Torsade de pointes: A nested case–control study in an integrated healthcare delivery system

Abstract Background TdP is a form of polymorphic ventricular tachycardia which develops in the setting of a prolonged QT interval. There are limited data describing risk factors, treatment, and outcomes of this potentially fatal arrhythmia. Objective Our goals were as follows: (1) to validate cases presenting with Torsade de Pointes (TdP), (2) to identify modifiable risk factors, and (3) to describe the management strategies used for TdP and its prognosis in a real‐world healthcare setting. Methods Case–control study (with 2:1 matching on age, sex, and race/ethnicity) nested within the Genetic Epidemiology Research on Aging (GERA) cohort. Follow‐up of the cohort for case ascertainment was between January 01, 2005 and December 31, 2018. Results A total of 56 cases of TdP were confirmed (incidence rate = 3.6 per 100,000 persons/years). The average (SD) age of the TdP cases was 74 (13) years, 55 percent were female, and 16 percent were non‐white. The independent predictors of TdP were potassium concentration <3.6 mEq/L (OR = 10.6), prior history of atrial fibrillation/flutter (OR = 6.2), QTc >480 ms (OR = 4.4) and prior history of coronary artery disease (OR = 2.6). Exposure to furosemide and amiodarone was significantly greater in cases than in controls. The most common treatment for TdP was IV magnesium (78.6%) and IV potassium repletion (73.2%). The in‐hospital and 1‐year mortality rates for TdP cases were 10.7% and 25.0% percent, respectively. Conclusions These findings may inform quantitative multivariate risk indices for the prediction of TdP and could guide practitioners on which patients may qualify for continuous ECG monitoring and/or electrolyte replacement therapy.


| INTRODUC TI ON
Torsade de Pointes (TdP), first described in 1966, (La, 1966) is a potentially life-threatening polymorphic ventricular tachycardia associated with prolongation of the QT interval on the 12-lead electrocardiogram (ECG). (Fontaine et al., 1982;Gupta et al., 2007) Because TdP is a rare diagnosis without specific International Classification of Diseases (ICD)-9 or ICD-10 codes, there are limited data describing clinical risk factors and management strategies for this malignant arrhythmia. (Kay et al., 1983).
The established risk factors for TdP include older age, female sex, heart disease, electrolyte disorders (especially hypokalemia and hypomagnesemia), renal or hepatic dysfunction, bradycardia or rhythms with long pauses, treatment with more than 1 QTprolonging drug, and genetic predisposition. (Drew et al., 2010;Faber et al., 1994).
The goal of this study was twofold: first, to validate a case series of patients presenting with TdP and compare these patients to age-, sex-, and race/ethnicity-2:1 matched controls in order to identify modifiable risk factors and secondly, to describe the management strategies used for TdP and the short-term survival and long-term survival in a real-world healthcare setting.

| Population and study design
Kaiser Permanente Medical Care Program (KPMCP) of Northern California is an integrated healthcare delivery system serving approximately 4.5 million members. The membership is stable with less than 10% turnover annually overall and <3%-5% among members 65 years of age and older and/or who have a chronic condition. The program delivers comprehensive inpatient and outpatient care to its members and captures many aspects of its care using its multiple comprehensive clinical and administrative databases. Kaiser's population is ethnically and socio-economically diverse and is broadly representative of the Northern California population. (Krieger, 1992).

Our study was based on the GERA (Genetic Epidemiology
Research on Aging) cohort. The goal of the GERA cohort was to create a large, multiethnic, and comprehensive population-based resource into the genetic and environmental basis of common agerelated diseases and their treatment, and factors influencing healthy aging and longevity. (Banda et al., 2015;Kvale et al., 2015) The GERA cohort consists of a diverse cohort of 110,266 members of the We used ICD-9 code 427.1 and ICD-10 code I47.2, augmented with natural language processing (NLP), to identify all cohort members who developed paroxysmal ventricular tachycardia between January 1, 2005, and December 31, 2018. We identified a total of 2,528 subjects with a code for paroxysmal ventricular tachycardia, and, among them, 132 had a mention of "torsade" or "torsades" in the medical record. Manual chart review, including examination of 12-lead ECGs, telemetry strips, ambulatory ECGs, and medical documentation, was performed for these 132 patients to confirm "true" TdP events. The final cohort of patients with a confirmed TdP event by a cardiologist (N.M.) comprised 56 persons. The 76 patients without validated TdP events typically mentioned "torsade" within the differential diagnosis given baseline risk factors for TdP (prolonged baseline QTc, electrolyte abnormalities, use of multiple QT-prolonging pharmacotherapies) or documented possible "torsade" based on inpatient telemetry monitoring. However, none of these patients had a confirmed "torsade" event documented in the electronic medical record by a physician. A second physician investigator (C.I.) reviewed 25 randomly selected eligible cases for review, and agreement with the first adjudicator was 100% (12 true cases out of 25). Thus, our code plus NLP algorithm to identify and validate TdP had a positive predictive value of 42% (56/132). Each of these TdP patients was matched with 2 controls with the same birth year, sex, and race/ethnicity. Moreover, control subjects were assigned an "index" date which was the hospitalization date of the corresponding matched case.
In both cases and controls, gender, date of birth, and race/ethnicity were obtained from the patient's demographic file. Diabetes was ascertained through linkage with the KPNC Diabetes Registry. (Karter, 2006) The prevalence of other comorbidities, laboratory values, body mass index (BMI), and smoking status was extracted from the EHR. A list of ICD-9, ICD-10, and current procedural terminology, fourth version (CPT4) codes is provided in (online-only) Appendix SA. Baseline QT c interval (with Bazett correction), available from a Regional ECG database, was from the most recent 12-lead ECG prior to index admission for TdP or assigned index date for controls, going back up to 5 years. Event QT c (only in cases) was determined from the baseline 12-lead ECG available during the hospitalization where the TdP event occurred. The methodology and validity of the QT measurement at KPNC can be found elsewhere. (Iribarren et al., 2013(Iribarren et al., , 2014 In brief, all ECGs in the KPNC system are obtained using cardiographs manufactured by Philips Medical Systems (Andover, Massachusetts) which have a proprietary algorithm for QT measurement. (Kligfield et al., 2006) Bazett's correction for heart rate was used because is widely utilized in clinical practice. (Bazett, 1920) Hypokalemia was defined as a potassium level of lower than 3.6 millimoles per liter (mmol/L). Hypomagnesemia was defined as a serum magnesium level of lower than 1.5 milligrams per deciliter (mg/dl).
Late-stage chronic kidney disease was defined by an estimated glomerular filtration rate (Levey et al., 2009) of less than 30 milliliter per min (ml/min) or by codes for chronic dialysis. Active prescriptions for QT-prolonging medications (as listed in Crediblemeds.org, accessed on 9/10/2020) as well as cholesterol-lowering drugs prescribed within a 60-day window prior to index hospitalization for TdP cases or index date for controls were identified utilizing the KPNC outpatient pharmacy database. Discharge disposition for cases and 30day, 6-month, and one-year mortality for both cases and controls were ascertained using complimentary validated sources of mortality in KPNC. (Arellano et al., 1984).

| Statistical analysis
Case-control differences were assessed using the t test for continuous variables and the chi-square test for categorical variables. To determine independent predictors of TdP, we used conditional logistic regression with stepwise entry of covariates and a p value threshold to stay in the model of 0.05. To retain all subjects in the model, we created a dummy variable for missing LVH. All the statistical analyses were performed with SAS version 9.4. (SAS Institute Inc., North Carolina, USA).

| RE SULTS
Our TdP case validation over a 14-year period provides a populationbased estimate of 3.6 cases of TdP per 100,000 persons per year.
The 56 patients with confirmed TdP were, on average, 74 years old (SD = 13 years); 25 (45%) were men and 31 (55%) women. The majority of TdP cases were white (84%), and a small portion were black/African American (3.6%), Asian & Pacific Islander (3.6%), or mixed race (8.9%) ( Table 1). There were no significant differences between cases and controls in body mass index, smoking status, or cholesterol-lowering medication use. The baseline QT c was 458 ± 40 ms in cases and 432 ± 35 ms in controls (p < .0001).
Prolonged QTc (i.e., >480 ms) was present in 30 percent of cases and 10 percent of controls (p = .0008). The mean (SD) delta (event baseline) QTc in cases was 48 (53) ms. The average serum potassium level was lower in cases than in controls (3.7 vs. 4.3 mEq/L; p < .0001), and the proportion with low potassium (<3.6 mEq/L) was greater in cases than in controls (30 vs. 4.5%; p < .0001). Average magnesium concentrations did not significantly differ between TdP cases and controls, but the proportion with low magnesium was higher in cases than in controls (10 vs. 0%; p = .01).
The ten most common QT-prolonging prescriptions administered to TdP cases within 60 days of index hospitalization and overall percentages of "other" and "any" QT-prolonging medications (along with corresponding prescription utilization in controls also within 60 days prior to their index date) are shown in Table 3. TdP cases were significantly more likely to have been prescribed furosemide (17.7% vs. 6.5%; p = .007), amiodarone (7.3% vs. 0.0%; p = .001), and "any" QT-prolonging drugs (80.2% vs. 57.6%; p = .0003). On the other hand, controls were more likely to have been prescribed "other" (i.e., other than the 10 most common among cases) QT-prolonging drugs (43.8% vs. 69.8%; p < .0001). The full list of QT-prolonging drugs is provided as (online-only) Appendix SB.
Stepwise conditional logistic regression analysis was performed to determine independent risk factors for TdP development within Six out of 56 TdP patients (10.7%) died during the hospitalization. The underlying cause of death was aspiration pneumonia (n = 2), cardiogenic shock (n = 1), severe sepsis (n = 1), ventricular tachycardia (n = 1), and myocarditis (n = 1). Eight died within 1 year after discharge, yielding a cumulative mortality of 25.0% at 1 year.
After hospital discharge, 1 death occurred within 30 days, 1 death between 30 days and 6 months, and 6 deaths between 6 months and 1 year. All controls were alive one year after their "index date" hospitalization.

| DISCUSS ION
Our study confirms that TdP is a rare occurrence (3.6 cases per 100,000 per year) with serious consequences: 55 percent developed sustained ventricular fibrillation, 11 percent died in the hospital, and 17 percent died within 12 months of discharge. Our rate (which was derived from a population-based cohort) is lower than the incidence reported in a Belgian study (34 per 100,000 per year) that was based on patients admitted to a tertiary hospital in a 3-year period (Vandael et al., 2017a)     interval and terminating TdP. Given the known QT-prolonging side effects of amiodarone, anti-arrhythmic treatment with intravenous amiodarone was used less commonly.
We highlight one example from our case series describing the development and management of TdP. He is an 82-year-old man with history of coronary artery bypass grafting, dual-chamber permanent pacemaker for complete heart block, ischemic cardiomyopathy with a left ventricular ejection fraction of 45%, and prostate cancer on chronic ketoconazole therapy who presented with syncope at home. Upon arrival to the emergency department, he suffered recurrent loss of consciousness and 12-lead ECG was performed demonstrating polymorphic ventricular tachycardia (Figure 1). He was immediately treated with initiation of intravenous lidocaine, intravenous magnesium sulfate, and intravenous potassium chloride.
All QT-prolonging medications, including his chronic ketoconazole treatment, were stopped. He was successfully treated by increasing the lower rate limit on his dual-chamber permanent pacemaker to shorten the QT interval and prevent future TdP. He was discharged from the hospital alive with no recurrent polymorphic ventricular tachycardia.
We recognize several limitations in our case-control study.
First, our sample size was small given the rare nature of TdP, limiting statistical power. Second, although we leveraged a real-world population with ethnic diversity, all patients included were insured, and thus, our results may not generalize to uninsured populations.
Third, we did not consider genetic data since this is part of a separate effort and is beyond the scope of this manuscript. Fourth, we did not develop RISQ-PATH scores because of missing hs-CRP values. (Vandael, Vandenberk, Vandenberghe, Spriet, et al., 2017). F I G U R E 1 Illustrative example of a 12-lead ECG of a patient presenting with TdP TdP is an uncommon but potentially fatal arrhythmia due to selective potential of action potential durations in certain layers of the ventricular myocardium leading to dispersion of repolarization and a prolonged QT interval. In light of the recent COVID-19 pandemic and the use of off-label treatment strategies for SARS-CoV-2 which are known to prolong QTc, there is heightened importance of identifying risk predictors for TdP development reported showing intravenous potassium and magnesium repletion and overdrive pacing to be the most commonly utilized strategies.
Further research is needed to understand the combined effects of predisposing genetic makeup, medication effects, and clinical risk factors. Thereafter, appropriate mitigation and management strategies can be fully understood.

ACK N OWLED G EM ENT
This work was supported by the National Heart, Lung, and Blood Institute (RO1HL140924; PI, Iribarren C).

CO N FLI C T O F I NTE R E S T
The authors have no conflicts to disclose.

This study was approved by the Kaiser Foundation Research Institute
Institutional Review Board and the requirement for informed consent was waived.

DATA AVA I L A B I L I T Y S TAT E M E N T
No individual data are available, but authors would contemplate reasonable requests.