Genetic and pharmacological relationship between P-glycoprotein and increased cardiovascular risk associated with clarithromycin prescription: An epidemiological and genomic population-based cohort study in Scotland, UK.

BACKGROUND: There are conflicting reports regarding the association of the macrolide antibiotic clarithromycin with cardiovascular (CV) events. A possible explanation may be that this risk is partly mediated through drug-drug interactions and only evident in at-risk populations. To the best of our knowledge, no studies have examined whether this association might be mediated via P-glycoprotein (P-gp), a major pathway for clarithromycin metabolism. The aim of this study was to examine CV risk following prescription of clarithromycin versus amoxicillin and in particular, the association with P-gp, a major pathway for clarithromycin metabolism. METHODS AND
FINDINGS: We conducted an observational cohort study of patients prescribed clarithromycin or amoxicillin in the community in Tayside, Scotland (population approximately 400,000) between 1 January 2004 and 31 December 2014 and a genomic observational cohort study evaluating genotyped patients from the Genetics of Diabetes Audit and Research Tayside Scotland (GoDARTS) study, a longitudinal cohort study of 18,306 individuals with and without type 2 diabetes recruited between 1 December 1988 and 31 December 2015. Two single-nucleotide polymorphisms associated with P-gp activity were evaluated (rs1045642 and rs1128503 -AA genotype associated with lowest P-gp activity). The primary outcome for both analyses was CV hospitalization following prescription of clarithromycin versus amoxicillin at 0-14 days, 15-30 days, and 30 days to 1 year. In the observational cohort study, we calculated hazard ratios (HRs) adjusted for likelihood of receiving clarithromycin using inverse proportion of treatment weighting as a covariate, whereas in the pharmacogenomic study, HRs were adjusted for age, sex, history of myocardial infarction, and history of chronic obstructive pulmonary disease. The observational cohort study included 48,026 individuals with 205,227 discrete antibiotic prescribing episodes (34,074 clarithromycin, mean age 73 years, 42% male; 171,153 amoxicillin, mean age 74 years, 45% male). Clarithromycin use was significantly associated with increased risk of CV hospitalization compared with amoxicillin at both 0-14 days (HR 1.31; 95% CI 1.17-1.46, p < 0.001) and 30 days to 1 year (HR 1.13; 95% CI 1.06-1.19, p < 0.001), with the association at 0-14 days modified by use of P-gp inhibitors or substrates (interaction p-value: 0.029). In the pharmacogenomic study (13,544 individuals with 44,618 discrete prescribing episodes [37,497 amoxicillin, mean age 63 years, 56% male; 7,121 clarithromycin, mean age 66 years, 47% male]), when prescribed clarithromycin, individuals with genetically determined lower P-gp activity had a significantly increased risk of CV hospitalization at 30 days to 1 year compared with heterozygotes or those homozygous for the non-P-gp-lowering allele (rs1045642 AA: HR 1.39, 95% CI 1.20-1.60, p < 0.001, GG/GA: HR 0.99, 95% CI 0.89-1.10, p = 0.85, interaction p-value < 0.001 and rs1128503 AA 1.41, 95% CI 1.18-1.70, p < 0.001, GG/GA: HR 1.04, 95% CI 0.95-1.14, p = 0.43, interaction p-value < 0.001). The main limitation of our study is its observational nature, meaning that we are unable to definitively determine causality.
CONCLUSIONS: In this study, we observed that the increased risk of CV events with clarithromycin compared with amoxicillin was associated with an interaction with P-glycoprotein.

Introduction

Clarithromycin is a widely prescribed macrolide antibiotic, comprising around 15% of all primary care antibiotic prescriptions in the United Kingdom, recommended for treatment of patients with lower respiratory tract infections either as monotherapy or in combination [1-3]. There has been growing concern regarding increased cardiovascular (CV) risk of clarithromycin. The Effect of Clarithromycin on Mortality and Morbidity in Patients With Ischemic Heart Disease (CLARICOR) trial in patients with high CV risk [4], designed to test the hypothesis that clarithromycin would reduce CV risk, actually found that 2 weeks of clarithromycin caused a 45% relative risk increase in CV mortality compared with placebo. These results were supported by a number of observational studies [5-8] and meta-analyses [9, 10] suggesting that clarithromycin and other macrolide antibiotics were associated with adverse outcome, not only in the short term during and after exposure, but also in the longer term after drug discontinuation, leading to a recent United States Food and Drug Administration (FDA) safety alert on the use of clarithromycin in patients with heart disease [11, 12]. Nonetheless, recently, studies in lower-risk community populations have not found this association. An emerging hypothesis is that these conflicting results suggest that the high CV risk with clarithromycin may only be present in a subset of individuals [13-16].

A potential risk modifier is concurrent medication use. Clarithromycin is both a substrate and inhibitor of both cytochrome p4503A4 (CYP3A4) [17] and permeability-glycoprotein (P-gp) [18], meaning that circulating clarithromycin levels are affected by alterations in CYP3A4 or P-gp activity [19]. Several studies have investigated the possibility of an interaction between clarithromycin, concomitant CYP3A4 medication use, and CV risk; however, no robust association has been found [6, 14, 20]. To the best of our knowledge, there have been no original research studies evaluating whether the association between clarithromycin use and CV risk is modulated via P-gp. Both animal [19] and clinical studies [21, 22] suggest that co-administration of P-gp inhibitors such as verapamil, omeprazole, nelfinavir [23], or ketoconazole with macrolide antibiotics leads to an increase in the oral bioavailability of macrolides and increased plasma levels. Further support for an interaction between P-gp inhibition and macrolides comes from genetic studies showing that patients with genetic variants associated with low P-gp activity also have higher levels of macrolides when exposed to these antibiotics [24]. The use of pharmacogenomics may help overcome some of the limitations of observational studies and help shed light on potential drug interactions.

We hypothesized that the increased CV risk associated with clarithromycin may be linked with concurrent use of P-gp inhibitors or substrates and that individuals with genotypes associated with low P-gp activity, a proxy for P-gp inhibition, would also have an increased CV risk when prescribed clarithromycin.

Methods

The study consisted of 2 parts: a traditional observational cohort study and a pharmacogenomic study.

Observational cohort study

The prospective analysis plan for this section of the study can be found in S1 Text. The study population was all approximately 400,000 residents of the Tayside region of Scotland registered with an NHS Tayside general practice at any point in the study period (2004–2014). Demographic and community prescribing data were obtained through the Health Informatics Centre (HIC), University of Dundee, which provides anonymized linked individual patient data, including prescribing of antibiotics as previously described [25]. These datasets were linked to other datasets including demographic, clinical, hospital admission, and mortality data that are linked by a unique 10-digit patient identifier (the Community Health Index number) used for all healthcare activities in Scotland. All research data are robustly anonymized and approved by the Tayside National Health Service Caldicott Guardian under an overarching ethical approval for anonymized data research using HIC. Prescribing data between 2004 and 2014 were used to identify all patients over 18 years old who were prescribed clarithromycin (alone or in combination with amoxicillin or another antibiotic) over this period. A control group of individuals prescribed amoxicillin as a sole antibiotic was also identified. A propensity-score model for clarithromycin exposure was estimated using baseline demographic and clinical covariates hypothesized to be relevant shown in Table 1. Outcome models were weighted using inverse propensity of treatment weights (IPTWs) to account for baseline differences between exposure cohorts and robust sandwich covariance estimation to account for multiple exposures for each individual.

Table 1

Baseline characteristics of observational cohort study.

	Clarithromycin	Amoxicillin	Standardized Difference
Total Number of Unique Patients	11,489	36,537
Total Number of Prescriptions	34,074	171,153
Age at Prescription (years) [mean ± SD]	73.3 ± 12.3	74.2 ± 13.2	0.070
Male	14,280 (41.9)	76,521 (44.7)	0.056
Type 2 Diabetes	5,555 (16.3)	27,942 (16.3)	<0.001
Chronic Obstructive Pulmonary Disease	8,647 (25.4)	25,282 (14.8)	0.267
Prior Myocardial Infarction	1,289 (3.8)	7,888 (4.6)	0.040
Prior Heart Failure	1,111 (3.3)	6,433 (3.8)	0.027
Clinically Indicated Echocardiography within previous year	3,151 (9.2)	13,913 (8.1)	0.039
History of Left Ventricular Systolic Impairment	359 (1.1)	1,743 (1.0)	0.010
Angiotensin Converting Enzyme Inhibitor	11,583 (34.0)	62,533 (36.5)	0.052
Angiotensin II Receptor Blocker	167 (0.5)	950 (0.6)	0.014
Aspirin	14,660 (43.0)	79,532 (46.4)	0.068
Beta Blocker	7,505 (22.0)	48,197 (28.2)	0.141
Clopidogrel	3,621 (10.6)	17,316 (10.1)	0.016
Dihydropyridine Calcium Channel Blocker	8,661 (25.4)	47,993 (28.0)	0.059
Loop Diuretic	15,292 (44.9)	66,654 (38.9)	0.122
Mineralocorticoid Receptor Antagonist	3,099 (9.1)	14,140 (8.3)	0.028
Nondihydropyridine Calcium Channel Blocker	527 (1.5)	2,547 (1.5)	<0.001
Statin	14,105 (41.4)	76,308 (44.6)	0.065
Thiazide Diuretic	7,678 (22.5)	42,877 (25.0)	0.059
Warfarin	3,620 (10.6)	20,217 (11.8)	0.038
CYP3A4 inhibitor/substrate	7,902 (23.2)	31,664 (18.5)	0.116
P-glycoprotein inhibitor/substrate	15,080 (44.2)	78,285 (45.7)	0.030
Nonsteroidal anti-inflammatory drug	21,973 (64.5)	112,831 (65.9)	0.029

Figures represent mean ± SD or number with percentage in parentheses. Percentages are reported as a proportion of the number of prescriptions. CYP3A4, cytochrome p4503A4.

Pharmacogenomic cohort study: Association of genetically determined P-gp activity, clarithromycin, and CV mortality

We did not have a prespecified analysis plan for the pharmacogenomic study; however, the analysis was informed by the observational cohort study. Patients with available prescribing data were obtained from the Genetics of Diabetes Audit and Research in Tayside Scotland study (GoDARTS), the details of which have been published previously [26]. In brief, this is a longitudinal cohort study comprising 18,306 individuals, 10,149 with type 2 diabetes (T2D) and 8,157 controls without T2D at the time of recruitment, of which genotype data were available for 8,564 T2D individuals and 4,586 controls. Genotyping data have been previously described in full [26]. A blood sample for genotyping was obtained from individuals at baseline, and all patients consented to electronic record linkage, allowing details on prescriptions from 1989 to present and outcome data on deaths and hospitalizations. In this part of the study, we again only included individuals who had received a prescription for either amoxicillin or clarithromycin. Collection and analysis of data in GoDARTS was approved by the East of Scotland Research and Ethics Committee. All participants had given written consent for their data to be linked and analyzed for research purposes.

We selected 2 single-nucleotide polymorphisms (SNPs) within the human multidrug-resistance MDR1 gene (ABCB1), which codes for P-gp, which have been shown to be associated with P-gp activity in white healthy volunteers and for which there was a reasonably high prevalence of each genotype—rs1045642, rs1128503 [27-31]. Patients were categorized into 2 groups based on genotype with those individuals homozygous for the risk allele (the allele associated with reduced P-gp activity—A for both SNPs), i.e., individuals with low genetically predicted P-gp activity compared with heterozygous individuals and those homozygous for the nonrisk allele (i.e., intermediate and high genetically determined P-gp activity).

Study endpoints

The primary endpoint for both studies was CV hospitalization. In our initial funding proposal, we planned to evaluate CV mortality as the primary endpoint; however, we felt we would be likely to be underpowered based on the number of patients and changed this to CV hospitalization before the preliminary analysis. Following the date of prescription, patients were followed up for CV hospitalizations, myocardial infarction, and death based on International Classification of Diseases (ICD)-10 coding for 1 year. The following codes were used: CV hospitalization or death –I00-I99; myocardial infarction—I21, I22, and I23. As previous studies have suggested that the CV risk may be different in the short term versus longer time periods, we stratified outcomes at 0–14 days, 15–30 days, and 30 days to 1 year post-prescription. Following the 1-year period of follow-up post-prescription, or after a censored event, surviving patients could be re-entered into the study if they had a further prescription of either antibiotic, similar to the methodology used in other large studies of this type [8, 32].

Statistical analysis

Continuous variables are reported as mean ± standard deviation, and categorical variables are reported as number and percentage. Standardized mean differences (SMDs) between the clarithromycin and amoxicillin groups are reported, with an SMD > 0.1 considered a significant difference between the groups.

In the observational cohort study, Cox proportional hazards regression was performed for the outcome of CV hospitalization or hospitalization for MI at 0–14 days, 15–30 days, and 30 days to 1 year. Additionally, we evaluated the endpoints of all-cause and CV mortality. Adjusted HRs in the observational study are reported adjusting for the IPTW, using the propensity score for likelihood of prescription of clarithromycin based on baseline variables reported in Table 1 [33]. Analysis for the pharmacogenomic study was informed by the design and results of the observational study. In the pharmacogenomic study, a multivariable Cox regression analysis was performed for the association between clarithromycin use versus amoxicillin on CV hospitalization with adjustment for age at the time of antibiotic prescription, sex, history of prior myocardial infarction (MI), chronic obstructive pulmonary disease (COPD), and T2D. Interaction testing was performed to determine whether there was a significant difference between amoxicillin and clarithromycin prescribing depending on P-gp activity. All tests were 2-sided. Statistical analysis was performed using SAS (version 9.4, SAS Institute, https://www.sas.com/en_gb/software/stat.html), STATA (version 14.0, StataCorp, https://www.stata.com/), and R (version 3.5.1, The R Project for Statistical Computing, https://www.r-project.org/).

This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist).

Results

Observational longitudinal cohort study

Baseline characteristics

Over the duration of the study, there were 34,074 prescriptions for clarithromycin for 11,489 unique individuals and 171,153 amoxicillin prescriptions for 36,537 unique individuals. The mean age at prescription was 73.3 ± 12.3 (mean ± standard deviation) years in the clarithromycin group, and 41.9% were male, whereas in the amoxicillin group, the mean age was 74.2 ± 13.2 years, and 44.7% were male. Baseline characteristics are summarized in Table 1. Overall, individuals prescribed clarithromycin more likely to have COPD and be prescribed loop diuretics and CYP3A4 inhibitors or substrates and were less likely to be taking beta-blockers (SMD > 0.1).

Clinical Outcomes

Clinical outcomes of CV hospitalization, hospitalization for MI, CV mortality, and all-cause mortality at 0–14 days, 15–30 days, and 30 days to 1 year are summarized in Table 2. Within the first 14 days following clarithromycin prescription, there were 559 CV hospitalizations (1.6% of all prescriptions) and 289 deaths from any cause, compared with the amoxicillin group in which there were 2,355 CV hospitalizations (1.2% of all prescriptions) and 1,601 deaths from any cause.

Table 2

Clinical outcomes in the observational cohort study (unadjusted counts and percentages).

		Clarithromycin (n = 34,074)	Amoxicillin (n = 171,153)
Cardiovascular Hospitalization	0–14 Days	559 (1.6)	2,355 (1.4)
	15–30 Days	431 (1.3)	2,105 (1.2)
	30 Days–1 Year	1,828 (5.4)	11,321 (6.6)
Hospitalization for MI	0–14 Days	34 (0.1)	169 (0.1)
	15–30 Days	23 (0.07)	131 (0.08)
	30 Days–1 Year	164 (0.5)	1,076 (0.6)
Cardiovascular Mortality	0–14 Days	73 (0.2)	532 (0.3)
	15–30 Days	69 (0.2)	558 (0.3)
	30 Days–1 Year	508 (1.5)	3,925 (2.3)
All-Cause Mortality	0–14 Days	289 (0.8)	1,601 (1.0)
	15–30 Days	277 (0.8)	1,722 (1.0)
	30 days–1 Year	1,530 (4.5)	9,991 (5.8)

Figures in parentheses refer to the number of events as a percentage of the total number of prescriptions. MI, myocardial infarction

In our propensity-weighted analysis, despite there being a higher number of absolute events in the amoxicillin group, clarithromycin use was significantly associated with an increased risk of CV hospitalization compared with amoxicillin only at both 0–14 days (unadjusted absolute risk 1.6% versus 1.4%; propensity-weighted hazard ratio (HR) 1.31; 95% CI 1.17–1.46, p < 0.001) and 30 days to 1 year (unadjusted absolute risk 1.3% versus 1.2%; propensity-weighted HR 1.13; 95% CI 1.06–1.19, p < 0.001) but not at 15–30 days (unadjusted absolute risk 5.4% versus 6.6%; propensity-weighted HR 1.11; 95% CI 0.98–1.26, p = 0.09) (Table 3). A higher proportion of those taking clarithromycin had MI requiring hospitalization within 14 days, though this difference was not statistically significant (adjusted HR 1.37; 95% CI 0.89–2.11, p = 0.12). There was no significant difference in all-cause or CV mortality between clarithromycin and amoxicillin at any time point.

Table 3

Association of clarithromycin and cardiovascular risk versus amoxicillin in the observational study.

	0–14 Days			15–30 Days			30 Days–1 Year
Outcome	Crude Hazard Ratio (95% CI)	Adjusted Hazard Ratio (95% CI)	p-value	Crude Hazard Ratio (95% CI)	Adjusted Hazard Ratio (95% CI)	p-value	Crude Hazard Ratio (95% CI)	Adjusted Hazard Ratio (95% CI)	p-value
Cardiovascular Hospitalization	1.22 (1.10–1.34)	1.31 (1.17–1.46)	<0.001	1.08 (0.97–1.21)	1.11 (0.98–1.26)	0.09	1.05 (1.00–1.11)	1.13 (1.06–1.19)	<0.001
Hospitalization for Myocardial Infarction	1.04 (0.72–1.49)	1.37 (0.89–2.11)	0.16	0.89 (0.57–1.40)	0.94 (0.54–1.62)	0.82	1.00 (0.84–1.18)	1.02 (0.85–1.24)	0.82
Cardiovascular Mortality	0.85 (0.77–0.93)	0.93 (0.81–1.06)	0.25	0.66 (0.51–0.85)	0.82 (0.62–1.10)	0.18	0.85 (0.71–1.01)	0.96 (0.87–1.07)	0.46
All-Cause Mortality	0.86 (0.77–0.96)	1.05 (0.92–1.20)	0.43	0.80 (0.71–0.90)	0.93 (0.81–1.06)	0.25	0.95 (0.91–1.00)	0.97 (0.92–1.02)	0.18

p-values were estimated using robust covariance sandwich estimation and refer to the adjusted hazard ratio, which was adjusted using the likelihood of clarithromycin prescription as a covariate (inverse probability of treatment weighting). This included the following variables: age at prescription, sex, prior history of chronic obstructive pulmonary disease, prior myocardial infarction, history of type 2 diabetes, left ventricular systolic function impairment, and all medications listed in Table 1. CI, confidence interval

When stratified by baseline clinical characteristics, the only significant interaction with clarithromycin use and the primary outcome was in patients with a concomitant prescription for medications metabolized through P-gp (Figs 1 and 2, unadjusted results in S1 Table). Patients prescribed clarithromycin who were also taking P-gp inhibitors were significantly more likely to have a CV hospitalization within the first 14 days (HR 1.97; 95% CI 1.16–3.36), whereas in those who were not taking P-gp inhibitors, there was no significant increase in CV hospitalization with use of clarithromycin versus amoxicillin (HR 1.43; 95% CI 0.90–2.26, interaction p-value 0.029). This interaction was not seen at 15–30 days or 30 days to 1 year (15–30 days: P-gp, HR 0.93, 95% CI 0.49–1.76; no P-gp, HR 0.98, 95% CI 0.56–1.71, interaction p value 0.74; 30 days–1 year: P-gp, HR 0.95, 95% CI 0.73–1.25; no P-gp, HR 0.99, 0.77–1.28, interaction p-value 0.53).

Fig 1

Subgroup analysis of risk of CV hospitalization at 14 days associated with clarithromycin use versus amoxicillin.

Hazard ratio adjusted using the likelihood of clarithromycin prescription as a covariate (inverse probability of treatment weighting)—this included the following variables: age at prescription, sex, prior history of COPD, prior MI, history of type 2 diabetes, left ventricular systolic function impairment, and all medications listed in Table 1. CI, confidence interval; COPD, chronic obstructive pulmonary disease; CV, cardiovascular; CYP3A4, cytochrome P450 3A4; HF, heart failure; MI, myocardial infarction; PGP, P-glycoprotein.

Fig 2

Risk of CV hospitalization at 14 days stratified by concomitant P-glycoprotein medication prescription.

Hazard ratio adjusted using the likelihood of clarithromycin prescription as a covariate (inverse probability of treatment weighting)—this included the following variables: age at prescription, sex, prior history of chronic obstructive pulmonary disease, prior myocardial infarction, history of type 2 diabetes, left ventricular systolic function impairment, and all medications listed in Table 1. CCB, calcium channel blocker; CI, confidence interval; CV, cardiovascular.

Pharmacogenomic cohort study

In total, there were 37,497 amoxicillin prescriptions from 8,513 unique individuals and 7,121 clarithromycin prescriptions from 5,031 unique individuals. Patients who were prescribed clarithromycin were older (66.4 ± 12.4, [mean ± standard deviation] versus 63.2 ± 13.2 versus years, SMD 0.246) and more likely to be male (3,959/7,121 prescriptions [55.6%] versus 17,549/37,497 [46.8%], SMD 0.177). Individuals prescribed clarithromycin were also more likely to have had a prior MI (933/7,121 [13.1%] versus 1,687/37,497 [4.5%], SMD 0.307) and a history of COPD (2,135/7,121 [29.9%] versus 6,712/37,497 [17.9%], SMD 0.284). There was no significant difference in the percentage of patients with diabetes (6,472/7,121 [90.9%] versus 33,822/37,497 [90.2%], SMD 0.024).

The numbers of prescriptions and unique individuals available for analysis stratified by genotype are summarized in Table 4. The heterozygous (intermediate genetically predicted P-gp activity) phenotype was most common for both SNPs. There were no significant differences in prescribing of amoxicillin versus clarithromycin based on genotype.

Table 4

Pharmacogenomic cohort population stratified by genotype and number of prescriptions.

	Amoxicillin (n = 37,497)	Clarithromycin (n = 7,121)	p-value
rs1045642			0.29
High P-gp (GG)	7,995 (21.3)	1,575 (22.1)
Intermediate P-gp (GA)	18,156 (48.4)	3,433 (48.2)
Low P-gp (AA)	11,346 (30.3)	2,113 (29.7)
rs1128503			0.15
High P-gp (GG)	11,516 (30.7)	2,163 (30.3)
Intermediate P-gp (GA)	18,468 (49.3)	3,599 (50.4)
Low P-gp (AA)	7,513 (20.0)	1,373 (19.3)

*Allele dependent on genotyping platform used. P-gp, permeability-glycoprotein.

Association between clarithromycin use and CV hospitalization stratified by genetically predicted P-gp activity

Overall outcomes are summarized in Table 5. In total, there were 952 CV hospitalizations within 1 year of antibiotic prescription in the clarithromycin group (13.3% of all clarithromycin prescriptions) compared with 3,160 (8.4%) in the amoxicillin group. In this cohort, irrespective of genotype, clarithromycin prescription was associated with increased risk CV hospitalization compared with amoxicillin at 15–30 days and 30 days to 1 year (crude HRs: 0–14 days 1.88; 95% CI 1.48–2.38, p < 0.001, 15–30 days 2.27; 95% CI 1.74–2.95, p < 0.001, 30 days–1 year 1.57; 95% CI 1.45–1.70, p < 0.001; adjusted HRs: 0–14 days 1.24; 95% CI 0.96–1.60, p = 0.10, 15–30 days 1.50; 95% CI 1.14–1.99, p = 0.004, 30 days–1 year 1.10; 95% CI 1.01–1.19, p = 0.027).

Table 5

Cardiovascular events in the pharmacogenomic cohort study.

	Total Number of Prescriptions	CV Hospitalization
		0–14 days	15–30 days	30 days to 1 year
Amoxicillin	Total Number of Prescriptions	37,497	37,238	37,053
	Number of Events (%)	259 (0.7)	185 (0.5)	2,716 (7.3)
Clarithromycin	Total Number of Prescriptions	7,121	7,030	6,951
	Number of Events (%)	91 (1.3)	79 (1.1)	782 (11.3)

When stratified by genotype, there was a significant gene-clarithromycin interaction for risk of CV hospitalization at 30 days to 1 year. Individuals who were homozygous for the allele associated with lower P-gp levels (AA) had significantly increased risk of CV hospitalization between 30 days and 1 year when prescribed clarithromycin compared with amoxicillin (rs1045642 AA: HR 1.39, 95% CI 1.20–1.60, p < 0.001, GG/GA: HR 0.99, 95% CI 0.89–1.10, p = 0.85, interaction p-value < 0.001 and rs1128503 AA 1.41, 95% CI 1.18–1.70, p < 0.001, GG/GA: HR 1.04, 95% CI 0.95–1.14, p = 0.43, interaction p-value < 0.001) (Table 6).

Table 6

Association of clarithromycin and cardiovascular hospitalization (versus amoxicillin) stratified by P-glycoprotein genotype.

SNP			rs1045642					rs1128503
	Number of events (%)	Crude Hazard Ratio	Adjusted Hazard Ratio	p-value	Interaction p-value	Number of events (%)	Crude Hazard Ratio	Adjusted Hazard Ratio	p-value	Interaction p-value
0–14 days					0.49					0.66
GG/GA	243 (7.8)	2.07 (1.56–2.73)	1.34 (0.99–1.82)	0.06		280 (7.8)	1.97 (1.52–2.57)	1.31 (0.99–1.74)	0.06
AA	107 (8.0)	1.47 (0.93–2.34)	1.03 (0.63–1.67)	0.92		70 (7.9)	1.50 (0.85–2.65)	1.00 (0.54–1.84)	0.99
15–30 days					0.51					0.11
GG/GA	206 (6.7)	2.44 (1.82–3.27)	1.54 (1.12–2.11)	0.008		234 (6.6)	2.11 (1.59–2.81)	1.36 (1.00–1.84)	0.049
AA	58 (4.3)	1.72 (0.94–3.13)	1.29 (0.69–2.42)	0.42		80 (3.4)	3.67 (1.77–7.63)	2.77 (1.29–5.97)	0.009
30 days–1 year					<0.001					<0.001
GG/GA	2,435 (7.9)	1.47 (1.33–1.62)	0.99 (0.89–1.10)	0.85		2,830 (8.0)	1.50 (1.37–1.64)	1.04 (0.95–1.14)	0.43
AA	1,062 (8.0)	1.81 (1.58–2.08)	1.39 (1.20–1.60)	<0.001		667 (7.6)	1.88 (1.58–2.24)	1.41 (1.18–1.70)	<0.001

AA, lowest genetically predicted P-glycoprotein levels; GA, intermediate genotype; GG, highest genetically predicted P-glycoprotein levels; SNP, single-nucleotide polymorphism. Adjusted hazard ratio adjusted for age at prescription, sex, history of type 2 diabetes, myocardial infarction, and chronic obstructive pulmonary disease.

Restricting the analysis only to those individuals prescribed clarithromycin, for both the rs1045642 and rs1128503 genetic variants, individuals with the AA genotype were more likely to have a CV hospitalization between 30 days and 1 year than those with the GA or GG genotypes (rs1045642: HR 1.34, 95% CI 1.15–1.56, p < 0.001; rs1128503: HR 1.22, 95% CI 1.02–1.45, p = 0.025) (S2 Table).

Discussion

We performed a cohort study in 2 parts, combining a traditional epidemiological approach with a pharmacogenomic study to evaluate the association of the use of the macrolide antibiotic clarithromycin with CV risk. Our study has 2 key findings. First, not only was clarithromycin use associated with increased CV risk compared with amoxicillin, as other studies have reported, but we have identified that the risk is particularly increased in those taking P-gp inhibitors concurrently. This finding was strengthened by use of propensity-score weighting to adjust the results, and a similar finding was observed in the observational analysis of the genomic cohort. Second, in order to strengthen our findings, we have shown in a genomic study (which should reduce the risk of confounding by indication) that the association of clarithromycin with CV hospitalization between 30 days and 1 year was significantly increased in individuals with lower genetically predicted levels of P-gp activity. To the best of our knowledge, this is the first time that such an approach has been used to examine this subject. These results may suggest that, at least in part, the association of clarithromycin with increased CV risk may be modified via P-gp and particular caution may be needed when prescribing clarithromycin in individuals taking P-gp inhibitors.

What this study adds to existing research

To the best of our knowledge, our study is the first to specifically examine a treatment interaction with P-gp inhibitors and substrates. Although clarithromycin is commonly recognized as a P-gp inhibitor [18], it is also a substrate for P-gp, and intracellular levels are increased when another P-gp inhibitor is co-administered [34, 35]. Supporting this, in our genomic analysis, we found that individuals with lower genetically predicted P-gp activity had a higher risk of CV hospitalization between 30 days and 1 year of clarithromycin prescription. Because of the random allocation of genotype at birth, pharmacogenomic studies are less likely to be affected by indication bias than traditional observational studies [36]. In a traditional observational study, this might account for an increase in CV risk seen with P-gp use. There was a higher crude number of events in the amoxicillin group than those prescribed clarithromycin (Table 2); however, after propensity-score adjustment, clarithromycin prescription was associated with worse outcome. This has been previously reported and is because patients prescribed amoxicillin alone tend to be older and more unwell; hence, methods such as propensity-score weighting are required to account for this [37], and this is likely to contribute to the nonsignificant unadjusted increase in mortality seen with amoxicillin in our study.

Concerns regarding the CV risk of macrolide antibiotics have been present for several years [38] and were strengthened by the results of the CLARICOR randomized trial, which, contrary to the authors’ original hypothesis, demonstrated an increased risk of CV mortality at both 3 and 10 years [4, 16]. These results have been further supported by several large observational studies [5–7, 39] and meta-analyses [9, 10], which have reported increased CV risk of myocardial infarction and CV hospitalization up to 1 year after macrolide prescription. Other macrolides such as azithromycin have also been linked with increased CV risk [8]. Nevertheless, there have been alternative large observational studies that have suggested that there is no significantly increased CV risk with macrolide use [13, 14]. These alternative results suggest that there may be specific patient groups who are at particular risk when prescribed clarithromycin. The mechanism of increased CV risk with clarithromycin is not completely clear. The short-term increased risk of sudden cardiac death has been attributed to the effect of macrolides on the QT interval leading to arrhythmia, but this does not explain the increased long-term CV risk observed in CLARICOR and other studies that persisted after drug discontinuation. Alternative theories include macrophage activation leading to coronary plaque destabilization and acute coronary syndrome [40]. With these proposed mechanisms, any pathway by which metabolism of clarithromycin is impaired might lead to increased CV risk. Most studies have so far focused on the CYP3A4 enzyme; however, the interaction with macrolides has not been consistent [6, 35]. We also did not find a significant interaction between clarithromycin and CYP3A4 inhibition.

Strengths and limitations

The key strength of our study is our use of both inverse probability of treatment weighting in our longitudinal cohort and a pharmacogenomic study. By using both of these methods, we strengthen the evidence supporting our finding that the increase in CV risk following clarithromycin prescription is associated with P-gp. Our study has some limitations. First, there are inherent limitations with any observational study, although our use of propensity weighting for likelihood of prescription and genomics do obviate some of these. Nevertheless, it is possible that unmeasured confounding could affect our results. Second, our pharmacogenomic cohort was mainly white, and further studies are required to determine whether our findings in this group are applicable to other ethnicities. As there have been no large genome-wide association studies looking at P-gp activity, we were unable to construct a weighted genetic risk score to evaluate the cumulative effect of P-gp SNPs. A large genome-wide association study would provide more precision around the effect of genetic variants. We used electronic health records and ICD coding to determine prescribing and outcomes; thus, we could not determine adherence. We could also not robustly ascertain prescribing indication and adjust for this, although the majority of prescriptions are likely to have been for suspected lower respiratory tract infection. Nevertheless, current coding accuracy in Scotland is considered reliable [41-43]. Furthermore, any inconsistency would affect both clarithromycin and amoxicillin cohorts equally. We used amoxicillin as a comparator as it is the most-commonly prescribed antibiotic for this indication in the UK, and we did not evaluate other antibiotic classes. We did not have any allergy information, which might also provide another source of indication bias.

Clinical implications and next steps

Macrolides are primarily used to treat respiratory infections; however, alternatives such as amoxicillin or tetracyclines are not associated with increased CV risk, are not metabolized through P-glycoprotein, and have been shown in randomized trials to be noninferior to treatment regimens including macrolides [44, 45]. The most recent American Thoracic Society guidelines for community-acquired pneumonia recommend the use of macrolide monotherapy only in areas where macrolide resistance is low and in patients in whom alternative antibiotics are contraindicated [46]. Given that drugs such as calcium channel blockers are widely used (in 1 cohort, concurrently prescribed in up to 20% of patients with respiratory infections) [8] and such patients may already be at increased CV risk because of their underlying drug indication, advice to be cautious with clarithromycin seems justified as there are equally efficacious alternatives that do not carry this increased risk. This is consistent with advice from the US FDA that proposes caution with clarithromycin use in those with increased CV risk. Our study heralds the possibility of “precision” prescribing, in which patients are prescribed alternative antibiotics if they are taking P-gp inhibitors or if they have a particular genotype. Recent work has suggested the potential for P-gp-associated SNPs to be used in pharmacogenomic strategies for prescribing in other settings [47]. Mechanistic studies evaluating the pathophysiology of P-gp-associated gene and drug interactions in detail would further our understanding and inform future clinical practice.

Conclusion

We found that clarithromycin use was associated with an increased risk of CV hospitalization up to 1 year post-prescription compared with amoxicillin. There appears to be an effect modification via P-gp, with a particularly increased risk of adverse CV events with clarithromycin in patients also taking drugs that are P-gp substrates or those with lower genetically predicted levels of P-gp activity.

STROBE Checklist.

STROBE, Strengthening the Reporting of Observational Studies in Epidemiology.

(DOCX)

Click here for additional data file.

Initial statistical analysis plan submitted at the time of funding application.

(DOCX)

Click here for additional data file.

Subgroup analysis of unadjusted hazard ratios for the association of clarithromycin with cardiovascular hospitalization at 14 days in the longitudinal cohort study.

(DOCX)

Click here for additional data file.

Association of AA genotype (lowest genetically predicted P-glycoprotein levels) with CV hospitalization compared with other GG or GA genotype in patients prescribed clarithromycin.

CV, cardiovascular.

(DOCX)

Click here for additional data file.

13 May 2020

Dear Dr. Mordi,

Thank you very much for submitting your manuscript "Drug Interactions Through P-Glycoprotein Increase Cardiovascular Risk Associated with Clarithromycin: An Epidemiological and Genomic Population-Based Cohort Study" (PMEDICINE-D-19-03443) for consideration at PLOS Medicine.

We apologize for the lengthy review process, and appreciate your patience. Your paper was evaluated by a senior editor and discussed among all the editors here. It was also sent to two independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Jun 03 2020 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see http://journals.plos.org/plosmedicine/s/submission-guidelines#loc-methods.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

1.Title: Please revise the title to avoid implications of causality, and also indicate the population/setting in the title, we suggest: “Genetic and pharmacological relationship between P-Glycoprotein and increased cardiovascular risk associated with clarithromycin prescription: An Epidemiological and Genomic Population-Based Cohort Study in Scotland, UK” or similar.

2. Prospective Analysis Plan: Did your study have a prospective protocol or analysis plan? Please state this (either way) early in the Methods section.

a) If a prospective analysis plan (from your funding proposal, IRB or other ethics committee submission, study protocol, or other planning document written before analyzing the data) was used in designing the study, please include the relevant prospectively written document with your revised manuscript as a Supporting Information file to be published alongside your study, and cite it in the Methods section. A legend for this file should be included at the end of your manuscript.

b) If no such document exists, please make sure that the Methods section transparently describes when analyses were planned, and when/why any data-driven changes to analyses took place.

c) In either case, changes in the analysis—including those made in response to peer review comments—should be identified as such in the Methods section of the paper, with rationale

3. Data Availabilty Statement: PLOS Medicine requires that the de-identified data underlying the specific results in a published article be made available, without restrictions on access, in a public repository or as Supporting Information at the time of article publication, provided it is legal and ethical to do so. Please see the policy at

http://journals.plos.org/plosmedicine/s/data-availability

and FAQs at

http://journals.plos.org/plosmedicine/s/data-availability#loc-faqs-for-data-policy

Additionally, please provide a link or file containing the STATA and R code as requested by reviewer 1.

4. Abstract: Please provide some of the relevant summary demographics for the observational cohort study in Scotland. Please include some information on the population and years during which the study took place for the GoDARTS participants.

5. Abstract: Please quantify the main results with both 95% CIs and p values. Where applicable, please include the important dependent variables that are adjusted for in the analyses.

6. Abstract: In the last sentence of the Abstract Methods and Findings section, please describe the main limitation(s) of the study's methodology.

7. Author Summary: At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary

8. Results: Please present p values as p<0.001 where applicable.

9. Results: At the bottom of page 9: Please revise the following sentence, as the term "trend" is used to refer to a nonsignificant p value. The term trend should be used only when the test for trend has been conducted. “There was also a non-significant trend to higher likelihood of hospitalisation for MI within the first 14 days of clarithromycin.”

10. Results: Please present the numerators and denominators when reporting percentages (at least in a table if not in the text); for example, describing the pharmacogenetic cohort: “Individuals prescribed clarithromycin were also more likely to have had a prior MI (13.1% vs. 4.5%, p<0.001) and a history of COPD (29.9% vs. 17.9%, p<0.001).”

11. Results: Please present both the unadjusted and adjusted results where applicable, at least in the tables if not in the text. For example, please present the unadjusted results for the relationship between clarithromycin and CV hospitalization risk described on page 13.

12. Results: Page 13: Please clarify this sentence, as it seems like there was also an increased risk at 15-30 days. “After adjustment for age, sex, history of myocardial infarction and history of COPD, clarithromycin prescription was associated with increased risk of CV hospitalisation between 30 days and 1 year (0-14 days: HR 1.23, 95% CI 0.95-1.58, p=0.12; 15-30 days: HR 1.50, 95% CI 1.13-1.99, p=0.005; 30 days-1 year: HR 1.10, 95% CI 1.01-1.19, p=0.031).”

13. Discussion, page 15: Please temper the following sentence with “To the best of our knowledge” or similar: “Our study is the first to specifically examine a treatment interaction with P-glycoprotein inhibitors and substrates.”

14. Discussion: Please present and organize the Discussion as follows: a short, clear summary of the article's findings; what the study adds to existing research and where and why the results may differ from previous research; strengths and limitations of the study; implications and next steps for research, clinical practice, and/or public policy; one-paragraph conclusion. You touch on implications for clinical practice throughout the discussion, but please consolidate together in a paragraph if possible.

15. Discussion (and throughout): Please replace "Caucasian" with "white" throughout the paper.

16. Table 1: For age, please indicate in the table or legend if you are reporting mean +/- SD, or some other measure here.

17. Table 2: Please clarify what the values are representing in the legend “Figure in brackets refers to the number of events as a percentage of the total number of prescriptions.” as there are parentheses in the table but no brackets. If the values in parentheses are the percentages of the total number of prescriptions, please place the number of prescriptions in the column headers, for reference. In the legend, please define the abbreviation for ‘MI’.

18. Table 3: Please present the unadjusted results in addition to the adjusted HRs. Please define abbreviations for MI, CI, and IPTW in the legend.

19. Figures 1 and 2: Please present the unadjusted results as well. Please specify the variables controlled for in the legend, and define abbreviations for CV, MI, COPD, HF, PGP, CCB, and CI.

20. References: Please place in-text citations in square brackets, like this [1].

21. Checklist: Thank you for including the STROBE checklist as Supporting Information. Please revise the checklist, using section and paragraph numbers, rather than page numbers to refer to locations in the manuscript.

Comments from the reviewers:

Reviewer #1: The paper examines the effect of possible drug-drug interactions between clarithromycin and P-glycoprotein (P-gp) inhibitors on cardiovascular risk. The authors consider one observational and one pharmacogenomic cohort in a setting that contrasts how prescriptions of clarithromycin and amoxicillin impact future hospitalizations due to cardiovascular events. The results highlight a P-gp-centric molecular mechanism explaining the bioavailability of clarithromycin and suggest the existence of an at-risk population with a genetic variant coding for low P-gp activity.

Overall, the paper is well-written and easy to follow. The hypothesis is clearly formulated and placed in the context of published literature. The usage of observational and pharmacogenomic cohorts strengthens the findings, which, if confirmed, may have an immediate impact on the prescription and dosage of clarithromycin in the clinic. However, several aspects of the statistical analysis currently give pause and must be revisited to ensure that the findings and conclusions are robust.

Since the focus is on drug-drug interactions, the authors need to be more clear about when clarithromycin is considered by itself vs. in combination with P-gp inhibitors. The Methods section states that the study contrasts "all patients over 18 years old who were prescribed clarithromycin (alone or in combination with another antibiotic) over this period, with those prescribed amoxicillin only as a control group", which gives the impression that clarithromycin monotherapy and combination therapy are always considered as a single group. If this is the case, then how do we know that the clinical outcomes reported in Tables 2 and 3 are not confounded by whether clarithromycin was prescribed alone or in combination with a P-gp inhibitor since amoxicillin is always considered as a monotherapy? (In fact, the subsequent analysis reported in Figures 1 and 2 shows that such a confounder is indeed likely.)

Along the same vein, I am not sure if amoxicillin monotherapy is a proper control for evaluating the effect of prescribing clarithromycin in combination with a P-gp inhibitor. It seems that either 1) amoxicillin should also be paired with a P-gp inhibitor, or 2) the control group should consist of concomitant prescriptions of clarithromycin and "another drug that is not a P-gp inhibitor". Otherwise, one could argue that all the observed differences in clinical outcomes are simply due to the number of prescribed drugs.

The pharmacogenomic analysis is well formulated and cleanly executed, other than the possible ramifications of the control group considerations above.

All reported p-values need to be adjusted for multiple hypothesis testing. The authors state that "a p value <0.05 was considered statistically significant", but without adjustment this significance threshold implies that approximately one out of every 20 tests may produce a false positive.

Minor comments:

1. The statement "To the best of our knowledge, there have been no studies evaluating the association between clarithromycin use, CV risk and P-gp." is a bit too strong. The study by Wessler and colleagues (which is cited in the manuscript) has certainly considered clarithromycin in the context of CV risk and P-gp, as part of a larger study.

2. The STATA and R code used for the analysis should be released alongside the manuscript (e.g., as a GitHub repository) to increase reproducibility.

3. What is the purpose of reporting a p value for "Total Number of Unique Patients" and "Total Number of Prescriptions" in Table 1? Is the goal to demonstrate that overall one drug is prescribed significantly more often than the other? I don't think that's very surprising...

Reviewer #2: This is an interesting analysis of an important but poorly understood phenomenon, the long-term cardiovascular risks following use of clarithromycin. I believe it would be a useful addition to the literature but there are several ways the manuscript could be improved and there are also some items that need to be clarified.

My major concern with the longitudinal cohort study is the possibility of residual confounding; in particular, indication was not accounted for, a limitation which should be noted in the Discussion. The only two significant HRs are modest in magnitude (1.31, 1.13) and so are in the range that could be the result of confounding. Were the p-values in Table 3 adjusted for multiple comparisons? Also, when you discuss the impact of weighting in the Discussion you note that before weighting the risk of CV events was higher with amoxicillin. It wasn't clear what result that refers to; was it the 15-30 day window?

The most persuasive evidence for a role of P-gp comes from the pharmacogenomic study and the interaction seen with genetically determined low P-gp activity, which as point out is unlikely to be due to confounding. I think you should seriously consider revising the paper to highlight the pharmacogenomic study rather than the longitudinal cohort study, perhaps by switching the order in which the studies are described in the manuscript. Have you compared the risk within clarithromycin-exposed patients alone; i.e., clarithromycin users with GG/GA genotype versus clarithromycin users with AA genotype? This comparison, if significant, would be even more persuasive, but may not be adequately powered in your sample, however.

Regarding Table 1 (observational longitudinal cohort study): Use of a t-test or chi-square test to assess baseline differences between groups results in many statistically significant differences which are small in magnitude, but statistically significant because of the large sample size. A better approach would be to compare the groups with standardized differences, for which a difference of 0.1 or more is considered consequential. After IPTW, the standardized differences can be compared again to determine if the weighting successfully balanced the baseline covariates. Your manuscript does not address the issue of whether the IPTW succeeded in balancing the baseline covariates. See Mamdani et al., at https://www.bmj.com/content/330/7497/60.long

In the observational longitudinal cohort study, it is puzzling why the mean age is so high (about 73 years for clarithromycin users), if any patient over 18 years of age was included. Can you please comment?

Study endpoints: Was ICD-10 in use in Scotland during the entire study period (2004-2014)?

Minor comments

Table 2: suggest showing total number of patients per group in the header row

Figure 2: is the risk window here also 14 days? Are all these drugs P-gp substrates? If so, why does there only appear to be an interaction with dihydropyridine CCBs and maybe amiodarone?

Table 4: the percentages for rs1045642 don't sum properly (total = 110.5%)

Discussion: You state that amoxicillin and tetracyclines are non-inferior to macrolides for respiratory infections. Some have argued that macrolides reduce mortality from community acquired pneumonia, however. See for example Asadi et al., https://academic.oup.com/cid/article/55/3/371/613119

Final sentence: It is probably more appropriate to speak of drugs that are substrates for P-gp rather than "metabolized" by P-gp.

Any attachments provided with reviews can be seen via the following link:

[LINK]

3 Jul 2020

Submitted filename: PLOS Medicine Reviewer Response IM 030720.docx

Click here for additional data file.

13 Aug 2020

Dear Dr. Mordi,

Thank you very much for re-submitting your manuscript "Genetic and pharmacological relationship between P-Glycoprotein and increased cardiovascular risk associated with clarithromycin prescription: An Epidemiological and Genomic Population-Based Cohort Study in Scotland, UK" (PMEDICINE-D-19-03443R1) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by two reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

Our publications team (plosmedicine@plos.org) will be in touch shortly about the production requirements for your paper, and the link and deadline for resubmission. DO NOT RESUBMIT BEFORE YOU'VE RECEIVED THE PRODUCTION REQUIREMENTS.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Aug 20 2020 11:59PM.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

1. Data availability statement: As mentioned by reviewer 1, please include the GitHub link specific for your dataset used in this study.

2. Abstract: Background: First sentence: Please change the word “and” to “with”

3. Abstract: Background: Please revise to temper this claim, we suggest: “To the best of our knowledge, no studies have examined whether this association might be mediated via P-glycoprotein (P-gp), a major pathway for clarithromycin metabolism.”

4. Abstract: Methods and Findings: There appears to be a missing p value for the pharmacognetic report of the AA allele associated with lower p-gp activity (rs1045642 AA: HR 1.39, 95% CI 1.20-1.60)

5. Abstract: Conclusions: We suggest revising to: “In this study, we observed that that increased risk of CV events with clarithromycin compared to amoxicillin were associated with an interaction with P-glycoprotein.” or similar to temper the causal implications.

6. Author summary: What did the researchers do and find?: We suggest combining the third and fourth bullet points as follows:

“-In this analysis we found that that patients prescribed clarithromycin were significantly more likely to have a cardiovascular hospitalisation at 0-14 days and 30 days to 1 year after prescription than those prescribed amoxicillin, and that individuals who were co-prescribed P-glycoprotein substrates or inhibitors and clarithromycin had significantly higher risk of cardiovascular hospitalisation.”

7. Author summary: What do these findings mean: We suggest revising the second point as follows:

“These results suggest implications for clarithromycin use patients taking P-glycoprotein inhibitors or with low genetically-predicted P-glycoprotein activity.” or similar.

8. Data availability statement: Please remove this section from the body of the manuscript and ensure it is entered accurately (with updated GitHub information) where appropriate in the manuscript submission system form.

9. Methods: Page 7: Regarding your prospective analysis plan, thank you for noting in your response to comments that “We did not have a pre-specified plan for the pharmacogenomic section of the study, however, the analysis was informed by the cohort study.” Please include a statement such as this in Methods.

10. Methods: Study Endpoints: First sentence, and throughout: Please consistently use the abbreviation “CV” for cardiovascular, defining the abbreviation at first use in the text and using the abbreviation throughout, for the sake of consistency: “The primary endpoint for both studies was CV hospitalisation. In our 1 initial funding proposal we planned to evaluate cardiovascular mortality as the primary endpoint…”

11. Methods: Please add the following statement, or similar, to the Methods: "This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist)."

12. Results Page 11: Clinical outcomes for Observational Cohort Study: Please provide the result with 95% CIs and p values here: “A higher proportion of those taking clarithromycin had MI requiring hospitalisation within 14 days, though this difference was not statistically significant.”

13. Results: Page 11: Please also present the HR and 95% CIs for the interactions with P-gp inhibitors at 30 days and 1 year to accompany these p values? “interaction was not seen at 30 days or 1 year (interaction p values 0.74 and 0.53 respectively).”

14. Discussion: Please start the discussion with 1-2 sentences briefly summarizing what was done in the study.

15. Discussion: Page 13: Please revise to: “These results may suggest that, at least in part, the association of clarithromycin with increased CV risk may be modified via P-glycoprotein and particular caution may be needed in prescribing clarithromycin in individuals taking P-gp inhibitors” or similar

16. Discussion: Page 13: Please revise to: “This has been previously reported and is because patients prescribed amoxicillin alone tend to be older and more unwell, hence methods such as propensity-score weighting are required to account for this [37], and this is likely to contribute to the non-significant unadjusted increase in mortality seen with amoxicillin in our study” or similar.

17. Discussion: page 14: Please change “was” to “is” in the following sentence: “With these proposed mechanisms, any pathway by which metabolism of clarithromycin was impaired might lead to increased CV risk.”

18. Discussion: page 15: Please replace the term “compliance” with “adherence” where it is used to refer to treatment adherence.

19. Discussion: Page 15: Please remove the word “which” to clarify: “The most recent American Thoracic Society guidelines for community acquired pneumonia recommend the use of macrolide monotherapy only in areas where macrolide resistance is low and in patients in whom alternative antibiotics are contraindicated [46].”

20. Sections: Declaration of Interests, Funding, Role of the Funding Source: Please remove these sections from the main text of the manuscript, and ensure the information is accurately entered into the relevant locations within the manuscript submission system.

21. Table 1: Please indicate in the table or legend that the values represent numbers and percentages.

22. Table 3: Please define the abbreviation “CI” in the legend.

23. Table 6: Please also define GA/GG in the legend. Please indicate in the column headers that the hazard ratios are followed by the 95% CIs.

24. Figure 1 and Figure 2: Please also define abbreviation for CV in the legend.

25. S1 Table: Please define abbreviations for pgp, CCB, and CI in the legend.

26. Supplementary File 2: Analysis Plan: Thank you for including your analysis plan. If you have a dated version of this plan, please update accordingly.

Comments from Reviewers:

Reviewer #1: Overall, I am satisfied with the revisions submitted by the authors and have no other concerns. In particular, I thank the authors for clarifying the nuances of clarithromycin prescription. With the new description, I no longer have reservations about the control group. Very minor comments are below.

I agree with the authors that the Bonferroni correction is too stringent and would generally recommend something like the recently-proposed harmonic mean p-value (PMID: 30610179), which is able to handle groups of dependent tests without explicit access to what the dependency structure is. However, I think that reporting raw p-values without placing an artificial "significance" threshold is also a reasonable strategy, albeit one that softens the main message.

Minor note regarding GitHub organization: The GitHub link in the manuscript (https://github.com/ifymordi) refers to a user, not a repository. As this user (I hope) publishes additional code in future studies, it will become increasingly difficult to find the code associated with the current clarithromycin study by following the link in the paper. Consider renaming your "Research" repository to "Clarithromycin" and updating the link to github.com/ifymordi/Clarithromycin. It also seems that all code is in a branch that is not immediately visible to a user. Consider merging this branch into master, since that's what readers will see when they first follow your link.

Reviewer #2: I appreciate the opportunity to review the revised manuscript. I thank the authors for their careful and complete consideration of my comments. The manuscript is much improved and while I have a couple of suggestions I don't feel that any further changes are mandatory. However, for the authors' consideration, I will provide some further thoughts regarding two of my previous comments.

1. Standardized differences--I would point out that it is possible to calculate standardized differences in baseline characteristics on the weighted population, just as you have now done for before IPTW. This is similar to calculating standardized differences on a propensity-score matched sample both before and after matching. If you show that the post-IPTW standardized differences are minimal (<0.1) that would strengthen the findings of the study.

2. Regarding my suggestion for a within-clarithromycin analysis of risk by P-gp genotype: I take the point that the test for the treatment X genotype interaction essentially controls for the effect of amoxicillin. However, I wonder if a within-clarithromycin analysis might provide a more straightforward demonstration of this effect, although one would have to assume that P-gp genotype varies randomly among clarithromycin users. This is just a suggestion for the authors as the existing analysis does address the effect of genotype.

Any attachments provided with reviews can be seen via the following link:

[LINK]

17 Sep 2020

Submitted filename: Editor Response 210820.docx

Click here for additional data file.

21 Sep 2020

Dear Dr. Mordi,

On behalf of my colleagues and the academic editor, Dr. Sanjay Basu, I am delighted to inform you that your manuscript entitled "Genetic and pharmacological relationship between P-Glycoprotein and increased cardiovascular risk associated with clarithromycin prescription: An Epidemiological and Genomic Population-Based Cohort Study in Scotland, UK" (PMEDICINE-D-19-03443R2) has been accepted for publication in PLOS Medicine.

PRODUCTION PROCESS

Before publication you will see the copyedited word document (in around 1-2 weeks from now) and a PDF galley proof shortly after that. The copyeditor will be in touch shortly before sending you the copyedited Word document. We will make some revisions at the copyediting stage to conform to our general style, and for clarification. When you receive this version you should check and revise it very carefully, including figures, tables, references, and supporting information, because corrections at the next stage (proofs) will be strictly limited to (1) errors in author names or affiliations, (2) errors of scientific fact that would cause misunderstandings to readers, and (3) printer's (introduced) errors.

If you are likely to be away when either this document or the proof is sent, please ensure we have contact information of a second person, as we will need you to respond quickly at each point.

PRESS

A selection of our articles each week are press released by the journal. You will be contacted nearer the time if we are press releasing your article in order to approve the content and check the contact information for journalists is correct. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact.

PROFILE INFORMATION

Now that your manuscript has been accepted, please log into EM and update your profile. Go to https://www.editorialmanager.com/pmedicine, log in, and click on the "Update My Information" link at the top of the page. Please update your user information to ensure an efficient production and billing process.

Thank you again for submitting the manuscript to PLOS Medicine. We look forward to publishing it.

Best wishes,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org