid treatment. The availability of predictors of response to either form of treatment may be of help in guiding this decision.

The aim of this study was to identify factors that may predict clinical response and need of colectomy immediately prior to cyclosporine administration in a series of consecutive patients from a referral university hospital, and validate the results in an independent series from another institution.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Patients were identified from prospectively established databases in 2 Spanish university hospitals. Thirty-four patients from the Hospital Clínic of Barcelona and 38 patients from the Hospital Germans Trias i Pujol of Badalona that had been treated with cyclosporine between 1998 to 2005 were included in the study. Demographic and clinical data including age and gender, UC history (time from diagnosis, disease extension, previous medical treatment), characteristics of current UC flare, immediately prior to cyclosporine administration, including clinical parameters of disease severity, biological data (stool cultures albumin, hemoglobin, C-reactive protein [CRP], erythrocyte sedimentation rate, leukocyte, and neutrophil counts), radiological data (signs of colonic dilatation [transverse colon diameter ≥6 cm] or gas in small bowel at plain x-ray), and composite indexes of disease severity (modified Truelove and Witts disease activity score,4 Ho index16 (Table 1), and Lindgren index17), were retrieved from the medical records in all patients.

Table 1. Ho Index: Risk Score Attributable to Each Category
VariablesScore
  1. Overall risk score = [score attributable to mean stool frequency (0,1,2 or 4)] + [presence of colonic dilatation (0 or 4)] + [presence of hypoalbuminemia (0 or 1)]. Minimum score = 0, maximum score = 9.

Mean stool frequency 
<40
4 ≤ 61
6 ≤ 92
>94
Colonic dilatation4
Hypoalbuminemia 
< 30 g/L1

All patients met the criteria of steroid resistance defined as failure to respond to treatment with 1 mg/kg/day prednisolone or the equivalent for at least 5 days. Patients with cytomegalovirus infection were excluded. Cyclosporine was administrated orally or intravenously. Starting doses of cyclosporine were 10 mg/kg/day for the oral route and 4 mg/kg/day for the intravenous (i.v.) route. Subsequent doses were adjusted according to blood levels, targeting 200–400 ng/mL. Patients who achieved clinical remission under cyclosporine treatment, defined as absence of blood in stool and no diarrhea, were shifted to azathioprine at a dose of 2.5 mg/kg following a treatment scheme previously recommended.18 Colectomy was performed when the clinical condition deteriorated during cyclosporine treatment, a clinical response was not obtained after cyclosporine administration during 14 days, or the clinical condition deteriorated within 3 months after treatment with cyclosporine.

Outcomes

The principal endpoint of the study was the need of early surgery within 3 months since cyclosporine treatment. Response to cyclosporine was defined as avoidance of colectomy at 3 months. Additional exploratory analyses were also performed considering colectomy during the index admission as the endpoint.

Statistical Analyses

χ2 analyses for qualitative variables and Student's t-test for quantitative variables were used. Those variables with significance P < 0.1 were introduced in a stepwise multiple logistic regression analysis to identify independent predictive factors for response to cyclosporine avoiding early colectomy. A characteristic receiver operating curve (ROC) analysis was carried out to determine the optimal cutoff point of the calculated predictable function. Sensitivity and specificity were calculated with a confidence interval (CI) of 95%. To compare the ROC curves we used the formula: z = (AUC1-AUC2) / √ standard error (1)2 + standard error (2)2 α N (0,1). A result >1.96 means that the 2 curves are different.19 All statistical analysis was performed using SPSS 11.0 package for Windows (Chicago, IL).

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The baseline characteristics of the study population are described in Tables 2 and 3.

Table 2. Demographic, Clinical, and Biological Baseline Characteristics of the Study Population
 Derivation Cohort (N = 34)Validation Cohort (N = 38)Significance
  1. NS, not significant.

Sex (M/F)21/1322/16NS
Age (years)36.3 ± 15.5534.13 ± 11.61NS
Disease duration30.90 ± 36.2855.24 ± 77.08NS
Disease location  P = 0.033
 Proctitis1 (2.9%)9 (23.7%) 
 Left sided8 (23.5%)9 (23.7%) 
 Extensive25 (73.5%)20 (52.6%) 
C-reactive protein (mg/L)8.53 ± 8.155.49 ± 5.00P = 0.05
Erythrocyte sedimentation rate (mm/h)51.47 ± 32.5750.72 ± 26.08NS
Hemoglobin (g/L)11.91 ± 2.039.80 ± 1.57P = 0,000
Albumin (g/L)34.84 ± 5.9930.06 ± 5.89P = 0.002
Leukocyte count (x106)10766 ± 301812920 ± 5812NS
Antibiotic use1719NS
Positive stool culture33NS
Colonic dilatation105NS
Ho index3.16 ± 2.652.59 ± 1.96NS
Lindgren index15.45 ± 9.0610.19 ± 7.25P = 0.006
Truelove17.45 ± 2.5812.84 ± 2.24P = 0.000
Corticosteroids duration (days)17 ± 31.3937.28 ± 51.37NS
Table 3. Characteristics of the Derivation and Validation Cohorts According to Colectomy
 Derivation Cohort (n = 34)Validation Cohort (n = 38)
Colectomy (n = 11)No Colectomy (n = 23)Colectomy (n = 9)No Colectomy (n = 29)
  • a

    Bowel movements, categorical variable (1 = <4 bowel movements, 2 = 4-6 bowel movements, 3 = >6 bowel movements).

  • b

    ESR (erythrocyte sedimentation rate).

Sex (M/F)4/717/65/317/13
Age (years)39.10 ± 14.1835.46 ± 15.5929.0 ± 8.6535.5 ± 12.03
Disease duration (months)44.10 ± 50.5026.33 ± 30.749.5 ± 9.367.43 ± 82.63
Disease location.    
 Proctosigmoditis0127
 Left-sided2636
 Extensive916416
Modified Truelove17.60 ± 1.4317.42 ± 2.8713.5 ± 2.9713.0 ± 2.03
Ho index5.5 ± 3.212.3 ± 1.494.75 ± 3.322.55 ± 1.76
Lindgren index21.8 ± 11.5115.09 ± 9.7014.57 ± 11.2710.00 ± 6.93
Bowel movements (1–3)a32.792.752.31
C-reactive protein level (mg/L)14.01 ± 8.376.62 ± 6.919.06 ± 7.014.82 ± 4.40
Albumin (g/dL)33.10 ± 6.5434.96 ± 5.6929.57 ± 3.8229.86 ± 6.15
Leukocyte count (x106)9125 ± 213611521 ± 368513712 ± 476112155 ± 5818
Hemoglobin (g/L)11.04 ± 1.3712.20 ± 2.1510.28 ± 0.899.62 ± 1.63
ESRb (mm/h)58.40 ± 31.4448.58 ± 33.2668.75 ± 32.6248.14 ± 24.65
Duration of corticosteroids (days)25.00 ± 47.6119.25 ± 35.1828.75 ± 40.8536.10 ± 50.04
Duration of cyclosporine (days)17.30 ± 35.1733.38 ± 41.9210.13 ± 6.0315.67 ± 4.46
Antibiotic use611613
Parenteral nutrition5424
Positive stool culture2112
Colonic dilatation6432

Derivation Cohort

The derivation cohort included 34 patients with steroid-refractory UC. Of these, 21 patients (64.7%) were male and 16 patients (35.3%) were female. Mean age at admission was 36.5 (SD 15.1). Six patients were under treatment with azathioprine prior to admission. All patients had a moderate to severe flare according to the modified Truelove–Witts activity index and fulfilled the criteria of steroid resistance defined as persistent disease activity after treatment with 1 mg/kg/day of oral or i.v. steroids for at least 5 days.

Twenty-four patients were initially treated with oral cyclosporine and 10 patients with i.v. cyclosporine. The major criteria used to prescribe i.v. treatment were the presence of colonic dilatation or significant ileus. Patients received cyclosporine for a mean of 28.65 days (SD 35.96). Mean levels were of 386 ± 133 ng/mL (95% CI, 339–433).

A response to treatment with cyclosporine and avoidance of colectomy was obtained in 23/34 (67.64%) patients; the remaining 11 patients (32.36%) were colectomized during the first 3 months due to lack of response (n = 6) or early relapse of disease activity (n = 5). The proportion of patients responding to cyclosporine was similar in those in whom the drug was given i.v. (60%) and those treated with oral cyclosporine (75%).

No serious adverse events were observed in this cohort. Four patients had an infectious complication associated to cyclosporine, but none was severe: 1 was herpes simplex, and 3 were oral candidiasis. Two patients had tremor and 1 patient suffered hypertension and elevation of creatinine level associated with high levels of cyclosporine, but both events resolved after cyclosporine discontinuation.

Validation Cohort

The validation cohort included 38 patients with steroid-refractory UC; 22 were male (57.89%) and 16 were female (42.11%). Mean age at admission was 34.13 (SD 11.13). Similar to the derivation cohort, all patients had a moderate to severe flare according to the modified Truelove–Witts activity index and fulfilled the criteria of steroid resistance according to the above-mentioned definition. In this group all patients were treated with i.v. cyclosporine at a dose of 4 mg/kg/day with a mean duration of 14.5 days (SD 5.26).

Clinical characteristics of the derivation and validation cohorts are detailed in Table 2. The 2 cohorts differed significantly in disease location (more patients had distal colitis in the validation cohort) and some parameters of disease severity.

A response to treatment with cyclosporine and avoidance of colectomy was obtained in 29/38 (76.3%) patients; the remaining 9 patients (23.7%) were colectomized during the first 3 months due to lack of response (n = 7) or early relapse of disease activity (n = 2).

Table 3 shows the baseline characteristics of the 2 cohorts of patients related to the need of colectomy.

Predictors of Response

Univariate analysis in the derivation cohort demonstrated a significant association of early colectomy with the 2 following baseline parameters: CRP (P = 0.012) and Ho index (P = 0.013). The number of stools per day (P = 0.057), leukocyte counts (P = 0.077), hemoglobin (P = 0.072), Lindgren index (P = 0.092), and colonic dilatation (P = 0.08) had borderline significance. Other variables were not different between the group of patients who underwent colectomy and those who did not (Table 3).

Variables with a significance of P ≤ 0.1 were included in the regression analysis (CRP, Ho index, leukocyte counts, and hemoglobin level), prioritizing the level of significance in the univariate analysis, and avoiding the duplication of variables contained in composite indexes. The number of stools and colonic dilatation were not included because they are contained in the Ho index, and the Lindgren index was not included since it contains CRP in its parameters. The regression analysis showed that the Ho index was the only independent predictive factor of response (P = 0.011) and no other variable significantly improved the prediction function.

The model correctly predicted response to cyclosporine avoiding colectomy in 87% of cases in the derivation cohort and in 82% of patients in the validation cohort.

The Ho-based index predicted failure to cyclosporine and need of colectomy, with an area under the characteristic ROC of 0.79 (95% CI, 0.59–0.99) in the derivation cohort and 0.74 (95% CI, 0.53–0.96) in the validation cohort (Fig. 1). When these 2 curves were compared they were not significantly different (z = 0.03).

thumbnail image

Figure 1. Characteristic receiver operating curve of the 2 cohorts for prediction of colectomy based on the Ho index in the derivation cohort (A) and validation cohort (B).

Download figure to PowerPoint

The Ho-based index value with the best specificity and sensitivity ratios to predict failure to cyclosporine and the need of colectomy was determined to be ≥5. We chose a cutoff point with a high specificity to predict response to cyclosporine treatment avoiding colectomy, and thus aid in the indication of this therapeutic option for management of steroid-resistant UC. Specificity was 91%, with a sensitivity of 55%, positive predictive value of 66.6%, and negative predictive value of 80% in the derivation cohort, whereas specificity was 82%, sensitivity 55.5%, positive predictive value 50%, and negative predictive value 85% in the validation cohort.

In an exploratory analysis combining the derivation and validation cohorts the factors significantly associated with a higher need of colectomy in the univariate analysis included CRP (11.8 ± 7.7 versus 5.5 ± 5.8 P < 0.01), number of stools per day (2.82 ± 0.5 versus 2.51 ± 0.7 P < 0.05), dilatation of the colon (45% versus 12%, P < 0.05), the Ho index (4.9 ± 2.9 versus 2.3 ± 1.5 P < 0.01), and the Lindgren index (19.5 ± 11.0 versus 11.7 ± 8.5 P < 0.01). The multivariate analysis confirmed that the Ho index was the only independent predictor of colectomy in the combined cohort.

When the analysis was restricted to colectomies performed during the initial hospitalization the Ho index also performed as the best predictor, with an optimum cutoff point of 6. Among patients with a Ho index <6, 93.1% (27/29) avoided colectomy in the derivation cohort and 96.7% (29/30) in the validation cohort, whereas in the group of patients with a Ho index ≥6, 60% (3/5) of patients in the derivation cohort and 57.1% (4/7) in the validation cohort required surgery during the initial hospitalization. The area under the characteristic ROC was 0.87 (95% CI, 0.73–0.99) in the derivation cohort and 0.82 (95% CI, 0.65–0.99) in the validation cohort.

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

A considerable body of evidence supports the efficacy of cyclosporine in the treatment of patients with steroid-refractory UC. Nevertheless, previous studies indicate that up to one-third of patients fail to respond to this treatment. In that regard, our data are in keeping with previous reports with 68% of patients in the derivation cohort and 76% in the validation cohort responding to cyclosporine and avoiding early colectomy (within 3 months).

In that scenario, a reliable predictor of drug failure could be of invaluable help to improve management of steroid-refractory UC, avoiding delays in indication of surgery or administration of other second-line therapies. Ho et al16 found that the mean stool frequency, colonic dilatation, and hypoalbuminemia were independent predictors of outcome in patients receiving steroid treatment for a UC attack. In the Ho et al study, patients with a high index (≥6) failed to respond to corticosteroid treatment and required second-line therapies or colectomy. In the current study we demonstrate that a low Ho index (<5) before commencing cyclosporine treatment in a condition of steroid refractoriness is a reliable predictor of drug response. Our data show that a Ho index ≥5 is associated with a high treatment failure rate (66.7%) whereas among patients with an index < 5 the majority (80.0%) respond to cyclosporine. The cutoff value of 5 provides an elevated specificity for treatment response and is therefore reliable for decision-making when other therapeutic modalities have to be considered, including surgery or anti-TNF biologicals. It is also interesting to note that a Ho index ≥6 is associated with a significant risk of needing urgent colectomy during the initial hospitalization, and these patients should be closely monitored.

The derivation and validation cohorts come from different hospital centers, and differ in some clinical characteristics such as disease location and disease severity as assessed by clinical indices and biomarkers (CRP). This fact probably reflects the reality of variation in population characteristics according to centers, and also different thresholds for hospitalization and introduction of cyclosporine treatment. In spite of these differences, the specificity of a Ho index <5 was high in both the derivation and the validation cohorts, and the area under the ROC obtained was also similar in both cohorts.

Our results also support the notion that response rates to oral cyclosporine are similar to those observed with intravenous treatment. In fact, in the derivation cohort, in which both forms of administration were used, the response rates tended to be higher in patients treated with oral cyclosporine, but this trend is probably related to the fact that patients treated with i.v. cyclosporine were those with more severe disease, since the factors used to decide i.v. administration (presence of ileus or colonic dilatation) are related to disease severity.

Travis et al20 described in a prospective study an index to predict the outcome in hospitalized patients with active UC, showing that after 3 days of treatment patients with frequent stools (>8/day), or raised CRP (>45 mg/L) are highly likely to require colectomy at the same admission. However, in this study most of the patients were under treatment with steroids, and only 28% received cyclosporine. When Travis' index was applied to the population of the derivation cohort, a high index of responses (8/9) was observed in patients without risk factors, whereas in patients with risk factors the colectomy rate was 40%.20 Carbonnel et al21 could not find any predictive factor of response to corticosteroid treatment in hospitalized patients with severe UC.

Prior studies have addressed the possibility to predict response to cyclosporine in steroid-refractory patients. A recent study by Moskovitz et al22 found that factors predicting failure of i.v. cyclosporine and the need for colectomy were left-sided disease (as opposed to extensive disease) and prior treatment with azathioprine. In the present study only 6 patients in the derivation cohort were under azathioprine treatment at the time of cyclosporine administration, and we did not find any trend toward different response rates related to azathioprine pretreatment (P = 1), or disease location (P = 0.66). In another study analyzing factors associated with response to cyclosporine in steroid-refractory UC, Cacheux et al23 reported that body temperature >37.5°C, heart rate >90 bpm, CRP >45 mg/L, and the presence of a severe lesion at colonoscopy were independent predictors of colectomy. However, in this study stool frequency, colonic dilatation, or albumin levels, which are included in the Ho-based index, were not analyzed. The current study is coincident in the finding of the association of high CRP with cyclosporine failure, but this variable was not significant in the multivariate logistic regression analysis. One of the limitations of the current study is its retrospective nature and the lack of endoscopy evaluation in some cases, preventing the inclusion of this variable in the predictive function analysis. Finally, in another retrospective study Rowe et al24 evaluated potential predictors of response to cyclosporine. They observed that only a high percentage of band neutrophils was a predictor of response to cyclosporine, although this study did not evaluate commonly used clinical variables such as number of stools, colon dilatation, or CRP. In our study neither leukocyte counts nor neutrophil counts were related to the need of colectomy.

A direct comparison between cyclosporine and infliximab in steroid-resistant UC patients is highly needed. Meanwhile, our results indicate that patients with a Ho-based index <5 can be treated with cyclosporine avoiding colectomy in the majority of cases. Whether patients with a Ho index ≥5 should be treated with anti-TNFs or should undergo surgery remains to be determined, since in the study by Jarnerot et al15 patients with fulminant disease had a considerably lower response rate to infliximab than patients with milder forms of steroid-resistant UC.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The authors thank the nurses and staff at the participating hospitals.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES