Risk of Nonmelanoma Skin Cancer in Association With Use of Hydrochlorothiazide-Containing Products in the United States

Abstract

Background

European studies reported an increased risk of nonmelanoma skin cancer associated with hydrochlorothiazide (HCTZ)-containing products. We examined the risks of basal cell carcinoma (BCC) and squamous cell carcinoma (SCC) associated with HCTZ compared with angiotensin-converting enzyme inhibitors (ACEIs) in a US population.

Methods

We conducted a retrospective cohort study in the US Food and Drug Administration’s Sentinel System. From the date of HCTZ or ACEI dispensing, patients were followed until a SCC or BCC diagnosis requiring excision or topical chemotherapy treatment on or within 30 days after the diagnosis date or a censoring event. Using Cox proportional hazards regression models, we estimated the hazard ratios (HRs), overall and separately by age, sex, and race. We also examined site- and age-adjusted incidence rate ratios (IRRs) by cumulative HCTZ dose within the matched cohort.

Results

Among 5.2 million propensity–score matched HCTZ and ACEI users, the incidence rate (per 1000 person-years) of BCC was 2.78 and 2.82, respectively, and 1.66 and 1.60 for SCC. Overall, there was no difference in risk between HCTZ and ACEIs for BCC (HR = 0.99, 95% confidence interval [CI] = 0.97 to 1.00), but there was an increased risk for SCC (HR = 1.04, 95% CI = 1.02 to 1.06). HCTZ use was associated with higher risks of BCC (HR = 1.09, 95% CI = 1.07 to 1.11) and SCC (HR = 1.15, 95% CI = 1.12 to 1.17) among Caucasians. Cumulative HCTZ dose of 50 000 mg or more was associated with an increased risk of SCC in the overall population (IRR = 1.19, 95% CI = 1.05 to 1.35) and among Caucasians (IRR = 1.27, 95% CI = 1.10 to 1.47).

Conclusions

Among Caucasians, we identified small increased risks of BCC and SCC with HCTZ compared with ACEI. Appropriate risk mitigation strategies should be taken while using HCTZ.

Hydrochlorothiazide (HCTZ) is indicated for the treatment of hypertension and edema resulting from various etiologies (1). In 2017, an estimated 20.9 million patients received a dispensed prescription for HCTZ-containing products in the United States (2). Single-ingredient and 25 mg strength HCTZ-containing products contributed to the largest proportion of use at 43% (9 million) and 55% (11.5 million) of the total patients, respectively (2).

Recently, the US Food and Drug Administration and European Union Pharmacovigilance Risk Assessment Committee were alerted to 2 Danish population-based case-control studies (3,4) that found an elevated risk of nonmelanoma skin cancer (NMSC), which includes basal cell carcinoma (BCC) and squamous cell carcinoma (SCC), associated with HCTZ-containing products. Both studies also observed a dose-response effect. An earlier observational study (5), conducted in 1 Danish county, also found an elevated risk of SCC associated with ever use of HCTZ.

The biological plausibility for an association between HCTZ and NMSC is likely related to its photosensitizing properties (6). Drug-induced photosensitivity can occur as photoallergic or phototoxic reactions (7). Although photoallergic reactions are less common than phototoxic reactions, they are not dependent on the dose of the medication or intensity of sun exposure and usually occur hours after the initial exposure as a delayed hypersensitivity reaction (7). Occurring at a higher frequency, phototoxic reactions are usually dose-dependent and require greater drug exposure (7). Several in vitro studies suggested that HCTZ and other sulfonamide-containing drugs may trigger drug-induced phototoxic reactions (8–11). The ensuing pathogenesis of phototoxic and photoallergic reactions is different. Phototoxic reactions occur following the activation of drug or its metabolite by light in viable skin cells, resulting in generation of reactive oxygen intermediates, which leads to nuclear cytoplasm or cell membrane alteration and ultimately cell damage (12,13). Photoallergic reactions do not result in cell damage. Instead, it is believed that these reactions are a type of cell-mediated hypersensitivity response (14). It is likely that the ensuing sunburns due to phototoxic reactions result in photodamage, as observed with high-dose oral psoralen and SCC (15).

Although the positive associations observed in the Danish studies were supported by biological plausibility, some aspects of these studies limited the generalizability of their findings to the US population. The Danish population, of which 93% are Caucasians, may have different NMSC susceptibility genes, racial composition, and sun exposure behaviors than the US population. One-third of HCTZ-containing products in the Danish studies included amiloride, which also has photosensitizing properties and could have contributed to the observed increased risk. In the United States, HCTZ-amiloride combinations make up less than 1% of the HCTZ market share.

We examined the risk of NMSC among patients exposed to HCTZ-containing products compared with patients who received angiotensin-converting enzyme inhibitors (ACEIs) in a large, demographically and geographically diverse US population.

Methods

Data Source

We conducted a retrospective, new-user cohort study using data from January 1, 2000, to August 31, 2018, from 17 health plans (ie, sites) participating in the Sentinel System (16). The health plans included large national insurers, integrated delivery care networks, a state Medicaid, and the 100% Medicare fee-for-service plan. Each plan regularly updates and transforms its data into standardized formats to facilitate queries (17). This study was conducted as part of public health surveillance activities and therefore not under the purview of institutional review boards (18,19).

Study Cohort

We included patients with continuous enrollment in plans with both medical and drug coverage for at least 183 days prior to the index date (defined below). The exposure cohort included new users of any HCTZ-containing products (referred to as HCTZ henceforth), and the comparator cohort included new users of ACEI monotherapy or non–HCTZ ACEI combination products (referred to as ACEI henceforth). In both cohorts, new use was defined as no use of HCTZ or ACEI in the 183-day baseline period prior to the date of index dispensing (index date). We required eligible patients to have no diagnosis of any cancer type, no use of any chemotherapeutic agent in any formulation, and no radiation therapy during the baseline period or on the index date. We further excluded patients who initiated both HCTZ and ACEI or those with a diagnosis of BCC or SCC on the index date (Supplementary Figure 1, available online). We chose ACEI as our comparator because these drugs are also indicated for hypertension and do not appear to be associated with the risk of NMSC based on previous studies (3,4).

Outcomes

The study outcomes included BCC and SCC, defined by an International Classification of Diseases, 9th revision, Clinical Modification diagnosis code (BCC: 173.x1; SCC: 173.x2) or International Classification of Diseases, 10th revision, Clinical Modification diagnosis code (BCC C44.x1x; SCC: C44.x2x) plus a Current Procedural Terminology code for excision (11600-11646), destruction (laser surgery, electrosurgery, cryosurgery, chemosurgery, or surgical curettement) (17260-17286), Mohs microscopic chemosurgery (17304-17310), Mohs excision technique (17311-17315), or a National Drug Code for a topical chemotherapy treatment on or within 30 days after the diagnosis date (Supplementary Table 1, available online). The date of diagnosis was used to establish the date of the outcome. A previous study (in a large health maintenance organization health plan) showed a positive predictive value of 98.7% for this outcome algorithm (20,21).

Covariates

We estimated the baseline propensity scores (22) using logistic regression to predict the probability of initiating HCTZ compared with ACEI. The propensity score model adjusted for year of treatment initiation; age; sex; presence of high-risk skin conditions for NMSC (see Table 1 for conditions); use of photosensitizing medications (Table 1); use of drugs with potential chemoprotective effects; immunosuppressive medications; diabetes mellitus, human papillomavirus, arsenic exposure, and alcohol use or abuse; health-care utilization metrics; a comorbidity score; and ultraviolet exposure. The description of how the health-care utilization metrics and comorbidity score are calculated is available elsewhere (23). We estimated ultraviolet exposure using data from the National Weather Service (24). The annual maximum median clear sky ultraviolet index was obtained for each state over the study period. For states with multiple cities (Florida, Texas, Pennsylvania, and New York), we selected the highest median to represent the state. We categorized ultraviolet index as low (0, 1, 2), moderate (3–5), high (6,7), very high (8–10), and extremely high (≥11) according to the World Health Organization ultraviolet index reporting categories (25).

Table 1.

Baseline characteristics of new users of HCTZ and new users of ACEI, before and after 1:1 propensity score matching

Characteristics	Before matching			After matching
	HCTZ	ACEI	SMD	HCTZ	ACEI	SMD
No. of patients	6 959 086	7 417 942		5 211 321	5 211 321
Demographics
Mean age, y	59.8	61.6	−0.137	60.7	60.7	0.001
Age category, %
<50 y	27.7	23.7	0.092	25.6	25.4	0.005
50-59 y	22.1	21.3	0.020	21.2	21.8	−0.013
60-74 y	33.2	34.5	−0.028	34.8	34.1	0.013
≥75 y	17.1	20.6	−0.091	18.4	18.7	−0.008
Sex, %
Female	61.5	46.8	0.298	52.5	53.1	−0.012
Male	38.5	53.2	−0.298	47.5	46.9	0.012
Recorded history, %
Actinic keratosis	2.8	2.7	0.005	2.9	2.9	0.001
Arsenic exposure	0.0	0.0	0.001	0.0	0.0	0.000
Diabetes	18.7	35.8	−0.391	24.3	24.3	0.001
Human papillomavirus	0.2	0.1	0.022	0.2	0.2	0.000
Immunosuppressive conditions
Chronic pulmonary disease	11.8	15.7	−0.115	12.6	12.6	0.000
Connective tissue disease	1.7	1.8	−0.005	1.7	1.7	0.000
Cardiovascular disease	11.2	26.0	−0.389	14.5	14.4	0.003
Moderate/severe renal disease	0.5	2.9	−0.184	0.7	0.9	−0.015
Transplant	0.2	0.7	−0.076	0.2	0.3	−0.005
Human immunodeficiency virus	0.3	0.4	−0.0147	0.4	0.4	0.000
Immune disorders	0.1	0.1	−0.009	0.1	0.1	−0.000
Severe skin disease	0.1	0.3	−0.038	0.1	0.1	−0.002
White blood cell disease	0.0	0.1	−0.015	0.1	0.1	−0.001
Mole removal	0.9	0.8	0.011	0.8	0.8	0.000
Naevi	0.8	0.6	0.017	0.7	0.7	0.000
Xeroderma pigmentosum	0.0	0.0	0.004	0.0	0.0	0.000
Ultraviolet radiation exposure, %
Low	0.1	0.2	−0.012	0.1	0.1	0.000
Moderate	39.3	42.6	−0.069	41.5	41.4	0.001
High	43.1	39.1	0.082	40.1	40.1	0.000
Very High	16.0	16.7	−0.020	16.8	16.8	−0.001
Extreme	0.1	0.0	0.009	0.1	0.1	0.001
Unknown	1.4	1.4	0.008	1.4	1.4	0.000
Medication or behavioral history, %
Alcohol use or abuse	0.9	1.5	−0.051	1.1	1.1	0.000
Drugs with chemoprotective effectsa	45.0	53.5	−0.169	48.3	48.2	0.004
Immunosuppressive medications
Glucocorticoids	28.6	27.4	0.025	27.7	27.8	0.000
Treatment for immune disorders	8.9	8.2	0.028	8.5	8.5	0.000
Photosensitizing medications
Aminoquinolines	0.5	0.5	0.005	0.5	0.5	0.000
Amiodarone	0.5	1.8	−0.118	0.7	0.7	−0.004
Macrolides	12.3	11.7	0.017	11.9	11.9	0.000
Methoxypsoralen	0.0	0.0	0.000	0.0	0.0	0.000
Retinoids	0.0	0.0	0.010	0.0	0.0	0.000
Tetracycline	0.2	0.2	0.008	0.2	0.2	0.000
Mean combined comorbidity scoreb	0.0	0.9	−0.435	0.2	0.2	−0.008
Mean health service utilization intensity
Ambulatory encounters	6.7	7.4	−0.096	6.8	6.8	−0.002
Emergency room encounters	0.3	4	−0.113	0.3	0.3	−0.005
Inpatient hospital encounters	0.1	3	−0.326	0.1	0.1	−0.014
Nonacute institutional encounters	0.0	0.1	−0.144	0.0	0.0	−0.008
Other ambulatory encounters	1.6	3.2	−0.269	1.8	1.9	−0.013
Filled prescriptions	13.6	16.0	−0.182	14.2	14.2	0.001

a

Such as aspirin, nonsteroidal anti-inflammatory drugs, statins. ACEI = angiotensin-converting enzyme inhibitor-containing products; HCTZ = hydrochlorothiazide-containing products; SMD = standardized mean difference.

b

Gagne JJ, Glynn RJ, Avorn J, Levin R, Schneeweiss S. A combined comorbidity score predicted mortality in elderly patients better than existing scores. J Clin Epidemiol. 2011; 64(7):749-759.

Follow-up

We followed patients from the day after the index date until the earliest of the following events: outcome diagnosis, health plan disenrollment, study end date, end of data availability, diagnosis of nonoutcome cancer, use of chemotherapy or radiation therapy, or death.

Statistical Analysis

We matched each exposed patient with a comparator patient within the same site based on their estimated propensity scores using the nearest neighbor approach (caliper = 0.05) (26). We assessed covariate balance before and after matching using standardized mean differences, with absolute standardized differences of less than 0.1 to indicate negligible differences between the 2 cohorts (27).

We first used Cox proportional hazards regression models to estimate the hazard ratios (HRs) and 95% confidence intervals (CIs) separately for BCC and SCC, comparing new users of HCTZ with new users of ACEI in the matched population. We examined the Kaplan-Meier curves for each analysis to ascertain whether the assumption of proportionality was violated. We also estimated the hazard ratios by sex (male and female), age category (younger than 50, 50-59, 60-74, and 75 years and older), and race (Caucasian, Black, Asian, Pacific Islander, American Indian, and unknown). For all subgroup analyses, we rematched patients within their subgroups using their original propensity scores (28).

Next, we examined the dose-response relation between HCTZ use and NMSC using Poisson regression models (29). We created 5 cumulative HCTZ dose categories within the propensity score–matched population: less than 10 000 mg, 10 000-24 999 mg, 25 000-49 999 mg, 50 000-74 999 mg, 75 000-99 999 mg, and 100 000 mg or more. To be consistent with prior studies (3,4), we also created an additional category of 50 000 mg or more. For each patient, we multiplied the quantity supplied (eg, 60 tablets) by the strength of HCTZ dispensed (eg, 25 mg) for each dispensing, starting from the index date to the end of follow-up. We then summed the dose (eg, 1500 mg) across all HCTZ dispensings observed during follow-up to calculate the cumulative HCTZ dose. Each HCTZ user could contribute to multiple dose categories as cumulative dose accrued. For example, a user with a cumulative dose of 10 000-24 999 mg would contribute to less than 10 000 mg and 10 000-24 999 mg categories. Follow-up time in each cumulative dose category was reset when patients surpassed the previous cumulative dose group and moved to the next cumulative dose group. For each cumulative HCTZ dose category, we estimated the incidence rate ratio by comparing the incidence rate in that HCTZ dose category with the incidence rate in all ACEI users in the propensity score–matched population, using a Poisson regression model that adjusted for site and the age group defined at the beginning of that dose category. This allowed us to adjust for age (a strong risk factor for NMSC) in a time-varying fashion. We used the parameter estimates from the Poisson model to estimate the adjusted incidence rate differences and 95% confidence intervals (30,31).

Sensitivity and Additional Analysis

We performed sensitivity analyses to examine the robustness of our results. First, we ignored NMSC events in the first 2 years of HCTZ use to account for potential lag time in the effect of HCTZ use on NMSC. Second, we increased the allowable gap in days between the NMSC diagnosis and treatment from 30 days to 60 and 180 days. Lastly, we evaluated descriptively the time to accrue each cumulative dose category to examine treatment intensity and obtained the proportion of patients exposed to different HCTZ strengths on the last HCTZ prescription as a proxy for HCTZ-dispensing patterns within each cumulative dose category.

The analytic process, including propensity score estimation and effect estimation, was completed using the Sentinel Query Request Package Propensity Score Analysis Module, a pretested, validated analytic program (32–35), ad hoc programming (30). The analytic package is publicly available at https://dev.sentinelsystem.org/projects/AP/repos/sentinel-analytic-packages/browse.

Results

Cohort Characteristics

We identified 6 959 086 and 7 417 492 eligible new users of HCTZ and ACEI, respectively. ACEI users were slightly older, more likely male, and had a higher proportion of patients with cardiovascular disease, diabetes mellitus, chronic pulmonary disease, and an overall higher comorbidity score compared with the HCTZ users (Table 1). More HCTZ users resided in areas with high ultraviolet index than ACEI users. After propensity score matching, 5 211 321 (74.9%) and 5 211 321 (70.3%) of the HCTZ and ACEI users, respectively, were retained. Except for a lower proportion of cardiovascular disease and diabetes patients in the matched ACEI cohort, the matched HCTZ and ACEI cohorts were similar to the unmatched cohorts. All measured covariates were well-balanced after matching.

Incidence Rate Estimates

Within the propensity score–matched cohort, we identified 35 112 BCC cases and 20 976 SCC cases among HCTZ users during 12.6 million person-years of total follow-up time or an average follow-up of 2.4 years, as well as 35 429 BCC events and 20 139 SCC events among ACEI users during 12.6 million person-years of total follow-up time or an average follow-up of 2.4 years. The crude incidence rate per 1000 person-years for BCC was higher than SCC for both exposure cohorts: 2.78 and 1.66 for HCTZ, 2.82 and 1.60 for ACEI, respectively (Table 2). Males had higher BCC rates than females, whereas the rate of SCC by sex was more comparable. The incidence rate for both BCC and SCC increased with age, with the highest rate occurring within the oldest age group. Among racial groups, Caucasians had the highest rates of BCC and SCC. The incidence rate estimates in the unmatched population were similar to those observed in the matched population (BCC rates: 2.41 vs 2.78; SCC rates: 1.43 vs 1.66 for HCTZ and ACEI, respectively) (Supplementary Table 2, available online).

Table 2.

Incidence rates and hazard ratios (HRs) for basal cell carcinoma and squamous cell carcinoma in new users of HCTZ and new users of ACEI in the 1:1 propensity score–matched population^a

Study population	Matched pair, No.	Basal cell carcinoma			Squamous cell carcinoma
		Incidence rate per 1000 person-years		HR (95% CI)	Incidence rate per 1000 person-years		HR (95% CI)
		HCTZ	ACEI		HCTZ	ACEI
Overall	5 211 321	2.78	2.82	0.99 (0.97 to 1.00)	1.66	1.60	1.04 (1.02 to 1.06)
Sex
Male	2 427 149	3.33	3.35	0.99 (0.97 to 1.01)	1.89	1.78	1.06 (1.03 to 1.09)
Female	2 716 796	2.30	2.37	0.98 (0.95 to 1.00)	1.47	1.46	1.01 (0.98 to 1.04)
Age, y
<50	1 268 629	0.50	0.57	0.88 (0.82 to 0.94)	0.13	0.14	0.92 (0.80 to 1.00)
50-59	1 095 379	1.52	1.67	0.91 (0.88 to 0.95)	0.55	0.61	0.91 (0.85 to 0.98)
60-74	1 753 457	3.78	3.84	0.99 (0.97 to 1.01)	2.10	2.01	1.05 (1.02 to 1.08)
≥75	942 684	6.04	6.02	1.01 (0.98 to 1.03)	4.69	4.52	1.04 (1.01 to 1.07)
Race
Caucasian	2 246 784	4.48	4.12	1.09 (1.07 to 1.11)	2.92	2.55	1.15 (1.12 to 1.17)
Black	336 205	0.04	0.05	0.83 (0.55 to 1.27)	0.04	0.05	0.84 (0.54 to 1.30)
Asian	103 864	0.14	0.08	1.72 (1.13 to 2.61)	0.07	0.10	0.72 (0.44 to 1.15)
Pacific Islander	13 377	0.27	0.25	1.09 (0.54 to 2.21)	0.22	0.13	1.69 (0.70 to 4.09)
American Indian	15 997	1.31	0.81	1.58 (1.06 to 2.35)	0.44	0.48	0.95 (0.52 to 1.72)
Unknown	2 260 367	1.91	1.99	0.96 (0.93 to 0.99)	0.86	0.85	1.02 (0.98 to 1.06)

a

ACEI = angiotensin-converting enzyme inhibitor-containing products; CI = confidence interval; HCTZ = hydrochlorothiazide-containing products; HR = hazard ratio.

Hazard Ratio, Incidence Rate Ratio, and Incidence Rate Difference Estimates

Table 2 summarizes the hazard ratio estimates comparing HCTZ and ACEI in the 1:1 propensity score–matched population. There was no difference in the overall risk of BCC (HR = 0.99, 95% CI = 0.97 to 1.00) and only a slight increase in the risk of SCC (HR = 1.04, 95% CI = 1.02 to 1.06) in new users of HCTZ compared with new users of ACEI. An elevated risk of SCC associated with HCTZ was observed for males (HR = 1.06, 95% CI = 1.03 to 1.09) and older-aged patients (HR = 1.05, 95% CI = 1.02 to 1.08 for 60- to 74-year-olds, and HR = 1.04, 95% CI = 1.01 to 1.07 for patients aged 75 years and older). We also found an increased risk of BCC (HR = 1.09, 95% CI = 1.07 to 1.11) and SCC (HR = 1.15, 95% CI = 1.12 to 1.17) among Caucasians, an increased risk of BCC among Asians (HR = 1.72, 95% CI = 1.13 to 2.61), and an increased risk of BCC among American Indians (HR = 1.58, 95% CI = 1.06 to 2.35).

An increased SCC risk (incidence rate ratios [IRR] = 1.19, 95% CI = 1.05 to 1.35) for the overall population and among Caucasians (IRR = 1.27, 95% CI = 1.10 to 1.47) was only observed in users with cumulative HCTZ dose of 50 000 mg or more (Supplementary Table 3, available online), who comprised approximately 3% of all HCTZ users. A dose-response relation was present for SCC (Figure 1) but not for BCC (Figure 2). For SCC, the incidence rate ratio increased from 1.02 (95% CI = 1.00 to 1.04) for less than 10 000 mg to 1.39 (95% CI = 1.07 to 1.80) for 75 000-99 999 mg and 1.32 (95% CI = 0.90 to 1.94) for 100 000 mg or more cumulative doses. Among Caucasians, the incidence rate ratio increased from 1.13 (95% CI = 1.10 to 1.16) for less than 10 00 mg to 1.47 (95% CI = 1.09 to 1.99) for 75 000-79 999 mg and 1.49 (95% CI = 0.99 to 2.25) for 100 000 mg or more cumulative doses.

Figure 1.

Incidence rate ratio for squamous cell carcinoma by cumulative hydrochlorothiazide dose using angiotensin-converting enzyme inhibitor use as reference within the 1:1 propensity score–matched population. A) Incidence rate ratio for squamous cell carcinoma by cumulative hydrochlorothiazide-containing products (HCTZ) dose for all HCTZ users. B) Incidence rate ratio for squamous cell carcinoma by cumulative HCTZ dose for Caucasian HCTZ users. One small data partner did not contribute data to the cumulative dose analyses (18 409 patients, 133 basal cell carcinoma events, 55 squamous cell carcinoma events).

Figure 2.

Incidence rate ratio for basal cell carcinoma by cumulative hydrochlorothiazide dose using angiotensin-converting enzyme inhibitor use as reference within the 1:1 propensity score–matched population. A) Incidence rate ratio for squamous cell carcinoma by cumulative hydrochlorothiazide-containing products (HCTZ) dose for all HCTZ users. B) Incidence rate ratio for squamous cell carcinoma by cumulative HCTZ dose for Caucasian HCTZ users. One small data partner did not contribute data to the cumulative dose analyses (18 409 patients, 133 basal cell carcinoma events, 55 squamous cell carcinoma events).

The estimated adjusted absolute rate difference (per 1000 person-years) for SCC comparing HCTZ use with ACEI use was 0.02 (95% CI = 0.01 to 0.03) in the overall population and 0.07 (95% CI = 0.05 to 0.08) among Caucasians. For 50 000 mg or more cumulative HCTZ dose, the absolute rate difference was 0.09 (95% CI = 0.02 to 0.15) in the overall population and 0.15 (95% CI = 0.04 to 0.25) among Caucasians (Table 3). As expected, the mean age increased as patients accrued high cumulative HCTZ doses, but most baseline covariates remained relatively well balanced between patients with a given cumulative HCTZ dose category and new users of ACEI. (See https://www.sentinelinitiative.org/drugs/assessments/risk-non-melanoma-skin-cancer-patients-treated-hydrochlorothiazide.)

Table 3.

Adjusted incidence rate difference (per 1000 person-years) by cumulative hydrochlorothiazide dose for basal cell carcinoma and squamous cell carcinoma in new users of HCTZ and new users of ACEI in the 1:1 propensity score–matched population^a

Cumulative hydrochlorothiazide dose category	Basal cell carcinoma adjusted incidence rate difference per 1000 person-years (95% CI)		Squamous cell carcinoma adjusted incidence rate difference per 1000 person-years (95% CI)
	Overall	Caucasians	Overall	Caucasians
<10 000 mg	−0.02 (−0.04 to −0.00)	0.12 (0.09 to 0.14)	0.01 (0.00 to 0.02)	0.06 (0.05 to 0.08)
10 000-24 999 mg	−0.11 (−0.14 to −0.08)	−0.06 (−0.10 to −0.02)	−0.02 (−0.04 to −0.01)	0.01 (−0.02 to 0.03)
25 000-49 999 mg	−0.07 (−0.12 to −0.02)	−0.07 (−0.14 to 0.00)	−0.04 (−0.07 to −0.02)	−0.04 (−0.08 to −0.01)
50 000-74 999 mg	−0.17 (−0.28 to −0.05)	−0.28 (−0.45 to −0.11)	0.03 (−0.04 to 0.10)	0.07 (−0.04 to 0.17)
75 000-99 999 mg	−0.04 (−0.29 to 0.20)	0.03 (−0.32 to 0.38)	0.17 (0.01 to 0.33)	0.26 (0.02 to 0.50)
≥100 000 mg	0.13 (−0.23 to 0.50)	0.29 (−0.20 to 0.79)	0.14 (−0.08 to 0.37)	0.26 (−0.06 to 0.59)
≥50 000 mg	−0.08 (−0.19 to 0.02)	−0.11 (−0.26 to 0.04)	0.09 (0.02 to 0.15)	0.15 (0.04 to 0.25)

a

ACEI = angiotensin-converting enzyme inhibitor-containing products; CI = confidence interval; HCTZ = hydrochlorothiazide-containing products.

Sensitivity and Additional Analyses

Excluding outcome events in the first 2 years (the analysis included 5.6 million person-years of follow-up time among HCTZ users) or extending the time period between the NMSC diagnosis and treatment to 60 or 180 days did not materially change the results (Supplementary Table 4, available online). For patients with cumulative HCTZ doses of 100 000 mg or more, 75 000-99 999 mg, and 50 000-74 999 mg, 98%, 85%, and 79%, respectively, had a total HCTZ treatment duration of 5 years or more (Supplementary Table 5, available online). These patients exposed to high cumulative dose were also more likely to receive higher HCTZ strength on the final dispensed prescription (Supplementary Table 6, available online). More than one-half of the patients with a cumulative dose of 100 000 mg or more received a 50 mg strength product on their last HCTZ dispensing, compared with less than 4% of patients with cumulative dose of less than 10 000 mg.

Discussion

In this study of more than 5 million HCTZ users and more than 5 million ACEI users, we observed a small, dose-dependent increased risk of SCC associated with HCTZ use but not for BCC. The SCC risk associated with HCTZ use was elevated among males and Caucasians, whereas the BCC risk with HCTZ use was greater among Caucasians, Asians, and American Indians. The elevated SCC risk was observed only among patients exposed to cumulative HCTZ dose of 50 000 mg or more, who comprised approximately 3% of all HCTZ users and were high treatment-intensity users with a long duration of exposure.

Our findings are supported by biological plausibility and corroborate previous observational studies (3–5). Drug-induced phototoxicity reactions, which are the more common photosensitizing reaction, require greater exposure to the sensitizing agent and are dose-dependent. The evidence for the photocarcinogenic mechanism following phototoxic reactions is less well established but experimental animal studies and other in vivo studies offer some insight. Young et al. (36) demonstrated an increasing rate of skin tumors associated with higher concentrations of psoralen compounds, and Stern et al. (15) observed a dose-related increased risk of SCC in patients exposed to oral psoralen and ultraviolet A light. Severe skin photosensitizing reactions have been observed with HCTZ use (37). A survey by dermatologists reported prolonged exposure to HCTZ among patients with multiple SCC (38).

Compared with the Danish studies (3,4), our study observed smaller effect estimates for BCC and SCC associated with 50 000 mg or more of cumulative HCTZ exposure. An odds ratio (OR) of 1.29 (95% CI = 1.23 to 1.35) for BCC and 3.98 (95% CI = 3.68 to 4.31) for SCC was reported for 50 000 mg or more of cumulative HCTZ dose in a Danish study (3). The difference in the magnitude of risk estimates could be due to 2 main reasons. First, one-third of HCTZ-containing products in Denmark are available in combination with amiloride, also a photosensitizing drug (39), whereas HCTZ-amiloride combination products contribute only 0.1% of all HCTZ dispensed in the United States. A previous observational study (5) found similar magnitude of SCC risk with HCTZ (OR = 1.58, 95% CI = 1.29 to 1.93) and amiloride (OR = 1.80, 95% CI = 1.46 to 2.20). In the Danish study (3), the exclusion of HCTZ-amiloride combination products resulted in a more comparable odds ratio of 1.21 (95% CI = 1.09 to 1.34) for BCC and 1.89 (95% CI = 1.50 to 2.39) for SCC with 50 000 mg or more of cumulative HCTZ exposure. Second, there were differences in the length of follow-up between our study and other published studies. The Danish case-control studies required eligible patients to have at least 10 years of data prior to the date of cancer diagnosis, whereas another study conducted using UK databases reported an odds ratio of 1.33 (95% CI = 0.99 to 1.79) for BCC and 2.73 (95% CI = 1.65 to 4.51) for SCC when restricted to patients with at least 10 years of follow-up (40). Although the overall average follow-up in our study was shorter, more than 90% of patients with 50 000 mg or more of cumulative HCTZ dose had at least 7 years of follow-up data (Supplementary Table 7, available online). Other differences between our study and other studies include different racial composition, skin types, and sun exposure.

This is the largest observational cohort study that examined the association between HCTZ use and NMSC risk in the United States. The sample size allowed us to identify a large number of patients with various levels of cumulative HCTZ dose to examine the dose-response relation between HCTZ use and NMSC risk. The demographically diverse population allowed us to examine the association by key patient characteristics. We used a previously validated outcome algorithm to improve the accuracy of outcome assessment. We were also able to adjust for proxies for ultraviolet exposure, a major risk factor for NMSC.

Our study is subject to several limitations. First, the risk of NMSC among HCTZ users may be underestimated. HCTZ is a known photosensitizing agent, so patients may have taken sun-protective measures. The inclusion of only skin cancers that were excised or treated enhanced the validity of outcome assessment, but it could have also led to underestimation of the NMSC risk. Second, our main analysis may be limited by a relatively short follow-up time, although follow-up was much longer in patients with high cumulative HCTZ dose. Third, we were unable to not examine the risk by skin location. Fourth, there could be residual or unmeasured confounding by skin phototype, susceptibility genes, outdoor occupation or hobby, or diet. However, the use of an active comparator with a similar indication helped reduce measured and unmeasured confounding (41). Although we attempted to adjust for ultraviolet exposure using the zip codes where the individual resided, actual individual-level ultraviolet exposure cannot be ascertained, which could lead to residual confounding. There might be detection bias if HCTZ users visited the dermatologists more frequently, because of warnings on the dispensing bottles, than ACEI users. Finally, although we were able to assess the association by race, a considerable number of patients had missing race information. Similarly, we were unable to reliably examine the risk by Hispanic or Latinx ethnicity because of the large amount of missing ethnicity information.

We identified a small, dose-dependent increased risk of SCC among patients who received HCTZ. Patients should take risk mitigating steps including protection from prolonged sun exposure and clinical follow-up on any skin lesions while using HCTZ-containing products longer term.

Funding

The Sentinel Initiative is funded by the US Food and Drug Administration through the Department of Health and Human Services contract number HHSF223200910006I.

Footnotes

Role of the funder: The funder had no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; or the decision to submit the manuscript for publication.

Disclosures: The authors declare no conflicts of interest.

Author contributions: Conceptualization: EE, MB, ST. Data curation: AC, TZ, NH, AP, SA, TW, ST, ECD. Formal analysis: AC, TZ, NH, AP, SA. Investigation: all authors. Methodology: EE, ST. Project administration: EE, ECD, TW. Software: AC, TZ, NH, AP, SA, TW, ST, ECD. Supervision: EE, ST. Writing—original draft: EE, ST. Writing—review & editing: all authors.

Acknowledgments: Many thanks are due to those who participated in this project. Data partners who provided data/medical records used in the analysis: Aetna, a CVS Health company, Blue Bell, PA; Blue Cross Blue Shield of Massachusetts, Boston, MA; Duke University School of Medicine, Department of Population Health Sciences, Durham, NC, through the Centers for Medicare and Medicaid Services, which provided data; Harvard Pilgrim Health Care Institute, Boston, MA; HealthCore, Inc, Translational Research for Affordability and Quality, Alexandria, VA; HealthPartners Institute, Minneapolis, MN; Humana, Inc, Healthcare Research, Miramar, FL; Kaiser Permanente Colorado Institute for Health Research, Aurora, CO; Kaiser Permanente Center for Health Research Hawaii, Honolulu, HI; Kaiser Permanente Mid-Atlantic States, Mid-Atlantic Permanente Research Institute, Rockville, MD; Kaiser Permanente Northern California, Division of Research, Oakland, CA; Kaiser Permanente Northwest Center for Health Research, Portland, OR; Kaiser Permanente Washington Health Research Institute, Seattle, WA; Marshfield Clinic Research Institute, Marshfield, WI; Meyers Primary Care Institute, Worcester, MA; OptumInsight Life Sciences Inc, Boston, MA; Vanderbilt University Medical Center, Department of Health Policy, Nashville, TN, through the TennCare Division of the Tennessee Department of Finance and Administration, which provided data. The authors thank April Duddy, MS, at Harvard Pilgrim Health Care Institute and Laurel Habel, PhD, at Kaiser Permanente Northern California for their programming and clinical review assistance.

Disclaimer: The views expressed in this paper are those of the authors and are not intended to convey official US Food and Drug Administration policy or guidance.

Prior presentations: The findings from this study were presented at the 36th International Conference on Pharmacoepidemiology & Therapeutic Risk Management Virtual Event (September 16 and 17, 2020).

Data Availability

The data underlying this article cannot be shared publicly due to confidentiality agreement.