Pre-operative use of anti-TNF-α agents and the risk of post-operative complications in patients with ulcerative colitis – a nationwide cohort study

Authors

  • B. M. Nørgård,

    Corresponding author
    • Centre for National Clinical Databases, South, Odense University Hospital, and Research Unit of Clinical Epidemiology, Institute of Clinical Research, University of Southern Denmark, Odense C, Denmark
    Search for more papers by this author
  • J. Nielsen,

    1. Centre for National Clinical Databases, South, Odense University Hospital, and Research Unit of Clinical Epidemiology, Institute of Clinical Research, University of Southern Denmark, Odense C, Denmark
    Search for more papers by this author
  • N. Qvist,

    1. Department of Surgical Gastroenterology A, Odense University Hospital, and, Research Unit of Surgical Gastroenterology, Institute of Clinical Research, University of Southern Denmark, Odense C, Denmark
    Search for more papers by this author
  • K. O. Gradel,

    1. Centre for National Clinical Databases, South, Odense University Hospital, and Research Unit of Clinical Epidemiology, Institute of Clinical Research, University of Southern Denmark, Odense C, Denmark
    Search for more papers by this author
  • O. B. S. de Muckadell,

    1. Department of Medical Gastroenterology S, Odense University Hospital, and Research Unit of Medical Gastroenterology, Institute of Clinical Research, University of Southern Denmark, Odense C, Denmark
    Search for more papers by this author
  • J. Kjeldsen

    1. Department of Medical Gastroenterology S, Odense University Hospital, and Research Unit of Medical Gastroenterology, Institute of Clinical Research, University of Southern Denmark, Odense C, Denmark
    Search for more papers by this author

Correspondence to:

Dr B. M. Nørgård, Centre for National Clinical Databases, South. Odense University Hospital. Sdr. Boulevard 29, entrance 1014th. DK-5000 Odense C, Denmark.

E-mail: bente.noergaard@ouh.regionsyddanmark.dk

Summary

Background

It is still controversial whether pre-operative anti-tumour necrosis factor-alpha (anti-TNF-α) agents increase post-operative complications in patients with ulcerative colitis (UC).

Aim

In a nationwide Danish cohort of patients with UC, we aimed to examine the impact of pre-operative use of anti-TNF-α agents on post-operative adverse outcomes after colectomy for UC. Outcomes (within 30 and 60 days after surgery) were reoperation, anastomosis leakage, intra-abdominal abscess, bacteremia and death.

Methods

Based on the Danish National Patient Registry we identified all UC patients, aged ≥15 years, having their first surgery for UC in the period of 1 January 2003–31 December 2010 (n = 1226). Patients were classified according to use of anti-TNF-α agents within 12 weeks before surgery or not. Outcome data were obtained from Danish registries. Logistic regression analyses were used to estimate adjusted risks [with 95% confidence intervals (CI)] of post-operative outcomes among patients treated with anti-TNF-α agents, relative to those not treated.

Results

A total of 199 UC patients were exposed to anti-TNF-α agents within 12 weeks before colectomy, and 1027 were not. Among exposed, the adjusted odds ratio of reoperation and anastomosis leakage within 30 days after colectomy was 1.07 (95% CI: 0.71–1.59) and 0.52 (95% CI: 0.06–4.11) respectively. No deaths, cases of abscess drainage or bacteremia occurred among exposed within 30 days. Furthermore, no increased relative risks were found within 60 days after colectomy.

Conclusions

Based on nationwide data on UC patients having colectomies, pre-operative use of anti-TNF-α agents did not increase the risk of post-operative complications.

Introduction

Previous studies have demonstrated the efficacy of anti-tumour necrosis factor-alpha (anti-TNF-α) agents (infliximab, adalimumab) in patients with severe ulcerative colitis (UC),[1-6] leading also to an increased use of anti-TNF-α agents to avoid colectomy.[7-10] When bowel resection comes into question surgeons will therefore be increasingly faced with abdominal surgery on patients who have failed therapy with potent immunosuppressants - i.e. also failed therapy with TNF-α blocking agents.

Little is still known about the risk of post-operative complications in UC patients receiving anti-TNF-α agents pre-operatively. Studies on UC patients in this field have been small-modest (47 exposed in the largest study[11]) and reporting conflicting results.[1, 11-14] Initial studies suggested an increased risk of post-operative complications (e.g., infectious complications) associated with pre-operative use of infliximab in UC patients,[1, 11] but this has not been found in other studies.[12-14] A recent meta-analysis concluded that pre-operative infliximab was associated with an approximately two-fold increased risk of short-term post-operative complications and that subgroup analyses were underpowered to assess the exact nature of complications.[15] Thus, further studies in this area are warranted as it is still controversial whether or not infliximab (or other anti-TNF-α agents) increases post-operative complications in UC patients.[9, 15, 16]

Pharmacokinetics of infliximab after intravenous infusion are characterised by a half-life of 10 days, and in week 12 after infusion of infliximab levels are no longer detectable with a dose of 5 mg/kg.[7, 16, 17] Thus, when examining post-operative complications after use of anti-TNF-α agents, a time-window of 12 weeks before surgery should be applied for patients to be treated. So far, all studies (also those included in the meta-analysis[15]) on post-operative complications in UC patients included patients with a much longer history of anti-TNF-α agent therapy before surgery,[1, 11, 12] making it possible that a potential negative role of anti-TNF-α agent therapy could not be attributed to an effect by anti-TNF-α agents themselves. Only two studies have included solely patients exposed to infliximab within 12 weeks before surgery; comprising 22 and 29 exposed patients respectively.[13, 14]

In this nationwide Danish cohort study of UC patients having colectomy, we examined the risk of 30 and 60 days post-operative complications after temporary suppression of the immune response by use of anti-TNF-α agents within 12 weeks before surgery.

Materials and methods

Study population and setting

In Denmark (population approximately 5.5 million people), all citizens have free access to a tax supported health care system. Its uniform organisation allowed us to use a population-based study design. In addition, the availability of nationwide Danish registries made it possible to assess data from Statistic Denmark on (i) patients with UC; (ii) treatment with anti-TNF-α agents; (iii) therapeutic drug use before date of first operation of UC and (iv) post-operative complications (death, reoperation, anastomosis leakage, percutaneous abscess drainage and bacteremia). The unique civil registration number (assigned to all Danish residents from the Central Personal Registration system) is used in all Danish registries and ensures accurate linkage between registries.[18]

Infliximab was first licenced for treatment of Crohn's disease (in Denmark since the end of 1999) and after some years also for treatment of UC, together with other kinds of anti-TNF-α agents. A recent Danish study showed that biological therapy has been used in patients with inflammatory bowel disease since 1999[19]; but for UC patients biological therapy was first used within 12 weeks before surgery in 2003. Thus, we decided the study period to begin on 1 January 2003.

To be eligible for the study, patients had to (i) have a valid civil registration number; (ii) have a discharge diagnosis of UC during the period from 1 January 1977 through 2010; (iii) have a first-time operation (a colectomy) for UC within the study period of 1 January 2003–31 December 2010 and (iv) be available for follow-up in Denmark 30 and 60 days after first UC surgery.

Identification of UC patients with a first-time operation for UC

The Danish National Patient Registry (NPR) contains records of all discharges from Danish hospitals since 1977 and all outpatient visits since 1994. Information in the NPR includes patients’ unique civil registration numbers, hospital, department, dates of admission and discharge, procedures performed and up to 20 discharge diagnoses based on the International Classification of Diseases (ICD-8 before 1994 and ICD-10 from 1994 onward).[20, 21]

All eligible patients had a discharge history of UC from any hospital in Denmark since 1 January 1977 (ICD-8 codes: 563.19, 569.04; ICD-10 codes: DK51.0, DK51.1, DK51.2, DK51.3); and included in the study were those patients with a first-time bowel operation for UC at age ≥15 years in the period 1 January 2003–31 December 2010 (in the ‘Nordic Classification of Surgical Procedures’ codes of KJFH; i.e. all types of colectomies; and a contemporary diagnosis of UC, K51.x, in the same record in the NPR). Thus, patients with a surgery for UC before 1 January 2003 were excluded (versions of the ‘Nordic Classification of Surgical Procedures’: colectomy codes of 45020, 45060, 45080, 45840, 45880 or KJFH). If more than one operation for UC occurred in the study period we used the first registered operation.

Pre-operative therapeutic drug use

In Denmark, anti-TNF-α agents are administered only in hospitalised patients or in hospital-based outpatient settings in accordance with the National Board of Health. Data on anti-TNF-α agent therapy were obtained from the NPR, where date of procedures for anti-TNF-α agent therapy is coded by BOHJ18 [or specific codes for infliximab (BOHJ18A1), adalimumab (BOHJ18A3) or certolizumab pegol (BOHJ18A5), or procedures coded with additional relevant Anatomical Therapeutical Chemical classification (ATC) codes (infliximab L04AA12 or L04AB02, adalimumab L04AA17 or L04AB04 or certolizumab pegrol L04AB05)].

Other therapeutic drug treatment

Data on other therapeutic drug treatment, besides anti-TNF-α agents, were obtained by out-patient drug prescriptions from the nationwide prescription database maintained by the Danish Medicine Agency. Since 1 January 1995, data on prescriptions have been available from all pharmacies in Denmark due to computerised accounting systems which send key data on out-patient drug prescriptions directly to the database. Thus, we were able to obtain the prescription history of each UC patient, and could classify the time of drug prescriptions before colectomy in the study period. We identified all prescriptions (including prescription date) with the following ATC-codes: L04A X01 (azathioprine); L01B B02 (6-mercaptopurine); L01BA01 (methotrexate); L04AD01 (ciclosporin); H02A B06 (prednisolone); H02A B07 (prednisone); A07E A01 (local prednisolone); A07E A02 (local hydrocortisone); and A07E A06 (local budesonide).

Outcome data

Data on reoperations were obtained from the NPR; we identified (i) all kinds of operations on the digestive organs and spleen within the procedure code KJ; (ii) specifically the subgroup of anastomosis leakage, procedure code KJWF and (iii) percutaneous abscess drainage, UXUD (ultrasound examination of the abdomen) together with KTJA40 (percutaneous needle biopsy of the intestine). Data on death were obtained from the Civil Registration Number system.

In a sub-analysis examining the risk of bacteremia we restricted our study population to UC patients having their colectomy at Aalborg and Hvidovre Hospital. This sub-cohort was linked to data from a bacteremia registry retrieved from the departments of clinical microbiology at these two Danish hospitals. The registry captures geographical areas (The North Region Denmark and Hvidovre Hospital in the Capital Region of Denmark) with a total background population of approximately 1.14 million, covering the period 1 January 2003–1 November 2010.[22, 23] We omitted positive blood cultures considered to be contaminated and defined bacteremia if the isolated pathogen was given aetiological significance based on clinical and microbiological criteria.[24, 25]

We evaluated all outcomes occurring within 30 and 60 days after UC colectomy.

Data on covariates

From the NPR we obtained data on patient gender, age at time of first UC surgery in the study period, duration of UC at the time of first operation (calculated from the date of the first time the UC diagnosis appears in the NPR until the date of first operation), calendar period of first UC surgery and number of inpatient days at hospital before surgery. Data from the NPR were also used to assess the Charlson Index (comorbidity index score) for each patient.[26] The Charlson Index, covering 19 major disease categories weighted according to their prognostic impact on patient survival, has been adapted for use of hospital discharge registry data. We computed the index based on diagnoses recorded during all previous hospitalisations since 1977. Three index levels were defined to capture increasing degrees of comorbidity: no co-morbidity (Charlson Index 0), intermediate co-morbidity level (Charlson Index 1–2) and high co-morbidity level (Charlson Index >2). From the nationwide prescription database we obtained data on prescriptions for steroids (prednisolone, prednisone, local prednisolone, local hydrocortisone or local budesonide) before colectomy.

Statistical analysis

Classification of exposed and unexposed

The associations between pre-operative anti-TNF-α agent therapy and adverse post-operative outcomes (reoperations, anastomosis leakage, intra-abdominal abscess, bacteremia and death) were studied in sub-cohorts of patients with colectomy in the study period - classified according to therapeutic drug treatment pre-operatively as recorded in the NPR and the nationwide prescription database.

  1. Exposed cohort of UC patients were patients treated at least once with anti-TNF-α agents within 12 weeks before colectomy. Patients in the exposed cohort were allowed to be treated with other kinds of drugs for UC treatment (for instance 5-aminosalicylic acid (5-ASA), steroids and/or azathioprine/6-mercaptopurine).
  2. Unexposed cohort 1 was UC patients not treated with anti-TNF-α agents within 12 weeks before colectomy, but allowing all kinds of other treatments (for instance 5-ASA, steroids and/or azathioprine/6-mercaptopurine).
  3. Unexposed cohort 2 was UC patients not treated with anti-TNF-α agents within 12 weeks before colectomy, but as recorded by the national prescription database the patients had at least one filled prescription for steroids, azathioprine/6-mercaptopurine, methotrexate or ciclosporin within 12 weeks before colectomy. Thus, this group constitutes a subgroup of the unexposed cohort 1.

The purpose of designing unexposed cohort 2 was to give the possibility to counter a potential impact of disease activity.

Adverse post-operative complications (death, reoperation, anastomosis leakage and intra-abdominal abscess) in the exposed cohort were compared to outcomes in unexposed cohort 1 and unexposed cohort 2. Bacteremia was only analysed in a sub-cohort of the study population with a first UC surgery in the North Denmark Region and Hvidovre Hospital in the Capital Region of Denmark.

Crude and adjusted analyses

We constructed contingency tables for the main study variables, and computed the relative risk estimates (prevalence odds ratio, [OR]), with 95 percentage confidence intervals (95% CI). We used logistic regression analyses to compute relative risk estimates for adverse post-operative outcomes (reoperations, anastomosis leakage, intra-abdominal abscess, bacteremia and death within 30 days and 60 days after the first UC operation) associated with anti-TNF-α agent therapy pre-operatively, adjusted for potential confounders. Adjustment was made for age (years 15–25 as reference, 26–55, >55), gender (females as reference), co-morbidity (Charlson Index 0 as reference, Charlson Index 1–2, Charlson Index >2), calendar period (2003–2004 as reference, 2005–2006, 2007–2008, 2009–2010), duration of UC (<5 years as reference, ≥5 years), use of steroids within 4 weeks before surgery (no as reference, yes) and number of inpatient days at hospital within 4 weeks before surgery (<2 weeks as reference, ≥2 weeks) in a logistic regression model. The purpose of including use of steroids and number of inpatient days at hospital within 4 weeks before surgery was to counter in a possible impact of disease activity. Confounders were included in the models based on the ‘change-in-estimate’ method. When estimating the relative risk of re-operations, anastomosis leakage, intra-abdominal abscess and bacteremia, only patients available for full follow-up were included.

All analyses were conducted using Stata 12 software (StataCorp LP, College Station, TX, USA). The study was approved by the Danish Data Protection Agency (j.nr. 2010-41-5660).

Results

During the study period of 8 years, 1226 colectomies were registered in UC patients aged ≥15 years. The exposed cohort comprised 199 (16.2%) patients treated with anti-TNF-α agents within 12 weeks before surgery, and the remaining were not treated with anti-TNF-α agent within 12 weeks before surgery, i.e. unexposed cohort 1 = 1027 (83.8%). Among exposed patients, infliximab counted for the vast majority of anti-TNF-α agent in the study period, 167/199 = 83.9% (and adalimumab for 7.5% and unspecified anti-TNF-α agents 8.5%). The pre-operative use of anti-TNF-α agents increased with calendar period; 10.6% of the use was registered before 2007, and 89.4% from 2007 until the end of the study period.

Table 1 gives the characteristics of study cohorts of UC patients with colectomy in the study period. The most commonly used procedure among exposed, unexposed cohort 1 and unexposed cohort 2 was KJFH10 (colectomy and ileostomy); 60.8%, 60.1% and 62.8% respectively. The median age among exposed, unexposed cohort 1 and unexposed cohort 2 was 37, 42 and 40 years respectively. In all cohorts, the vast majority of patients had (i) duration of UC of less than 5 years at the time of surgery; (ii) no co-morbid diseases; (iii) stayed at hospital for less than 2 weeks within a period of 4 weeks before surgery and (iv) no prescriptions of steroids within 4 weeks before surgery (Table 1). Numbers of post-operative complications and crude and adjusted relative risk estimates (with unexposed cohort 1 as the reference) are given in Table 2. Among the 1226 colectomies there were 30 deaths within 30 days after surgery, 30/1226 = 2.4%, and none of these deaths occurred among the exposed. One death among the exposed occurred within 60 days (0.5% among exposed vs. 3.1% among unexposed; adjusted OR = 0.27, 95% CI: 0.03–2.17), and the one who died had no preceding adverse surgical outcomes. Among those exposed, 21.6% had a reoperation within 30 days vs. 23.1% among unexposed; adjusted OR = 1.07, 95% CI: 0.71–1.59. Moreover, no significantly increased risk of reoperation was found within 60 days after colectomy. The two most often used groups of reoperations, which were performed among exposed within 30 days after colectomy, were (i) revision of enterostomy or colostomy with/without laparotomy (10/43 = 23.3%) and (ii) laparoscopy/explorative laparotomy (9/43 = 20.9%). We found no increased risk of anastomosis leakage or percutaneous abscess drainage. In general, adjusting for potential confounders had only little-moderate impact.

Table 1. Characteristics of study cohorts of patients with ulcerative colitis (UC) with a first-time operation from 1 January 2003 through 2010
 Exposed cohort, n = 199 (receiving anti-TNF-α agents within 12 weeks before surgery)Unexposed cohort 1, n = 1027 (no anti-TNF-α agents within 12 weeks before surgery)Unexposed cohort 2a, n = 403 (no anti-TNF-α agents within 12 weeks before surgery but other kind of immunosuppressants)
  1. a

    Unexposed cohort 2 is a sub-group of unexposed cohort 1.

  2. b

    No co-morbidity (Charlson Index 0), intermediate co-morbidity level (Charlson Index 1–2), high co-morbidity level (Charlson Index >2).

Type of UC surgery, number (%)
KJFH00, colectomy and ileorectostomy2 (1.0)29 (2.8)18 (2.9)
KJFH01, laparoscopic colectomy and ileorectostomy2 (1.0)3 (0.3)0 (0.0)
KJFH10, colectomy and ileostomy121 (60.8)617 (60.1)392 (62.8)
KJFH11, laparoscopic colectomy and ileostomy39 (19.6)141 (13.7)86 (13.8)
KJFH20, proctocolectomy and ileostomy18 (9.0)112 (10.9)59 (9.5)
KJFH21, laparoscopic proctocolectomy and ileostomy2 (1.0)4 (0.4)2 (0.3)
KJFH30, colectomy and mucosal rectotomy with ileoanal anastomosis without ileostomy1 (0.5)19 (1.9)9 (1.4)
KJFH31, laparoscopic colectomy and mucosal rectotomy with ileoanal anastomosis without ileostomy0 (0.0)0 (0.0)0 (0.0)
KJFH33, colectomy and mucosal rectotomy with ileoanal anastomosis with ileostomy14 (7.0)86 (8.4)50 (8.0)
KJFH40, proctocolectomy and continent ileostomy0 (0.0)2 (0.2)1 (0.2)
KJFH96, other colectomy0 (0.0)14 (1.4)7 (1.1)
Sex
Females, number (%)104 (52.3)500 (48.7)299 (47.9)
Males, number (%)95 (47.7)527 (51.3)325 (52.1)
Age at time of first UC operation,
Median (min–max), in years37 (15–82)42 (15–95)40 (15–91)
Age 15–25, number (%)90 (45.2)353 (34.4)240 (38.5)
Age 26–55, number (%)66 (33.2)347 (33.8)215 (34.5)
Age >55, number (%)43 (21.6)327 (31.8)169 (27.1)
Duration of UC at the time of first operation
<5 years with UC, number (%)134 (67.3)656 (63.9)406 (65.1)
≥5 years with UC, number (%)65 (32.7)371 (36.1)218 (34.9)
Co-morbidity levelb
No co-morbidity, number (%)168 (84.4)728 (70.9)458 (73.4)
Intermediate co-morbidity level, number (%)24 (12.1)223 (21.7)126 (20.2)
High co-morbidity level, number (%)7 (3.5)76 (7.4)40 (6.4)
Number of inpatient days within 4 weeks before surgery
<2 weeks175 (87.9)812 (79.1)303 (75.2)
≥2 weeks24 (12.1)215 (20.9)100 (24.8)
Use of steroids within 4 weeks before surgery
Yes, number (%)49 (24.6)300 (29.2) 
No, number (%)150 (75.4)727 (70.8) 
Table 2. Post-operative outcomes in study cohorts of patients with ulcerative colitis with a first-time operation from 1 January 2003 through 2010
 Exposed cohort, n = 199a (anti-TNF-α agents within 12 weeks before operation)Unexposed cohort 1, n = 1027b (no anti-TNF-α agents within 12 weeks before operation)Crude OR (95% confidence interval)Adjusted OR c (95% confidence interval)
  1. a

    One patient who died within the 60 day time-window was excluded in the analyses of reoperation, anastomosis leakage and abscess drainage within 60 days.

  2. b

    Thirty patients who died within the 30 day time-window were excluded in the analyses of reoperation, anastomosis leakage and abscess drainage within 30 days, and 32 patients who died within the 60 day time-window were excluded in the analyses of reoperation, anastomosis leakage and abscess drainage within 60 days.

  3. c

    Adjusted for age (years 15–25, 26–55, >55), gender, co-morbidity (no co-morbidity, co-morbidity level 1, co-morbidity level 2), calendar period (2003–2004, 2005–2006, 2007–2008, 2009–2010), duration of UC (<5 years, ≥5 years), steroid prescriptions (no/yes) within 4 weeks before surgery and number of inpatient days at hospital within 4 weeks before surgery (<2 weeks, ≥2 weeks) in a logistic regression model.

Post-operative outcomes
Death within 30 days after first operation
Yes, number (%)0 (0)30 (2.9)
No, number (%)199 (100)997 (97.1)
Death within 60 days after first operation
Yes, number (%)1 (0.5)32 (3.1)0.16 (0.02–1.16)0.27 (0.03–2.17)
No, number (%)198 (99.5)995 (96.9)
Reoperation within 30 days after first operation
Yes, number (%)43 (21.6)230 (23.1)0.92 (0.64–1.33)1.07 (0.71–1.59)
No, number (%)156 (78.4)767 (76.9)
Reoperation within 60 days after first operation
Yes, number (%)46 (23.2)242 (24.3)0.94 (0.66–1.35)1.10 (0.74–1.63)
No, number (%)152 (76.8)753 (75.7)
Anastomosis leakage within 30 days after first operation
Yes, number (%)1 (0.5)16 (1.6)0.31 (0.04–2.35)0.52 (0.06–4.11)
No, number (%)198 (99.5)981 (98.4)
Anastomosis leakage within 60 days after first operation
Yes, number (%)1 (0.5)16 (1.6)0.31 (0.04–2.36)0.51 (0.06–4.11)
No, number (%)197 (99.5)979 (98.4)
Percutaneous abscess drainage within 30 days after first operation
Yes, number (%)0 (0.0)2 (0.2)
No, number (%)199 (100.0)995 (99.8)
Percutaneous abscess drainage within 60 days after first operation
Yes, number (%)0 (0.0)2 (0.2)
No, number (%)198 (100.0)993 (99.8)

The results of post-operative outcomes, with unexposed cohort 2 as the reference, were virtually unchanged (data not shown); i.e. no increased adjusted relative risks of adverse post-operative outcomes.

In the sub-analysis of bacteremia, we found no cases within 30 days after colectomy among anti-TNF-α agent exposed (0/49 = 0%), vs. 6/214 = 2.8% in the unexposed cohort 1. Moreover, we found no cases of bacteremia within 60 days after colectomy among anti-TNF-α agents exposed.

Discussion

This study showed no increased risk of 30 or 60 days post-operative complications, after use of anti-TNF-α agent therapy within 12 weeks before colectomy for UC.

Our study has several strengths. A population-based approach is possible in Denmark due to a unique availability of nationwide registries and has no loss to follow-up. The NPR records more than 99% of all hospital discharges for somatic diseases,[27] and we thus had access to the obligatory registration from Danish hospitals since 1977 and all out-patient visits since 1994. Hence, we used the NPR to identify patients with UC having their first UC surgery in the study period, based on hospital discharge diagnoses and specific procedure codes for UC surgery. The completeness and the validity of UC diagnoses in the NPR are of high quality. The completeness of diagnoses of UC has been examined in a Danish study using the pathology system as a reference standard – showing that of all patients with a confirmed diagnosis of UC, 94% were included in the NPR.[28] Furthermore, the overall validity of diagnosis of UC in the NPR was 90%, and with most patients diagnosed at specialised departments the validity of UC diagnosis increased to 94%.[28] Moreover, we have more reasons to believe that there is no major misclassification between patients with diagnosis of UC and Crohn's disease. First, patients were selected with a discharge history of UC in the NPR since 1977; secondly, patients were only included if coded as having specific procedure codes for UC colectomy in the study period; and thirdly, patients had to have a discharge diagnosis of UC in the same record as the code for UC colectomy occurred. Data on patients treated with anti-TNF-α agents were based on procedure codes in the NPR, which are usually of high quality with positive predictive values of 94–100%.[29-32] Some misclassification of anti-TNF-α agents (the exposure) cannot be ruled out if the coding in the NPR has not been entirely accurate, but a possible misclassification of anti-TNF-α agent therapy is unrelated to the outcomes examined, and will bias our risk estimates towards acceptance of the null hypothesis.

In addition, we had access to high quality data on out-patient drug prescriptions from the nationwide prescription database, making it possible to identify unexposed cohort 2. It is a strength that this group is based on prescriptions and not on recall, as drug exposure based on self-reported use may lead to severe recall bias or under-ascertainment.[33] The data from the prescription database are of high quality as a result of direct computerised transfer of information when a prescribed drug is dispensed at a pharmacy. However, a redeemed drug is only a proxy measurement of drug intake, but patient noncompliance would tend to underestimate our risk estimates. Prescription databases have thus been found to be of great value for drugs prescribed for serious diseases that need continuous treatment.[34]

Outcome data were obtained independently of the hypotheses investigated and exposure measurement, preventing differential misclassification of the outcome measurements. Our outcome data from the NPR are reported of high quality for surgical procedure codes with positive predictive values of 94–100%.[29-32]

Finally, we had information on several confounders, which actually turned out to have only little impact on the results. In an observational study like this it is not possible to control all potential confounders, but we countered in several factors that could have impacted the results. We included an impact of calendar period, patient and disease related details and proxy measurements of disease activity. When it comes to patient and disease related details we countered in an effect of age, gender and duration of UC. Duration of UC was calculated from the first time the patient was registered in the NPR, i.e. the first time the patient was discharged with the diagnosis or had been seen as an out-patient. We believe that this is a very reasonable approach to estimate the duration of disease as the general practitioners in Denmark refer a patient to hospital as soon as UC is suspected. Furthermore, we took into consideration co-morbidity based on complete discharge histories for all patients; actually the vast majority of UC patients were healthy individuals with no other diseases than UC. Especially, one could consider whether disease activity could confound our results, and therefore we had different approaches to counter in such a possible impact. First, in our main regression analyses using unexposed cohort 1, we adjusted for the proxy measurements of disease activity (use of steroids before colectomy and number of inpatient hospital days before colectomy) and found no important impact of these factors. Secondly, another approach to reveal an effect of disease activity was the design of unexposed cohort 2. This cohort was designed to counter possible arguments that patients in the exposed cohort might have higher disease activity than those in the unexposed cohort 1; i.e. if we had found increased risks of post-operative outcomes some might have argued that disease activity, and not anti-TNF-α agent therapy, was responsible for the adverse outcomes. However, with our negative findings we did not expect the analyses using unexposed cohort 2 to change our results – and they did not. Thus, overall, we have no indication that disease activity should have confounded our results; and with our negative results it is most likely that disease activity had no major impact on our post-operative outcomes or disease severity was equally distributed between exposed and unexposed (all patients were so severely diseased that they underwent colectomy).

The study also had limitations as we could not counter in surgery-specific factors, acute or elective colectomies and factors related to post-operative care. Regarding acute or elective colectomies it is extremely rare in Denmark that colectomies are performed as acute procedures, i.e. almost all colectomies are performed as sub-acute or elective surgery; and therefore we have no reason to believe that the mode of surgery should have confounded our results. Theoretically, one could consider a potential impact of surgery-specific factors and factors related to post-operative care; however, a confounding effect of such factors is only speculative as we have no reason to believe that these factors should differ among exposed and unexposed UC patients.

Also, our use of two different unexposed cohorts supports the validity of our results, as the two set of analyses give virtually the same relative risk estimates.

This study is the largest to date on the risk of post-operative outcomes after pre-operative exposure to anti-TNF-α agents in patients with UC. Existing data on this subject have been sparse, and most published studies included patients in the exposed cohorts that had been treated with anti-TNF-α agents so long before surgery that a potential negative role of anti-TNF-α agent therapy most probably could not be attributed to an effect by anti-TNF-α agents themselves.[15-17] Only two studies included solely UC patients exposed within 12 weeks before surgery in the exposed cohorts. Based on 22 infliximab exposed patients, Ferrante et al. found no mortality within 30 days after primary surgery and found that infliximab was not associated with an increased risk of short-term complications (anastomotic leaks, pelvic abscesses and wound infections).[13] Based on 29 infliximab exposed patients, Gainsbury et al. also found no mortality within 30 days after surgery and an OR = 0.78, 95% CI: 0.26–2.38 for overall post-operative complications (leaks, abscess, pouch-related complications, wound infections and others).[14] Our nationwide data on post-operative complications support these results. Especially, our results on death are reassuring as we found no deaths within 30 days among those exposed to anti-TNF-α agents pre-operatively. Overall, our result on mortality in UC patients undergoing colectomy is comparable to other findings of in-hospital mortality of 0.7–5.4%, depending on urgent or elective surgery.[35]

This first nationwide study on short-term post-operative complications provides reassuring results after preoperative use of anti-TNF-α agents (primarily infliximab) in UC patients. As anti-TNF-α agent therapy has been introduced recently it is of course especially important to monitor a possible negative effect on post-operative outcomes, and thus, these first results from nationwide data should be confirmed in other settings.

Acknowledgements

Declaration of personal interests: None. Declaration of funding interests: The study was funded in part by the Danish Colitis-Crohn Society (Colitis-Crohn Foreningen).

Ancillary