Weber, Nicholas K. MD*; Bruining, David H. MD*; Loftus, Edward V. Jr MD*; Tremaine, William J. MD*; Augustin, Jessica J. BS*; Becker, Brenda D. CCRP*; Kammer, Patricia P.*; Harmsen, William Scott MS†; Zinsmeister, Alan R. PhD†; Pardi, Darrell S. MD, MS†
Inflammatory bowel disease (IBD) is generally considered a disease of young adults, but 15% of patients are 65 years and older at diagnosis.1,2 In addition, because IBD is not associated with substantial mortality, many patients diagnosed at an earlier age will carry their diagnosis into later life. Therefore, given the aging of the general population, the percentage of older patients with IBD will likely increase.3 Differences in physiology, comorbidities, and concomitant medications between younger and older patients make it essential to understand the unique treatment circumstances for older patients with IBD. For example, Cottone et al,4 recently demonstrated an increased risk of death, severe infection, and neoplasm with anti–tumor necrosis factor (TNF) use in the elderly compared with younger patients with IBD. In keeping with these data, physicians may take a conservative approach to the treatment of older patients with IBD. For example, in one recent study, only a small percentage of older patients with IBD were treated with an immunomodulator or an anti-TNF agent.5
Data on the inpatient management of older patients with IBD are limited. A recent population-based study demonstrated increased mortality in hospitalized elderly patients with IBD compared with younger patients.2 Older patients who did not undergo surgery had higher mortality than those who had surgical intervention (4.8% versus 2.6%),2 perhaps because the patients chosen for surgery were generally healthier and thus better surgical candidates. Alternatively, perhaps medical rather than surgical therapy was attempted more often in older rather than younger patients, and this was associated with the increased mortality.2,6,7 To our knowledge, the treatments and outcomes in patients hospitalized with an IBD flare stratified by age have not been evaluated previously. Therefore, we aimed to assess differences in response, safety, and subsequent treatments in older compared with younger inpatient adults with IBD treated with corticosteroids.
MATERIALS AND METHODS
Medical records were abstracted to identify all patients 60 years and older who were hospitalized at our institution between 2003 and 2011 for worsening symptoms of an IBD flare. They were matched 1:1 to younger patients aged 18 to 50 years based on geography and grouped into 4 categories—Olmsted County, the rest of Minnesota, the 5-state area (Iowa, Wisconsin, Illinois, South Dakota, or North Dakota) and all other locations, gender, IBD subtype, admission date (2003–2007 or 2008–2011), and disease duration (±5 yr). Geography was matched to minimize potential referral bias. For older patients who could not be matched to a younger patient on all 5 criteria, matches were made after eliminating no more than 2 variables. If geography was relaxed, this was collapsed to 1 level (Olmsted County relaxed to the rest of Minnesota, the 5-state area relaxed to the rest of Minnesota, etc.), and for disease duration, this was relaxed to ±10 years. Baseline comorbidities were assessed using the Charlson Comorbidity Index.8 Other baseline characteristics, including IBD treatment, nonsteroidal anti-inflammatory drug use, history of malignancy, and presenting stool infections, were recorded.
Disease severity was assessed by evaluating endoscopic, histologic, and clinical parameters. Endoscopic and histologic severity was based on documentation in the medical record as mild, moderate, or severe, and when a range was reported, the highest severity was used. When the severity was not clearly stated on an endoscopic report, 3 IBD experts independently reviewed the report, and the most common designation was recorded. In addition, the highest severity of histologic inflammation was recorded (based on available biopsy data because all segments were not routinely biopsied). Clinical disease severity for ulcerative colitis (UC) was assessed according to the Truelove and Witts criteria9,10: patients were classified as having clinically severe disease if they had >6 bowel movements per day and at least one of the following: temperature >37.5°C, heart rate >90 per minute, hemoglobin <75% of normal, or erythrocyte sedimentation rate >30 mm/h. Clinical disease severity for Crohn's disease (CD) was assessed using the American College of Gastroenterology practice guidelines.11 To be classified as having severe clinical disease, patients had to present with at least 2 of the following: temperature >37.5° C, ≥10 pound weight loss, abdominal pain, nausea/vomiting, hemoglobin <75% of normal, or requiring a packed red blood cell transfusion.11
We included patients who were naive to anti-TNF therapy, and who either started corticosteroids during hospitalization or had the dose of corticosteroids increased from the dose taken before hospitalization. A flare was determined by clinical presentation (Table 1), which generally refers to patients with increasing abdominal pain, diarrhea, and/or hematochezia and supported by findings from a flexible sigmoidoscopy or colonoscopy performed either during the index admission or no greater than 4 weeks before the date of admission. Patients with UC with previous colon resections were excluded. Patients who could not be matched for at least 3 of the matching criteria were excluded as well.
Clinical, demographic, and outcomes data were compared between cohorts. The primary endpoints included inpatient corticosteroid response and subsequent treatments. Patients were classified as being corticosteroid-responsive, if there was documented clinical improvement and corticosteroid-refractory, if there was failure to improve or clinical worsening as reported in the medical record. In general, clinical response was based on change in abdominal pain, diarrhea, and/or hematochezia. For corticosteroid-refractory patients, the rates of inpatient colectomy and anti-TNF initiation were assessed. For corticosteroid-responsive patients, the percentage of patients requiring initial anti-TNF treatment and/or colectomy at 1 year were assessed and compared between each cohort. Details of inpatient corticosteroid use, including corticosteroid type, starting dose, and total days of corticosteroid use before hospital discharge, were collected. In addition, data regarding parenteral corticosteroid use before admission to our institution were recorded. Secondary endpoints included hospital length of stay, corticosteroid-related side effects, IBD medical management at discharge for corticosteroid-responsive patients, 30-day postsurgical complications, 30-day readmission to our hospital, and 30-day postdischarge mortality. Hyperglycemia was defined as a blood glucose level ≥126 mg/dL recorded on at least 2 occasions after starting corticosteroids and before colectomy. Elevated blood pressure was defined as at least 1 systolic blood pressure ≥140 mm Hg after starting or increasing corticosteroids during hospitalization and before colectomy. Psychosis, insomnia, and other suspected adverse reactions to inpatient corticosteroid use were recorded as well.
For statistical analysis, data were summarized as mean (SD), median (range), or frequency (%) as warranted. The association of primary and secondary endpoints with age cohort (older versus younger) was assessed using conditional logistic regression, which accounted for the matching for the overall cohorts. When restricting to subsets of the cohorts (e.g., corticosteroid-refractory patients), this matching could not be retained, and thus unmatched analyses were used. Odds ratios (OR) and 95% confidence intervals (CI) were computed from the regression coefficients (and their standard errors). Survival estimates were obtained using the Kaplan–Meier method and hazard ratios from proportional hazards regression models. These latter 2 analyses ignored matching.
After initial review, 73 older patients met the criteria for study inclusion. Matching was unsuccessful for 8 older patients, and these patients were excluded from additional analysis. A total of 65 older patients (34 men) were thus included. They were matched to 65 younger patients, and all 5 matching criteria were satisfied for 54 of these patients (83%). One matching criterion was eliminated for 5 patients, and 2 criteria were eliminated for 6 patients. Demographic data are reported in Table 1. The most common symptoms at admission in each cohort included diarrhea, hematochezia, abdominal pain, and weight loss (Table 2).
For the 115 patients who were tested for an enteric infection during the flare, 25% of older patients and 31% of younger patients were positive. Clostridium difficile was the most common infection in both groups (Table 1). Other bowel infections in the older cohort included cytomegalovirus (n = 5), Aeromonas (n = 1), Proteus (n = 1), and Clostridium perfringens (n = 1). In the younger cohort, Aeromonas (n = 1) and cytomegalovirus (n = 1) were present.
There were no significant differences in the prevalence of severe clinical, endoscopic, or histologic disease at presentation in older compared with younger adults (Table 1). In addition, a similar number of older and younger patients with CD presented with fistulizing disease and/or partial small bowel obstruction (Table 1). All these patients responded to corticosteroid treatment during hospitalization.
Parenteral corticosteroid use before admission to our hospital occurred in 8% and 20% of the older and younger patients, respectively (P = 0.06). For the 18 patients who were treated with parenteral corticosteroids before transfer to our facility, the median duration of previous treatment was 2 days (range, 1–10) for older patients and 4 days (range, 1–12) for the younger cohort (P = 0.37). Additional data regarding outpatient medical management before admit are shown in Table 1.
Methylprednisolone was the most common corticosteroid used during hospitalization for both older (86%) and younger patients (85%) (Table 1). Regarding the primary endpoints, 23% (n = 15) of older patients were refractory to inpatient corticosteroids compared with 38% (n = 25) of the younger cohort (OR, 0.5; 95% CI, 0.2–1.1). For these corticosteroid-refractory patients, more aggressive treatment for the IBD flare was deemed necessary by the clinician as noted in the medical record. Inpatient management with surgery (80% versus 72%) or infliximab (20% versus 12%), which was the only anti-TNF used during hospitalization, was not significantly different in older compared with younger patients. In the older and younger cohorts, there were 1 and 2 patients, respectively, who underwent surgery after failing to respond to inpatient infliximab. For statistical purposes, these patients were included in the colectomy group. In the 4 younger patients refractory to corticosteroids who did not undergo colectomy or initiate anti-TNF therapy during hospitalization, 1 entered a placebo-controlled clinical trial of visilizumab with subsequent clinical improvement. Two patients were discharged on oral prednisone and scheduled for outpatient colectomy 4 and 18 days after hospital discharge, respectively. The fourth patient was offered infliximab and declined. However, this patient subsequently was started on infliximab 178 days after hospital discharge and then required colectomy 22 days later. Reasons for not receiving anti-TNF therapy before colectomy are demonstrated in Figures 1 and 2 and Table 3.
When reviewing these primary outcomes for the subgroup of patients with CD, a similar percentage of older (8.3%, n = 1) and younger patients (18.2%, n = 2) were corticosteroid refractory. All 3 patients were treated with infliximab during hospitalization. For older and younger patients with UC, 26.4% (n = 14) and 42.6% (n = 23) were corticosteroid refractory. For older corticosteroid-refractory patients with UC, 14.3% (n = 2) were treated with infliximab, whereas 85.7% (n = 12) were treated with surgical intervention. For younger corticosteroid-refractory patients with UC, infliximab therapy was used in just 4.3% (n = 1), surgery in 78.3% (n = 18), and an alternative approach (i.e., study protocol or no further aggressive inpatient therapy) in 17.4% (n = 4) of patients. Primary outcomes based on duration of IBD at time of index admission are also reported in Table 4.
All postoperative patients had at least 30 days of follow-up in each cohort. The cumulative probability of a first postsurgical complication within 30 days was similar in the 2 cohorts (Table 5). For older patients who had a complication, 40% (n = 4/10) had more than 1 complication as compared with 56% of younger patients (n = 9/16). Postsurgical complications are listed in Table 6.
Corticosteroid-responsive patients were managed medically as outlined in Table 5. Only 1 patient in each cohort did not receive corticosteroids at hospital discharge; 1 older patient was discharged on mesalamine monotherapy, and 1 younger patient was treated with azathioprine alone.
Corticosteroid monotherapy was common at hospital discharge in both older (n = 21) and younger (n = 21) corticosteroid-responsive cohorts (Table 5); maintenance therapy decisions for most of these patients (n = 20 and n = 21) were deferred to the outpatient gastroenterology provider. Subsequent recommended therapies for older corticosteroid-responsive patients at outpatient follow-up included 5-aminosalicylate (n = 9), azathioprine (n = 7), azathioprine + 5-aminosalicylate (n = 1), prednisone monotherapy (n = 1), infliximab (n = 1), and unclear recommendation (n = 1). Recommended therapies for younger corticosteroid-responsive patients at outpatient follow-up included 5-aminosalicylate (n = 11), azathioprine (n = 5), prednisone monotherapy (n = 1), anti-TNF (n = 2), and surgery (n = 2). Medical therapy at hospital discharge for corticosteroid-responsive patients was also assessed based on IBD subtype, and this is reported in Table 7.
In corticosteroid-responsive patients, the 1-year cumulative probability of starting an anti-TNF agent was 7% (95% CI, 0%–16%) for older patients and 31% (95% CI, 13%–45%) for younger patients. Thus, an older patient was significantly less likely to be started on an anti-TNF agent (hazard ratio, 0.2; 95% CI, 0.1–0.6). A variety of initial anti-TNF agents were used in this subgroup. In the older cohort, 4 patients received infliximab, and in the younger cohort, 11 patients received infliximab, 3 patients received adalimumab, and 1 patient received certolizumab. The 1-year cumulative probability of colectomy was 27% (95% CI, 12%–42%) for older patients and 28% (95% CI, 10%–42%) for younger patients (hazard ratio, 0.8; 95% CI, 0.4–1.8).
Except for differences in adverse effects from corticosteroids and median follow-up (87% of patients had at least 30 d of follow-up data available), secondary endpoints, including average length of stay, 30-day readmission rates, and 30-day postdischarge mortality, were similar between cohorts (Table 5).
In this parallel cohort study, corticosteroid response rates and inpatient treatment approaches for corticosteroid-refractory patients (both anti-TNF use and surgery) were similar among older and younger patients hospitalized for an IBD flare. For corticosteroid-responsive patients, older patients were less likely to be started on an anti-TNF agent than younger patients.
There was low utilization of inpatient anti-TNF therapy (12%–20%) observed in our study. However, if we include inpatients who received anti-TNF therapy followed by inpatient colectomy and those who declined anti-TNF agents (i.e., selected colectomy or a study protocol), our adjusted anti-TNF use for older (40%) and younger (68%) cohorts would likely be comparable with other institutions' experience. Low outpatient use of anti-TNF agents has been reported previously by Juneja et al,5 in which only 2.6% of 393 geriatric patients with IBD (62% UC and 38% CD) were being treated with anti-TNF medications. In addition, we report that older corticosteroid-responsive patients were significantly less likely to be started on anti-TNF therapy during the first year of follow-up compared with younger patients. The explanation for such infrequent use in this population may be related to an increase in comorbidities, including malignancies, in the older IBD population, as evident in our study. In addition, there is a growing appreciation that anti-TNF use in older patients with IBD may be associated with more frequent side effects.3,4 Our finding of similar rates of inpatient surgery between the 2 cohorts is in agreement with a recent study that showed no difference in colectomy rates based on age (OR, 1.03; 95% CI, 0.70–1.18).2
With regard to secondary endpoints, initial data review suggested possible under-treatment of corticosteroid-responsive patients at hospital discharge; older and younger patients were 5 times and 3 times, respectively, more likely to be treated with a 5-aminosalicylate or corticosteroid compared with treatment with an immunomodulator. In addition, only 16% and 25% of older and younger corticosteroid-responsive patients, respectively, were discharged from the hospital on an immunomodulator. However, approximately 40% and 50% of each cohort started maintenance therapy after hospital discharge; immunomodulator use increased to 32% and 38%, respectively, when including these patients.
We found a statistically significant increase in corticosteroid-related hypertension for the older cohort compared with younger patients, as previously reported.3,12 Other secondary endpoints, including postsurgical complications, inpatient mortality, 30-day readmission rates, and 30-day mortality rates, were similar between the 2 groups. However, significantly higher inpatient mortality among older compared with younger patients with IBD has been previously reported, and this discrepancy is likely explained by our insufficient power to detect such differences.2
In addition, compared with a previous series on the topic,13 we demonstrated a significant rate of enteric infections in both the older (25%) and younger (31%) cohorts; this decreased to 18.2% and 24.6% when assessing for only those patients with a C. difficile and/or cytomegalovirus infection. Our C. difficile infection rates also were greater in the older (12.7%) and younger (23%) cohorts as compared with the 1% to 5% reported in other series.14–18 This may be partly due to the high rate of corticosteroid use before admission in our older (52.3%) and younger (64.6%) cohorts. Moreover, this rate of corticosteroid use and enteric infections at admission may explain in part our high 30-day surgical complication rate (83%–89%) as compared with other recent series (38%–48%).19 Varying definitions may also play a role in the differences. For our older and younger surgical patients who had a postoperative complication, 90% and 100% were either on corticosteroids at admission or had an enteric infection documented.
Additional limitations of our study are worth mentioning. This was a retrospective study with highly selective inclusion criteria to minimize bias. As a result, our study population was small, and the clearly decreased power prevented multivariate analysis of our primary endpoints. The small sample size may have prevented us from detecting a significant difference in the corticosteroid response rate between older and younger patients. We reported that older patients were corticosteroid refractory less frequently (OR, 0.5; 95% CI, 0.2–1.1) than the younger cohort, and a larger sample size may have resulted in statistical significance. To achieve a significant OR, we would need 37 corticosteroid-refractory older patients (rather than 15), and 61 corticosteroid-refractory younger patients (rather than 25). The reason for such differences in corticosteroid responsiveness may be related to referral bias in which younger patients may have had more severe illness. Although there was no significant difference in severe clinical, endoscopic, or histologic disease at presentation between the 2 cohorts, there was a greater percentage (nearly significant) of younger compared with older patients treated with parenteral corticosteroids before transfer to our facility. Despite the power limitations, the goal of this study was to serve as a descriptive pilot study and to evaluate for novel trends in treatment approach to patients with IBD stratified by age that might be potentially worth additional prospective analysis.
There were a number of other potential confounders. The inclusion of patients with CD in a study with mostly UC is one such example. Such patients were included to prevent further decreases in sample size. However, there were few patients with CD with transmural complications (Table 1), and this may explain why the corticosteroid-refractory rate in this subgroup of older and younger patients (8.3% and 18.2%, respectively) was much lower than for patients with UC (26.4% and 42.6%, respectively). As a result, the inclusion of patients with CD and minimal transmural complications may have decreased the overall corticosteroid-refractory rates for the study population. In addition, endoscopic and histologic scoring, definitions of a flare, and subsequent corticosteroid response were subjective given the retrospective design. A repeat prospective study would minimize this potential confounding.
Lastly, the relaxation of disease duration from 5 to 10 years could result in further confounding because this could select outpatients more likely to have mild disease. In our study population, however, only 4 patients required relaxation of disease duration to be included in the analysis.
In summary, there are limited data on how older age affects treatment response and management decisions for inpatients with an IBD flare. We found similar corticosteroid response rates in older and younger patients, and similar treatment approaches, with either anti-TNF therapy or surgical management, were used for corticosteroid-refractory inpatients. To our knowledge, this is the first study to compare inpatient corticosteroid response rates and subsequent treatment for older and younger corticosteroid-refractory patients hospitalized for an IBD flare. Because of our aging population and the potential for a greater proportion of patients diagnosed with IBD later in life, these issues warrant broader study in larger, prospective population-based studies.
Author contributions: Study concept and design: N. K. Weber, D. S. Pardi; Analysis and interpretation of data: N. K. Weber, D. S. Pardi, A. R. Zinsmeister, W. S. Harmsen; Drafting of manuscript: N. K. Weber, D. H. Bruining, E. V. Loftus, W. J. Tremaine, D. S. Pardi, A. R. Zinsmeister, W. S. Harmsen; Statistical analysis: A. R. Zinsmeister, W. S. Harmsen; Administrative, technical, or material support: J. J. Augustin, B. D. Becker, P. K. Kammer; Study supervision: D. S. Pardi.
© Crohn's & Colitis Foundation of America, Inc.