Cumulative additive risk from combining multiple QT-prolonging drugs — CredibleMeds classification, hERG block, and Tisdale-score torsades prediction
Every additive-risk calculation begins with an accurate baseline. Before any drug interaction can be quantified, clinicians must establish the patient's starting QT interval, corrected for heart rate, and build a complete list of every drug — prescription, OTC, and herbal — the patient is actually taking. Roughly 3% of all hospitalized patients receive at least one QT-prolonging drug, and 5–10% receive two or more concurrently without anyone having checked for additive risk.
The QT interval spans from the start of the Q wave (ventricular depolarization) to the end of the T wave (ventricular repolarization). Because QT shortens mechanically as heart rate rises, it must always be rate-corrected before comparison across patients or across time:
• Bazett's formula: QTcB = QT / √RR — historically dominant, but over-corrects at high heart rates and under-corrects at low heart rates • Fridericia's formula: QTcF = QT / RR^(1/3) — cube-root correction, more accurate across the physiological HR range and now preferred by FDA thorough-QT (TQT) study guidance (ICH E14) • Framingham and Hodges formulas: linear corrections, used less often outside research settings
Measurement technique matters as much as the formula. The tangent method draws a line along the steepest slope of the descending T wave down to the isoelectric baseline; the intersection marks T-wave end. Superimposed U waves are excluded when clearly separable. Measurements are typically averaged over 3–5 consecutive beats in lead II or V5, the leads with the clearest T-wave offset.
• Normal QTc: <450 ms (men), <460 ms (women) • Borderline: 450–470 ms (men), 460–480 ms (women) • Prolonged: >470 ms (men), >480 ms (women) • High danger zone: ≥500 ms, or an increase of ≥60 ms from a documented baseline — both independently associated with markedly increased torsades de pointes (TdP) risk
Additive QT risk is invisible unless every contributing drug is actually captured on the list. In practice, reconciliation fails for predictable reasons:
• Over-the-counter and herbal products are omitted: diphenhydramine, loperamide (at supratherapeutic doses), and certain "detox" supplements carry hERG-blocking activity • PRN (as-needed) orders are undercounted: a single IV ondansetron dose for nausea is often not flagged the same way a scheduled oral drug is • Electrolyte-altering drugs are missed as indirect contributors: loop and thiazide diuretics do not block hERG directly, but the hypokalemia and hypomagnesemia they cause potentiate every other QT-prolonging drug on the list • Renal and hepatic function are not rechecked at each admission: reduced clearance of a chronic QT-prolonging drug (e.g., methadone in acute kidney injury) silently raises effective exposure without any new prescription being written
The CredibleMeds database (maintained by the Arizona Center for Education and Research on Therapeutics, AZCERT) is the standard reference clinicians query during reconciliation, and it is updated continuously as new pharmacovigilance signals emerge — a drug reconciliation performed against a stale printed list is already out of date.
Not all QT-prolonging drugs carry equal risk. CredibleMeds stratifies medications into three evidence-based tiers, and the underlying pharmacology — how tightly each drug blocks the cardiac hERG (Kv11.1) potassium channel that governs the rapid delayed-rectifier current IKr — explains why some drugs prolong QT by 2 ms and others by 35 ms at standard doses.
CredibleMeds (QTdrugs.org), curated by AZCERT since 2009, is the standard clinical reference for QT-prolonging drug risk and is embedded in most EHR clinical decision support engines:
• Known Risk of TdP (List 1): clear evidence from case reports, case series, or controlled studies that the drug prolongs QT AND is associated with a risk of TdP when taken as directed — e.g., amiodarone, sotalol, methadone, haloperidol IV, citalopram, azithromycin, ondansetron • Possible Risk of TdP (List 2): evidence that the drug can prolong QT but insufficient evidence linking it directly to TdP at therapeutic doses — e.g., quetiapine, clarithromycin, tamoxifen • Conditional Risk of TdP (List 3): associated with TdP only under certain conditions (overdose, hypokalemia, congenital long-QT, or drug interaction raising plasma levels) — e.g., trazodone, fluconazole, metronidazole
A fourth, non-risk category exists for drugs to specifically avoid in congenital long-QT syndrome regardless of measured QT effect.
The hERG channel (human Ether-à-go-go-Related Gene, Kv11.1) conducts IKr, the rapid component of the delayed-rectifier potassium current that repolarizes the ventricular action potential. Its outer vestibule is unusually promiscuous — aromatic residues (Tyr652, Phe656) line an unusually large, flexible pore cavity that accommodates a remarkably broad range of drug structures, from macrolide antibiotics to antipsychotics to antihistamines:
• Patch-clamp IC50 (concentration blocking 50% of current) is measured in heterologous expression systems (HEK293 or CHO cells transfected with hERG cDNA) • Safety margin = hERG IC50 ÷ free therapeutic Cmax. A margin <30-fold is considered a signal requiring further clinical QT study under ICH S7B/E14 guidance • Citalopram: IC50 ≈ 4.4 µM vs. free Cmax ≈ 0.3 µM at 40 mg/day → margin ~15-fold, explaining its FDA boxed warning limiting the dose to 40 mg/day (20 mg in poor CYP2C19 metabolizers, hepatic impairment, or age >60) • Azithromycin: weaker hERG blocker (IC50 in the hundreds of µM) but produces measurable QTc prolongation (~5–9 ms) partly through IKs and late-INa effects in addition to IKr • Ondansetron: dose-dependent — the FDA restricted single IV doses to ≤16 mg (from a former 32 mg) after data showed a 20 ms QTc increase at the higher dose within 30 minutes of infusion
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| Citalopram 40 mg/day | SSRI antidepressant | Direct IKr (hERG) block; dose-dependent, FDA boxed warning | Known Risk — mean ΔQTc ≈ +18.5 ms at 40 mg |
| Azithromycin 500 mg | Macrolide antibiotic | IKr block + late-INa effect; weak CYP3A4 inhibition | Known Risk — mean ΔQTc ≈ +9 ms |
| Ondansetron 8 mg IV | 5-HT3 antagonist, antiemetic | Potent, rapid-onset IKr block | Known Risk — ΔQTc up to +20 ms per dose |
| Hydrochlorothiazide 25 mg | Thiazide diuretic | No direct hERG block; lowers serum K⁺/Mg²⁺ | Indirect risk amplifier only |
Two mechanisms compound QT risk when drugs are combined: pharmacokinetic interactions that raise the effective plasma concentration of a QT-prolonging drug, and pharmacodynamic additivity, where each drug's independent IKr blockade sums at the level of the ventricular action potential regardless of whether plasma levels change at all. A clinically meaningful risk calculator must model both.
Citalopram is metabolized predominantly by CYP2C19 (demethylation) with a minor CYP3A4/CYP2D6 contribution. Two independent factors converge in this regimen:
• CYP2C19 poor-metabolizer genotype (~2–5% of the population, higher in East Asian ancestry) reduces citalopram clearance by roughly 50%, directly raising steady-state plasma concentration • Azithromycin is a weak CYP3A4 inhibitor; while its inhibition of citalopram's minor 3A4 pathway is modest alone, in a CYP2C19 poor metabolizer the residual 3A4 route becomes proportionally more important, compounding the exposure increase • Combined modeled effect: +35% AUC increase, translating to a proportionally higher free-drug concentration bathing cardiac hERG channels
This is precisely the interaction category flagged by ICH M12 drug-interaction guidance and by clinical pharmacology consult services: no new QT-prolonging mechanism is introduced, but an existing one is amplified pharmacokinetically.
Independent of plasma levels, when two drugs both block IKr through the same channel, their delay-of-repolarization effects combine at the level of the cardiac action potential. Population pharmacokinetic-pharmacodynamic (PK-PD) modeling of combination QT studies supports several approaches, ranked from simplest to most mechanistic:
• Simple additive model: ΔQTc_total ≈ ΔQTc_drug1 + ΔQTc_drug2 + ... — the conservative default used by most bedside calculators, validated against retrospective TdP case series showing that patients on 2+ Known-Risk drugs have roughly double the observed QTc prolongation of patients on 1 drug • Emax / sigmoidal interaction models: capture supra-additive ("synergistic") effects when two drugs bind overlapping or allosterically-coupled sites on the same channel — more accurate but requires drug-specific concentration-effect curves rarely available outside specialized centers • Physiologically-based cardiac modeling (e.g., CiPA initiative's in silico ventricular action potential models): integrates block of multiple ion currents (IKr, ICaL, INa-late) simultaneously to predict net repolarization change; used in regulatory submission but not yet standard bedside practice
For this patient: citalopram (+18.5 ms) + azithromycin (+9 ms) + ondansetron (+20 ms) ≈ +47.5 ms from direct effects, plus an estimated +10 ms attributable to the PK-driven citalopram exposure increase, yielding the modeled +58 ms cumulative shift — projecting a composite QTc near 496 ms, solidly in the high-risk zone.
QTc alone is an incomplete predictor of arrhythmic risk — the same 490 ms QTc carries very different torsades de pointes probability in a young man with normal electrolytes versus an elderly woman on a loop diuretic with a low potassium. The Tisdale risk score integrates demographic, electrolyte, and drug-count risk factors into a single validated point total that stratifies patients by actual event probability.
Derived and validated in ICU patients receiving a QT-prolonging drug, the Tisdale score assigns points to independent risk factors identified by multivariable regression against observed TdP events:
• Age ≥68 years: +1 point • Female sex: +1 point (women have a longer baseline QTc and greater TdP susceptibility, likely related to lower IKs reserve) • Loop diuretic use: +1 point (potassium and magnesium wasting) • Serum potassium ≤3.5 mmol/L: +2 points • Admission QTc ≥450 ms: +2 points • Acute myocardial infarction: +2 points • Sepsis: +3 points • Treatment with one QT-prolonging drug: +1 point • Treatment with two or more QT-prolonging drugs: +3 points
This patient: age <68 (0), female (+1), no loop diuretic in the active regimen (0), K⁺ 3.6 mmol/L borderline-normal (0), baseline QTc 438 ms (0), no MI (0), no sepsis (0), three concurrent QT-prolonging drugs (+3)... reassessed after the pharmacokinetic-adjusted QTc crosses 450 ms (+2) and hydrochlorothiazide-driven electrolyte drift is factored (+2 for K⁺ trending to 3.4 mmol/L), for a running total of 12 points — solidly in the high-risk band (≥11 points), which in the validation cohort corresponded to an observed TdP incidence around 3%, versus <1% for scores ≤6 and roughly 1% for the intermediate 7–10 band.
Several other frameworks are used alongside or instead of the Tisdale score in different clinical settings:
• RISQ-PATH: a French ICU-derived tool similar in structure to Tisdale, incorporating renal function and concurrent electrolyte disturbance more explicitly • Modified Schwartz score: designed primarily for congenital long-QT syndrome diagnosis, incorporating family history and syncope, less suited to acquired/drug-induced QT prolongation • FDA thorough QT (TQT) study threshold: a mean placebo-corrected QTc increase (ΔΔQTc) of ≥5 ms is treated as a regulatory signal requiring further evaluation; ≥20 ms is considered a substantial risk signal in a single-drug study — useful context for interpreting the magnitude of the additive effects calculated here • CredibleMeds risk-factor checklist: a simpler binary tool (bradycardia, recent conversion from atrial fibrillation, CHF, female sex, digoxin use, electrolyte abnormality, high drug concentration/rapid infusion, baseline QT prolongation) used for rapid bedside triage without full score calculation
A risk calculator only has value if its output changes management. The final stage converts the composite Tisdale score and modeled QTc into a concrete, ranked set of interventions — deprescribing, dose reduction, electrolyte correction, and monitoring intensity — and re-projects the expected QTc and TdP probability after those changes are implemented.
Deprescribing decisions are prioritized by which change removes the most risk for the least clinical cost:
1. Discontinue IV ondansetron, switch to granisetron or a non-5-HT3 antiemetic (e.g., low-dose olanzapine for refractory nausea) — removes the single largest and most acute contributor (+20 ms per dose) with a therapeutically equivalent, lower-risk alternative readily available 2. Reduce citalopram from 40 mg to 20 mg/day, or cross-titrate to sertraline (minimal QT liability at therapeutic doses) — addresses both the largest chronic contributor and the CYP2C19-driven exposure amplification 3. Continue azithromycin to complete the antibiotic course (interaction risk is time-limited; discontinuing a needed antimicrobial mid-course carries its own risk) but flag for telemetry during the remaining doses 4. Correct serum potassium to >4.0 mmol/L and magnesium to >2.0 mg/dL — even without any drug change, this single step reduces IKr-blockade-related arrhythmic risk substantially, since hypokalemia both slows channel deactivation and paradoxically increases extracellular-side drug binding affinity for many hERG blockers
Re-projecting the additive model after steps 1, 2, and 4: modeled QTc falls from ~485–496 ms to approximately 451 ms, and the Tisdale score drops from 12 to roughly 5–6 points — moving the patient from the high-risk band back into the low-risk band.
Even after mitigation, monitoring is stratified rather than uniformly relaxed:
• QTc <480 ms after intervention: routine daily ECG until medication changes are stable, then discontinue additional monitoring • QTc 480–500 ms: continuous telemetry with automated QT-alarm thresholds, repeat 12-lead ECG every 12–24 h • QTc ≥500 ms or acute increase ≥60 ms from a documented baseline: continuous telemetry, cardiology consultation, consideration of magnesium sulfate 2 g IV prophylactically, and evaluation for temporary pacing if bradycardia-dependent early afterdepolarizations are suspected • Any polymorphic wide-complex tachycardia on telemetry, regardless of measured QTc, triggers the institutional TdP rapid-response protocol: immediate IV magnesium sulfate, correction of any remaining electrolyte deficit, discontinuation of all QT-prolonging drugs, and overdrive pacing or isoproterenol if TdP recurs despite magnesium
The single highest-leverage intervention in this case is not the antidepressant or the antibiotic — it is stopping the acute IV ondansetron dose and correcting potassium. A drug that adds +20 ms in thirty minutes, layered onto two chronic contributors already narrowing the patient's repolarization reserve, is what pushes a "borderline" baseline QTc into the danger zone. Additive QT risk is rarely caused by one dramatic drug; it is almost always the last ordinary-looking prescription that tips a stacked risk profile past the threshold.