ABSTRACT
Background
Hypokalemia is a risk factor for drug-induced QT prolongation. Larger serum-to-dialysate potassium gradients during hemodialysis (HD) may augment the proarrhythmic risks of selective serotonin reuptake inhibitors (SSRIs).
Methods
We conducted a cohort study using 2007–2017 data from the United States Renal Data System and a large dialysis provider to examine if the serum-to-dialysate potassium gradient modifies SSRI cardiac safety. Using a new-user design, we compared 1-year sudden cardiac death (SCD) risk among HD patients newly treated with higher (citalopram, escitalopram) versus lower (fluoxetine, fluvoxamine, paroxetine, sertraline) QT-prolonging potential SSRIs, overall and stratified by baseline potassium gradient (≥4 versus <4 mEq/l). We used inverse probability of treatment-weighted survival models to estimate weighted hazard ratios (HRs) and 95% confidence intervals (CIs) and conducted a confirmatory nested case–control study.
Results
The study included 25 099 patients: 11 107 (44.3%) higher QT-prolonging potential SSRI new users and 13 992 (55.7%) lower QT-prolonging potential SSRI new users. Overall, higher versus lower QT-prolonging potential SSRI use was not associated with SCD [weighted HR 1.03 (95% CI 0.86–1.24)]. However, a greater risk of SCD was associated with higher versus lower QT-prolonging potential SSRI use among patients with baseline potassium gradients ≥4 mEq/l but not among those with gradients <4 mEq/l [weighted HR 2.17 (95% CI 1.16–4.03) versus 0.95 (0.78–1.16)]. Nested case–control analyses yielded analogous results.
Conclusions
The serum-to-dialysate potassium gradient may modify the association between higher versus lower QT-prolonging SSRI use and SCD among people receiving HD. Minimizing the potassium gradient in the setting of QT-prolonging medication use may be warranted.
Keywords: hemodialysis, potassium gradient, SSRI, sudden cardiac death, USRDS
Graphical Abstract
Graphical Abstract.
KEY LEARNING POINTS.
What is already known about this subject?
Larger serum-to-dialysate potassium gradients may augment the proarrhythmic risks of QT-prolonging medications. Use of higher (citalopram, escitalopram) versus lower (fluoxetine, fluvoxamine, paroxetine, sertraline) QT-prolonging potential selective serotonin reuptake inhibitors (SSRIs) is associated with an increased risk of sudden cardiac death (SCD) in the hemodialysis (HD) population.
What this study adds?
Among patients with serum-to-dialysate potassium gradients ≥4 mEq/l, we found that the risk of SCD with higher QT-prolonging potential SSRIs was approximately two times greater than the risk with lower QT-prolonging potential SSRIs. When the gradient was <4 mEq/l, the risk of SCD comparing SSRIs with different QT-prolonging potential was similar.
What impact this may have on practice or policy?
Minimizing the serum-to-dialysate potassium gradient may be warranted when individuals receiving maintenance HD are using QT-prolonging medications such as citalopram and escitalopram.
INTRODUCTION
The rate of sudden cardiac death (SCD) among people with hemodialysis (HD)-dependent kidney failure is 20–30 times higher than the rate of SCD among people in the general population [1]. This differential risk relates in part to the high prevalence of structural heart disease among individuals with HD-dependent kidney failure, which renders them uniquely vulnerable to arrhythmias provoked by electrolyte derangements, treatment-related electrolyte shifts and medications. More than 250 medications have QT-interval prolonging effects [2] and many of these drugs are frequently prescribed to patients treated with maintenance HD [3].
Selective serotonin reuptake inhibitors (SSRIs) are the most common class of antidepressants used by the HD population [4], and all available SSRIs can prolong the QT-interval [5–7]. Existing data show that among people receiving HD, treatment with citalopram and escitalopram, SSRIs with higher QT-prolonging potential, associates with an increased risk of SCD compared with SSRIs with lower QT-prolonging potential [8]. Pharmaceutical regulatory agencies, such as the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA), have issued drug safety communications warning that citalopram and escitalopram can cause fatal arrhythmias, particularly in people with risk factors for QT-prolongation such as hypokalemia [9–12]. Dialyzing against larger serum-to-dialysate potassium gradients increases the extent of potassium removal during HD and can result in transient intra- or postdialysis hypokalemia [13–15]. Numerous studies have linked larger serum-to-dialysate potassium gradients to more pronounced HD-induced QT prolongation [16–20] and QT dispersion [16, 21–23].
Exposure to larger serum-to-dialysate potassium gradients in the setting of QT-prolonging medications, such as citalopram and escitalopram, may augment the proarrhythmic effects of these drugs [2], further enhancing SCD risk. Therefore, minimizing the serum-to-dialysate potassium gradient may be one strategy to mitigate the risk of SCD among patients receiving HD, who are taking QT-prolonging medications, but corroborative data are lacking. We undertook this study to examine if the serum-to-dialysate potassium gradient modifies the association between higher versus lower QT-prolonging potential SSRI use and the risk of SCD in the HD population.
MATERIALS AND METHODS
This study was approved by the University of North Carolina Institutional Review Board (18-0297) and a waiver of consent was granted because of the study's large size, data anonymity and retrospective nature.
Data source
We used data from the United States Renal Data System (USRDS) linked at the patient level with data from the electronic healthcare records of a large US dialysis provider organization (DaVita). The USRDS database includes the End Stage Renal Disease (ESRD) Medical Evidence Report, a registration form nephrologists complete for all patients at the outset of maintenance dialysis to establish Medicare eligibility for individuals <65 years of age and to reclassify previously eligible Medicare beneficiaries as having ESRD; the ESRD Death Notification form, a reporting form submitted by dialysis providers to notify the Centers for Medicare & Medicaid Services of patient deaths; and Medicare standard analytic files, including enrollment information and final action hospital (Part A), physician/supplier (Part B) and prescription drug (Part D) insurance claims [24]. The dialysis provider operates >2500 outpatient dialysis clinics throughout the USA and its electronic health record captures detailed demographic, clinical, laboratory and dialysis treatment data. In general, laboratory tests are performed biweekly or monthly and analyzed at a central laboratory and HD treatment parameters are recorded on a treatment-to-treatment basis.
Study overview and design
We conducted a retrospective cohort study using an active-comparator new-user design (Fig. 1A) [25] to evaluate the comparative risk of SCD among patients prescribed SSRIs with higher (citalopram, escitalopram) versus lower (fluvoxamine, fluoxetine, paroxetine, sertraline) QT-prolonging potential [2, 8] who were and were not exposed to larger baseline serum-to-dialysate potassium gradients. We subsequently conducted a nested case–control study within the SSRI new-user cohort to confirm the findings from the cohort analyses and to consider the serum-to-dialysate potassium gradient closest in time to SCD events (Fig. 1B).
Figure 1:
Study designs. K+ gradient, serum-to-dialysate potassium gradient; SCD, sudden cardiac death; SSRI, selective serotonin reuptake inhibitor.
Study population
To construct the SSRI new-user cohort, we identified patients in the USRDS database receiving in-center maintenance HD with continuous baseline Medicare coverage (Parts A, B and D) who newly initiated SSRI therapy from 1 January 2007 to 30 December 2017. We then applied the following exclusion criteria: age <18 years at the beginning of the baseline period, time on dialysis ≤90 days at the beginning of the baseline period, receipt of hospice care during the baseline period, presence of an implantable cardioverter defibrillator, receipt of fewer than three HD treatments at the large dialysis organization during the last 30 days of the baseline period, missing baseline serum and/or dialysate potassium values and missing covariate information.
Study exposure and effect modifier
We used Medicare Part D prescription drug claims to identify SSRI new users and defined the index date as the date of the first prescription for an SSRI after a 180-day washout period free of SSRI prescription fills. We classified citalopram and escitalopram as SSRIs with higher QT-prolonging potential and fluvoxamine, fluoxetine, paroxetine and sertraline as SSRIs with lower QT-prolonging potential [2, 8].
The effect modifier of interest was the baseline serum-to-dialysate potassium gradient, calculated from the baseline serum and dialysate potassium values closest to, but preceding or on, the index date. For each patient, we computed the baseline potassium gradient (mEq/l) as predialysis serum potassium (mEq/l) minus dialysate potassium (mEq/l). In primary analyses we examined a potassium gradient as a binary variable (≥4 and <4 mEq/l) based on existing data showing associations between gradients ≥4 mEq/l and higher risks of death and hospitalization [26]. Because the risk of adverse outcomes associated with the potassium gradient may be incremental, we also considered the potassium gradient as a multilevel variable: ≥4 mEq/l, 3–<4 mEq/l and <3 mEq/l.
Study outcome
We obtained dates and causes of death from the ESRD Death Notification form. The primary outcome of interest was SCD within 1 year of the index date. We defined SCD using the established USRDS definition, death due to a cardiac arrhythmia or cardiac arrest listed as the primary cause of death on the Death Notification form (Supplementary data, Table S1) [24]. We also considered two broader cardiac outcomes in secondary analyses: a composite of SCD or hospitalized ventricular arrhythmia and, separately, cardiovascular mortality (Supplementary data, Table S1).
Study covariates
We identified covariates in the 180 days prior to the index date, including patient demographics, comorbid conditions, laboratory values, dialysis treatment parameters, prescription medication use and metrics of healthcare utilization (Supplementary data, Tables S2–S4).
Statistical analysis
All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA). Baseline characteristics of higher and lower QT-prolonging potential SSRI initiators stratified by baseline potassium gradient are displayed as count and percentage for categorical variables and as mean ± standard deviation (SD) or median [interquartile range (IQR)] for continuous variables. We compared baseline covariate distributions between groups using absolute standardized differences and considered a standardized difference >0.10 as indicating an imbalance between groups [10].
We used an on-treatment analytic approach and followed individuals from the index date to the first occurrence of an outcome, censoring or competing event. Censoring events included: a change of dialysis modality to home HD or peritoneal dialysis; kidney transplantation; recovery of kidney function; loss of Medicare Part A, B or D coverage; loss to follow-up; index SSRI discontinuation (using a 7-day grace period); switch to a nonindex SSRI; completion of 1 year of follow-up and study end (31 December 2017). We treated death due to a cause other than SCD (i.e. non-SCD) as a competing event.
We used Fine and Gray proportional subdistribution hazards models [11] to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for SCD, comparing higher versus lower QT-prolonging potential SSRIs. We then assessed if the baseline potassium gradient had a modifying effect by including a multiplicative interaction term between the study exposure (new use of an SSRI with higher versus lower QT-prolonging potential) and the baseline gradient. We used inverse probability of treatment (IPT) weighting, a propensity score method, for confounding control and estimated weighted (i.e. adjusted) HR by applying IPT weights in our regression models [27].
Secondary nested case–control study
Because serum-to-dialysate potassium gradients may vary across time and the gradient value closest to an outcome event may be the most physiologically relevant value to the risk of SCD, we conducted a prespecified secondary nested case–control study within our SSRI new-user cohort. Briefly, we identified all SCD cases that occurred in our new-user cohort after SSRI initiation. Then, using risk-set sampling, we identified up to 10 controls for each SCD case, matching on age ±5 years, sex and time since maintenance dialysis initiation (<1.0, 1.0–1.9, 2.0–2.9 and ≥3.0 years). Controls were identified on the SCD event date of their corresponding SCD case. Sampled controls could subsequently become cases. After SCD cases and their matched controls were identified, we determined which SSRI type (i.e. higher or lower QT-prolonging potential) cases and controls were used on the event date and identified the serum potassium and dialysate potassium values closest to, but preceding or on, the event date. For each patient, we computed the corresponding potassium gradient (mEq/l) as predialysis serum potassium (mEq/l) minus dialysate potassium (mEq/l). To estimate the association between SSRI type and SCD, we used multivariable conditional logistic regression to estimate odds ratios (ORs) and 95% CIs, adjusting for patient demographics as well as baseline comorbid conditions, laboratory values, dialysis treatment parameters, prescription medication use and metrics of healthcare utilization. We then assessed if the potassium gradient had a modifying effect by including a multiplicative interaction term between the study exposure (use of an SSRI with higher versus lower QT-prolonging potential) and the gradient value closest to the event date.
Post hoc analyses
We conducted two post hoc analyses to evaluate the robustness of our primary study findings. First, rather than excluding the 1989 patients with missing baseline serum-to-dialysate potassium gradient data and/or missing covariate data, we imputed these missing data using the Markov chain Monte Carlo multiple imputation method [28] and then evaluated if the baseline potassium gradient modifies the association between higher versus lower QT-prolonging potential SSRI use and the risk of SCD. Second, in a separate post hoc analysis, we created a four-level exposure variable that captured both the SSRI type and serum-to-dialysate potassium gradient category and examined SCD associations. The exposure groups were new users of higher QT-prolonging potential SSRIs with baseline potassium gradients ≥4 mEq/l, new users of higher QT-prolonging potential SSRIs with baseline potassium gradients <4 mEq/l, new users of lower QT-prolonging potential SSRIs with baseline potassium gradients ≥4 mEq/l and new users of lower QT-prolonging potential SSRIs with baseline potassium gradients <4 mEq/l (referent). Because we considered a four-level exposure in these analyses, we estimated propensity scores and generated IPT weights using standard methods for multicategorical exposures [27, 29].
RESULTS
Characteristics of the new-user cohort
Figure 2A displays the flow diagram of the new-user cohort creation. Overall, 25 099 patients receiving HD were included: 11 107 (44.3%) higher QT-prolonging potential SSRI new users and 13 992 (55.7%) lower QT-prolonging potential SSRI new users. The average patient age was 60 ± 15 years, 52.4% were women, 35.3% were Black, 19.7% were Hispanic and the most common cause of kidney failure was diabetes (50.3%). In total, 2244 (8.9%) SSRI new users had a baseline serum-to-dialysate potassium gradient of ≥4 mEq/l and 22 855 (91.1%) SSRI new-users had a baseline serum-to-dialysate potassium gradient <4 mEq/l. The mean serum potassium concentration was 6.0 ± 0.6 mEq/l among patients with baseline gradients ≥4 mEq/l and 4.7 ± 0.6 mEq/l among patients with baseline gradients <4 mEq/l. The use of dialysate potassium concentrations <2 mEq/l was more common among patients with baseline potassium gradients ≥4 mEq/l than among those with baseline potassium gradients <4 mEq/l (57.4% versus 3.4% of patients). Across the follow-up period, the majority of patients remained in the same serum-to-dialysate potassium gradient category as their baseline gradient category (Supplementary data, Table S5).
Figure 2:
Flow diagrams depicting assembly of the study cohorts. aThis exclusion criteria resulted in the exclusion of patients who were receiving their maintenance dialysis treatments at non-DaVita dialysis provider organizations and DaVita-treated patients who received less than three treatments at an outpatient DaVita dialysis clinic during the last month of the baseline period. K+ gradient, serum-to-dialysate potassium gradient; LDO, large dialysis organization; SCD, sudden cardiac death; SSRI, selective serotonin reuptake inhibitor; USRDS, United States Renal Data System.
The propensity score distributions of the SSRI groups exhibited substantial overlap, indicating the groups were highly comparable. Table 1 and Supplementary data, Tables S6–S7 show the characteristics of the exposure groups stratified by potassium gradient before and after IPT weighting. Before IPT weighting, baseline covariates were generally well balanced between treatment groups within potassium gradient strata (standardized differences ≤0.10), with some exceptions (e.g. race, Hispanic ethnicity, depression). After IPT weighting, all baseline covariates were well balanced.
Table 1.
Select characteristics of higher and lower QT-prolonging potential SSRI new users stratified by baseline serum-to-dialysate potassium gradient before IPT weighting
| Baseline K+ gradient ≥4 mEq/l | Baseline K+ gradient <4 mEq/l | |||||
|---|---|---|---|---|---|---|
| Characteristic | Higher QT-prolonging potential SSRIs (n = 951) | Lower QT-prolonging potential SSRIs (n = 1293) | Std diff | Higher QT-prolonging potential SSRIs (n = 10 156) | Lower QT-prolonging potential SSRIs (n = 12 699) | Std diff |
| Age (years), mean ± SD | 56 ± 16 | 55 ± 14 | 0.07 | 61 ± 15 | 60 ±15 | 0.06 |
| Female | 493 (51.8) | 624 (48.3) | 0.07 | 5415 (53.3) | 6623 (52.2) | 0.02 |
| Race | ||||||
| Black | 239 (25.1) | 301 (23.3) | 0.04 | 3758 (37.0) | 4550 (35.8) | 0.02 |
| White | 668 (70.2) | 898 (69.5) | 0.02 | 5984 (58.9) | 7468 (58.8) | <0.01 |
| Other | 44 (4.6) | 94 (7.3) | 0.11 | 414 (4.1) | 681 (5.4) | 0.06 |
| Hispanic | 233 (24.5) | 399 (30.9) | 0.14 | 1772 (17.4) | 2542 (20.0) | 0.07 |
| Cause of ESRD | ||||||
| Diabetes | 499 (52.5) | 661 (51.1) | 0.03 | 5160 (50.8) | 6302 (49.6) | 0.02 |
| Hypertension | 196 (20.6) | 272 (21.0) | 0.01 | 2654 (26.1) | 3413 (26.9) | 0.02 |
| Glomerular disease | 116 (12.2) | 161 (12.5) | 0.01 | 947 (9.3) | 1282 (10.1) | 0.03 |
| Other | 140 (14.7) | 199 (15.4) | 0.02 | 1395 (13.7) | 1702 (13.4) | 0.01 |
| Time on maintenance dialysis (years) | ||||||
| <1.0 | 115 (12.1) | 138 (10.7) | 0.05 | 1696 (16.7) | 2014 (15.9) | 0.02 |
| 1.0–1.9 | 162 (17.0) | 199 (15.4) | 0.05 | 1879 (18.5) | 2377 (18.7) | 0.01 |
| 2.0–2.9 | 134 (14.1) | 202 (15.6) | 0.04 | 1519 (15.0) | 1888 (14.9) | <0.01 |
| ≥3.0 | 540 (56.8) | 754 (58.3) | 0.03 | 5062 (49.8) | 6420 (50.6) | 0.01 |
| Anxiety | 248 (26.1) | 301 (23.3) | 0.07 | 2384 (23.5) | 2933 (23.1) | 0.01 |
| Depression | 382 (40.2) | 443 (34.3) | 0.12 | 3674 (36.2) | 4151 (32.7) | 0.07 |
| Arrhythmia | 270 (28.4) | 353 (27.3) | 0.02 | 3213 (31.6) | 3843 (30.3) | 0.03 |
| Conduction disorder | 81 (8.5) | 105 (8.1) | 0.01 | 977 (9.6) | 1181 (9.3) | 0.01 |
| Heart failure | 467 (49.1) | 616 (47.6) | 0.03 | 4938 (48.6) | 5953 (46.9) | 0.04 |
| Ischemic heart disease | 455 (47.8) | 580 (44.9) | 0.06 | 4963 (48.9) | 5969 (47.0) | 0.04 |
| Peripheral artery disease | 355 (37.3) | 446 (34.5) | 0.06 | 3733 (36.8) | 4241 (33.4) | 0.07 |
| Corrected calcium (mg/dl) | ||||||
| <8.5 | 148 (15.6) | 223 (17.2) | 0.03 | 1387 (13.7) | 1800 (14.2) | 0.02 |
| 8.5–10.2 | 750 (78.9) | 1006 (77.8) | 0.05 | 8287 (81.6) | 10 305 (81.1) | 0.01 |
| >10.2 | 53 (5.6) | 64 (4.9) | 0.03 | 482 (4.7) | 594 (4.7) | <0.01 |
| Hemoglobin (g/dl) | ||||||
| <9.5 | 101 (10.6) | 158 (12.2) | 0.05 | 1218 (12.0) | 1446 (11.4) | 0.02 |
| 9.5–12.0 | 608 (63.9) | 797 (61.6) | 0.05 | 6925 (68.2) | 8779 (69.1) | 0.02 |
| >12.0 | 242 (25.4) | 338 (26.1) | 0.02 | 2013 (19.8) | 2474 (19.5) | 0.01 |
| Predialysis systolic BP (mmHg) | ||||||
| <130 | 123 (12.9) | 153 (11.8) | 0.03 | 1882 (18.5) | 2198 (17.3) | 0.03 |
| 130–149 | 286 (30.1) | 394 (30.5) | 0.01 | 3216 (31.7) | 4072 (32.1) | 0.01 |
| 150–169 | 333 (35.0) | 466 (36.0) | 0.02 | 3260 (32.1) | 4028 (31.7) | 0.01 |
| ≥170 | 209 (22.0) | 280 (21.7) | 0.01 | 1798 (17.7) | 2401 (18.9) | 0.03 |
| Vascular access | ||||||
| Fistula | 563 (59.2) | 797 (61.6) | 0.05 | 6101 (60.1) | 7988 (62.9) | 0.06 |
| Graft | 192 (20.2) | 274 (21.2) | 0.03 | 2447 (24.1) | 3017 (23.8) | 0.01 |
| Catheter | 196 (20.6) | 222 (17.2) | 0.09 | 1608 (15.8) | 1694 (13.3) | 0.07 |
| Treatment time >240 min | 152 (16.0) | 215 (16.6) | 0.02 | 1716 (16.9) | 2129 (16.8) | <0.01 |
| Ultrafiltration volume ≥2 l | 734 (77.2) | 1025 (79.3) | 0.05 | 5994 (59.0) | 7638 (60.1) | 0.02 |
| Total baseline hospital admissions | ||||||
| 0 | 317 (33.3) | 450 (34.8) | 0.03 | 3830 (37.7) | 5327 (41.9) | 0.09 |
| 1–2 | 346 (36.4) | 498 (38.5) | 0.04 | 3935 (38.7) | 4750 (37.4) | 0.03 |
| 3–4 | 165 (17.4) | 216 (16.7) | 0.02 | 1569 (15.4) | 1757 (13.8) | 0.05 |
| ≥5 | 123 (12.9) | 129 (10.0) | 0.09 | 822 (8.1) | 865 (6.8) | 0.05 |
| ≥1 med with any TdP riska | 513 (53.9) | 693 (53.6) | 0.01 | 5202 (51.2) | 6601 (52.0) | 0.02 |
| ≥1 med with known TdP riska | 85 (8.9) | 123 (9.5) | 0.02 | 1035 (10.2) | 1357 (10.7) | 0.02 |
Values are presented as n (%) unless stated otherwise. All covariates were measured during the 180-day baseline period. Because effect modification was examined, assessments of covariate balance comparing new users of higher versus lower QT-prolonging potential SSRIs were made within each serum-to-dialysate potassium strata. Supplementary data, Tables S6 and S7 display the full list of baseline covariates considered in our analyses for both the unweighted and IPT-weighted cohorts.
The CredibleMeds website (https://crediblemeds.org) is a reliable online clinical resource with up-to-date information about medications that can cause QT-prolongation and/or TdP. CredibleMeds classifies QT-prolonging medications as having a known, possible or conditional TdP risk. Lists of medications falling into each category are provided in Supplementary data, Table S4. Medications classified as having any level of TdP risk are those falling into any of the three CredibleMeds categories.
BP, blood pressure; med, medication; K+ gradient, serum-to-dialysate potassium gradient; Std dff, standardized difference.
Cohort study findings
A total of 443 SCDs occurred during the 1-year follow-up period (26 events at a rate of 102.1/1000 person-years among higher QT-prolonging potential SSRI new users with a baseline serum-to-dialysate potassium gradient ≥4 mEq/l, 186 events at a rate of 66.9/1000 person-years among higher QT-prolonging potential SSRI new users with a baseline gradient <4 mEq/l, 15 events at a rate of 42.8/1000 person-years among lower QT-prolonging potential SSRI new users with a baseline gradient ≥4 mEq/l and 216 events at a rate of 62.0/1000 person-years among lower QT-prolonging potential SSRI new users with a baseline gradient <4 mEq/l).
Figure 3A and Supplementary data, Table S8, display the results from analyses evaluating the associations between higher versus lower QT-prolonging SSRI new-use and SCD risk, overall and stratified by the binary serum-to-dialysate potassium gradient. Higher versus lower QT-prolonging potential SSRI use was not significantly associated with SCD risk in the overall cohort [weighted HR 1.03 (95% CI 0.86–1.24)]. However, an increased risk of SCD was associated with higher versus lower QT-prolonging potential SSRI use among patients with a baseline potassium gradient ≥4 mEq/l, but not among those with a gradient <4 mEq/l [weighted HR0 2.17 (95% CI 1.16–4.03) versus 0.95 (0.78–1.16)].
Figure 3:
Higher versus lower QT-prolonging SSRIs and SCD risk, overall and stratified by serum-to-dialysate potassium gradient. (a) In the cohort study, Fine and Gray proportional subdistribution hazards models were used to estimate HRs and corresponding 95% CIs and IPT weighting was used for confounding control. (b) In the nested case–control study, conditional logistic regression models were used to estimate ORs and corresponding 95% CIs, and multivariable adjustment was used for confounding control. CI, confidence intervals; HR, hazard ratio; K+ gradient; serum-to-dialysate potassium gradient; OR, odds ratio; ref., referent; SCD, sudden cardiac death; SSRI, selective serotonin reuptake inhibitor.
Secondary analyses evaluating alternative cardiac outcomes produced consistent results. New use of a higher versus lower QT-prolonging potential SSRI was only associated with an increased risk of the composite outcome and cardiovascular mortality among patients with a baseline serum-to-dialysate potassium gradient ≥4 mEq/l and not among patients with a gradient <4 mEq/l (Table 2). Additionally, analyses considering the baseline potassium gradient as a multilevel variable suggested that the gradient may incrementally modify the association between SSRI type and SCD (Table 3).
Table 2.
Higher versus lower QT-prolonging SSRIs and alternative cardiac outcome risk, overall and stratified by baseline serum-to-dialysate potassium gradient in the cohort study
| Sudden cardiac death or hospitalized ventricular arrhythmia—secondary outcome | ||||
|---|---|---|---|---|
| Exposure | n | Events, n | Crude HR (95% CI) | Weighted HR (95% CI) |
| Overall | ||||
| Lower QT-prolonging-potential | 13 992 | 244 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 11 107 | 224 | 1.15 (0.96–1.38) | 1.03 (0.86–1.24) |
| K+ gradient ≥4 mEq/l | ||||
| Lower QT-prolonging-potential | 1293 | 16 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 951 | 26 | 2.22 (1.19–4.15) | 2.02 (1.10–3.70) |
| K+ gradient <4 mEq/l | ||||
| Lower QT-prolonging-potential | 12 699 | 228 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 10 156 | 198 | 1.08 (0.89–1.31) | 0.96 (0.79–1.16) |
| Cardiovascular death—secondary outcome | ||||
| Exposure | n | Events, n |
Crude
HR (95% CI) |
Weighted
HR (95% CI) |
| Overall | ||||
| Lower QT-prolonging-potential | 13 992 | 329 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 11 107 | 294 | 1.12 (0.95–1.31) | 0.99 (0.84–1.16) |
| K+ gradient ≥4 mEq/l | ||||
| Lower QT-prolonging-potential | 1293 | 21 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 951 | 33 | 2.16 (1.24–3.73) | 1.88 (1.10–3.23) |
| K+ gradient <4 mEq/l | ||||
| Lower QT-prolonging-potential | 12 699 | 308 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 10 156 | 261 | 1.05 (0.89–1.24) | 0.92 (0.78–1.09) |
In the cohort study, Fine and Gray proportional subdistribution hazards models were used to estimate HRs and corresponding 95% CIs, and IPT weighting was used for confounding control.
CI, confidence interval; HR, hazard ratio; K+ gradient, serum-to-dialysate potassium gradient; No., number; ref., referent; SSRI, selective serotonin reuptake inhibitor.
Table 3.
Higher versus lower QT-prolonging SSRIs and SCD risk, overall and stratified by multilevel serum-to-dialysate potassium gradient
| Cohort study | ||||
|---|---|---|---|---|
| Exposure | n | Events, n | Crude HR (95% CI) | Weighted HR (95% CI) |
| Overall | ||||
| Lower QT-prolonging-potential | 13 992 | 231 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 11 107 | 212 | 1.15 (0.96–1.38) | 1.03 (0.86–1.24) |
| K+ gradient ≥4 mEq/l | ||||
| Lower QT-prolonging-potential | 1293 | 15 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 951 | 26 | 2.37 (1.25–4.48) | 2.17 (1.16–4.03) |
| K+ gradient 3 to <4 mEq/l | ||||
| Lower QT-prolonging-potential | 4063 | 53 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 3134 | 56 | 1.38 (0.94–2.01) | 1.24 (0.85–1.81) |
| K+ gradient <3 mEq/l | ||||
| Lower QT-prolonging-potential | 8636 | 163 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 7022 | 130 | 0.97 (0.77–1.22) | 0.86 (0.68–1.09) |
| Nested case–control study | ||||
| Exposure | n | SCD cases, n |
Crude
OR (95% CI) |
Adjusted
OR (95% CI) |
| Overall | ||||
| Lower QT-prolonging-potential | 2748 | 231 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 2111 | 211 | 1.20 (0.99–1.47) | 1.15 (0.93–1.43) |
| K+ gradient ≥4 mEq/l | ||||
| Lower QT-prolonging-potential | 208 | 15 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 132 | 22 | 2.59 (1.28–5.21) | 2.81 (1.30–6.07) |
| K+ gradient 3 to <4 mEq/l | ||||
| Lower QT-prolonging-potential | 789 | 52 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 561 | 49 | 1.36 (0.91–2.05) | 1.13 (0.73–1.76) |
| K+ gradient <3 mEq/l | ||||
| Lower QT-prolonging-potential | 1751 | 164 | 1.00 (ref.) | 1.00 (ref.) |
| Higher QT-prolonging-potential | 1418 | 140 | 1.06 (0.83–1.43) | 1.06 (0.82–1.37) |
In the cohort study, Fine and Gray proportional subdistribution hazards models were used to estimate HRs and corresponding 95% CIs, and IPT weighting was used for confounding control. In the nested case–control study, conditional logistic regression models were used to estimate ORs and corresponding 95% CIs, and multivariable adjustment was used for confounding control.CI, confidence interval; HR, hazard ratio; K+ gradient, serum-to-dialysate potassium gradient; No., number; OR, odds ratio; ref., referent; SSRI: selective serotonin reuptake inhibitor.
Finally, negative control outcome analyses evaluating the association between the new use of a higher versus lower QT-prolonging potential SSRI and death due to a cause other than SCD produced null results in the full cohort and in subgroups of patients with baseline potassium gradients ≥4 mEq/l and <4 mEq/l (Supplementary data, Table S9).
Nested case–control study findings
Of the 443 SCD cases in the SSRI new-user cohort, 442 SCD cases matched successfully to 4417 controls (Fig. 2B). Supplementary data, Table S10, displays characteristics of the case–control cohort. Matching variables (age, sex and time since maintenance dialysis initiation) were well balanced between cases and controls. Compared with controls, SCD cases had a higher prevalence of cardiovascular comorbidities, cardiovascular medication use and metrics of healthcare utilization. The median time from the date of potassium gradient determination to the event date was 1 day (IQR 0–2). In total, 37 (8.4%) cases and 303 (6.9%) controls had a serum-to-dialysate potassium gradient of ≥4 mEq/l immediately preceding or on the SCD event date. Cases and controls with serum-to-dialysate potassium gradients ≥4 mEq/l immediately preceding or on the SCD event date had a mean serum potassium concentration of 6.0 ± 0.5 mEq/l, and those with potassium gradients <4 mEq/l had a mean serum potassium of 4.6 ± 0.6 mEq/l. The use of dialysate potassium concentrations <2 mEq/l was more common among patients with potassium gradients ≥4 mEq/l than among those with potassium gradients <4 mEq/l (52.9% versus 2.1% of patients).
Nested case–control analyses considering the potassium gradient closest to the SCD event date produced results consistent with those from the cohort analyses. Figure 3B and Supplementary data, Table S8, display the SCD associations. Overall, the use of a higher versus lower QT-prolonging potential SSRI was not significantly associated with SCD [adjusted OR 1.15 (95% CI 0.93–1.43)]. However, increased odds of SCD were observed with higher versus lower QT-prolonging potential SSRI use among patients with a potassium gradient ≥4 mEq/l but not among those with a gradient <4 mEq/l [adjusted OR 2.82 (95% CI 1.31–6.09) versus 1.07 (0.86–1.34)]. Findings from case–control analyses considering a multilevel potassium gradient variable were also consistent with the corresponding cohort study analyses (Table 3).
Post hoc analyses
Post hoc analyses using imputed gradient and covariate data and separately considering a four-level exposure variable capturing both the SSRI type and serum-to-dialysate potassium gradient category produced results consistent with our primary cohort study analyses, higher QT-prolonging potential SSRI new users with baseline serum-to-potassium gradients ≥4 mEq/l had a greater risk of SCD (Supplementary data, Tables S11–S12).
DISCUSSION
In evaluating the cardiac safety of SSRIs, we found that the risk of SCD associated with the new use of an SSRI with higher (citalopram, escitalopram) versus lower (fluoxetine, fluvoxamine, paroxetine, sertraline) QT-prolonging potential is modified by the serum-to-dialysate potassium gradient. Specifically, in the setting of a baseline serum-to-potassium gradient ≥4 mEq/l, the SCD risk associated with higher QT-prolonging potential SSRIs was more than two times greater than the risk associated with lower QT-prolonging potential SSRIs. In contrast, among patients with a baseline potassium gradient <4 mEq/l, the risk of SCD associated with higher versus lower QT-prolonging potential SSRIs was similar.
A prior study linking higher QT-prolonging potential SSRIs to adverse cardiac outcomes in the HD population showed that the risk of SCD associated with citalopram and escitalopram was higher among patients with risk factors for drug-induced QT prolongation, including female sex, advanced age, preexisting cardiac conduction disorders and concomitant QT-prolonging medication use [8]. However, dialysis treatment–related factors, such as dialysate composition, were not considered. In the present study, we found that the risk of SCD associated with higher QT-prolonging potential SSRI use was augmented in the presence of larger serum-to-dialysate potassium gradients. These findings were consistent across two study designs—a cohort analysis that considered baseline potassium gradients and a nested case–control analysis that considered gradients immediately preceding SCD events. Our finding that SCD risk was not different between higher and lower QT-prolonging potential SSRIs in the setting of serum-to-potassium gradients <4 mEq/l may relate to more cautious prescribing of higher QT-prolonging potential SSRIs and enhanced electrocardiogram (ECG) monitoring in high-risk patients across time [30, 31]. In 2011 and 2012, pharmaceutical regulatory agencies, including the FDA and EMA, issued drug safety communications warning that citalopram and escitalopram can cause QT prolongation, potentially leading to fatal arrhythmias such as torsade de pointes (TdP) [9–12].
SSRI-induced QT prolongation occurs via drug blockade of myocardial potassium channels encoded by the human ether-a-go-go–related gene [32, 33], which causes a delay in ventricular repolarization. This delay in ventricular repolarization manifests as QT prolongation on an ECG. At therapeutic doses, citalopram and escitalopram prolong the QT interval more than other SSRIs [5–7], increasing the risk of TdP and SCD. Extracellular hypokalemia can cause QT prolongation on its own and also enhance the degree of drug-induced delays in ventricular repolarization [34]. During HD, the serum-to-dialysate potassium gradient drives the extent and speed of dialytic potassium removal from the extracellular space. The elevated risk of SCD associated with higher QT-prolonging potential SSRIs in the setting of larger potassium gradients may relate in part to the occurrence of gradient-induced intradialytic and/or postdialysis hypokalemia. Hypokalemia is a well-established risk factor for drug-induced QT prolongation [35, 36], with many reported cases of citalopram- and escitalopram-induced QT-prolongation, arrhythmias and SCD occurring among patients with this electrolyte abnormality [11, 12, 37]. Moreover, ECG monitoring studies in patients treated with HD have shown that larger serum-to-dialysate potassium gradients are associated with more pronounced QT prolongation [16–20] and QT dispersion [16, 21–23] immediately following the HD procedure. Thus it is plausible that rapid shifts in serum potassium that are induced by larger serum-to-dialysate potassium gradients may enhance the magnitude of drug-induced QT prolongation, ultimately increasing SCD risk.
While the data sources and study designs used in this investigation preclude definitive conclusions about the mechanism(s) underlying the observed association of an elevated risk of SCD with higher QT-prolonging potential SSRIs in the setting of larger serum-to-dialysate potassium gradients, these findings suggest the occurrence of a drug–dialysis treatment interaction. In drug–dialysis treatment interactions, the dialysis treatment either enhances or diminishes the pharmacologic effect(s) of a drug. Such interactions can occur through pharmacokinetic or pharmacodynamic mechanisms and can enhance either the adverse or the beneficial effects of a drug. In our study, it appears as though the QT-prolonging effects of higher QT-prolonging potential SSRIs were augmented by the QT-prolonging effects of larger serum-to-dialysate potassium gradients, elevating the risk of SCD. While this pharmacoepidemiologic study is one of the first explicit investigations of a potential drug–dialysis treatment interaction between QT-prolonging medications and potassium gradients, drug–dialysis treatment interactions could occur in other clinical scenarios as well. For example, the disorienting effects of insomnia medications could be amplified by the cerebral ischemic effects of intradialytic hypotension occurring in the setting of overly rapid fluid removal during HD.
Clinicians should be mindful of the potential for drug–dialysis treatment interactions, especially in scenarios when the risk of medication-related adverse events could be augmented by the physiologic effects of the dialysis treatment. In the case of higher QT-prolonging potential SSRIs, prior to prescribing citalopram or escitalopram, clinicians should determine if patients have histories of frequent exposure to serum-to-potassium gradients ≥4 mEq/l and, if so, make efforts to reduce the potassium gradient by lowering predialysis serum potassium levels and/or raising dialysate potassium concentrations. Options for lowering predialysis serum potassium levels include reduction of dietary potassium intake and/or the use of potassium binders. In addition, it is important to consider if other risk factors for drug-induced QT prolongation, such as advanced age, female sex, structural heart disease and concurrent QT-prolonging medication use, are present since patients with these risk profiles may have an even higher risk of drug-induced SCD [38]. In cases where there are no appropriate alternatives and prescribing QT-prolonging medications, such as citalopram and escitalopram, is unavoidable, clinicians should perform a baseline ECG to measure the QT interval prior to the initiation of a QT-prolonging drug and then monitor the interval with ECGs every 3–6 months in accordance with the American Heart Association guidance [38].
Strengths of the present study include the use of a linked dataset with detailed administrative claims and clinical information, which facilitated adjustment of our analyses for relevant biochemical indexes and dialysis treatment parameters not available in claims-only databases, as well as the use of two complementary study designs. However, our findings must be considered in light of certain study limitations. First, because this study was observational, residual confounding may remain. Reassuringly, negative control outcome analyses in our new-user cohort study yielded null results in both the ≥4 and <4 mEq/l groups, suggesting that the influence of unmeasured residual confounding was minimal. Second, although we were able to capture and control for ECG performance in the month prior to SSRI new use, we were unable to determine if ECG findings informed clinicians’ SSRI prescribing decisions. Third, even though we defined SCD using the established USRDS definition, it is possible that outcome misclassification may have occurred. Reassuringly, analyses considering two broader cardiac outcomes generated consistent results. Fourth, because postdialysis serum potassium levels were not available in our dataset, we were unable to determine if the rate of serum potassium decline during HD and/or the occurrence of postdialysis hypokalemia drove the observed elevated risk of SCD associated with higher QT-prolonging potential SSRIs among patients with potassium gradients ≥4 mEq/l. Future mechanistic studies are needed. Finally, it is possible that our study results may not generalize to other QT-prolonging medications.
In conclusion, we showed that the serum-to-dialysate potassium gradient modifies the association between higher versus lower QT-prolonging potential SSRI use and the risk of SCD among people with HD-dependent kidney failure. The risk of SCD associated with higher versus lower QT-prolonging SSRIs was elevated among patients with potassium gradients ≥4 mEq/l but not in patients with potassium gradients <4 mEq/l, suggesting that minimization of the serum-to-dialysate potassium gradient in the setting of QT-prolonging medication use may be warranted. Future research examining the underlying mechanisms of the observed association as well as the impact of serum-to-dialysate potassium gradients on the risk of SCD with other QT-prolonging medications is needed.
Supplementary Material
ACKNOWLEDGEMENTS
Some of the data reported here have been provided by the USRDS under Data Use Agreement 2018-23d (to J.E.F.). This manuscript underwent privacy review by a National Institute of Diabetes and Digestive and Kidney Diseases project officer and received clearance. The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as official policy or interpretation of the US government. Additionally, some of the data reported here have been supplied by DaVita Clinical Research. DaVita Clinical Research had no role in the design or implementation of this study or in the decision to publish.
Contributor Information
Magdalene M Assimon, University of North Carolina Kidney Center , Division of Nephrology and Hypertension, Department of Medicine, UNC School of Medicine, Chapel Hill, NC, USA.
Patrick H Pun, Division of Nephrology, Department of Medicine, Duke University School of Medicine, Durham, NC, USA; Duke Clinical Research Institute, Durham, NC, USA.
Sana M Al-Khatib, Duke Clinical Research Institute, Durham, NC, USA; Division of Cardiology, Duke University Medical Center, Durham, NC, USA.
Maurice Alan Brookhart, Department of Population Health Sciences, Duke University School of Medicine, Durham, NC, USA.
Bradley N Gaynes, Department of Psychiatry, UNC School of Medicine, Chapel Hill, NC, USA; Department of Epidemiology, UNC Gillings School of Global Public Health, Chapel Hill, NC, USA.
Wolfgang C Winkelmayer, Selzman Institute for Kidney Health, Section of Nephrology, Baylor College of Medicine, Houston, TX, USA.
Jennifer E Flythe, University of North Carolina Kidney Center , Division of Nephrology and Hypertension, Department of Medicine, UNC School of Medicine, Chapel Hill, NC, USA; Cecil G. Sheps Center for Health Services Research, University of North Carolina, Chapel Hill, NC, USA.
FUNDING
M.M.A. and J.E.F. were supported by R03 HS026801 awarded by the Agency for Healthcare Research and Quality. M.M.A., P.H.P., S.M.A., M.A.B., W.C.W. and J.E.F. are supported by R01 HL152034 awarded by the National Heart, Lung, and Blood Institute of the National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality or the National Institutes of Health.
DATA AVAILABILITY STATEMENT
Because of contractual data use agreements, the authors cannot make the data and materials used in this study available to other investigators. Interested parties can contact the USRDS Coordinating Center to obtain USRDS data and DaVita Clinical Research to obtain electronic healthcare records data from the large dialysis organization.
CONFLICT OF INTEREST STATEMENT
M.M.A. has received investigator-initiated research funding from the Renal Research Institute, a subsidiary of Fresenius Medical Care North America, and honoraria from the American Society of Nephrology and the International Society of Nephrology. P.H.P. has received investigator-initiated research funding unrelated to this project from Medtronic, honoraria from the American Society of Nephrology and the National Kidney Foundation and consulting fees from Fresenius Kidney Care North America, AstraZeneca, Janssen, Relypsa and Ardelyx. S.M.A. has received research funding from Medtronic, Boston Scientific and Abbott and has received speaking fees from Medtronic. M.A.B. serves on scientific advisory boards for American Academy of Allergy, Asthma, and Immunology; Amgen; Atara Biotherapeutics; Brigham and Women's Hospital; Gilead; Merck; National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) and Vertex and receives consulting fees and owns equity in Target RWE. W.C.W. has received honoraria for consultancy or scientific advice to Akebia/Otsuka, AstraZeneca, Bayer, Boehringer Ingelheim/Lilly, GlaxoSmithKline, Janssen, Merck, Pharmacosmos, Reata and Relypsa. In the last 3 years, J.E.F. has received speaking honoraria from the American Society of Nephrology and multiple universities, as well as investigator-initiated research funding unrelated to this project from the Renal Research Institute, a subsidiary of Fresenius Kidney Care North America; serves on a medical advisory board for Fresenius Kidney Care North America and a scientific advisory board and Data and Safety Monitoring Committee for the NIDDK and has received consulting fees from Fresenius Kidney Care North America and AstraZeneca.
REFERENCES
- 1. Makar MS, Pun PH.. Sudden cardiac death among hemodialysis patients. Am J Kidney Dis 2017;69:684–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Woosley RL, Heise CW, Romero KA.. QTdrugs List. www.Crediblemeds.org (10 January 2022, date last accessed). [Google Scholar]
- 3. Assimon MM, Wang L, Pun PHet al. Use of QT prolonging medications by hemodialysis patients and individuals without end-stage kidney disease. J Am Heart Assoc 2020;9:e015969. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. United States Renal Data System . USRDS annual data report: epidemiology of kidney disease in the United States. Bethesda, MD: National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, 2021. [Google Scholar]
- 5. Castro VM, Clements CC, Murphy SNet al. QT interval and antidepressant use: a cross sectional study of electronic health records. BMJ 2013;346:f288. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Funk KA, Bostwick JR.. A comparison of the risk of QT prolongation among SSRIs. Ann Pharmacother 2013;47:1330–41. [DOI] [PubMed] [Google Scholar]
- 7. Beach SR, Kostis WJ, Celano CMet al. Meta-analysis of selective serotonin reuptake inhibitor-associated QTc prolongation. J Clin Psychiatry 2014;75:e441–9. [DOI] [PubMed] [Google Scholar]
- 8. Assimon MM, Brookhart MA, Flythe JE.. Comparative cardiac safety of selective serotonin reuptake inhibitors among individuals receiving maintenance hemodialysis. J Am Soc Nephrol 2019;30:611–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. US Food and Drug Administration. FDA Drug Safety Communication: revised recommendations for celexa (citalopram hydrobromide) related to a potential risk of abnormal heart rhythms with high doses. https://www.fda.gov/drugs/drug-safety-and-availability/fda-drug-safety-communication-revised-recommendations-celexa-citalopram-hydrobromide-related (10 January 2022, date last accessed). [Google Scholar]
- 10. US Food and Drug Administration . FDA Drug Safety Communication: abnormal heart rhythms associated with high doses of Celexa (citalopram hydrobromide). https://www.fda.gov/drugs/drug-safety-and-availability/fda-drug-safety-communication-abnormal-heart-rhythms-associated-high-doses-celexa-citalopram (10 January 2022, date last accessed). [Google Scholar]
- 11. European Medicines Agency . The Pharmacovigilance Working Party (PhVWP): November 2011 plenary meeting–safety concerns. https://www.ema.europa.eu/documents/report/monthly-report-pharmacovigilance-working-party-phvwp-november-2011-plenary-meeting_en.pdf (10 January 2022, date last accessed). [Google Scholar]
- 12. European Medicines Agency . The pharmacovigilance working party of the european medicines agency: october 2011 plenary meeting–safety concerns. https://www.ema.europa.eu/documents/report/monthly-report-pharmacovigilance-working-party-phvwp-october-2011-plenary-meeting_en.pdf (10 January 2022, date last accessed). [Google Scholar]
- 13. Blumberg A, Roser HW, Zehnder Cet al. Plasma potassium in patients with terminal renal failure during and after haemodialysis; relationship with dialytic potassium removal and total body potassium. Nephrol Dial Transplant 1997;12:1629–34. [DOI] [PubMed] [Google Scholar]
- 14. Ohnishi T, Kimachi M, Fukuma Set al. Postdialysis hypokalemia and all-cause mortality in patients undergoing maintenance hemodialysis. Clin J Am Soc Nephrol 2019;14:873–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Correa S, Scovner KM, Tumlin JAet al. Electrolyte changes in contemporary hemodialysis: a secondary analysis of the monitoring in dialysis (MiD) study. Kidney360 2021;2:695–707. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Lorincz I, Matyus J, Zilahi Zet al. QT dispersion in patients with end-stage renal failure and during hemodialysis. J Am Soc Nephrol 1999;10:1297–1302. [DOI] [PubMed] [Google Scholar]
- 17. Genovesi S, Dossi C, Vigano MRet al. Electrolyte concentration during haemodialysis and QT interval prolongation in uraemic patients. Europace 2008;10:771–7. [DOI] [PubMed] [Google Scholar]
- 18. Di Iorio B, Torraca S, Piscopo Cet al. Dialysate bath and QTc interval in patients on chronic maintenance hemodialysis: pilot study of single dialysis effects. J Nephrol 2012;25:653–60. [DOI] [PubMed] [Google Scholar]
- 19. Severi S, Vecchietti S, Cavalcanti Set al. Electrocardiographic changes during hemodiafiltration with different potassium removal rates. Blood Purif 2003;21:381–8. [DOI] [PubMed] [Google Scholar]
- 20. Severi S, Grandi E, Pes Cet al. Calcium and potassium changes during haemodialysis alter ventricular repolarization duration: in vivo and in silico analysis. Nephrol Dial Transplant 2008;23:1378–86. [DOI] [PubMed] [Google Scholar]
- 21. Yetkin E, Ileri M, Tandogan Iet al. Increased QT interval dispersion after hemodialysis: role of peridialytic electrolyte gradients. Angiology 2000;51:499–504. [DOI] [PubMed] [Google Scholar]
- 22. Buemi M, Aloisi E, Coppolino Get al. The effect of two different protocols of potassium haemodiafiltration on QT dispersion. Nephrol Dial Transplant 2005;20:1148–54. [DOI] [PubMed] [Google Scholar]
- 23. Delanaye P, Krzesinski F, Dubois BEet al. A simple modification of dialysate potassium: its impact on plasma potassium concentrations and the electrocardiogram. Clin Kidney J 2021;14:390–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. United States Renal Data System. ESRD: analytical methods. https://adr.usrds.org/2021/end-stage-renal-disease/esrd-analytical-methods (10 January 2022, date last accessed). [Google Scholar]
- 25. Ray WA. Evaluating medication effects outside of clinical trials: new-user designs. Am J Epidemiol 2003;158:915–20. [DOI] [PubMed] [Google Scholar]
- 26. Brunelli SM, Spiegel DM, Du Mond Cet al. Serum-to-dialysate potassium gradient and its association with short-term outcomes in hemodialysis patients. Nephrol Dial Transplant 2018;33:1207–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Brookhart MA, Wyss R, Layton JBet al. Propensity score methods for confounding control in nonexperimental research. Circ Cardiovasc Qual Outcomes 2013;6:604–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Schafer JL. Multiple imputation: a primer. Stat Methods Med Res 1999;8:3–15. [DOI] [PubMed] [Google Scholar]
- 29. Imbens GW. The role of the propensity score in estimating dose-response functions. Biometrika 2000;87:706–10. [Google Scholar]
- 30. Gerlach LB, Kim HM, Yosef Met al. Assessing responsiveness of health systems to drug safety warnings. Am J Geriatr Psychiatry 2018;26:476–83. [DOI] [PubMed] [Google Scholar]
- 31. Lalji HM, McGrogan A, Bailey SJ.. An analysis of antidepressant prescribing trends in England 2015-2019. J Affect Disord Rep 2021;6:100205. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Witchel HJ, Pabbathi VK, Hofmann Get al. Inhibitory actions of the selective serotonin re-uptake inhibitor citalopram on HERG and ventricular L-type calcium currents. FEBS Lett 2002;512:59–66. [DOI] [PubMed] [Google Scholar]
- 33. Chae YJ, Jeon JH, Lee HJet al. Escitalopram block of hERG potassium channels. Naunyn Schmiedebergs Arch Pharmacol 2014;387:23–32. [DOI] [PubMed] [Google Scholar]
- 34. Yang T, Roden DM.. Extracellular potassium modulation of drug block of IKr. Implications for torsade de pointes and reverse use-dependence. Circulation 1996;93:407–11. [DOI] [PubMed] [Google Scholar]
- 35. Trinkley KE, Page RL 2nd, Lien Het al. QT interval prolongation and the risk of torsades de pointes: essentials for clinicians. Curr Med Res Opin 2013;29:1719–26. [DOI] [PubMed] [Google Scholar]
- 36. Turner JR, Rodriguez I, Mantovani Eet al. Drug-induced proarrhythmia and torsade de pointes: a primer for students and practitioners of medicine and pharmacy. J Clin Pharmacol 2018;58:997–12. [DOI] [PubMed] [Google Scholar]
- 37. Kogut C, Crouse EB, Vieweg WVet al. Selective serotonin reuptake inhibitors and torsade de pointes: new concepts and new directions derived from a systematic review of case reports. Ther Adv Drug Saf 2013;4:189–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Tisdale JE, Chung MK, Campbell KB.et al. Drug-induced arrhythmias: a scientific statement from the American Heart Association. Circulation 2020;142:e214–33. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Because of contractual data use agreements, the authors cannot make the data and materials used in this study available to other investigators. Interested parties can contact the USRDS Coordinating Center to obtain USRDS data and DaVita Clinical Research to obtain electronic healthcare records data from the large dialysis organization.