SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendix
  10. Supporting Information

Background  There remain concerns about the safety of infliximab therapy in patients with inflammatory bowel disease (IBD).

Aim  To assess the association between the initiation of infliximab and other immunomodulating drugs and the risk of serious bacterial infection in the treatment of IBD.

Methods  We assembled a cohort study of patients with IBD, including Crohn’s disease (CD) and ulcerative colitis (UC). All patients initiating an immunomodulating drug between January 2001 and April 2006 were identified in British Columbia from linked health care utilization databases. Exposure of interest was initiation of infliximab or corticosteroids compared with initiation of other immunosuppressive agents, including azathioprine, mercaptopurine (MP) and methotrexate (MTX). Outcome of interest was serious bacterial infections requiring hospitalization, including Clostridium difficile.

Results  Among 10 662 IBD patients, the incidence rate of bacteriaemia ranged from 3.8 per 1000 person-years (95% confidence interval 2.1–6.2) for other immunosuppressive agents to 7.4 (3.3–19.3) for infliximab with slightly higher rate for serious bacterial infections resulting in an adjusted relative risk 1.4 (0.47–4.24). Clostridium difficile infections occurred in 0/1000 (0–5.4) among 521 infliximab initiations and 14/1000 (10.6–18.2) for corticosteroids. Corticosteroid initiation tripled the risk of C. difficile infections (RR = 3.4; 1.9–6.1) compared with other immunosuppressant agents. This corticosteroid effect was neither dose-dependent nor duration-dependent. Bacteriaemia and other serious bacterial infections were not increased by corticosteroids or infliximab (5 events).

Conclusions  In a population-based cohort of patients with IBD, we found no meaningful association between infliximab and serious bacterial infections, although some subgroups had few events. Corticosteroid initiation increased the risk for C. difficile infections in these patients.


Background

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendix
  10. Supporting Information

Infliximab is a chimeric IgG4 monoclonal antibody targeting tumour necrosis factor α (TNF-α). More than 800 000 individuals and more than two million patient-years were estimated to have received treatment with Infliximab.1 Infliximab’s potential to alter the natural history of CD has created significant interest to widen the use of this medication for treating patients earlier in their disease course, thus potentially exposing many more patients to this medication. In a pivotal study examining the use of Infliximab as a maintenance therapy for treatment of CD, there was a significant reduction in disease-related hospitalizations and disease-related intra-abdominal surgeries in CD patients who received scheduled maintenance infusions of infliximab compared with those who only received episodic therapy.2 However, several reports linked infliximab use with the incidence of bacterial and opportunistic infections.3–6 More clarity on the potential risk of serious bacterial infections is necessary for a more informed benefit-risk assessment for use of infliximab in patients with inflammatory bowel disease (IBD). Clostridium difficile infections (C. difficile) are of particular concern in light of the impact this particular infection has on patients with IBD, including hospitalization rates and even potentially mortality, although there is currently no evidence of an increased risk for C. difficile infections by infliximab.

Outside the clinical trial setting, the first study to address the adverse event profile of Infliximab in IBD was a retrospective review of 500 consecutive patients treated with Infliximab at the Mayo Clinic in Rochester, Minnesota between October 1998 and October 2002.3 The median follow-up for this cohort was 17 months (range, 0–48 months). Forty-three patients (8.6%) experienced a serious adverse event, 30 (6%) of which were thought to be related to Infliximab. Forty-eight patients (8.2%) acquired an infection although not all were classified as a serious adverse event. In a prospective, observational, multicentre registry of patients in North America treated with Infliximab (the TREAT registry), the rate of serious infections associated with Infliximab was 1.37 infections per 100 patient-years.4

The only population-based study addressing this important issue of safety with Infliximab consisted of a cohort of 217 patients with inflammatory bowel disease (IBD) in Stockholm County, Sweden between January 1999 and April 2001.5 A total of 11 medical centres contributed to the study; these centres were the only ones administering Infliximab during this time period. Forty-one patients (18.9%) experienced a serious event from January 1999 to April 2001. Eighteen of these patients’ serious events were infections, 7 of them were in the post-operative period. Two patients died of sepsis while on infliximab. As the use of Infliximab is restricted to specific centres in Sweden, their complication rates may differ from that in other countries where such a restriction is not in place.

The risk of Infliximab treatment to cause serious infections, including C. difficile infections, in routine care of patients with IBD compared to other immunomodulating medications remains largely unknown. We sought to evaluate the comparative safety of infliximab in patients with IBD using linked administrative databases from British Columbia, Canada.

Methods

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendix
  10. Supporting Information

Data source and patients

We conducted a cohort study of British Columbia (BC) residents 18 years and older with Crohn’s disease (CD) or ulcerative colitis (UC) who initiated use of a range of immunomodulating drugs, including Infliximab, glucocorticoids and other drugs, from January 2001 to April 2006. Patients were identified in linked health care utilization databases from the BC Ministry of Health. This database contains information on all physician services (Medical Services Plan), hospitalizations with up to 25 diagnostic codes, and all prescription drug dispensings were recorded by the province-wide PharmaNet database. We further linked vital status information from the BC vital statistics agency. Underreporting and misclassification appear minimal because of the electronic data entry of all drug dispensings and hospital diagnoses showed good specificity and completeness.7

Patients with IBD were identified by having at least five health care encounters with a recorded diagnosis of IBD (ICD-9 codes 555.x and 556.x). This definition was relaxed to three recorded diagnoses of IBD if patients were residing in BC for 730 days or shorter. This definition has been developed and validated for identifying IBD cases in administrative databases using the Manitoba Health Database that is very similar to the British Columbia linked administrative databases in its population coverage and data completeness.8 Patients were considered for this study only after the fifth (or third) occurrence of a diagnostic code for IBD. Patients with any cancer (except non-melanoma skin cancer) or HIV/AIDS were identified by at least two physician services with a corresponding ICD-9 diagnosis in the prior 365 days and were excluded from the analysis.

Within this cohort of patients with IBD, we identified each patient’s date of initiating immunomodulating therapy (cohort entry date). To be categorized as initiating a specific therapy, we required at least 365 days without use of that agent before the cohort entry date. Patients could initiate use of more than one immunomodulating therapy as well as multiple courses of the same agent during the study period if each initiation fulfilled the 365 days washout period. Within each course of therapy, follow-up time started at the date of cohort entry. Follow-up ended at the earliest of (1) discontinuation of the drug, (2) first occurrence of a study infection, (3) end of the study period, or (4) death.

All traceable personal identifiers were removed from the dataset prior to analysis to protect patient confidentiality. The Institutional Review Board of the Brigham and Women’s Hospital approved this study and signed data use agreements were in place.

Drug exposure

Study drug categories included initiation of infliximab therapy, oral glucocorticoids and ‘other immunosuppressive agents’ (methotrexate, azathioprine, mercaptopurine), which served as the reference drug exposure. We further identified a minimally treated patient group consisting of initiators of aminosalicylate or patients with a qualifying diagnosis of IBD, but without any of the treatments listed above. This group was not included in the analysis because patients may have milder expression of their disease and including these patients would increase the potential for confounding by discrepancies in disease severity. The index date for each drug was the date of first dispensing. If patients started two of the study drugs on the same day, each would be considered as episodes of treatment initiation and the other drug would be considered a concurrently used immunomodulating agent.

To allow for continued immunosuppressive effects of these drugs even after dosing may have stopped, we calculated the end of exposure based on approximate half-lives. MTX, azathioprine, mercaptopurine and glucocorticoids were assigned 21 days of exposure beyond the days supply.9 We assumed that each infliximab infusion was affecting the immune system for up to 126 days, which we defined as our exposure-risk window.9

To assess whether duration of use modifies the effect of immunomodulating treatment on serious infections, we stratified the analysis into short-term effects (0–90 days after initiation) and longer-term use (more than 90 days). For this analysis, all infection outcomes were combined to increase the number of outcomes in each time period. The daily amount of oral glucocorticoid use was grouped into categories of low-dose (<20 mg) and high-dose (≥20 mg per day).

Concomitant use of other immunomodulating agents in addition to the index drug was assessed at the date of cohort entry. If, for example, a patient added infliximab to an ongoing MTX regime, the initiation of infliximab would be the study exposure while concomitant MTX therapy would be adjusted in the analysis. Conversely, if a patient would add MTX to an ongoing infliximab therapy, the initiation of MTX would be considered the study drug.

Outcomes

The study outcome was the first occurrence of a hospitalization for a serious bacterial infection during a course of therapy, with the following diagnoses as the primary or first listed diagnosis: bacteriaemia/septicaemia, pneumonia, osteomyelitis, pyelonephritis, meningitis, encephalitis or endocarditis. For reasons of the small numbers of events, only bacteriaemia/septicaemia, was considered as a separate event; all events were also grouped into a composite outcome ‘serious bacterial infection’. For each endpoint, the event date was the date of hospital admission.

In a validation study of 158 patients hospitalized for any of the above diagnoses in Veterans Affairs hospitals in the Northeast US between 2001 and 2004, we found a positive predictive value of >80% for a set of predefined ICD-9 codes (Appendix 1) using detailed medical records as the gold standard method. The predictive value increased to >90% if the outcome definition was any bacterial infection leading to hospitalization rather than a specific infection diagnosis.10

An additional outcome was C. difficile infection defined as either a hospital discharge diagnosis of C. difficile or an outpatient diagnostic code for C. difficile followed by a metronidazole or vancomycin antibiotic dispensing within 14 days. C. difficile infections and their treatment are specific enough to be identified using outpatient diagnoses with subsequent treatment in addition to inpatient diagnoses. Recognizing the different strategies to identify our study outcomes, we avoided direct comparisons of incidence rates of C. difficile and other serious infections.

Potential confounders

A set of potential confounders were measured based on health care utilization data within 365 days before cohort entry. These included socio-demographic characteristics (age, gender); number of other immunomodulating study drugs that were used concomittantly at cohort entry; generic markers of comorbidity that have shown good validity in British Columbia11,12 (hospitalization for any reason, number of physician visits, number of distinct prescription drugs excluding study drugs, Charlson comorbidity score13); markers for IBD severity14 (number of previous surgical resections of large or small intestine, number of colonoscopies) as well as independent predictors of serious infections including diabetes, numbers of prior outpatient or inpatient visits for infections and any prior antibiotic use. To address the potentially non-linear age dependency of serious infections,15 we also included a quadratic age term in the model. Calendar time was modelled by including multiple indicator terms to adjust for possible changes in treatment choice during the study period and possible changes in the prevention strategies of serious infections in this population16

Statistical analysis

We calculated incidence rates and 95% confidence intervals for CD and UC over a 3-year period starting 2003. For each drug exposure group, we calculated incidence rates of serious bacterial infections leading to hospital admission. We used ‘other immunosuppressant drug use’ (methotrexate, azathioprine, mercaptopurine) as a common reference group for three types of comparisons that were made separately for the three study endpoints (bacteriaemia, serious bacterial infection, and C. difficile infection). Within each model, the exposure categories are mutually exclusive and collectively exhaustive:

  • (i)
     Model 1 compared any infliximab therapy and any glucocorticoids vs. any other immunosuppressant drug use.
  • (ii)
     Model 2 compared infliximab mono therapy, glucocorticoid mono therapy, MTX mono therapy and several combinations vs. other immunosuppressant drug mono therapy.
  • (iii)
     Model 3 compared infliximab therapy, glucocorticoid therapy, MTX therapy each in combination with at least one other ongoing immunomodulating drug vs. other immunosuppressant drug use with at least one other ongoing immunomodulating drug.

We adjusted for all patient characteristics listed above applying no further variable selection by fitting a multivariate Cox proportional hazard regression.17 The proportional hazards assumption was evaluated by visual inspection of log-cumulative hazard curves.

We also developed propensity score models for more efficient estimation.18,19 Propensity score estimates for each patient from logistic regression were entered into the Cox regression as a continuous variable and as quintiles, but both strategies did not result in meaningful changes in relative risk estimates or 95% confidence intervals and results of the propensity score analysis are not shown.

Less than 10% of patients were members of more than one drug initiator cohort as they were allowed to initiate more than one immunomodulating drug during the study period and less than 1% of events were attributed to the same patient twice for that reason; we therefore repeated analyses with an adjustment of standard errors for repeated observations.20

Important predictors of IBD severity may not be fully captured in health care utilization databases. If IBD severity were associated with an increased risk for infection and infliximab users were more likely to have severe disease, this could result in an overestimation of the relative risk of serious infections among infliximab users. We used sensitivity analyses to quantify the extent of such residual confounding as a function of these associations.21

Results

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendix
  10. Supporting Information

The incidence rate of Crohn’s disease in the years 2003 through 2005 was 7.1 per 100 000 (95% CI 6.2–8.0) and was 6.3 (6.4–7.2) for ulcerative colitis. Among all patients diagnosed with either CD or UC in BC during the study period, we identified 521 initiations of infliximab, 2883 initiations of immunosuppressant agents and 7258 initiations of oral corticosteroids (Table 1). Patients initiating infliximab were younger and were less likely to have diabetes than initiators of other IBD drugs, but had more physician visits in the past and were more likely to have received antibiotics or proton pump inhibitors. On average, infliximab initiators had a similar number of past colonoscopies compared with users of other immunosuppressants, but corticosteroid users had on average only half as many colonoscopies during the same time period.

Table 1.   Patient characteristics by drug exposure categories (n, % if not specified otherwise)*
 Newly initiated study drug
InfliximabMTXImmuno-suppressant drugs incl. MTX†Corticosteroids
  1. * Based on health care utilization data within 12 months prior to the initiation of index drug use (index date).

  2. † Immunosuppressant drugs incl. MTX = Azathioprine, mercaptopurine, methotrexate.

  3. ‡ Ongoing drug use means that a prior dispensing overlaps with the initiation of the study drug.

Number of treatment episodes (Total = 10 662)52136128837258
Follow-up years1.3 (1.1)1.2 (1.2)1.4 (1.3)0.6 (0.7)
Age, mean (standard deviation)36.0 (13.2)41.4 (14.5)38.6 (14.3)41.5 (15.4)
Gender, female285 (55)221 (61)1548 (54)3947 (54)
Crohn’s disease (Total = 6353)463 (89)262 (73)1860 (65)4030 (56)
Ulcerative colitis (Total = 4309)58 (11)99 (27)1023 (35)3228 (45)
Calendar year
 200148 (9)48 (13)428 (15)1222 (17)
 2002117 (22)66 (18)552 (19)1311 (18)
 2003110 (21)77 (21)548 (19)1403 (19)
 2004116 (22)94 (26)644 (22)1637 (23)
 2005 + first quarter of 2006130 (25)75 (21)588 (20)1660 (23)
Number of ongoing immunomodulating drug use:
 0192 (37)231 (64)2038 (71)6403 (88)
 1205 (39)105 (29)829 (29)750 (10)
 2115 (22)25 (7)16 (1)96 (1)
 3+9 (2)--9 (0)
Ongoing immunomodulating. drugs:
 Infliximab-32 (9)75 (3)83 (1)
 MTX41 (8)--73 (1)
 Immunosuppressant drugs incl. MTX†269 (52)58 (16)-813 (11)
 Corticosteroids152 (29)97 (27)786 (27)-
Mean number of MD visits (s.d.)15.1 (10.8)14.8 (10.2)12.8 (10.3)10.9 (10.2)
Mean number of different meds (s.d.)‡8.4 (4.3)6.5 (6.0)6.4 (4.4)6.5 (4.5)
Mean number of previous surgical intestinal resections (s.d.)0.47 (0.82)0.44 (0.85)0.49 (0.83)0.66 (0.34)
Mean number of colonoscopies within 1 year (s.d.)0.41 (0.66)0.35 (0.62)0.43 (0.66)0.24 (0.52)
Diabetes14 (2.7)12 (3.3)81 (2.8)241 (3.3)
At least 1 prior outpatient or inpatient visits for infections (%)12 (2.3)7 (1.9)69 (2.4)178 (2.5)
At least 1 hospitalization prior to index date306 (59)172 (48)1575 (55)2921 (40)
Any previous antibiotic dispensing320 (61)207 (57)1443 (50)3321 (46)
Mean comorbidity index (s.d.)0.34 (1.1)0.45 (1.1)0.26 (0.9)0.34 (1.1)
Any proton pump inhibitor use122 (23)50 (14)463 (16)1096 (15)

Half of patients who initiated infliximab therapy were already treated with immunosuppressive agents and 29% were already treated with corticosteroids (Table 1). Among initiators of immunosuppressants, 27% were already treated with corticosteroids, but only 3% were treated with infliximab at that point.

The rate of bacteriaemia and all serious bacterial infections leading to hospital admission ranged from 3.8 per 1000 (95% CI 2.1–6.2) to 8.9 per 1000 (95% CI 3.3–19.3) depending on treatment group and outcome (see Table 2). The rate of C. difficile infections ranged from 0 per 1000 (95% CI 0 to 5.4) for infliximab to 14 per 1000 (10.6 to 18.2) for corticosteroids. 57.2% of patients with a new C. difficile infection received antibiotics within the preceding 180 days.

Table 2.   Rates and rate ratios of serious infection outcomes for any use of the study drugs (Model 1)
Endpoints and drug exposuresPatient yearsEventsEvent rate per 1000 95% CI‡Unadjusted rate ratio 95% CIAdjusted rate ratio§ 95% CI
  1. * Immunosuppressant drugs incl. MTX = Azathioprine, mercaptopurine, methotrexate.

  2. † With or without ongoing use of other immunomodulating drugs at initiation of the study drug.

  3. ‡ 95% confidence interval.

  4. § Adjusted for all covariates listed in Table 1 plus 2nd degree polynomials for comorbidity index, number of visits, number of medications (generic entities).

Bacteriaemia:
 Immunosuppressive drug* use†3973.4153.782.11, 6.22Reference-Reference-
 Infliximab use†678.157.372.39, 17.21.950.71, 5.371.410.47, 4.24
 Corticosteroid use†4077.3174.172.42, 6.681.110.55, 2.211.120.52, 2.41
Serious bacterial infection:
 Immunosuppressive drug* use†3963.8276.814.49, 9.91Reference-Reference-
 Infliximab use†677.868.853.25, 19.31.300.54, 3.151.080.42, 2.74
Corticosteroid use†4065.8327.875.38, 11.11.160.69, 1.931.080.62, 1.87
Clostridium difficile infection:
 Immunosuppressive drug* use†3971.9143.521.93, 5.91Reference-Reference-
 Infliximab use†682.1000, 5.41----
 Corticosteroid use†4063.45714.0310.6, 18.23.982.22, 7.143.381.88, 6.10

Corticosteroid initiation with or without simultaneous use of other immunomodulating drugs was associated with a threefold increase in the risk of C. difficile infections (RR = 3.38; 1.88 to 6.10), but not other serious bacterial infections after adjusting for patient characteristics (Table 2). Initiators of corticosteroid monotherapy showed a slightly weaker association with C. difficile infections (RR = 2.65; 1.53 to 4.57) (Table 3). Initiation of infliximab showed a slightly increased relative risk for bacteriaemia based on five exposed events with wide confidence limits crossing 1.0 (RR = 1.41; 0.47 to 4.24).

Table 3.   Rates and rate ratios of serious infection outcomes for mono therapy with the study drugs (Model 2)
Endpoints and drug exposuresPatient yearsEventsEvent rate per 1000 95% CI‡Unadjusted rate ratio 95% CIAdjusted rate ratio§ 95% CI
  1. * Immunosuppressant drugs incl. MTX = Azathioprine, mercaptopurine, methotrexate.

  2. ‡ 95% confidence interval.

  3. § Adjusted for all covariates listed in Table 1 plus 2nd degree polynomials for comorbidity index, number of visits, number of medications (generic entities).

Bacteriaemia:
 Immunosuppressant* only1695.395.312.43, 10.1Reference-Reference-
 Infliximab only234.1000, 15.8----
 MTX only000---  
 Corticosteroids only3362.9154.462.50, 7.360.840.37, 1.921.380.63, 3.01
 Infliximab and ongoing MTX000-----
 Infliximab and ongoing corticosteroids105.5000, 34.97----
 Infliximab and ongoing immunosuppressants*137.217.290.18, 40.61.370.17, 10.82.130.27, 16.6
Serious bacterial infection:
 Immunosuppressant* only1690.3158.874.97, 14.6Reference-Reference-
 Infliximab only233.814.280.11, 23.80.480.06, 3.650.740.10, 5.53
 MTX only000-- --
 Corticosteroids only3353.2288.355.55, 12.10.940.50, 1.761.220.70, 2.13
 Infliximab and ongoing MTX000-----
 Infliximab and ongoing corticosteroids108.5000, 34.0----
 Infliximab and ongoing immunosuppressants*137.117.290.18, 40.60.820.11, 6.221.050.14. 7.81
Clostridium difficile infection:
 Immunosuppressant* only1699.910.590.01, 3.28Reference-Reference-
 Infliximab only234.1000,15.8----
 MTX only000-----
 Corticosteroids only3350.54714.0310.3, 18.723.853.29, 1732.651.53, 4.57
 Infliximab and ongoing MTX000-----
 Infliximab and ongoing corticosteroids108.5000, 34.0----
 Infliximab and ongoing immunosuppressants*138.5000, 26.6----

Based on few events, initiators of infliximab who were already treated by another immunomodulating medication were not any more likely to develop serious bacterial infections (RR = 0.70; 0.22 to 3.28) or bacteriaemia (RR = 0.84; 0.22 to 3.28) compared with initiators of immunosuppressants who also used at least one other immunomodulating drug (Table 4). Corticosteroid therapy in combination with other agents showed a slightly increased risk of having a C. difficile infection diagnosed (RR = 1.39; 0.69 to 2.80).

Table 4.   Rates and rate ratios of serious infection outcomes for combination therapy with the study drugs (Model 3)
Endpoints and drug exposuresPatient yearsEventsEvent rate per 1000 95% CI‡Unadjusted rate ratio 95% CIAdjusted rate ratio§ 95% CI§
  1. * Immunosuppressant drugs incl. MTX = Azathioprine, mercaptopurine, methotrexate.

  2. ‡ 95% confidence interval.

  3. § Adjusted for all covariates listed in Table 1 plus 2nd degree polynomials for comorbidity index, number of visits, number of medications (generic entities).

Bacteriaemia:
 Immunosuppressant* with at least 1 ongoing immunomod. Drug2078.162.891.06, 6.28Reference-Reference-
 Infliximab with at least 1 ongoing immunomod. Drug444.0511.263.66, 26.33.941.20, 12.90.840.22, 3.28
 MTX with at least 1 ongoing immunomod. Drug445.0000, 8.29----
 Corticosteroids with at least 1 ongoing immunomod. Drug714.322.800.34, 10.10.970.20, 4.800.350.07, 1.68
Serious bacterial infection:
 Immunosuppressant* with at least 1 ongoing immunomod. drug2273.5125.282.73, 9.22Reference-Reference-
 Infliximab with at least 1 ongoing immunomod. drug444.0511.263.66, 26.32.150.76, 6.110.700.22, 2.22
 MTX with at least 1 ongoing immunomod. drug444.512.250.06, 12.50.430.06, 3.280.120.02, 1.04
 Corticosteroids with at least 1 ongoing immunomod. drug712.545.611.53, 14.41.060.34, 3.300.500.17, 1.48
Clostridium difficile infection:
 Immunosuppressant* with at least 1 ongoing immunomod. drug2272.0135.723.05, 9.78Reference-Reference-
 Infliximab with at least 1 ongoing immunomod. drug448.0000, 8.23----
 MTX with at least 1 ongoing immunomod. drug443.549.022.46, 23.11.580.51, 4.830.780.24, 2.55
 Corticosteroids with at least 1 ongoing immunomod. drug712.91014.036.73, 25.82.451.08, 5.591.390.69, 2.80

The increased risk for C. difficile infection associated with corticosteroid initiation was constant over time since initiation (Table 5) with a multivariate adjusted rate ratio of 2.31 in the first 90 days (1.26 to 4.23) and 3.14 afterwards (1.37 to 7.19). This increased risk of corticosteroids was not dose-dependent (RR = 1.17; 0.63 to 2.18 comparing <20 mg vs. ≥20 mg daily (Table 6). No changes in rate ratios over study time were observed for the other infection outcomes (Tables S1 and S2 in the supplementary data document).

Table 5.   Duration-related effects on the incidence of clostridium difficile infection
Drug exposure† 1–90 days after treatment initiationMore than 90 days after treatment initiation
Number of EventsAdjusted rate ratio‡ 95% CI§Number of EventsAdjusted rate ratio 95% CI
  1. * Immunosuppressant drugs incl. MTX = Azathioprine, mercaptopurine, methotrexate.

  2. † For any combinations of study drugs, the first-listed drug was the newly initiated drug at cohort entry and other drugs were used subsequently.

  3. ‡ Adjusted for all covariates listed in Table 1 plus 2nd degree polynomials for comorbidity index, number of visits, number of medications (generic entities).

  4. § 95% confidence interval.

Any drug use:
 Immunosuppressive drug* use†14Reference-10Reference-
 Infliximab use†0--0--
 Corticosteroid use†572.311.26, 4.23233.141.37, 7.19
Mono-therapy:
 Immunosuppressant* only-Reference--Reference-
 Infliximab only0--0--
 MTX only0--0--
 Corticosteroids only471.891.11, 3.21172.050.93, 4.50
 Infliximab and ongoing MTX0--0--
 Infliximab and ongoing corticosteroids0--0--
 Infliximab and ongoing immunosuppressants*0--0--
Combination-therapy:
 Immunosuppressant* with at least 1 ongoing immunomod. drug13Reference-9Reference-
 Infliximab with at least 1 ongoing immunomod. drug0--0--
 MTX with at least 1 ongoing immunomod. drug40.820.24, 2.7430.750.16, 3.52
 Corticosteroids with at least 1 ongoing immunomod. drug101.150.56, 2.3662.160.79, 5.89
Table 6.   Dose-related effects of corticosteroid use on the incidence of serious bacterial infections* in the first 90 days after treatment initiation
Study endpoint<20 mg daily oral corticosteroids≥20 mg daily oral corticosteroidsAdjusted rate ratio†95% CI‡
Patient yearsNumber of EventsPatient yearsNumber of Events
  1. * ≥20 mg daily vs <20 mg daily oral corticosteroids; 3746 treatment episodes.

  2. † Adjusted for all covariates listed in Table 1 plus 2nd degree polynomials for comorbidity index, number of visits, number of medications (generic entities).

  3. ‡ 95% confidence interval.

Bacteriaemia473.83735.850.650.22, 1.96
Any bacterial infection473.56735.0110.890.41, 1.93
Clostridium difficile infect.472.813734.3171.170.63, 2.18

Sensitivity analyses to explore how strong confounder that remained unmeasured would have to be to fully explain our findings showed that an unmeasured or incompletely measured confounder like severity of IBD (in contrast to less severe IBD) that would be found in 50% of patients would need to elevate the risk for C. difficile infection six-fold, and would also need to have a six-fold greater prevalence among corticosteroid initiators to produce sufficient confounding to explain our findings of RR = 3.38. Less strong confounding is required to explain the lower bound of the 95% confidence limit (RR = 1.88) (Figure 1).

image

Figure 1.  Sensitivity analysis to identify the amount of confounding that would be necessary to explain the association between corticosteroid initiation in patients with IBD and C. difficile infection. [Plotted is the strength of the associations between an unmeasured confounder and treatment choice (corticosteroids vs. immunosuppressive drug, OREC) and the association between an unmeasured confounder and C difficile infection (RRCD) that are required to explain fully the observed association (ARR = 3.38) or its lower 95% confidence limit (ARR = 1.88). We further assumed a 68% prevalence of exposure equivalent to our study population and a 50% prevalence of the unmeasured confounder. Any factor (a single factor or combination of multiple factors) that has a combination of RRCD and OREC values resulting in points higher than and to the right of the plotted lines will be able to explain our observed results fully.]

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendix
  10. Supporting Information

In a large cohort study based on health care utilization data of British Columbia patients with IBD who initiated treatment with infliximab, we did not find an increased rate of serious bacterial infections requiring hospitalizations, including bacteriaemia and C. difficile infections compared to patients treated with other immunosuppressive agents. These results support previous studies suggesting no increased risk of infections in patients on infliximab compared to those on immunosuppressants,4,6 but are in contrast to others.3–5 We did observe a three-fold increase in C. difficile infections among new initiators of oral corticosteroids that were used either in combination with other immunomodulating drugs or alone suggesting that confounding is unlikely to explain our null-finding for infliximab. The association persisted as a slightly weaker 2.5 fold increase in risk in patients who used corticosteroids without other immunomodulating drugs. Overall, the rate of serious infections in this population-based cohort was relatively low for IBD patients.

Combining medications that lead to immunosuppression in patients with IBD has been shown to increase risk of developing opportunistic infections by as much as 14 fold.6 A possible bias in these results may be patients with more advanced disease are more likely to be on combination therapy. The corticosteroid association was reduced after multivariate adjustment for disease severity markers confirming that sicker patients were treated, but a clinically meaningful association remained. A quantitative sensitivity analysis showed only very strong unmeasured patient predictors of the risk for serious infections could explain our findings, which is unlikely, but cannot be ruled out entirely.

Two recent studies have shown a significant rise in C. difficile. associated diarrhoea in patients with IBD since 2001.22,23 Having IBD appears to be its own risk factor to acquiring C. difficile infection irrespective of antibiotic exposure. In a Milwaukee study, a retrospective observational study from an IBD referral centre, only 61% of patients acquiring C. difficile infection had antibiotic exposure.7 Our study supports this finding in that 57.2% of patients acquiring a C. difficile infection received antibiotics in the previous 6 months. This association is becoming more of a concern as the negative impact of this infection on the course IBD is being appreciated. Recent evidence demonstrates a four-fold increase in mortality in patients admitted to hospital with C. difficile infection and IBD compared to those admitted to hospital only with IBD.24 The C. difficile and IBD admitted patients were also in hospital 3 days longer.23 Based on our findings, a possible way to reduce the risk of patients with IBD acquiring C. difficile infection could be to reduce steroid exposure in these patients. This is the first study to demonstrate an association between corticosteroid use in IBD patients and the development of C. difficile infections. This finding requires further attention.

Epidemiological studies using health care utilization data are particularly scrutinized for their limited control of confounding and their potential for misclassifying diagnoses.25 We found some evidence for confounding; adjusting for a number of covariates describing patients’ underlying risk for infections reduced a strong, almost four-fold spurious unadjusted association between infliximab and serious bacterial infections to a null effect. This is demonstration of the successful multivariate adjustment strategy. Despite the same adjustment, the corticosteroid effect persisted with regard to C. difficile infections. Although this study included all infliximab dispensing independent of health plan coverage, it is worth noting that in British Columbia, infliximab is usually only covered after earlier treatment failure. PharmaCare, the provincial pharmacy benefit programme that covers about 70% of patients with Crohn’s disease requires before full coverage for infliximab that patients must have failed appropriate antibiotic therapy for a minimum of 3 weeks, plus failed or be intolerant to azathioprine or MP, or have been refractory to 5-ASA medications, plus either refractory or developed a serious adverse event to corticosteroids, plus failed or intolerant to azathioprine or MP. These criteria may influence concomitant medications patients received in our cohort; however, these steps prior to being eligible to receive infliximab with PharmaCare coverage closely parallels treatment strategies used in routine care and were adjusted as much as possible. It is also important to note that this study also includes the 30% Crohn’s patients not covered by PharmaCare. A quantitative sensitivity analysis revealed that very strong confounder would have to be omitted to explain our findings fully.

We limited our outcome definition of serious infection to primary diagnoses of a bacterial infection that required hospitalization. This reduced subjectivity in the assessment of severity, although it does not fully rule out individual differences in physicians’ threshold to admit such patients. Our validation study of diagnostic codes showed a robust positive predictive value for specific diagnoses,10 which further increased when the definition of infection outcome was relaxed to any infectious activity as the primary reason for the hospitalization. New infections with C. difficile were alternatively defined as outpatient visits in combination with antibiotic therapy that is specific to C. difficile treatment. Therefore, we avoided direct comparisons of incidence rates of C. difficile and other serious infections that may be misleading. Nevertheless, high positive predictive values or high specificity minimizes bias in ratio measures like the reported rate ratios.26 The reported data were generated prospectively in routine care settings without knowledge of the study hypothesis, minimizing the potential for observer bias.

In summary, in a large cohort of patients with IBD, we found that corticosteroid initiation but not infliximab use increased the risk for C. difficile infections. No association between infliximab and bacteriaemia and other serious infections was observed. The analysis is limited by the small number of serious bacterial infections that required hospitalization.

Acknowledgements

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendix
  10. Supporting Information

Declaration of personal interests: Dr Schneeweiss has received an investigator-initiated research grant from Pfizer on the safety of selective cox-2 inhibitors. Dr Korzenik has been a consultant for Procter and Gamble, Shire Pharmaceuticals, Eurand, Cytokine Pharma, Zymogenetics and receives research support from Procter and Gamble. Dr. Solomon receives research support from Amgen and Abbott. Dr Bressler has served as a speaker, a consultant and an advisory board member for Abbott, UCB and Schering-Plough. Declaration of funding interests: This study was funded by the Division of Pharmacoepidemiology, Brigham and Women’s Hospital. Dr Schneeweiss is Principal Investigator of the Brigham and Women’s Hospital DEcIDE Research Center funded by the Agency for Healthcare Research and Quality.

References

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendix
  10. Supporting Information
  • 1
    Hsia EC, Ruley KM, Rahman MU. Infliximab (Remicade): from bench to clinical practice. A paradigm shift in rheumatology practice. APLAR Journal of Rheumatology 2006; 9: 107118.
  • 2
    Rutgeerts P, Feagan BG, Lichtenstein GR, et al. Comparison of scheduled and episodic treatment strategies of infliximab in Crohn’s disease. Gastroenterology 2004; 126: 40213.
  • 3
    Colombel JF, Loftus EV Jr, Tremaine WJ, et al. The safety profile of infliximab in patients with Crohn’s disease: the Mayo Clinic experience in 500 patients. Gastroenterology 2004; 126: 1931.
  • 4
    Lichtenstein GR, Feagan BG, Cohen RD, et al. Serious infections and mortality in association with therapies fro Crohn’s disease: TREAT registry. Clin Gastroenterol Hepatol 2006; 4: 62130.
  • 5
    Ljung T, Karlen P, Schmidt D, et al. Infliximab in inflammatory bowel disease: clinical outcome in a population based cohort from Stockholm County. Gut 2004; 53: 84953.
  • 6
    Toruner M, Loftus EV, Harmsen WS, et al. Risk factors for opportunistic infections in inflammatory bowel diseases. Gastroenterology 2008; 134: 92936.
  • 7
    Williams JI, Young W. Inventory of Studies on the Accuracy of Canadian Health Administrative Databases. Technical report: Institute for Clinical Evaluative Sciences (ICES), Toronto, Canada, December 1996.
  • 8
    Bernstein CN, Blanchard JF, Rawsthorne P, et al. The epidemiology of Crohn’s disease and ulcerative colitis in a central Canadian province: a population-based study. Am J Epidemiol 1999; 149: 91624.
  • 9
    Schneeweiss S, Setoguchi S, Weinblatt M, et al. Anti-TNFα therapy and the risk of serious bacterial infections in elderly patients with rheumatoid arthritis. Arthritis Rheum 2007; 56: 175464.
  • 10
    Schneeweiss S, Robicsek A, Scranton R, Zuckerman D, Solomon DH. Coding accuracy of serious bacterial infections and opportunistic infections in health care utilization databases. J Clin Epidemiol 2007; 60: 397409.
  • 11
    Schneeweiss S, Seeger J, Maclure M, Wang P, Avorn J, Glynn RJ. Performance of comorbidity scores to control for confounding in epidemiologic studies using claims data. Am J Epidemiol 2001; 154: 85464.
  • 12
    Schneeweiss S, Wang PS, Avorn J, Maclure M, Levin R, Glynn RJ. Consistency of performance ranking of comorbidity adjustment scores in Canadian and U.S. health care utilization data. J Gen Intern Med 2004; 19: 44450.
  • 13
    Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol 1993; 46: 10759.
  • 14
    Ting G, Schneeweiss S, Katz JN, et al. Performance of a medical chart-based severity index for rheumatoid arthritis. J Rheum 2005; 32: 167987.
  • 15
    Gavazzi G, Krause KH. Ageing and infection. Lancet Infect Dis 2002; 2: 65966.
  • 16
    Carmona M, Go′mez-Reino JJ, Rodrı′guez-Valverde V, et al. Effectiveness of recommendations to prevent reactivation of latent tuberculosis infection in patients treated with tumor necrosis factor antagonists. Arthritis Rheum 2005; 52: 176672.
  • 17
    Cox DR. Regression models and life tables. J Roy Statist Soc B 1972; 34: 187220.
  • 18
    Braitman LE, Rosenbaum PR. Rare outcomes, common treatments: analytic strategies using propensity scores. Ann Intern Med 2002; 137: 6935.
  • 19
    Cepeda MS, Boston R, Farrar JT, Strom BL. Comparison of logistic regression versus propensity score when the number of events is low and there are multiple confounders. Am J Epidemiol 2003; 158: 2807.
  • 20
    Lee EW, Wei LJ, Amato DA. Cox-Type Regression Analysis for Large Numbers of Small Groups of Correlated Failure Time Observations. Survival Analysis; State of the Art. Netherlands: Kluwer Academic Publishers, 1992: 23747.
  • 21
    Schneeweiss S. Sensitivity analysis and external adjustment for unmeasured confounders in epidemiologic database studies of therapeutics. Pharmacoepidemiol Drug Safety, 2006; 15: 291303.
  • 22
    Rodemann JF, Dubberke ER, Reske KA, Seo da H, Stone CD. Incidence of Clostridium difficile infection in inflammatory bowel disease. Clin Gastroenterol Hepatol 2007; 3: 33941.
  • 23
    Issa M, Vijayapal A, Graham MB, et al. Impact of Clostridium difficile on inflammatory bowel disease. Clin Gastroenterol Hepatol 2007; 3: 34551.
  • 24
    Ananthakrishnan AN, McGinley EL, Binion DG. Excess hospitalization burden associated with Clostridium difficile in patients with Inflammatory Bowel Disease. Gut 2008; 57: 20510.
  • 25
    Schneeweiss S, Avorn J. A review of uses of health care utilization databases for epidemiologic research on therapeutics. J Clin Epidemiol 2005; 58: 32337.
  • 26
    Kelsey JL, Whittemore AS, Evans AS, Thompson WD. Methods in Observational Epidemiology, 2nd edn. New York, NY: Oxford University Press, 1996: 34190.

Appendix

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendix
  10. Supporting Information

Appendix 1: ICD-9-CM codes for study outcomes

Discharge diagnosisICD-9-CM
Meningitis320. x, 049. x
Encephalitis323. x, 054.3
Endocarditis421.x
Pneumonia481. x, 482.x
Pyelonephritis590.x
Septic arthritis711.0x, excluding 711.08
Osteomyelitis730.0x, 730.1x, 730.2x
Septicaemia or Bacteriaemia038.x, 790.7
Clostridium difficile008.4

Supporting Information

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendix
  10. Supporting Information

Table S1. Duration-related effects on the incidence of bacteremia.

Table S2. Duration-related effects on the incidence of any bacterial infection.

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

FilenameFormatSizeDescription
APT_4037_sm_TablesS1-S2.doc70KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.