SEARCH

SEARCH BY CITATION

Keywords:

  • inflammatory bowel disease;
  • ulcerative colitis;
  • colectomy;
  • hospitalization;
  • infliximab

Abstract

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

Background: Patients who require hospitalization for the management of ulcerative colitis (UC) may represent a subset with severe disease. These patients may be more likely to require future colectomy. There are limited data examining whether medical hospitalization is predictive of subsequent colectomy.

Methods: This was a retrospective case–control study utilizing the inflammatory bowel disease center database at our academic referral center. Cases comprised UC patients who underwent colectomy for disease refractory to medical management. The control population was comprised of all patients with UC who had not undergone colectomy. Multivariate logistic regression was used to identify independent predictors of requiring colectomy.

Results: There were a total of 246 UC patients included in our study, with 103 being hospitalized sometime in their disease course (41.9%). A total of 27 patients underwent colectomy (11%). Colectomy patients were significantly more likely to have been on infliximab therapy (51.8% versus 22.4%, P = 0.001) but no more likely to have been on immunomodulator therapy (74.1% versus 59.4%, P = 0.14). Patients who required medical hospitalization for UC were more likely to require future colectomy (20.4% versus 4.2%, P < 0.001) than those who had not required hospitalization. On multivariate analysis, requiring medical hospitalization for management of UC (odds ratio [OR] 5.37, 95% confidence interval [CI] 2.00–14.46) and ever requiring infliximab therapy (OR 3.12, 95% CI 1.21–8.07) were independent predictors of colectomy.

Conclusions: Requiring medical hospitalization for the management of disease activity in UC is an independent predictor of the need for colectomy. Future studies will determine whether aggressive medical management may modify the need for colectomy in this cohort.

(Inflamm Bowel Dis 2008)

Ulcerative colitis (UC) is a chronic inflammatory disorder of the colon with a variable course ranging from mild to severe disease.1–3 The emergence of new therapies for UC4 makes it important to develop algorithms that stratify patients by disease severity and institute appropriate treatment regimens. While a majority of the patients achieve good control of disease activity with medical management, patients who fail medical therapy require surgical treatment with colectomy for management of intractable disease.

Several centers have reported that between 8%–30% of patients with UC eventually require colectomy.5–13 A majority of the colectomies early on in the natural history of disease are for intractable symptoms despite maximal medical therapy or for severe colitis. While surgery is preferable to persistent severe disease and is sometimes life-saving in severe colitis, it may also be associated with several postsurgical complications. These include need for pouch excision or diversion in 4%–10% of patients, pouch fistulae, strictures, afferent and efferent limb syndromes, and pouchitis, with cumulative rates of this complication ranging from 23%–46%.14–18 In addition to these complications it is also associated with patients' concerns about body image.19–22 Among patients who eventually require colectomy, elective hospitalization for resection following failed outpatient management of disease activity may be preferable to emergent resections that are associated with higher morbidity and mortality.23–25 In this scenario, it is important to identify early those cohorts of patients with UC who have a more severe course of disease and hence a higher likelihood of requiring future colectomy.26–29

There are several risk stratification scores that have been proposed to calculate the likelihood of colectomy in patients with UC. However, these scores are limited to the subset of patients hospitalized with severe disease and do not provide a longitudinal risk over the course of natural history of UC.27, 30 The risk of colectomy is also directly dependent on disease activity. Various scales have been used in clinical trials to assess and monitor disease activity in UC including clinical, endoscopic, and combined models.31 However, such scores may not be widely used in clinical practice and also may not be easily tracked longitudinally across healthcare providers. There has been limited examination of whether a need for medical hospitalization predicts a future need for colectomy in UC patients. Previous studies examining factors predicting colectomy in UC were either restricted to hospitalized patients with severe colitis,29, 32 or did not examine the prior need for medical hospitalizations.7, 8, 27 We proposed that requiring medical hospitalization for the control of disease activity may be a manifestation of a more severe natural history of disease, and thus predictive of future need for colectomy. This is a simple parameter that could be ascertained from the patient during the initial encounter by any healthcare provider and could provide important prognostic information in planning future therapeutic strategy.

Our study had the following aims: 1) To identify the rates of colectomy for management of intractable disease in patients with UC from a tertiary referral inflammatory bowel disease (IBD) center; 2) To examine the rates of medical hospitalizations and demographic variables among those patients who required colectomy compared to those who did not; and 3) To examine if requiring inpatient management for control of disease activity was an independent predictor of the subsequent need for colectomy in UC.

MATERIALS AND METHODS

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

Study Setting

This was a retrospective case–control study performed at a single academic IBD referral center. The clinical database maintained by our IBD center was used to identify eligible patients for this study, as described previously.33 All patients with a diagnosis of UC who were currently receiving care or had received care previously at our center were included in our study. The diagnosis of UC was made using the standard criteria including clinical features, laboratory and radiologic findings, and lower gastrointestinal endoscopy. For patients who had these tests at outside centers, we reviewed the records to confirm the diagnosis. Patients with indeterminate colitis or with coexisting primary sclerosing cholangitis (PSC) were excluded from this analysis.

Variables and Outcomes

Demographic information including age and gender were available for each patient. In addition, information was also recorded about the age of diagnosis of UC and the duration of disease. Both age of onset and duration of disease were analyzed as continuous variables. However, sensitivity analysis was performed by analyzing these variables as dichotomous variables at cutoffs of 50 years and 10-year duration of disease, respectively. Extent of disease was characterized per the Montreal classification as pancolitis if the inflammation extended proximal to the splenic flexure, left-sided disease for inflammation restricted to the colon distal to the splenic flexure, and proctitis for disease confined to the rectum.34

Our primary predictor variable of interest was requiring medical hospitalization for the management of disease activity. This was identified through review of medical records including clinic notes, hospital discharge summaries, operative reports, and endoscopy records. Only hospitalizations directly related to UC were included. For patients who had received part of their care at other facilities prior to being evaluated at our center, we reviewed their initial clinic encounters for any reference to prior hospitalizations for UC. Patients who had colectomy for dysplasia or cancer, or who had undergone colectomy prior to their first encounter at our IBD with no information on preceding hospitalizations (n = 9) were excluded from our analysis.

Statistical Analysis

Data were analyzed using Stata 9.2 (StataCorp, College Station, TX). The chi-square test was used to perform between-group comparisons for categorical variables. Linear variables were compared using t-tests. Univariate logistic regression was performed with colectomy as the outcome of interest with the individual predictor variables. Variables that were significant in this analysis at P < 0.1 were then included in our final multivariate model. Stepwise multivariate logistic regression was performed to identify independent predictors of colectomy in UC. The analysis was repeated by serially excluding variables that were not statistically significant in the final model to identify any changes in the odds ratios (ORs) with a >10% change being considered significant. Independent predictors were considered to be significant in the final model at P < 0.05. Kaplan–Meier survival curves were plotted for time to colectomy since diagnosis of disease, comparing the cohort of hospitalized patients to those who did not require hospitalization. The log-rank test was used to compare the survival curves.

The study was approved by the Institutional Review Board of the Medical College of Wisconsin.

RESULTS

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

There were a total of 246 patients with UC included in our study, among which 27 underwent colectomy for management of severe disease (27/246, 11.0%). The mean age of the entire cohort was 43.6 years (range 19–87 years) with women comprising just over half the cohort (131/246, 53.3%). Table 1 presents the characteristics of patients in the colectomy and no colectomy groups. There was no difference in the gender distribution or mean age of the patients between the 2 groups. Requiring prior medical hospitalization for management of UC was twice as common in patients who subsequently required colectomy than those who did not (77.8% versus 37.4%, P = 0.000). Conversely, among patients who required medical hospitalization for management of disease, one-fifth (21/103, 20.4%) subsequently required colectomy, while only 6 out of 143 patients (4.2%) who had never required hospitalization underwent colectomy. The median time to colectomy after hospitalization was 2.0 months.

Table 1. Characteristics of the Study Population
CharacteristicsColectomy (n = 27)No Colectomy (n = 219)
  • a

    Includes methotrexate and azathioprine;

  • +

    P < 0.05,

  • *

    P < 0.001.

Number of women (n (%))15 (55.6)116 (53.0)
Mean age (in years)40.044.1
Mean age at diagnosis (in years)29.035.6+
Mean duration of disease (in years)5.68.6+
Pancolitis26 (96.3)204 (94.0)
Smoking status  
 Nonsmokers24 (88.9)168 (76.7)
 Ever smokers3 (11.1)51 (23.3)
Standard immunomodulator usea  
 Never7 (25.9)89 (40.6)
 Ever20 (74.1)130 (59.4)
Biologic therapy (infliximab) use  
 Never13 (48.2)170 (77.6)
 Ever14 (51.9)49 (22.4)*
Medical hospitalization  
 Never6 (22.2)137 (62.6)
 Ever21 (77.8)82 (37.4)*

On univariate regression analysis, requiring medical hospitalization for control of disease was significantly associated with future colectomy (OR 5.85, 95% confidence interval [CI] 2.27–15.08) (Fig. 1). Older age at disease diagnosis was associated with a lower risk of colectomy (OR 0.97, 95% CI 0.93–0.99), as was a longer disease duration (OR 0.92, 95% CI 0.84–0.99), possibly suggesting a more mild disease course. Requiring biologic therapy (OR 3.74, 95% CI 1.65–8.48), but not standard immunomodulator therapy (OR 1.96, 9% CI 0.78–4.82), was strongly associated with colectomy. We then repeated the analysis including only those patients who did not undergo a colectomy during their initial hospitalization but subsequently required the procedure during their disease course (n = 15). This subgroup also had a 4 times greater risk of colectomy (OR 4.18, 95% CI 1.56–11.19) than the cohort that had never required hospitalization.

thumbnail image

Figure 1. Relationship between history of medical hospitalization and future need for colectomy in ulcerative colitis.

Download figure to PowerPoint

On multivariate analysis, requiring medical hospitalization remained an independent predictor of future colectomy (OR 5.37, 95% CI 2.00–14.46) as did ever requiring biologic therapy (OR 3.12, 95% CI 1.21–8.07). Neither disease duration nor age at diagnosis were independent predictors of colectomy. Figure 2 demonstrates the Kaplan–Meier survival curves for time to colectomy among patients who did or did not require medical hospitalization. This illustrates the shorter time to colectomy since diagnosis of disease in patients who required medical hospitalization (P = 0.0002).

thumbnail image

Figure 2. Kaplan–Meier survival analysis of time to colectomy in patients with UC who required hospitalization compared with UC controls who did not require hospitalization. Log-rank test for statistical significance between the curves: P = 0.0002. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Download figure to PowerPoint

DISCUSSION

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

UC is a chronic IBD often requiring hospitalization or surgery.1, 3, 7, 35 Approximately 8%–30% of patients will eventually require colectomy for intractable disease, severe colitis, or the development of dysplasia or colon cancer.5–8, 36 Despite being effective for control of disease, colectomy with pouch reconstruction is associated with a high incidence of complications such as pouchitis or pouch failure,15, 16 thus underscoring the importance of identifying potential predictors of surgery. In the present study we identify that patients who have required medical hospitalization for management of UC disease activity have a 5 times greater risk of requiring subsequent colectomy and comprise a cohort with more severe natural history of disease. We believe this information is important because early identification of this at-risk population may help to direct treatment strategies toward more potent therapy.

One of the largest population-based studies of the rates and predictors of colectomy was from Hoie et al,7 who examined the colectomy rate in UC between 1991–93 from 9 centers in Europe. Their cumulative rate of colectomy at 10 years was 8.7%, significantly lower than previously reported rates. The only independent predictors identified in their study were residence in northern Europe (compared to southern Europe) and pancolitis. An earlier study by Leijonmarck et al8 examined 1586 patients with UC in Stockholm County between 1955–84 and identified a 10-year colectomy rate of 42% among those with pancolitis. More recently, data from the IBSEN study suggested a 7.5% rate of colectomy 5 years after diagnosis of UC,12 while Vind et al13 had a 6% colectomy rate 1 year after diagnosis. The lower overall rate of colectomy (11%) in our study is likely a reflection of the availability of more effective therapy in the form of immunomodulators or biologic agents and is consistent with these recent data.7

To our knowledge, there has been no prior examination of the role of medical hospitalizations in predicting colectomy. We found on multivariate analysis, after adjusting for disease extent, various treatment regimens, and demographic factors, a history of requiring inpatient management of UC disease activity was associated with 5 times greater odds of requiring colectomy, either during the index hospitalization or later on in the disease course. This suggests that history of medical hospitalizations may be an important factor in predicting the natural history of UC. This is further strengthened by our finding that 20% of those who required medical hospitalization subsequently underwent colectomy compared to less than 5% of those who had never needed inpatient management.

Disease activity directly influences the risk of colectomy in patients with UC. There is also likely a high degree of colinearity between higher disease activity scores and need for medical hospitalization for management. Despite the availability of a number of clinical and endoscopic scores to assess and monitor disease activity, these scores may not be routinely ascertained in clinical practice, especially outside of tertiary referral centers. Moreover, longitudinal tracking of these scores across different healthcare institutions may also be limited. In this scenario, we suggest that a history of medical hospitalization can serve as a simple red flag to identify UC patients at a higher risk of colectomy. We suggest that this risk factor could be ascertained easily in every UC patient at the initial encounter. This history, along with disease activity scores, where available, can then be used to identify the cohort with more severe disease. This information can then be used accordingly to stratify patients into different treatment regimens beyond 5-ASA compounds that are appropriate for mild-to-moderate disease.

In parallel, not surprisingly, ever requiring biologic therapy for control of UC activity was also associated with a 3 times greater risk of future colectomy. Jarnerot et al,37 in a randomized controlled trial of infliximab as rescue therapy in UC, documented success rates of around 70% in severe UC. However, despite the use of these agents, 30% of patients do not respond and subsequently require colectomy.37 In our institution, infliximab has been routinely used as a rescue agent over the past several years. However, because the current study was not designed prospectively to evaluate the effectiveness of infliximab as a rescue therapy, we cannot comment on whether its use modified the risk of colectomy in the subset of hospitalized patients with severe colitis.

Requiring immunomodulator therapy was not associated with colectomy. This lack of difference may be because the UC cohort seen at tertiary referral centers such as ours is not representative of the spectrum of disease seen at hospitals in the community. The rates of immunomodulator use in our cohort are higher than in community practices, and hence may be a reason for the lack of statistical difference between the groups that subsequently required colectomy and those that did not. It is also possible that the retrospective design of our study as well as relatively small sample size may have resulted in inadequate power to identify this association. Our results are consistent with a previous analysis by Hoie et al,7 who also found no association between azathioprine/6-MP use and risk of colectomy (OR 2.8, 95% CI 0.6–12.9). We suggest that prospective intervention trials in this high-risk subgroup of patients may be better poised to identify if immunomodulator or biologic therapy may exert a disease-modifying effect, postponing or avoiding the need for colectomy.

In our study, none of the demographic parameters were found to be significant predictors of colectomy, although disease duration longer than 10 years had a trend of inverse association with colectomy (OR 0.29, 95% CI 0.08–1.11). This is similar to the analyses by Hoie et al7 and Leijonmarck et al,8 who found no association between age at diagnosis, gender, or smoking status with the risk of colectomy. The trend toward lower colectomy rates in patients with longer disease duration may be explained by our excluding those who underwent colectomy for dysplasia, which is likely the major indication of colectomy in those with long-standing UC. Those with more aggressive disease phenotype, on the other hand, have a shorter history of disease, thus explaining the association with colectomy rates.

There are several limitations to our study. Studies from tertiary centers are vulnerable to referral bias resulting from a sicker patient population. The retrospective design of the study is also subject to bias, although we relied on objectively collected data that was confirmed by chart review. The relatively small number of colectomy patients in our study may have decreased our power to identify smaller associations. Since the majority of our patients had pancolitis, our findings may not be generalizable to those with more limited extent of disease. The large number of patients with pancolitis may be representative of a sicker tertiary center cohort. It is possible that a small number of patients with more limited colitis may have been misclassified as pancolitis. However, this misclassification is likely to be nondifferential. We also did not have information on patient compliance with therapy, a factor that may modify outcomes.

The implications of our study are that patients who have required inpatient management of disease for UC may have demonstrated a progression beyond mild-to-moderate disease. Medical hospitalization for UC should serve as a red flag to identify a cohort at significantly increased risk for colectomy. Close follow-up of this cohort, identification and treatment of possible precipitating or complicating factors such as Clostridium difficile infection,38–40 and early escalation to more potent agents including immunomodulator and biologic therapy should be considered. Our findings also suggest that future intervention trials aimed at preventing colectomy may need to focus on the UC population that has required hospitalizations.

In conclusion, we demonstrate that requiring medical hospitalization for the management of disease activity in UC is an independent predictor of the need for colectomy. Future studies will determine whether aggressive medical management with biologic and immunomodulator therapy may help decrease the likelihood of needing colectomy in this cohort.

REFERENCES

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