Article Text
PDF
Abstract
Objective To determine the risk of diverticular perforation associated with current and ever use of corticosteroids, opiate analgesics, non-steroidal anti-inflammatory drugs, aspirin, cyclo-oxygenase-2 inhibitors, statins and calcium antagonists.
Design, setting and participants Case–control analysis using conditional logistic regression analysis of data from the UK General Practice Research Database. The study involved 899 cases of incident diverticular perforation and 8980 population controls from 1990 to 2005.
Main outcome measures Odds ratios (ORs) are presented for perforation associated with use of corticosteroids, opiate analgesics, non-steroidal anti-inflammatory drugs, aspirin, cyclo-oxygenase-2 inhibitors, statins and calcium antagonists. Data were adjusted for smoking, comorbidity, prior abdominal pain and body mass index.
Results A total of 899 patients with an incident diagnosis of perforated diverticular disease were identified. Current use of opiate analgesics (OR=2.16; 95% CI 1.55 to 3.01) and oral corticosteroids (OR=2.74; 95% CI 1.63 to 4.61) was associated with a two- and threefold increase in the risk of diverticular perforation, respectively. Current use of a calcium antagonist and aspirin were not associated with an increased risk of diverticular perforation. Current statin use was associated with a reduction in the risk of perforation (OR=0.44; 95% CI 0.20 to 0.95).
Conclusion Perforated diverticular disease is a serious surgical emergency with current opiate analgesics and oral corticosteroids being strongly associated with an increased risk of diverticular perforation.
- Perforation
- opiate
- steroid
- diverticular disease
Statistics from Altmetric.com
Significance of this study
What is already known about this subject?
The incidence of diverticular perforation is increasing.
Diverticular perforation is associated with a twofold increase in mortality compared with the general population.
Small case–control studies have demonstrated an association between current medication use and diverticular perforation but these studies were unable to account for other confounding factors, relied on self-reported medication use and were not population based.
What are the new findings?
Current opiate analgesics use is associated with a twofold increase in the risk of diverticular perforation (OR=2.16; 95% CI 1.55 to 3.01).
Current oral corticosteroid use is associated with a threefold increase in the risk of diverticular perforation (OR=2.74; 95% CI 1.63 to 4.61).
Current use of a calcium antagonist and aspirin were not associated with an increased risk of diverticular perforation.
How might they impact on clinical practice in the foreseeable future?
Clinicians should avoid prescribing opiates and corticosteroids to patients with diverticular disease, where possible.
Introduction
Background
Perforated diverticular disease is an important surgical emergency with a high mortality and increasing incidence.1 2 Although we know that there is an increasing burden on healthcare services owing to increase of this disease and the ageing population, we have little idea of the aetiology of the condition and what factors may precipitate perforation. Clearly, the identification of risk factors is an important research area as doing so will allow risk reduction through use of appropriate preventive strategies.
Drugs—notably, corticosteroids, opiate analgesic and non-steroidal anti-inflammatory drugs (NSAIDs), have been suggested in previous studies as potential risk factors for perforation,3–7 while calcium antagonists have been shown to be potentially protective against perforation.4 There are biologically plausible explanations for the development of perforation with exposure to NSAIDs8 and aspirin with a decrease in prostaglandin synthesis and alterations in large intestine permeability9 which serve to decrease colonic wall integrity. Similarly, corticosteroids may inhibit peptic ulcer healing and increase the risk of perforation of such ulcers3 4 with a reduction in collagen turnover associated with chronic use, which further reduces colonic wall integrity. These mechanisms are believed to be responsible for the associations demonstrated in prior case–control studies.3 It has been proposed that owing to increased constipation associated with opiate analgesics that the resultant high-pressure colonic contractions may predispose to diverticular perforation.3 Opiates are used for the treatment of pain, which is a common symptom among people with diverticular disease, so any association found needs to take into account the indication for the prescribed drug. Calcium channel blockers have been reported to decrease the frequency of high-pressure colonic contractions and improve mucosal blood flow and this potential mechanism was thought responsible for the prior finding of a reduced association between users of these drugs and diverticular perforation.3 Statins have emerged as potential anti-inflammatory agents as they apparently reduce the proinflammatory tendencies of macrophages and neutrophils.10 Therefore we aimed to investigate the role of statins in diverticular perforation.
The previous studies that have investigated the relationships between medication use and perforated diverticular disease have in general been small, principally retrospective and used self-reported medication use or reports from hospital admission clerking to identify the exposure. They have also relied on hospital-based controls for comparison, which in itself often leads to biased estimates.11 In addition to these limitations in population selection and exposure measurement, there has been little attempt to measure and adjust for potential confounding factors such as cigarette smoking, comorbid illness and body mass index (BMI). This is partly because the design of the studies did not allow it but also because the comprehensive collection of such information requires prospective recording of risk factors. Similarly, to take account of confounding by indication, a prospective record of symptoms is required to avoid recall bias.
Objective
The objective of our study was to determine if current or ever use of a corticosteroid, opiate analgesic, NSAID, aspirin, cyclo-oxygenase-2 inhibitors (coxibs), calcium channel blocker or statin was associated with diverticular perforation using a large administrative primary care database from the UK to identify cases of perforation and compare them with the general population.
Materials and methods
General practice research database
Study design
The General Practice Research Database (GPRD) contains diagnostic and prescription data for over 13 million people of the general population in the United Kingdom, with 3.4 million active patients contributing high-quality validated data. The GPRD has been extensively used for epidemiological research. Originally developed in 1987, the database represents a prospectively collected source of continuous data on illness and prescribing habits in general practice in the UK, where all patient care is coordinated by the general practitioners (GPs). The data collected are audited regularly and the participating general practices are subjected to a number of quality checks to ensure they are ‘up to standard’ for research purposes. This audit process requires that 95% of the data entered into the database are sufficient for epidemiological research. The participating practices receive financial incentives and penalties to ensure that the data entered are correct. The database consists of observations, diagnoses made by, and treatments prescribed by, GPs, as well as information sent to them from hospitals, such as pathology and radiology reports as well as discharge letters. Diseases are coded within the GPRD using Read or Oxford Medical Indexing System (OXMIS) codes.
Study population
The patients were incident cases of free diverticular perforation (excluding acute diverticulitis, diverticular abscess and diverticular bleeding) who were identified by our previously validated diagnosis of perforated diverticular disease.2 All patients with an incident diagnosis between 1990 and 2005 were included in the study except those who had died at the index date. We used the definition for free perforation as any patient with a code specifically for perforated colonic diverticular disease present in the ‘up to standard’ data period or if a code for diverticular disease (excluding other complications of diverticular disease) was present at any time during follow-up plus a code for a colostomy or Hartmann's procedure was present during ‘up to standard’ follow-up. This prevented misclassification of patients with other complications of diverticular disease such as abscess formation and inflammation as cases. The date at which these codes first appeared in their record is known as the index date. The controls consisted of up to 10 patients within the same practice matched by age and gender.2 Each of the population controls had to be alive and contributing data to the GPRD at the index date of the case, which is effectively matching on calendar period. These controls were matched by the GPRD to the patients with perforation. A case was defined as incident if no prior record of perforated colonic diverticular disease was entered into the record within 90 days of original registration at the practice. This time period has been shown to be valid for other acute diagnoses such as acute diverticulitis.12 We excluded potential controls if they had any recorded codes for diverticular disease within their medical record so as to prevent any null bias and minimise confounding by indication that might occur by having potential cases misclassified as controls. The validation of the diagnosis of perforated diverticular disease has previously been reported, with over 90% of cases having evidence from secondary care of perforation.2
Exposures
Drug exposure was determined for each drug as current or ever use with reference to the date of perforation (for cases) or the corresponding date among matched controls. To account for any possible delay in diagnosis being received by the GP and subsequent recording in the patient clinical electronic record we excluded the first 30 days before perforation from the analysis to prevent overascertainment of prescriptions that might have occurred following hospital discharge. We therefore categorised current users as those who had a prescription of a drug in the period 6 months before this 1 month exclusion. Ever users were those who had a prescription at any time before this 6 month window but did not have a current prescription, thus forming two mutually exclusive groups (figure 1). This definition of current and ever users has been used previously with data extracted from the GPRD for pharmacoepidemiological analysis.13 Data were extracted for NSAIDs (excluding aspirin), opiate analgesics, aspirin (for prophylaxis of cardiovascular disease), oral corticosteroids, calcium antagonists, statins and coxibs.
Timeline of study and exposure windows. GPRD, General Practice Research Database.
Potential confounders
The potential confounding factors of comorbid illness, cigarette smoking and BMI were identified using data from 1 month before the index date to the beginning of ‘up to standard’ data. We also identified people who had had abdominal pain before the index date. Comorbidity was determined using the Charlson index as scored 1 month before the index date.14 Smoking status was determined as never and ever smoking based on codes recorded before the index date with missing data accounted for as a separate group. BMI was classified as <25 kg/m2, 25–29.9 kg/m2 and ≥30 kg/m2. The BMI recorded before the index date was used for this purpose with missing data being analysed as a separate category. A history of prior abdominal pain was identified using codes for abdominal pain for the 6 months before the index date. This was used as a proxy for symptomatic diverticular disease and as a measure of confounding by indication—that is, that those with pain might be more likely to perforate and also receive analgesics.
Statistical analysis
After a basic description of the case and control groups a univariate analysis was undertaken to identify factors associated with perforation. Adjusted ORs and 95% CIs were estimated using conditional logistic regression. A multivariate analysis was undertaken controlling for comorbid illness, BMI and smoking. A stratified analysis (for the presence or absence of abdominal pain) adjusted for age and sex was undertaken to assess confounding by indication with the analgesics. Interaction between comorbidity (Charlson index) and drug use was assessed using likelihood ratio tests. We planned a stratified analysis using logistic regression (ie, not taking account of the matching) to estimate within-strata ORs adjusted for age and sex for those exposures which demonstrated an interaction with comorbidity. Stata 10 was used for all analysis and data management.
Power calculation
Studies on the prevalence of NSAID and opioid analgesic use suggest that up to 30% of these patients may be taking an NSAID/opioid analgesic in this age group. The population-attributable risk of colonic perforation was 16% for any use of opioids, 20% for NSAIDs and 12% for corticosteroids.3 Using previously published studies on perforated colonic diverticular disease,3 we calculated that we required approximately 60 patients to detect ORs of >2.5 for medication use at the 5% significance level with a 90% power.
The study had approval from the Independent Scientific Advisory Committee approval board (ISAC) who provide scientific advice to the Medicines and Healthcare products Regulatory Agency (MHRA) on study design and advise if further approval is required from the Multi-centre Research Ethics Committee outside the MHRA's current approval for observational studies.
Results
Participants
We identified 899 incident cases of perforated diverticular disease and 8980 matched general population controls. The median age was 71.3 years (Interquartile range 62.1–78.9 years). The age and sex distribution of the cases and controls are shown in table 1. The cases were more likely to be ever smokers, obese (BMI ≥30 kg/m2) and have a greater comorbidity score before perforation than the controls (table 2). The cases were also more likely to have had a consultation for abdominal pain in the 6 months before the episode of perforation than the controls, though this was true for only a minority of cases 9% versus 0.8% controls.
Demographic characteristics of cases and controls
Univariate analysis of possible confounders
Corticosteroids
Overall, just over 20% of cases had ever used oral corticosteroids before perforation compared with just over 10% of controls. Current corticosteroid use was associated with a threefold increased risk of perforation, unadjusted OR=3.13 (95% CI 1.88 to 5.20) and this was somewhat confounded by comorbidity and smoking, giving an adjusted OR=2.74 (95% CI 1.63 to 4.61). Ever use of oral corticosteroids was also associated with approximately a 70% increased risk of perforation (adjusted OR=1.69; 95% CI 1.41 to 2.04) (table 3).
Univariate and multivariate analysis for corticosteroids and cardiovascular drugs
Analgesics
Any use of an analgesic was extremely common among our population with the majority having used either an opiate or NSAID medication. Both opiate and NSAID use was more common among cases (63% and 65% vs 47% and 52%, respectively) and compared with non-users of opiate analgesics current use was associated with a threefold increased risk of perforation (OR=2.75; 95% CI 2.03 to 3.73). This relationship was confounded moderately by comorbidity, smoking, BMI and prior recording of abdominal pain (adjusted OR=2.16; 95% CI 1.55 to 3.01). Most of the attenuation of this relationship occurred after adding abdominal pain to the model. There was also evidence of an association between ever use of opiates and perforation with an adjusted OR=of 1.88 (95% CI 1.60 to 2.21) (table 4). Stratifying by prior abdominal pain demonstrated an increase in the point estimate in those with no prior history of abdominal pain and current opiate use (adjusted OR=2.46; 95% CI 1.77 to 3.43).
Univariate and multivariate analysis for analgesics
Compared with non-users, current NSAID use conferred a small (non-statistically significant), increase in the risk of perforation, unadjusted OR=1.46 (95% CI 0.93 to 2.2.1) and this relationship did not alter substantially after adjustment for potential confounders. However, ever users of NSAIDs compared with non-users were at increased risk of perforation OR=1.62 (95% CI 1.39 to 1.90). Current use of a coxib compared with non-use did confer a risk of perforation, unadjusted OR=1.85 (95% CI 1.01 to 3.35) but this relationship was not maintained after controlling for confounding. The total number of exposures to coxibs was small being <10% in cases and 4% in controls (table 4).
Cardiovascular drugs
Overall aspirin use was more common among cases than among controls (approximately 30% vs 20%). Analysis of current users did not show a statistically significant increase in risk; however, there was an increased risk of perforation in ever users of about 40% (adjusted OR=1.37; 95% CI 1.14 to 1.66)(table 3) but this was confounded by comorbidity and smoking. Current use of a calcium antagonists showed no association with perforation, adjusted OR=0.54 (95% CI 0.24 to 1.24), although the point estimate indicated a potentially protective role. However, the adjusted results for ever use were centred around no increase or a decrease in risk. Current use of a statin was associated with a lower risk of perforation, adjusted OR=0.44 (95% CI 0.20 to 0.95) with no discernible risk increase or decrease among ever users (adjusted OR=0.98; 95% CI 0.76 to 1.27).
Interaction between comorbidity and drug use
There was evidence of interaction between comorbidity and steroid use (likelihood ratio test (LRT) p=0.03) and between comorbidity and ever use of an NSAID (LRT p=0.04) but not between comorbidity and either aspirin or opiate use (LRT p>0.05). We have therefore presented stratum-specific estimates by Charlson group for steroid and NSAID use in table 5. For both current and ever corticosteroid use there was a graded relationship across comorbidity strata. Those with no recorded serious comorbidity and a current prescription for corticosteroids had a sixfold increased risk of perforation compared with non-users, while those with two or more comorbidities had only a 1.5-fold increase in risk. Ever use of corticosteroids showed a similar trend, although not as strong. In contrast, although there was an interaction between NSAID use and comorbidity this showed no discernible trend and rather inconsistent variation through the strata.
Stratum-specific OR (95% CI) by Charlson index for steroid and NSAID use adjusted for age and gender
Discussion
Summary of findings
We found strong and consistent relationships between the current use of both opiate analgesics and oral corticosteroids and the occurrence of perforated diverticular disease. Current use of both these drugs conferred approximately a two- to threefold increase in risk of perforation compared with non-use, even after accounting for various confounding factors including indication for the medication. Ever use of both drug groups was also associated with perforation. In those people with no serious comorbidity recorded there was a particularly strong association between corticosteroid use and perforation (OR=6.45; 95% CI 2.92 to 14.21). These findings taken together with a lack of clear relationships for NSAIDs and various cardiovascular drugs suggest that current use of corticosteroids and opiate analgesics is associated with the development of perforation of the colon in diverticular disease.
Strengths and limitations
This is a large study of the risk of medication use for perforated diverticular disease and has 10 times more cases than previous studies. This has enabled us to determine precisely associations with various drugs while taking account of important confounding factors. The accuracy of prescribing data recorded in the GPRD allows for correct ascertainment of prescription drugs at the time immediately before perforation (and previous use), which is crucial in establishing a putative causal relationship.15 In addition, as the recording of both exposure and confounders was carried out concurrently within routine medical care there is no danger of recall bias affecting the results. Both the cases and controls are general population-based being from a widely, geographically dispersed population drawn from primary care in the UK so these findings are widely generalisable. The strengths of primary care data have also allowed us to examine (at least partially) confounding by indication in the case of analgesics by evaluation of prior consultations for abdominal pain.
A possible limitation of the study when analysing data with respect to some opiate analgesics and NSAIDs is that some of these drugs are available as over-the-counter drugs in the UK. Previous studies have identified an increased risk of lower gastrointestinal events with current use of NSAIDs with possible biological explanations for their effect being alterations in mucosal integrity.9 16 17 Our study has not shown an increased risk of perforation in those currently using an NSAID but ever use was associated. One reason therefore that we may not have observed a relationship is that we have studied the less frequent event of perforation.16 17 Another may be because over-the-counter use is not captured in the data we used and that ever use is a marker of this, perhaps indicating that there may be a relationship with this drug class. However, the majority of the patients studied in this analysis were over the age of 65 when perforation occurred and would be eligible at this point for free prescriptions. We have conducted a post hoc power calculation for the current NSAID exposure and estimate that with a sample size of 899 with 2% of controls exposed that with an α value of 5% we had 46% power to detect an OR of 1.5. It is also possible that the association between current NSAID use and perforation is an underestimate owing to the reduction in NSAID prescriptions which occurred over the study period because of the increased use of coxibs.18 The concept of ‘over-matching’ cannot also be discounted.11 This is the situation where practices in which both cases and controls reside have a common approach to prescribing that reduces observed variation and leads to underestimates of any associations. In the case of corticosteroids we may not necessarily have taken account of comorbidity that is serious enough to require prescription of these drugs yet does not form part of the Charlson index—that is, it may well be that the Charlson index failed to capture completely the conditions in which steroids are prescribed to in this group.
The diagnosis of perforated diverticular disease in the GPRD has previously been validated. In this study there may be under recording of incident cases of perforation by up to 20%; however, the diagnosis used within the GPRD is highly specific with 90% of those with case notes having an accurate diagnosis. This under recording will have led to a decrease in the power of the study to show an association with certain drug exposures but is unlikely to have had an effect on the magnitude of the estimates themselves or to have led to a particular bias.
Interpretation
Associations between diverticular perforation and use of NSAIDs, opiate analgesics and steroids have been previously reported.3 5–7 19 These studies have reported higher ORs than found in our study but owing to the use of hospital-based controls and their inability to control for confounding factors, it is likely that they have overestimated the risk associated with these drugs. Morris et al reported a six- to eightfold increase in perforation associated with steroid use and a two- to threefold increase associated with use of opiate analgesics dependent on the control group they used—either ophthalmology or dermatology controls. Clearly, the use of two hospital-based control groups has resulted in a wide variation in the estimate of association in this study as well as the use of only 120 cases.3 Further studies by Wilson et al and Campbell and Steele did not calculate measures of association—that is, ORs to quantify the excess risk of perforation.5 6 Also, none of the previous studies have recorded either prescription information or symptoms for a long period before perforation as we did from available data and so provide less evidence of a temporal relationship. The total number of exposures to coxibs was small; however, both current and ever use were associated with an increased risk of perforation. We have not reproduced the finding that calcium antagonists confer a reduction in risk of perforation4; however, our findings for statins of a 50% reduction in risk, albeit only in current users, is consistent with contemporary views of this class of drugs as potential anti-inflammatory agents and the observation of a reduction in risk of sepsis among statin users.20
Smoking, obesity and comorbidity have previously been associated with an increased risk of symptomatic diverticular disease and acute diverticulitis; however, no study has previously reported on the interaction between them and their role in perforation. Smoking has been shown in a small retrospective study to be associated with hospitalisation and complicated diverticular disease21 but ours is the first report of an association between smoking and perforation with a 50% increase in the risk of diverticular perforation. Increased BMI and obesity have been shown to be associated with increased episodes of hospitalisation in patients with diverticular disease and also in promoting symptomatic diverticular disease.22 23 Our findings of a higher risk with increasing BMI fit with these observations.
Clinical significance
Overall, we found marked increases in risk of perforated diverticular disease among current users of opiate and corticosteroid drugs. These drugs are commonly prescribed in the population aged over 60 years and constitute important modifiable risk factors for this life-threatening disease. It seems reasonable to suggest that clinicians should avoid prescribing opiates and corticosteroids for people with diverticular disease, where possible.
References
Footnotes
Funding JW and KMF are funded by a National Institute of Health Research/Department of Health Clinician Scientist Fellowship for JW. The dataset was acquired through the Medical Research Council scheme for academic use of General Practice Research Database data. The validation study was funded in part by the Mason Medical Foundation and Institute of Clinical Research at Nottingham University. The funders had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; and preparation, review, or approval of the manuscript.
Competing interests None.
Ethics approval This study was conducted with the approval of the Independent Scientific Advisory Committee approval board (ISAC) who provide scientific advice to the MHRA on study design and advise if further approval is required from the Multi-centre Research Ethics Committee outside the MHRA's current approval for observational studies.
Provenance and peer review Not commissioned; externally peer reviewed.