Predictive factors of response to intravenous ciclosporin in severe ulcerative colitis: the development of a novel prediction formula

Authors


Correspondence to:

Dr T. Katsuno, Department of Medicine and Clinical Oncology (K1), Graduate School of Medicine, Chiba University, 1-8-1 Inohana, Chuo-ku, Chiba-shi 260-8670, Japan.

E-mail: katsuno@faculty.chiba-u.jp

Summary

Background

When treating patients with severe ulcerative colitis (UC), accurate prediction of drug efficacy contributes to early clinical decision-making.

Aim

To identify predictive factors and to develop a reliable prediction formula and a decision tree of response to intravenous ciclosporin treatment for severe UC.

Methods

Patients included in this study were those diagnosed with refractory severe UC who had undergone ciclosporin treatment between December 2004 and March 2011 at a tertiary referral centre in Japan. Demographic and clinical parameters from all patients were analysed by multivariate statistics.

Results

Fifty-two patients were included in this study (36.5% men with an average age of ciclosporin initiation of 40.2 ± 15.6 years). Thirty-four patients (65.4%) were responders to the treatment with ciclosporin and avoided colectomy, 18 patients (34.6%) were nonresponders and underwent colectomy. Stepwise multiple logistic regression analysis identified four independent predictive factors of response to intravenous ciclosporin: age at hospitalisation (AGE), platelet count (×104/μL) on the first day (PLA), Lichtiger score on the third day (LIC) and total protein (g/dL) on the third day minus total protein on the first day (ΔTP). The calculation formula (8.5 − 0.16 × AGE + 0.21 × PLA − 0.61 × LIC + 2.3 × ΔTP < 0) predicted colectomy with an accuracy of 88.5% and the decision tree predicted colectomy with an accuracy of 90.4%.

Conclusion

The novel calculation formula and the decision tree effectively predict the clinical outcome of ciclosporin treatment for severe ulcerative colitis as early as on day 3 after starting ciclosporin treatment.

Introduction

Ulcerative colitis (UC) is a chronic colonic inflammatory disease characterised by a life-long chronic course with remissions and exacerbations. Approximately 25% of patients have an acute severe colitis requiring hospitalisation during their illness.[1, 2] Patients with severe symptoms are treated with intravenous corticosteroids as a first-line therapy with a response rate of approximately 60%.[3-6] Previous studies have determined some predictive indices of response to intravenous corticosteroids in patients with UC: Oxford Index consisting of stool frequency and C-reactive protein on the third day after corticosteroid administration, Sweden Index (fulminant colitis index) consisting of stool frequency and C-reactive protein on the third day after corticosteroid administration, and Ho Index consisting of serum albumin on the first day, transverse colon diameter on plain X-ray, and average stool frequency on the first to third days after corticosteroid administration (Table S1).[7-9] In other reports, the presence of severe endoscopic lesions and long duration of present flare-up were also shown to be useful for predicting response to intravenous corticosteroids.[10-12] It is noteworthy that all of the factors needed to calculate these indices are composed of the clinical data obtained within 3 days after starting intravenous corticosteroids. Therefore, these indices are helpful for early prediction of the effectiveness and earlier clinical decision-making in patients with UC treated with corticosteroids.

Treatment with ciclosporin, a calcineurin inhibitor, has been recommended for patients with severe UC refractory to intravenous steroids. Ciclosporin is efficacious with initial response rates ranging from 62.5% to 83.7%.[13-16] That is, 20–40% of steroid-refractory UC patients treated with ciclosporin still require colectomy. Although a few studies determined predictive factors of response to ciclosporin, there is no preferred index of response to intravenous ciclosporin in patients with severe UC.[17-20] Early prediction of the therapeutic response of ciclosporin is crucial because ciclosporin has been considered one of the final treatments for steroid-refractory severe UC patients before colectomy is considered.

Recently, alternative drugs such as infliximab and adalimumab have emerged for the treatment of severe UC.[21, 22] In a prospective clinical study on patients with acute severe UC refractory to intravenous steroids, ciclosporin was not more effective than infliximab for achieving short-term remission and avoiding urgent colectomy (CySIF study).[23] This study suggested that ciclosporin and infliximab are equivalently efficacious in patients with severe UC refractory to intravenous corticosteroid. There have been some reports of switching from ciclosporin to infliximab and vice versa if one treatment is not effective.[24-27] However, higher rates of adverse events and mortality due to long-term stronger immunosuppression have warned that administering the sequential therapy should be carefully decided on an individual basis. If we could predict the efficiency of the individual treatment for severe refractory UC with high accuracy, an early clinical decision-making would be made and ineffective medications would be avoided before patients undergo long-term stronger immunosuppression. The present study was planned to determine a formula and a decision tree capable of predicting the effectiveness of ciclosporin for severe refractory UC. Reliable predictors of response of ciclosporin would be helpful for improving management of steroid-refractory severe UC by avoiding delays in administration of other second-line therapies or decision to perform surgery.

Methods

Patients

Fifty-two sequential patients with refractory severe UC who had undergone ciclosporin (Novartis Pharma Inc., Basel, Switzerland) treatment between December 2004 and March 2011 at a tertiary referral centre in Japan were reviewed retrospectively. The diagnosis of UC was made based on clinical, radiological and pathological criteria. Disease severity was assessed using Lichtiger score (Table S2).[13, 28] We defined severe UC as Lichtiger score ≥10. Steroid resistance was defined as failure to respond to treatment with 1 mg/kg/day intravenous prednisolone for at least 5 days. Ciclosporin was administered intravenously to all the patients with severe steroid-resistant UC. Starting doses of ciclosporin were 3 mg/kg/day and subsequent doses were adjusted to a target plasma ciclosporin concentration of 400–450 ng/mL according to the guidelines of the Japanese Ministry of Health. Cytomegalovirus (CMV) reactivation was diagnosed when CMV antigenemia was detected, CMV inclusion bodies were detected in colonic mucosal biopsies histologically, or CMV DNA was detected in colonic mucosal biopsies by PCR analysis. Clostridial infection was diagnosed by detecting CD toxin in the stool or detecting Clostridium bacteria in the stool culture. For early detection of side effects and infection, stool culture, CD toxin test, cytomegalovirus antigenemia test and Chest X-ray were performed at least once a week and blood tests were carried out 2 or 3 times a week for all the patients. Oral sulfamethoxazole-trimethoprim was prophylactically used during treatment with ciclosporin.

Epidemiological, clinical and biological parameters and colonoscopic findings were collected as follows: Epidemiological data included age, gender, age at onset of UC, time from diagnosis, disease extension, duration of the current attack, previous medical treatment (steroids, immunomodulator, and leucocytapheresis), and number of previous hospitalisations. Clinical parameters included body temperature, pulse rate, stool frequency, stool frequency on the third day minus that on the first day (Δstool frequency), Lichtiger score and Lichtiger score on the third day minus that on the first day (ΔLichtiger score). Biological data included total protein, albumin, haemoglobin, C-reactive protein and leucocyte and platelet counts on the first and third days of ciclosporin administration. Total protein, albumin, haemoglobin, and C-reactive protein on the third day minus each corresponding value on the first day (ΔTP, Δalbumin, Δhemoglobin and ΔCRP respectively) were also collected. Colonoscopic findings were evaluated by Mayo score, EAI score, ulcer size, ulcer depth, redness, bleeding, mucosal oedema and mucous exudates just before ciclosporin administration.[29, 30] Diameter of the transverse colon on plain X-ray, cytomegalovirus infection, clostridial infection and plasma ciclosporin concentration were retrieved from the medical records for all patients. If the biological data on the third day were not available, we used biological data from the morning of the fourth day for analysis.

The primary endpoint of this study was colectomy within 3 months after ciclosporin treatment. Colectomy was performed in patients who were complicated with bleeding, perforation or toxic megacolon, clinical symptoms worsened during treatment with ciclosporin, a clinical response was not obtained after ciclosporin administration for 14 days, or no remission (Lichtiger score >4) within 3 months of medication including ciclosporin. Patients who avoided colectomy within 3 months after starting ciclosporin treatment were defined as responders.

Statistical analysis

Baseline characteristics were compared between the two groups using the Chi-squared test or Mann–Whitney U test. Logistic regression analysis was utilised to identify potential factors for inducing remission. Independent variables with significance < 0.25 were introduced in stepwise multiple logistic regression analysis to identify independent predictive factors for response to intravenous ciclosporin. Predictive factors were also analysed by means of decision-tree analysis with Chi-squared automatic interaction detection (CHAID). All statistical analyses were performed using IBM spss software version 20 (IBM Corporation, Somers, NY, USA).

Results

Figure 1 shows the clinical course of the assessed patients for this study as well as treatment that they received and their outcomes. Between December 2004 and March 2011, 142 patients with moderate-to-severe UC were hospitalised in our hospital. Three of them had undergone urgent colectomy because of serious haematochezia or colonic perforation. Other patients received intravenous prednisolone treatment, resulting in 3 patients having undergone colectomy, 58 patients with good response and 78 patients with poor response. Among 78 poor responders to the intravenous prednisolone treatment, 26 patients with moderate activity (defined by Lichtiger score <10 after intravenous prednisolone treatment) were then treated with oral tacrolimus (Astellas Pharma Inc., Tokyo, Japan) treatment that has been approved for refractory UC in Japan, while 52 patients with severe activity (defined by Lichtiger score ≥10 after intravenous prednisolone treatment) were recruited in this study.

Figure 1.

Clinical course of the assessed patients. Patients with moderate-to-severe UC hospitalised in our hospital between December 2004 and March 2011 were assessed for this study. Clinical outcomes of 142 patients assessed for this study are shown accompanied by the treatment they have received.

The clinical characteristics of the recruited patients are described in Table 1. Fifty-two sequential patients with severe UC refractory to steroids who received intravenous ciclosporin were included in the present study. All patients had a severe flare-up according to the Lichtiger score and fulfilled the criteria of steroid resistance. Nineteen patients (36.5%) were male and 33 patients (63.5%) were female. Mean age at admission was 40.15 ± 15.60 (95% CI, range 17–69). Six patients had received oral ciclosporin treatment for the previous flare-up, but discontinued it at least 3 months before the latest administration. One patient had received low-dose oral tacrolimus for 2 years but flared-up after stopping it 4 weeks before the ciclosporin administration. Two patients had started infliximab as induction therapy, but their symptoms had worsened before the third infusion of infliximab and discontinued it 4 weeks before the ciclosporin administration. Twenty-nine patients had received 2–10 sessions of leucocytapheresis that is approved in Japan, but they did not respond to it and discontinued it 2–14 days before ciclosporin administration. Four patients had been taking azathioprine and they continued taking it during ciclosporin treatment.

Table 1. Baseline characteristics of patients
Number52
  1. Data are expressed as means ± s.d.

GenderMale 19/Female 33
Age at hospitalisation (years)40.15 ± 15.60 (range 17–69)
Time since diagnosis of UC (months)64.44 ± 73.79 (range 0.3–240)
Disease extensionPancolitis 35/Left-sided 16/Proctitis 0/Right-sided 1
Number of previous hospitalisationsMedian 1 (range 0–6)
Body temperature (°C)37.29 ± 0.64 (range 35.9–39.6)
Pulse rate (/min)88.15 ± 12.30 (range 68–130)

Endoscopic findings before administration

(Mayo UC Endoscopic Score)

Severe disease 34/Moderate disease 18
Lichtiger score at the first day of ciclosporin administration13.69 ± 2.40 (range 10–18)
Leucocyte count (/μL)9761.54 ± 4134.46 (range 2800–25900)
Haemoglobin (g/dL)10.55 ± 1.99 (range 7.2–14.7)
Platelet count (×104/μL)34.39 ± 10.94 (range 10.8–59.4)
Total protein (g/dL)5.80 ± 0.66 (range 4.4–7.0)
C-reactive protein (mg/L)53.3 ± 44.7 (range 0.0–177)

Nineteen patients were diagnosed as CMV reactivation after a mean of 19.2 ± 10.1 days from steroid administration and after a mean of 8.8 ± 9.5 days from CsA administration. All of them were treated with intravenous ganciclovir as a pre-emptive therapy and CMV antigenemia became negative in all the patients, but one patient temporarily experienced reactivation after the discharge. Four patients were diagnosed as clostridial infection after starting ciclosporin (range 1–41 days). They were treated with oral metronidazole and we confirmed that CD toxin and Clostridium bacteria disappeared.

Patients received ciclosporin for a mean of 16.32 ± 4.43 days (95% CI, range 1–27). Mean plasma ciclosporin level was 427.53 ± 77.58 ng/mL (95% CI, range 203.3–800.0). Ciclosporin concentration showed no significant differences between colectomy and responder groups. Thirty-four patients (65.4%) were responders to the treatment with ciclosporin and avoided colectomy, whereas 18 patients (34.6%) were nonresponders and underwent colectomy throughout this study (Figure 1). Clinical response was achieved in 65.4% of patients in the present study, suggesting that the effectiveness of ciclosporin was similar to the previous reports in steroid-refractory severe UC.[13-16] In 14 days of ciclosporin treatment, 25 patients achieved clinical remission (Lichtiger score ≤4) and all of them remained in remission 3 months after ciclosporin administration. Nine patients underwent surgical therapy in 14 days. The other 18 patients had a partial response in 14 days and continued intravenous ciclosporin treatment for more several days and/or started leucocytapheresis sessions. Of 18 patients with partial response, 8 patients achieved clinical remission and one patient was still in a partial response (Lichtiger score = 5) at the time of discharge, whereas the other 9 patients underwent surgical therapy within 3 months after ciclosporin adminstration. Of 34 responders in this study, 24 patients were on thiopurines, but 10 patients did not agree to take thiopurines at the time of discharge. Twenty-six received oral ciclosporin and 8 received oral tacrolimus after stopping intravenous ciclosporin until thiopurines exert their therapeutic efficacy in 3 months. One patient who was in a partial response at the time of discharge achieved clinical remission in 3 months.

As a serious adverse event, renal dysfunction (elevated creatinine, hyperkalaemia), occurred in one patient and administration of ciclosporin was discontinued on the third day. One patient suffered from nasal septal abscess and another with periodontitis during receiving ciclosporin, but they were cured immediately by using antibiotics. There were no severe infections and no death in this study.

To determine which epidemiological, clinical or biological parameters were capable of predicting response to intravenous ciclosporin, univariate analyses of the parameters between patients who did and did not show a response were performed (Table 2). Older age at hospitalisation (= 0.000), lower platelet counts on the first day (= 0.011) and higher Lichtiger score on the third day (= 0.003) showed strong associations with colectomy. Figure 2 shows time course of changes in Lichtiger score, demonstrating that Lichtiger scores on the third day were associated with the outcomes on the 14th day. Moreover, lower platelet counts on the third day (= 0.031), higher stool frequency on the third day (= 0.042) and lower ΔLichtiger score (Lichtiger score on the third day minus that on the first day) (= 0.019) were shown to be significantly associated with colectomy by univariate analyses.

Figure 2.

Time course of Lichtiger scores. Lichtiger scores of the patients treated with intravenous ciclosporin on the first, third, and fourteenth days are shown. Data are expressed as means ± s.d. The solid line indicates the mean time course of Lichtiger scores in patients with response. The dotted line indicates mean time course of Lichtiger scores in the patients who underwent colectomy.

Table 2. Results of univariate analyses
 RespondersColectomyP
  1. Left, left-sided colitis; Pan, pancolitis; Pro, proctitis; Right, right-sided colitis.

  2. Univariate analyses on each factor were performed between the responder group and the colectomy group. Data are expressed as means ± s.d.

  3. a

    Statistically significant.

  4. b

    Differences between the data on the third and first days are shown.

Gender (M/F)14/205/130.148
Age at hospitalisation (years)34.12 ± 12.9651.56 ± 13.90.000a
Time from diagnosis of UC (months)63.82 ± 80.0365.62 ± 62.450.436
Disease extensionPan 24/Left 10/Pro 0/Right 0Pan 13/Left 4/Pro 0/Right 10.664
Number of previous hospitalisationsMedian 0 (range 0-3)Median 1 (range 0-6)0.091
Previous steroid treatment (Yes/No)21/1313/50.455
Body temperature (°C)37.26 ± 0.7637.36 ± 0.340.057
Pulse rate (/min)87.26 ± 13.0289.83 ± 10.990.345
Endoscopic findings before administration
Mayo UC endoscopic scoreSevere 21/moderate 13Severe 13/moderate 50.455
EAI scoreMedian 13 (range 10–15)Median 13 (range 9–16)0.806
Lichtiger score on the first day13.32 ± 2.2414.39 ± 2.620.138
Stool frequency on the first day (times/day)10.47 ± 6.6013.39 ± 6.980.103
Leucocyte count on the first day (/μL)10144.1 ± 4421.59038.9 ± 3533.60.430
Haemoglobin on the first day (g/dL)10.50 ± 2.1810.65 ± 1.640.637
Platelet count on the first day (×104/μL)37.32 ± 9.9328.86 ± 10.870.011a
Total protein on the first day (g/dL)5.76 ± 0.655.87 ± 0.700.413
Albumin on the first day (g/dL)3.00 ± 0.463.05 ± 0.630.839
C-reactive protein on the first day (mg/L)54.5 ± 43.250.9 ± 48.60.658
Lichtiger score on the third day10.76 ± 2.3913.11 ± 3.200.003a
Stool frequency on the third day (times/day)7.03 ± 3.8510.89 ± 6.430.042a
Leucocyte count on the third day (/μL)9588.2 ± 3547.77850.0 ± 2870.80.062
Haemoglobin on the third day (g/dL)10.04 ± 1.919.93 ± 1.361.000
Platelet count on the third day (×104/μL)37.31 ± 9.2129.68 ± 13.430.031a
Total protein on the third day (g/dL)5.64 ± 0.495.56 ± 0.680.370
Albumin on the third day (g/dL)2.96 ± 0.372.89 ± 0.560.750
C-reactive protein on the third day (mg/L)24.0 ± 19.343.9 ± 52.30.191
ΔLichtiger scoreb−2.56 ± 2.02−1.28 ± 1.960.019a
Δstool frequencyb−3.05 ± 3.39−2.50 ± 3.240.607
Δtotal proteinb−0.13 ± 0.48−0.31 ± 0.560.272
Δalbuminb−0.04 ± 0.32−0.16 ± 0.460.485
ΔCRPb−30.5 ± 33.9−7.1 ± 47.00.017a
Transverse colon diameter at plain X-ray35.57 ± 10.5433.41 ± 14.410.392
Presence of cytomegalovirus disease (Yes/No)11/238/100.394
Presence of clostridial infection (Yes/No)3/311/170.677
Ciclosporin concentration on the third day (ng/mL)389.82 ± 115.03379.32 ± 114.810.744

To identify subjects to be included in a subsequent multivariate analysis, variables with significance of < 0.25 by univariate analyses were introduced into stepwise multiple logistic regression analysis. The analysis identified four independent predictive factors of response to intravenous ciclosporin: age at hospitalisation (AGE), platelet counts (×104/μL) on the first day (PLA), Lichtiger score on the third day (LIC) and total protein (g/dL) on the third day minus total protein on the first day (ΔTP). We obtained a regression equation: 1/1 + e(−1 × score) (score = –8.471 + 0.160 × AGE + (–0.209) × PLA + 0.610 × LIC + (–2.347) × ΔTP) (Table 3). The value of the formula >0.5 predicted colectomy with an accuracy of 88.5%, positive predictive value of 83.3% and negative predictive value of 91.2% (Table 4). This formula is complicated and cannot be calculated easily at the bedside. Therefore, we have simplified it without reducing its accuracy: [8.5 + (–0.16) × AGE + 0.21 × PLA + (−0.61) × LIC + 2.3 × ΔTP < 0]. The simplified formula predicted colectomy with the same probability. This formula infers that older age, lower platelet count on the first day, higher Lichtiger score on the third day and a decrease in total protein level on the third day compared with that on the first day are associated with a higher rate of requiring colectomy. According to an ROC curve analysis (Figure 3), age at hospitalisation (AGE) and Lichtiger score on the third day (LIC) had moderate sensitivity and specificity for prediction (AUC = 0.815 and 0.750 respectively).

Figure 3.

ROC curves. ROC curves were created for the four factors extracted by the multivariate analysis in the present study. Abbreviations of the four factors are as follows: AGE, age at hospitalisation; LIC, Lichtiger score on the third day; PLA, platelet count (×104/μL) on the first day; ΔTP, total protein (g/dL) on the third day minus that on the first day.

Table 3. A predictive formula obtained by multivariate analysis
Multivariate analysis
  1. AGE, age at hospitalisation; LIC, Lichtiger score on the third day; PLA, platelet count (×104/μL) on the first day; ΔTP, total protein (g/dL) on the third day minus that on the first day.

Regression equation: [1/1 + e(−1 × *score) > 0.5] predicts colectomy
*score = −8.471 + 0.160 × AGE + (−0.209) × PLA + 0.610 × LIC + (−2.347) × ΔTP
Simplified formula: (8.5 − 0.16 × AGE + 0.21 × PLA − 0.61 × LIC + 2.3 × ΔTP < 0) predicts colectomy
Table 4. Verification of the predictive formula proposed in the present study
Predictions & outcomesPredictionsTotal
ResponderColectomy
  1. The predictive capacity of the formula proposed in the present study was assessed.

OutcomesResponder31334
Colectomy31518
Total341852

We finally performed decision-tree analysis (CHAID model) using the four factors revealed by the multiple logistic regression analysis (Figure 4). Lichtiger score on the third day (LIC) became the first splitting variable; in patients with LIC ≤13, 77.5% patients avoided colectomy, whereas in those with LIC >13, 75.0% patients underwent colectomy. Under the node of LIC ≤13, platelet counts (×104/μL) on the first day (PLA) became the second splitting variable and a best cut-off level of 33.1 was identified; in patients with PLA >33.1, 95.2% patients avoided colectomy. Age at hospitalisation (AGE) became the last splitting variable under the node of PLA ≤33.1 and a best cut-off level of 49 years old was identified; 73.3% of patients with AGE ≤49 avoided colectomy, whereas all the patients with AGE >49 underwent colectomy. Under the node of LIC >13, age at hospitalisation (AGE) became the second splitting variable and a best cut-off level of 44 years old was identified; in patients with AGE >44, 100% patients underwent colectomy. Total protein (g/dL) on the third day minus total protein on the first day (ΔTP) became the last splitting variable under the node of AGE ≤44 and a best cut-off level of −0.4 was identified; all the patients with ΔTP > −0.4 avoided colectomy, whereas all the patients with ΔTP ≤ −0.4 underwent colectomy. This decision-tree predicted colectomy with an accuracy of 90.4%, positive predictive value of 72.2%, and negative predictive value of 100.0%.

Figure 4.

Decision-tree analysis. Decision-tree analysis (CHAID model) using the four factors revealed by the multiple logistic regression analysis was performed. The nodes in which over 50% of patients underwent colectomy are shown in grey. This model as a whole predicts colectomy with an accuracy of 90.4%, positive predictive value of 72.2%, and negative predictive value of 100.0%. AGE, age at hospitalisation; LIC, Lichtiger score on the third day; PLA, platelet count (×104/μL) on the first day; ΔTP, total protein (g/dL) on the third day minus that on the first day.

To determine the association of colonoscopic findings before ciclosporin administration and the clinical outcomes, Mayo score, EAI score, ulcer size, ulcer depth, redness, bleeding, mucosal oedema and mucous exudates were analysed. However, we found no correlations between endoscopic findings before initiation of ciclosporin and clinical outcomes by univariate analyses or stepwise multiple logistic regression analysis.

Discussion

In the present study, stepwise multiple logistic regression analysis identified four independent predictive factors of response to intravenous ciclosporin in patients with severe UC refractory to corticosteroids: age at hospitalisation (AGE), platelet count (×104/μL) on the first day (PLA), Lichtiger score on the third day (LIC) and total protein (g/dL) on the third day minus total protein on the first day (ΔTP). These simple clinical criteria were shown to effectively predict the clinical outcome of ciclosporin treatment for steroid-refractory severe UC. Furthermore, we propose a novel calculation formula based on four factors (8.5 − 0.16 × AGE + 0.21 × PLA − 0.61 × LIC + 2.3 × ΔTP < 0) and a decision tree, both of which reliably predict colectomy and can be used in all hospitals because it consists of only simple clinical data.

There have been a few reports of predictive factors for the response to ciclosporin. Rowe et al. identified three predictive factors of absence of response to ciclosporin – tachycardia, hypoalbuminemia, and increased nonsegmented neutrophils – based on a retrospective study in 36 patients.[17] Cacheux et al. identified three variables (body temperature >37.5°C, heart rate >90 bpm, and C-reactive protein >45 mg/L) as predictive factors for colectomy.[18] Aceituno et al. reported that higher Ho index before initiation of ciclosporin is a predictor of colectomy, suggesting that lower serum albumin, wider transverse colon diameter on plain X-ray, and higher stool frequency predict colectomy.[19] In contrast, Rios et al. identified three factors (older age, thrombocytosis, and a history of previous treatment with ciclosporin) as predictive factors for colectomy.[20]

Multivariate analysis in our study first indicated that older age is an important factor predicting colectomy, similar to the report of Rios et al.[20] Ulcer healing in the colonic mucosa is delayed in elderly patients compared to younger patients. Furthermore, elderly patients often have complications in multiple organs and their general condition tends to be poor. Therefore, it is reasonable that older age at hospitalisation was extracted as one of the important background factors capable of predicting colectomy. Rios et al. also reported a history of previous treatment with ciclosporin as a factor that predicts colectomy. Univariate analysis in the present study indicated that a higher number of previous hospitalisations had a moderate association with colectomy, consistent with their reports, in that the repetition of severe attacks of UC brings about resistance to ciclosporin treatment.

As the second factor, lower platelet count on the first day was indicated as one of the predictors of colectomy by multivariate analysis. In the active phase of UC, activated monocytes, macrophages and lymphocytes produce inflammatory cytokines, such as IL-1, IL-6, and TNF-α, in the colonic mucosa and in the peripheral blood. The inflammatory cytokines, especially IL-6, are known to stimulate bone marrow leading to thrombocytosis. Ciclosporin is an immunosuppressant that inhibits production of IL-2 and other cytokines, including IL-6, in lymphocytes. Thrombocytosis indicates that there is continuous active inflammation in the colonic mucosa, suggesting the existence of a target of ciclosporin. Thrombocytosis should be considered one of the reasons for use of ciclosporin. Interestingly, Rios et al. suggested that thrombocytosis predicted colectomy, which was the converse of the results of the present study. This discrepancy was likely due to differences in blood concentration of ciclosporin. We maintained ciclosporin at 400–450 ng/mL, which was higher than that reported in other studies. Cacheux et al. reported that C-reactive protein (CRP) >45 mg/L on the first day predicts colectomy. However, CRP levels of most patients in the present study were >45 mg/L, and univariate analyses in the present study showed that higher CRP level on the first day had no association with colectomy. These discrepancies suggest that the UC severity of the patients in our cohort was higher than in previous studies.

Notably, multivariate analysis in the present study showed that Lichtiger score on the third day had a strong association with the clinical outcome, whereas the score on the first day had a much weaker association. Lichtiger score reflects both general condition and clinical symptoms, including stool frequency and abdominal pain. In the present cohort, a considerable number of patients whose Lichtiger scores were very high on the first day but were decreased on the third day of ciclosporin treatment finally showed clinical remission. These findings support the suggestion that signs of recovery appear as early as on the third day of ciclosporin treatment, and that Lichtiger score on the third day is a good marker for detecting the signs of recovery and reflects the efficacy of ciclosporin treatment in each patient. Although Aceituno et al. determined the stool frequency on the first day as one of the predictors and univariate analysis in our study showed that a greater stool frequency on the first day had a weak association with colectomy; Lichtiger score on the third day was extracted as a more significant factor for predicting the outcome by multivariate analysis. Rowe et al. reported that tachycardia on the first day, and Cacheux et al. reported that higher body temperature and higher heart rate on the first day were associated with colectomy, whereas the present study detected no significant associations of these parameters even on univariate analyses. In our recent clinical experience, body temperature and heart rate are easily compensated by antipyretics and treatment for dehydration, which probably explained why these were not extracted as predictive factors in the present study. The multivariate analysis and the decision-tree analysis in the present study extracted higher Lichtiger score on the third day as the best marker for colectomy from the factors representing the clinical severity and the early stage of clinical efficacy of ciclosporin treatment.

Finally, total protein level on the third day minus that on the first day (ΔTP) was extracted as a predictive factor associated with the outcome with lower ΔTP predicting colectomy. Malabsorption progresses in patients with severe UC due to maldigestion, gastrointestinal protein loss due to wide and deep ulcerations and increased energy expenditure. Decreasing levels of total protein are usually correlated with exacerbation of disease activity. Both Rowe et al. and Aceituno et al. reported that lower albumin level on the first day predicts colectomy, whereas univariate analyses in the present study showed no association of albumin levels on the first day with colectomy. However, our study demonstrated not only that total protein level reflects severity and duration of the current attack but also that their changes reflect the effectiveness of early stages of ciclosporin treatment, which can predict the final outcome. Interestingly, ΔCRP (CRP level on the third day minus that on the first day) also had a significant association with the outcome by univariate analysis with lower ΔCRP suggesting colectomy, but was not selected by multivariate analysis. ΔTP was considered to be extracted as a stronger marker for colectomy than ΔCRP by multivariate analysis among the blood data in the early stage of ciclosporin treatment.

Cacheux et al. demonstrated that severe endoscopic findings (extensive, deep ulcerations, large mucosal abrasions, weld-like ulcerations, or mucosal detachment on the edge of these ulcerations) were also independent predictive factors for colectomy.[18] Rowe et al. reported that an increase in nonsegmented neutrophil number before starting ciclosporin predicts colectomy. Aceituno et al. showed that wider transverse colon diameter on plain X-ray predicts colectomy. Conversely, Naganuma et al. reported that ciclosporin was effective even in patients with severe endoscopic findings (EAI 14–16), while patients with higher EAI scores did not respond to corticosteroids.[30] Similar to the results of Naganuma et al., we found no correlations between endoscopic findings before initiation of ciclosporin and clinical outcome.

Three of the patients in the group predicted for responders underwent colectomy. In two of these three patients, although Lichtiger score on the third day fell to <10, clinical symptoms worsened again until the fourteenth day of ciclosporin administration. Ciclosporin was thought to be less effective against long-term chronic inflammation with severe fibrosis and high levels of lymphocytic infiltration. In contrast, three of the patients who entered remission had been predicted for colectomy using the current predictive formula. Although they did not show recovery during the early stage of ciclosporin treatment, they showed marked improvement after the third day of ciclosporin treatment. These cases indicate that the clinical condition of individual patients should be observed carefully while taking into account the current formula. Finally, we considered the reason why we could not determine the association between the use of thiopurines and the outcome was because Walch et al. showed that previous failure under thiopurines was a strong predictor of nonresponsiveness to ciclosporin within the first 18 months of follow-up.[31] In the present study, only 4 patients received thiopurine therapy prior to administration of ciclosporin. The smaller number of patients receiving thiopurine therapy was considered responsible for the lack of statistical significance.

Here, we have proposed a novel calculation formula and a decision tree consisting of four simple clinical criteria of response to intravenous ciclosporin for patients with severe UC refractory to corticosteroids: age at hospitalisation, platelet count on the first day, Lichtiger score on the third day and total protein on the third day minus total protein on the first day. Age at hospitalisation and platelet counts on the first day represent background factors of each patient, whereas Lichtiger score on the third day and total protein on the third day minus that on the first day reflect the clinical severity and the early stage of clinical response to ciclosporin treatment. Utilising the predictive calculation formula and the decision tree for individual patients may facilitate earlier clinical decision-making in the treatment of severe refractory UC.

Acknowledgements

The authors express their gratitude to Atsushi Isono, Hirotsugu Uehara, Yasutaka Kato, Kiyoshi Furukawa, Kazuki Hatakeyama, Shin-ichiro Takeda, Yasushi Mandai, Tomoo Makita, Yoshie Takahashi and Yuhoko Furuya (Department of Gastroenterology and Hepatology, Chiba University Hospital, Japan) for their clinical assistance. Declaration of personal and funding interests: None.

Ancillary