Inhaled Long-acting Anticholinergics and Urinary Tract Infection in Individuals with COPD

ABSTRACT

Inhaled, long-acting anticholinergic medication (LAA), commonly used for moderate-to-severe chronic obstructive pulmonary disease (COPD), has been shown to decrease COPD hospitalizations, emergency department visits, and acute exacerbations but has also been associated with urinary tract infection (UTI) in a prior meta-analysis. The objective of this study was to verify if there was an association between LAA and UTI in older individuals with COPD. A population-based, real-world cohort study using health administrative data from Ontario, Canada was conducted. Incidence of UTI was compared between older people with physician-diagnosed COPD, who were new users of inhaled long-acting anticholinergics and new users of inhaled corticosteroids–a reference medication used in similar clinical settings that has no known association with UTI. Propensity score matching was used to minimize the effects of confounding. An overall association between LAA and various measures of UTI in older individuals was not found. However, in a priori defined stratified analyses, men newly initiated on LAA were 75% more likely to develop a UTI than men newly started on an inhaled corticosteroid (hazard ratio 1.75; 95% confidence interval 1.05–2.92). No significant association was seen in women. In conclusion, older men with COPD newly started on LAA are at increased risk of UTI. Men considering an inhaled LAA should be informed of this risk and, if they decide to take it, be provided with appropriate monitoring.

KEYWORDS:

• Medication
• adverse events
• infection
• health outcomes

Introduction

Chronic obstructive pulmonary disease (COPD) is the third leading cause of death worldwide and effects over 10% of adults over the age of 40 (1–3). Inhaled, long-acting anticholinergics (LAA), commonly used for first-line treatment for moderate-to-severe COPD, have been shown to decrease COPD hospitalizations, emergency department (ED) visits, and acute exacerbations, as well as improve quality of life (4). They work by blocking muscarinic receptors in the lung, relaxing airway smooth muscle and decreasing airflow obstruction (5, 6). Their effects are also experienced systemically. In a meta-analysis, Barr et al. found LAA to be associated with increased odds of urinary tract infection (UTI; odd ratio [OR] 1.6, 95% confidence interval 1.03–2.6) (7). To the best of our knowledge, this finding has not been confirmed in other studies. UTIs are the third most burdensome infectious disease syndrome and occur due to many reasons (8). Thus, an association between UTIs and LAA in everyday clinical practice might be unnoticed by physicians and patients alike.

We conducted this study to determine if there was an association between LAA and UTI. We also sought to determine if such an association was greater in men because previous studies have shown that men (but not women) with COPD using inhaled anticholinergics were more likely to experience acute urinary retention requiring medical attention (9, 10). To provide a point of reference and minimize unmeasured confounding, we compared LAA to inhaled corticosteroids (ICS), a class of medication used in similar clinical settings and that, to the best of our knowledge, is not associated with increased risk of UTI (11).

Methods

Study design and setting

A population-based, cohort study using health administrative data from Ontario, the largest province of Canada with a diverse, multicultural population of approximately 13 million, during 2003–2011, was conducted.

Ethics approval was obtained from the Sunnybrook Health Sciences Centre Research Ethics Board, which included a waiver of informed consent.

Data sources

Residents of Ontario have universal public health insurance where the province is the single payer for all medically necessary services across the full spectrum of residents, providers and hospitals. Service details are captured in health administrative databases, which can be linked on an individual level to provide a complete health services profile for each resident. The Registered Persons Database contains basic demographic information and date of death. The Ontario Drug Benefit Program database contains prescription claim records for all residents aged 65 years and older. These publicly funded prescriptions are subject to a small, means-tested co-payment, which does not affect the rate at which prescriptions are filled (12). The Canadian Institute of Health Information (CIHI) Discharge Abstract Database contains information on all discharges from acute care hospitals and the National Ambulatory Care Reporting System database includes data on same-day surgical procedures and ED visits. The Ontario Health Insurance Plan Physician Claims database contains information about all services provided by fee-for-service physician payments and “shadow-billings” for physicians paid under alternate payment plans. These datasets were linked using unique, encoded identifiers and analyzed at the Institute for Clinical Evaluative Sciences (ICES) (13).

Study population

Individuals who met a previously validated case definition of physician-diagnosed COPD based on health administrative data, were 66 years or older, and had been newly prescribed either an LAA or ICS between September, 2003 (when LAA were first commonly used in Ontario) and March 31, 2011 were included (14). The only LAA available in Ontario during the study period was tiotropium (Spiriva). The COPD case definition of one COPD ambulatory care visit and/or one COPD hospitalization has been previously shown to have a sensitivity of 85.0% and a specificity of 78.4% when compared to real-world clinical evaluation by a physician (14). Details and examples of its use can be found elsewhere (15, 16). The study was restricted to those aged 66 years and older to allow a one-year look back for previous medication use in order to identify new users. If individuals initiated LAA at more than one time during the study period, only the first was considered. New users were studied because individuals were presumably tolerating their medication if they continued to take it. Individuals were excluded if (i) they had initiated both medications on the same day; (ii) they were receiving palliative health services; or (iii) they were ineligible for health insurance.

Exposures

All new users of an LAA, the exposure of interest, were compared to all new users of ICS (fluticasone, budesonide, mometasone). ICS could have been prescribed alone or in a combination product with long-acting beta-agonists. Short-acting anticholinergics were not considered because their non-sustained effects would be less likely to lead to UTI and because they can be prescribed on an as-needed basis (17). The date the prescription was filled served as the index date. ICS were chosen as the reference medication because they are used in similar circumstances as LAA and they have not been associated with UTI or alterations in urinary tract function (11).

Outcomes

The primary outcome was a UTI within 30 days of the index date. A UTI was defined as an ambulatory care visit for UTI accompanied by filling a prescription for a systemic antibiotic within seven days (eTable. 1). Visits for UTI were identified using previously published UTI health administrative codes (8) (eTable. 2). A period of 30 days was chosen because LAA takes effect shortly after being started. If more than one UTI occurred, only the first was considered. Secondary outcomes were a UTI ambulatory care visits alone and a UTI ED visit. In secondary analyses, outcomes were examined at 180 days to determine if UTI was associated with longer-term LAA use.

Baseline characteristics

Demographic, COPD-related, co-morbidity, medication and UTI-related covariates were obtained from the health administrative data. COPD severity was determined using information from hospitalizations, ED visits, ambulatory care visits and prescription drug use. Socioeconomic status was inferred from neighborhood income derived from postal code and census data (18, 19). Overall co-morbidity was based on diagnostic information from health services records within two years prior to the index date and characterized using the Johns Hopkins Adjusted Clinical Groups Case-Mix System (20, 21). Anticholinergic burden was measured by summing the anticholinergic properties of all prescription medications the patients were receiving and expressed using an anticholinergic drug scale previously validated by Carnahan et al. (22).

Propensity score matching and analysis

All analyses were stratified according to place of residence (long-term care vs. community-dwelling) because these populations may have a different predisposition to UTI. For each cohort, propensity score matching was used to compare patients with similar observed characteristics, all of whom were potential candidates for LAA or ICS (23, 24). A multivariable logistic regression model was used to estimate the probability of receiving an LAA (as opposed to an ICS) using all the measured baseline characteristics as predictors. Within each of the two strata, individuals prescribed an LAA were matched 1:1 with individuals prescribed an ICS on age (+/−1 year), sex, index year, history of UTI, anticholinergic drug burden (low/high ) and the logit of the propensity score (+/−0.2 standard deviations) (25). Standardized differences between the groups of less than 10% were accepted as indicative of adequate balance (26).

Patients were followed until their first UTI, their death, or 30 days after their index date (180 days for some of the secondary analyses) at which time they were censored. Thirty days was chosen because the pharmacologic properties of LAA are expected to take effect shortly after they are started. To quantify the effect of an exposure on the risk of an outcome, a Cox proportional hazards model was fit in the matched sample, with exposure group as the sole predictor variable. A robust, sandwich-type variance estimator was used to account for the matched nature of the sample (27). Due to our interest in etiology and biologic association, we focused on estimating cause-specific models when accounting for the competing risk of death, rather than using subdistribution hazard models, which are better suited for questions of risk prediction and prognosis (28, 29). We determined the absolute difference in risk due to inhaled LAA compared to ICS, and their 95% confidence intervals and calculated numbers needed to harm (NNH). All statistical tests were 2-sided with statistical significance defined as a p value <0.05. Analyses were performed using SAS, version 9.3 (SAS Institute, Cary, North Carolina). Patients with missing data, which was minimal, were excluded from the analysis.

Subgroup analyses

The association of LAA compared to ICS on UTI was examined separately within subgroups of a priori interest. Sex was examined because of the different susceptibility of women and men to acute urinary retention as observed in previous studies (9, 10). History of previous UTI was considered because it is a risk factor for UTI. Anticholinergic drug burden (low or <2/high or ≥2) was considered because other medications may accentuate the anticholinergic effects of LAA (22).

Sensitivity analysis

A sensitivity analysis was conducted to examine the magnitude of the effects an unmeasured confounding variable would require on both exposure to LAA and on developing a UTI to nullify the statistically significant association observed. More details can be found in the online supplement (eText 1) and elsewhere (30).

Ethics approval

Ethics approval was obtained from the Sunnybrook Health Sciences Centre Research Ethics Board. A waiver of informed consent was obtained.

Results

Study population

Of 71,428 eligible people with physician-diagnosed COPD who were new users of LAA or ICS, 5140 (7.2%) who started both medications on the same date, 1474 (2.1%) who were receiving palliative health services and 531 (0.7%) who were ineligible for health insurance were excluded, leaving 64,283 for analysis: 59,470 in the community and 4,813 in the nursing home groups. In the community group, compared to new users of ICS, new users of LAA were more likely to be male, have received spirometry and have a shorter duration of COPD. They were also less likely to have asthma. About 2.8% and 1.9% died in the LAA and ICS groups, respectively (Table 1^{Table 1}). In the nursing home group, compared with new users of ICS, new users of LAA were also more likely to be male and have received spirometry. They also had less primary care ambulatory care visits in the previous year. There were 18.0% and 17.9% who died in the LAA and ICS groups, respectively (eTable 3).

Table 1. Baseline characteristics of new users of inhaled long-acting anticholinergic and inhaled corticosteroid medications before and after propensity score matching among people living in the community.

	Before matching			After matching
	Inhaled long-acting	Inhaled	Standardized	Inhaled long-acting	Inhaled	Standardized
Characteristic	anticholinergic	corticosteroid	difference	anticholinergic	corticosteroid	difference
N	23,200	36,270		18,774	18,774
Demographic characteristics
Age (Mean ± SD)	75.5 ± 6.8	74.2 ± 7.0	0.18	74.9 ± 6.6	74.9 ± 6.7	0
Female (%)	45.0	53.1	0.16	46.4	46.4	0
Socioeconomic status (%)
1 (lowest)	23.2	22.6	0.01	23.1	22.9	0
2	22.1	21.6	0.01	22.1	22.3	0
3	19.6	19.6	0	19.6	19.6	0
4	18.3	18.9	0.02	18.5	18.3	0.01
5 (highest)	16.8	17.3	0.01	16.7	16.9	0.01
Immigrant (%)	6.2	6.8	0.02	6.4	6.5	0
Urban residence (%)	80.8	83.7	0.08	81.7	81.7	0
Year of index drug claim (%)
2003	1.0	1.4	0.03	0.8	0.8	0
2004	3.5	3.8	0.02	3.4	3.4	0
2005	5.0	5.3	0.01	4.9	4.9	0
2006	7.3	7.3	0	7.4	7.4	0
2007	10.2	9.1	0.04	10.1	10.1	0
2008	11.9	12.0	0	12.1	12.1	0
2009	16.0	17.2	0.03	16.8	16.8	0
2010	45.2	43.9	0.03	44.5	44.5	0
Region of province (%)
1	8.1	6.8	0.05	7.6	7.8	0.01
2	10.1	8.3	0.06	9.4	9.5	0
3	5.3	4.4	0.04	5.0	5.0	0
4	12.1	14.1	0.06	12.8	12.8	0
5	2.6	4.1	0.08	3.0	2.8	0.01
6	5.6	5.9	0.01	5.7	5.8	0
7	6.0	6.9	0.04	6.2	6.2	0
8	7.2	9.7	0.09	7.7	7.6	0
9	12.2	12.3	0	12.4	12.6	0
10	5.7	4.9	0.04	5.4	5.4	0
11	9.2	8.8	0.01	9.2	9.1	0
12	5.1	3.9	0.06	4.7	4.5	0.01
13	8.0	7.4	0.02	8.1	8.1	0
14	2.7	2.5	0.01	2.8	2.7	0.01
COPD characteristics
COPD duration (%)
1 year or less	28.4	21.6	0.16	25.8	25.6	0
1 to 5yrs	38.0	35.3	0.06	37.3	38.4	0.02
over 5yrs	33.6	43.1	0.2	36.9	36.0	0.02
Previous spirometry (%)	58.6	46.9	0.24	55	56	0.02
Number of primary care visits in previous year, Median (interquartile range)	6 (3–10)	7 (4–10)	0.08	6 (4–10)	6 (4–10)	0.02
COPD Specialist visit^a (%)	65.0	65.2	0	65.1	65.7	0.01
Influenza vaccination (%)	57.2	57.0	0	56.8	56.9	0
Previous or co-existing medical conditions (%)
Asthma	16.3	27.7	0.27	18.2	18.2	0
Lung cancer	3.0	2.2	0.05	2.7	2.7	0
Heart failure	24.0	20.7	0.08	22.9	22.6	0.01
Arrhythmia	9.5	8.3	0.04	9.2	9.1	0
Coronary artery disease	13.0	11.4	0.05	12.7	12.8	0
Stroke	2.9	2.5	0.03	2.8	2.6	0.01
Hypertension	74.9	74.9	0	74.6	74.8	0
Non-lung cancer	9.6	9.0	0.02	9.0	9.3	0.01
Diabetes	28.8	30.9	0.05	29.7	29.8	0
Dementia	9.6	8.9	0.02	9.2	9.0	0.01
End-stage renal disease	5.0	4.3	0.03	4.7	4.6	0
Osteoporosis	6.2	5.6	0.03	5.9	5.7	0.01
Mental health encounter in previous year (%)
Hospitalizations	0.6	0.7	0.01	0.6	0.6	0
Ambulatory care visits	5.0	5.8	0.04	5.2	5.1	0
None	94.4	93.5	0.04	94.2	94.3	0
Overall level of co-morbidity (%)
Low	21.0	18.5	0.06	20.7	20.1	0.02
Moderate	39.3	39.6	0.01	39.9	40.2	0.01
High	39.7	41.9	0.04	39.4	39.7	0.01
(Continued on next page)

Table 1. (Continued)

Medication use in previous year
Number of courses of respiratory antibiotics (%)
None	48.6	41.9	0.13	47.6	47.4	0
1	24.9	26.3	0.03	25.5	25.5	0
2 or more	26.5	31.7	0.11	27.0	27.1	0
Number of courses of oral corticosteroids (%)
None	86.1	86.6	0.02	86.1	86.0	0
1	7.5	7.5	0	7.4	7.5	0
2 or more	6.4	5.8	0.02	6.5	6.5	0
Short-acting inhaled anticholinergics (%)	8.7	8.2	0.02	8.6	8.4	0.01
Short-acting inhaled beta-agonists (%)	40.2	38.6	0.03	39.4	39.7	0
Long-acting inhaled beta-agonists (%)	22.3	31.2	0.2	24.7	25.9	0.03
Methylxanthines (%)	0.8	0.7	0	0.8	0.8	0
Alpha-blockers (%)	4.8	4.2	0.03	4.4	4.5	0
Dutasteride (%)	1.6	1.5	0	1.5	1.5	0
Finasteride (%)	2.2	1.9	0.02	2.0	2.0	0
Number of different medications used, median (interquartile range)	4 (3–6)	4 (3–6)	0.03	4 (3–6)	4 (3–6)	0.01
Anticholinergic drug scale (20) (%)
0	28.0	27.9	0	27.5	28.2	0.02
1	21.9	21.9	0	21.5	20.8	0.02
2	15.9	15.7	0.01	16.4	16.2	0.01
3 or more	34.2	34.5	0.01	34.6	34.8	0
Acute events before index date
Most recent hospitalization for COPD (%)
In the past 6 months	6.9	4.5	0.11	5.9	5.5	0.02
Between 6 months and 5 years	10.2	7.1	0.11	9.3	9.3	0
Over 5 years or never	82.9	88.4	0.16	84.8	85.2	0.01
Most recent ED visit for COPD (%)
In the past 6 months	4.2	3.4	0.04	3.9	4.1	0.01
Between 6 months and 5 years	6.9	5.8	0.05	6.7	7.0	0.01
Over 5 years or never	88.9	90.8	0.06	89.3	89.0	0.01
Most recent hospitalization for COPD-related condition (%)
In the past 6 months	4.2	3.2	0.05	3.8	3.5	0.01
Between 6 months and 5 years	8.0	6.8	0.05	7.6	7.4	0.01
Over 5 years or never	87.8	90.0	0.07	88.6	89.1	0.02
Most recent ED visit for COPD-related condition
In the past 6 months	2.8	3.7	0.05	3.0	3.0	0
Between 6 months and 5 years	8.5	9.1	0.02	8.7	8.8	0
Over 5 years or never	88.7	87.1	0.05	88.3	88.3	0
Most recent hospitalization for any other reason
In the past 6 months	6.7	6.2	0.02	6.6	6.6	0
Between 6 months and 5 years	19.4	19.7	0.01	19.6	19.9	0.01
Over 5 years or never	73.8	74.1	0.01	73.8	73.5	0.01
Most recent ED visit for any other reason
In the past 6 months	14.6	15.0	0.01	14.4	14.5	0
Between 6 months and 5 years	35.3	36.7	0.03	36.0	35.5	0.01
Over 5 years or never	50.1	48.3	0.04	49.6	50.0	0.01
Other risk factors for urinary tract infections (%)
Urinary tract infections in previous year
Ambulatory	5.0	5.4	0.02	3.3	3.3	0
Requiring hospitalization or ED visit	4.5	4.0	0.02	2.7	2.7	0
Sexually transmitted disease	0.1	0.1	0	0.1	0.1	0
Antibiotic use in previous 90 days	28.2	34.1	0.13	29.0	29.3	0.01
Benign prostatic hyperplasia	3.6	3.8	0.01	3.2	3.2	0
HIV/AIDS	0.1	0.1	0	0.1	0.1	0
Nephrotic syndrome	0.0	0.1	0.01	0.0	0.0	0.01
Renal calculi	1.0	1.0	0	0.9	0.9	0
Inflammatory bowel disease	0.8	0.7	0.01	0.7	0.8	0
Splenectomy (since 1991)	0.2	0.2	0	0.2	0.2	0
Spinal cord injury	0.0	0.0	0	0.0	0.0	0.01
Multiple sclerosis	0.1	0.1	0.01	0.1	0.1	0
Paraplegia or quadriplegia	0.1	0.1	0.01	0.1	0.1	0
Previous organ transplant (except bone, cornea, and skin)	0.1	0.2	0.01	0.2	0.2	0
Urinary retention in previous year	1.5	1.1	0.03	1.1	1.1	0
Foley catheter in previous 90 days	0.6	0.6	0.01	0.6	0.6	0.01
Urogenital surgery in previous 3 months	2.2	2.1	0.01	1.9	1.9	0
General anesthetic in previous 3 months	23.2	22.5	0.02	22.8	23.1	0.01
Diabetic nephropathy	1.3	1.0	0.03	1.2	1.1	0.01

^aa COPD specialist was a pulmonologist, a general internist or a geriatrician

Propensity score matching

Matching produced 18,774 matched pairs in the community group and 962 matched pairs in the nursing home group. About 4357 (18.8%) and 1,135 (54.1%) patients newly started on LAA in the community and nursing home groups, respectively, were not matched. Standardized differences for all baseline characteristics were less than 10%, indicating an acceptable balance. (Tables 1, eTable 3)

In the propensity-score-matched community cohort, there was no statistically significant association between LAA exposure compared to ICS exposure and the primary outcome (adjusted hazard ratio (HR), 1.12; 95% confidence interval (CI), 0.81–1.56). Similar non-statistically significant effects were seen for secondary outcomes (Table 2). The HRs for the nursing home group were similarly non-significant as were the results at 180 days (data not shown).

Table 2. Associations of health outcomes in new users of inhaled, long-acting anticholinergics compared with new users of inhaled corticosteroids in the matched community group.

	Inhaled, long-acting	Inhaled corticosteroid		Propensity-score-matched regression
	anticholinergic cohort	cohort		Hazards ratio^a (95%
Outcomes	had outcome (%)	had outcome (%)	Absolute difference, %	confidence interval)
Outcomes at 30 days
Urinary tract infection ambulatory care visit followed by a prescription for an antibiotic	74 (0.39)	66 (0.35)	0.04	1.12 (0.81–1.56)
Urinary tract infection ambulatory care visit	110 (0.59)	87 (0.46)	0.13	1.27 (0.96–1.68)
Urinary tract infection emergency department visit	51 (0.27)	42 (0.22)	0.05	1.22 (0.81–1.83)
Outcomes at 180 days
Urinary tract infection ambulatory care visit followed by a prescription for an antibiotic	453 (2.42)	435 (2.32)	0.10	1.06 (0.93–1.21)
Urinary tract infection ambulatory care visit	493 (2.63)	483 (2.58)	0.05	1.04 (0.92–1.17)
Urinary tract infection emergency department visit	251 (1.34)	237 (1.26)	0.06	1.08 (0.90–1.28)

^aReflects the risk of outcome in new users of inhaled, long-acting anticholinergics compared with new users of inhaled corticosteroids.

Subgroup analyses

Stratification was not performed in the nursing home group because there were too few patients in each stratum for reliable analysis. In the community cohort, men who received an LAA were 75% more likely to have a UTI than men in the ICS cohort (adjusted HR 1.75, 95% CI 1.05–2.92; absolute risk increase 0.17% at 30 days; NNH 588). Hazards for ambulatory care and emergency department visits were also increased (Table 3). Non-statistically significant higher rates were observed in individuals with a history of UTI (adjusted HR 1.27, 95% CI 0.66–2.46) and with a high anticholinergic drug burden (adjusted HR 1.35, 95% CI 0.89–2.04) (eTable 5).

Table 3. Stratified associations of having an ambulatory care visit for a urinary tract infection ambulatory care visit followed by a prescription for an antibiotic in new users of inhaled, long-acting anticholinergics compared with new users of inhaled corticosteroids in the community group.

			Inhaled, long-acting	Inhaled corticosteroid		Propensity-score-matched
			anticholinergic cohort	cohort		regression Hazards ratio^a
Stratifying characteristics	Strata	n	had outcome (%)	had outcome (%)	Absolute difference, %	(95% Confidence interval)
Outcomes at 30 days
Sex	Male	20,134	40 (0.40)	23 (0.23)	0.17	1.75 (1.05–2.92)
	Female	17,414	34 (0.39)	43 (0.49)	0.10	0.79 (0.51–1.23)
Previous urinary tract infection	Yes	2,252	19 (1.69)	15 (1.33)	0.36	1.27 (0.66–2.46)
	No	35,296	55 (0.31)	51 (0.29)	0.02	1.08 (0.74–1.58)
Anticholinergic burden	High	19,166	51 (0.53)	38 (0.40)	0.13	1.35 (0.89–2.04)
	Low	18,382	23 (0.25)	28 (0.30)	0.05	0.82 (0.47–1.43)
Outcomes at 180 days
Sex	Male	20,134	217 (1.08)	188 (0.93)	0.15	1.18 (0.98–1.43)
	Female	17,414	236 (1.36)	247 (1.42)	0.04	0.97 (0.81–1.15)
Previous urinary tract infection	Yes	2,252	117 (5.20)	127 (5.64)	0.44	0.94 (0.74–1.19)
	No	35,296	336 (0.95)	308 (0.87)	0.08	1.11 (0.95–1.29)
Anticholinergic burden	High	19,166	269 (1.40)	262 (1.37)	0.03	1.04 (0.89–2.04)
	Low	18,382	184 (1.00)	173 (0.94)	0.06	1.08 (0.88–1.32)

^aReflects the risk of outcome in new users of inhaled, long-acting anticholinergics compared with the new users of inhaled corticosteroids.

Sensitivity analysis

Even if an unmeasured confounder increased the odds of exposure to LAA by 50% and the odds of developing a UTI by 60%, the observed association would still be statistically significant in men (etext 1).

Discussion

We conducted a population-based study of adults aged 66 or older with COPD and found that men, but not women, initiated on LAA were 75% more likely to have a UTI than men initiated on ICS although the absolute increase was low. To the best of our knowledge this is the first large, population study to examine the association of LAA and UTI and its results should be confirmed in prospective clinical studies. However, until that is done, we believe that men considering LAA should be informed of this potential side effect and provided with appropriate monitoring should they choose to take it.

Our results in men are consistent with a previous meta-analysis of randomized controlled trials designed to study both sexes, which revealed that patients taking tiotropium were more likely to experience a UTI (OR 1.6) (95% confidence interval 1.03 to 2.6) (7). This meta-analysis, however, in actuality predominantly studied men, which is why its results might have been so similar to our results in this sex. Our results in men are also consistent with previous studies demonstrating a statistically significant increased risk of acute urinary retention associated with inhaled LAA in men and literature that concedes that urinary retention due to oral anticholinergic medications likely leads to UTI (9, 10, 31). Our overall non-significant results are consistent with a large randomized controlled trial that did not find an increase in UTI in all subjects taking LAA compared to placebo (4). This study, however, did not stratify patients experiencing adverse events by sex.

It has been hypothesized that anticholinergics inhibit contraction of the detrusor muscle, which causes partial urinary obstruction and urinary stasis predisposing to UTI. This mechanism is likely compounded in older men by their predisposition to bladder outlet obstruction from, sometimes undiagnosed, benign prostatic hyperplasia. Women do not have this difficulty.

While a 75% increased relative risk of UTI in men initiating LAA is concerning, the low increased absolute risk and the fact that more than 580 individuals would need to be treated with LAA to cause one UTI is worth noting. In addition, the insignificant findings at 180 days suggest that the risk of UTI is transient. Statistically, one might question these findings as being spurious; however, they are consistent with previous clinical studies (7, 9).

Much higher rates of UTI in the population have been documented by other studies and there are a few reasons our results likely undercount UTI and underestimate the true impact of LAA (8). First, to ensure we were truly measuring UTI, we chose a definition with high specificity. However, to do this we had to compromise sensitivity. Thus, our definition likely missed many UTIs caused by LAA. As evidence of this, we found a similar number of ambulatory care visits for UTI as ED visits for UTI when clinical experience dictates that there should have been many more. Undercounting of UTI could have also occurred because physicians in Ontario can only enter one diagnostic code with each ambulatory care visit claim and they chose patients' COPD or another co-morbidity instead of UTI (32). Second, our results might have underestimated the true impact of LAA because we only considered patients' first UTI in the study period and few patients had had a UTI in the previous year. The impact of LAA on recurrent UTI was not measured. Finally, we both included current users of short-acting anticholinergics, who–by virtue of the fact they were voluntarily taking their medication–were more likely to tolerate LAA, and did not exclude previous users of short-acting anticholinergics, who might not have started LAA because of bad experiences with UTI in the past. These factors could have led us to detect fewer UTI than actually occurred in the real world.

We found trends toward, but not significant, associations between LAA and UTI, in people with a history of UTI and people who were taking other anticholinergic medications. The lack of significance was likely because our study was underpowered (as suggested by the wide confidence intervals) and because we could not divide these subgroups by gender because of small numbers. Future studies should target these, and other, potential high-risk groups.

The strengths of our study were its real-world population base, its comparison of new use of LAA to new use of another inhaled COPD medication in similar circumstances, its excellent follow-up, and its large numbers that allowed for the adjustment of many covariates and examination of subgroups. It also had limitations that merit emphasis. First, we did not have a validated ambulatory care case definition for UTI, which might have led to underestimation of this outcome, as discussed above. Second, we cannot exclude the possibility that an unmeasured confounder (or groups of confounders) was responsible for the results. However, such a confounder would have to have increased the odds of exposure to LAA by 50% and the odds of developing a UTI by 60% and have no correlation with any of the other variables (or arguably be present in women) to negate the findings seen. Further, such a confounder was not likely to be a result of indication bias as we used a control medication (ICS) that was used in similar circumstances as LAA. Third, it was possible that people with COPD were misclassified. However, because the mechanism of action by which LAA are believed to cause UTI is not unique to people with COPD, there is no reason to believe the effects of LAA on the urinary tract would be different in people with asthma, heart failure, or other diseases that might be misclassified as COPD. This is not likely to have biased the results. Fourth, future studies will be required to see if the increased risk observed also applies to inhaled short-acting anticholinergics and newer LAA like glycopyrronium.

In summary, we found that new use of LAA was associated with a 75% increased risk of UTI in older men with COPD although the absolute increased risk was low. This finding should be confirmed prospectively in clinical populations. However, in the meantime, patients and their physicians contemplating initiating these medications should be informed of this potential adverse event so they can take appropriate measures.

Acknowledgments

We thank Brogan Inc., Ottawa for permission to use their Drug Product and Therapeutic Class Database. Parts of this material are based on data and information compiled and provided by CIHI. However, the analyses, conclusions, opinions, and statements expressed herein are those of the author, and not necessarily those of CIHI.

Declaration of interest

Dr. Bell serves as a Medical Consultant to the Health Quality Branch of the Ontario Ministry of Health and Long-Term Care, Ontario, Canada. Otherwise, the authors have received no support from any organization for the submitted work; have no financial relationships with any organizations that might have an interest in the submitted work in the previous three years, and have no other relationships or activities that could appear to have influenced the submitted work.

Author contributors

All authors participated in the design and interpretation of the data and critical revision of the manuscript, all approved the final version, and all agree to be accountable for the aspects of the work he or she has completed. Additionally, ASG conceived of the study. ASG, PR acquired the health administrative data. AN and PA carried out the statistical analysis. ASG drafted the manuscript. PR obtained funding.

Funding

This study was funded by a Team Grant (OTG-88591) from the Canadian Institutes of Health Research Institute of Nutrition, Metabolism and Diabetes. The study was also supported by the ICES, which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC). The opinions, results and conclusions reported in this paper are those of the authors and are independent from the funding sources. The funders had no role in the study design, or in the collection, analysis, or in interpretation of data, and preparation, review, or approval of the manuscript. No endorsement by ICES or the Ontario MOHLTC is intended or should be inferred. Dr. Gershon is currently supported by the Physicians' Services Incorporated Foundation Graham Farquharson Fellowship in Translational Medicine and was supported by a New Investigator Award funded by team grant OTG-88591 from the Canadian Institutes of Health Research Institute of Nutrition, Metabolism and Diabetes while working on this study. Dr. Bell is supported by a Canadian Patient Safety Institute/Canadian Institutes of Health Research chair in patient safety and continuity of care. Dr. Austin is supported by a career investigator award from the Heart and Stroke Foundation.