Restraint‐Based ECG and Arterial Pressure Assessment Do Not Reliably Detect Drug Induced QTc Prolongation and Hypotension: Evidence From Case Studies

ABSTRACT

Telemetry‐based methods are recommended as best practice for monitoring changes in cardiovascular (CV) function in conscious nonrodent species during the development of new chemical entities. Such methods allow for unrestrained data collection in animals over extended time periods with high sensitivity to detect small effects. The CV profile of new drugs is also assessed in repeat‐dose toxicology studies, routinely utilizing non‐invasive blood pressure, heart rate, and ECG measurement methods that require physical or chemical restraint. These methods are limited to a short “snapshot” data collection period and incur physiological stress and behavioral excitement from handling/restraint during data collection. The resultant sympathetic activation impacts heart rate, blood pressure, and ECG intervals, causing increased variability and reduced sensitivity. Nonclinical best practices have been defined for standalone CV telemetry studies to support an integrated QTc risk assessment (ICH E14/S7B Q&A 5.1 & 6.1) using non‐restraint telemetry methods; however, the pharmacological and statistical sensitivity of restraint‐based methods used in repeat‐dose toxicology studies is a gap. This paper retrospectively analyzed two case studies (AMG 319 and AMG 337) in which proprietary small molecules were evaluated by both non‐restraint‐based telemetry and restraint‐based methods. AMG 319 and AMG 337 caused QTc interval prolongation and hypotension, respectively, in telemetry studies, which were also observed clinically with these compounds. However, in toxicology studies in which restraint‐based ECG and blood pressure methods were used, CV effects were missed, blunted, or directionally wrong. These case studies highlight the need for the utilization of unrestrained telemetry methods over restraint‐based methods.

Summary.

• What is the current knowledge on the topic?

  • ○
    Current best practices in preclinical cardiovascular (CV) safety pharmacology recommend telemetry‐based methods for CV monitoring in nonrodent species due to their ability to provide accurate, continuous data from unrestrained animals. Restraint‐based, snapshot methods, while commonly used in toxicology studies, are known to introduce variability and stress‐related artifacts, reducing sensitivity and potentially obscuring drug‐induced CV effects.
• What question did this study address?

  • ○
    This study sought to evaluate and compare the effectiveness of telemetry‐based versus restraint‐based CV monitoring methods in detecting drug‐induced CV effects in nonrodent subjects. Specifically, it addressed whether telemetry methods could better predict clinically relevant CV risks associated with investigational drugs, compared to snapshot measurements obtained using restraint‐based methods.
• What does this study add to our knowledge?

  • ○
    This study highlights the limitations of restraint‐based, snapshot CV monitoring methods in detecting large and clinically relevant CV effects, using two case studies (AMG 319 and AMG 337) to illustrate the differences. The results demonstrate that telemetry‐based methods offer superior sensitivity for detecting effects like QTc prolongation and hypotension, accurately reflecting the clinical CV risk profiles of the drugs. In contrast, restraint‐based methods missed, blunted, or even misrepresented these effects, which could lead to inaccurate safety conclusions in preclinical evaluations.
• How might this change clinical pharmacology or translational science?

  • ○
    These findings underscore the importance of using telemetry‐based methods as the primary approach for CV monitoring in nonrodent toxicology studies, especially when these studies are intended to serve as definitive CV safety assessments. By highlighting the risk of false negatives and misleading data from restraint‐based methods, this study supports a shift toward unrestrained telemetry as a standard for CV safety profiling in preclinical drug development. This shift may prompt updates to regulatory guidelines and industry standards, leading to more accurate and reliable safety assessments that better inform clinical drug development and ultimately improve patient safety.

1. Introduction

The ICH S7A and S7B guidelines require evaluation of new chemical entities (NCEs) for potential cardiovascular (CV) effects and pro‐arrhythmic risks. As a result, telemetry‐based methods are the quality standard to monitor NCE‐induced changes in cardiovascular function in conscious nonrodent species during preclinical development [1, 2, 3]. Telemetry‐based methods allow for continuous CV monitoring from unrestrained and unstressed animal subjects, which enables high‐density data collection over extended time periods and improves sensitivity to detect small effects [4, 5]. The recent release of the ICH E14/S7B Q&As for an integrated nonclinical and clinical pro‐arrhythmic assessment prompted the safety pharmacology community to develop best practice recommendations for in vivo ECG evaluation [6, 7, 8]. The Rossman et al. paper focused primarily on the acquisition and analysis of ECGs (particularly the QTc interval) collected from conscious and unrestrained nonrodent animals with implanted telemetry devices in a cross‐over study design. Pharmacological and statistical sensitivity studies with clinically relevant QTc positive controls clearly demonstrate that high‐quality telemetry studies have the sensitivity to detect small magnitude QTc changes (< 10 ms) with group sizes of 4 to 8 nonrodents [4, 9, 10, 11]. The utilization of telemetry best practices is critical to achieve the sensitivity necessary for nonclinical QTc data to support clinical safety [7, 12, 13].

In addition to sensitive telemetry‐based methods, the CV profile of new drugs is routinely assessed in repeat‐dose general toxicology studies by the inclusion of non‐invasive blood pressure, heart rate, and ECG measurements from physically or chemically restrained non‐rodent study animals [3, 14, 15]. Chemical and/or physical restraint is needed for these methodologies in order to accurately place surface ECG leads and blood pressure (tail) cuff, as well as to reduce artifacts caused by animal movement. In addition, the data acquisition period is limited to a short interval (30–120 s of collected data) occurring at select times during a typical toxicology study (e.g., pre‐dose, single day during week 1, and end of dosing phase occurring at the predicted Tmax). The short duration or “snapshot” data collection increases the likelihood that critical effects will be missed, including arrhythmias. The physiological stress and behavioral excitement induced by human presence and handling (e.g., manual restraint) increase plasma catecholamines, sympathetic tone, blood pressure, heart rate, and can alter ECG intervals, which introduces physiological variability that confounds the interpretation of drug effects and reduces overall detection sensitivity [16, 17, 18].

Rossman et al. (2023) clearly defined nonclinical best practices for standalone safety pharmacology CV telemetry studies to support an integrated QTc risk assessment (ICH E14/S7B Q&A 5.1 & 6.1); however, the pharmacological and statistical sensitivity of restraint‐based methods to detect drug‐induced CV effects in a repeat‐dose toxicology study has not been demonstrated and represents a significant gap [8, 19]. Some drug sponsors and contract research organizations consistently use a snapshot evaluation technique in chronic toxicology studies despite reports of low or poor sensitivity to detect CV effects, which can lead to erroneous conclusions about the safety of a new drug [15, 20, 21]. An overview of CV evaluation practices across regulatory submissions indicated that restraint‐based ECG collections are used routinely for longer non‐rodent duration studies and highlighted several factors indicating that restraint‐based data can have lower sensitivity and higher variability [19]. The impact of restraint‐based ECG data collection methods on sensitivity and variability has been demonstrated in published examples in which moxifloxacin‐induced QTc prolongation was undetected by snapshot ECG in restrained dogs in a chronic toxicology study at doses that produced significant QTc prolongation using telemetry in unrestrained dogs [10, 22]. Likewise, PNU‐142093 caused dose‐dependent heart rate reduction, QTc prolongation, and T‐wave morphology changes indicative of abnormal ventricular repolarization using implanted telemetry (N = 4 NHP); however, these effects were missed under restraint (N = 4 NHP/dose group) on Days 1 and 7 of a 13‐day toxicology study [23]. On Day 13, QTc prolongation of a lower magnitude than observed with unrestrained telemetry was eventually detected at the mid‐ and high‐dose, whereas unrestrained telemetry detected QTc prolongation beginning at the low dose. The bradycardia and altered ECG morphology remained undetected due to the high heart rate (> 200 bpm) caused by restraint‐induced sympathetic activation. Lastly, torcetrapib caused dose‐dependent BP elevation in a dog telemetry study, but peak BP effects were blunted or missed when time‐matched tail cuff BP measurements were taken from the same animals, most likely due to handling‐induced stress [18]. These examples underscore the sub‐optimal detection sensitivity of restraint‐based CV data, likely due to the profound impact of restraint‐induced stress on the dog and NHP cardiovascular system [17, 24].

Given the scant examples of telemetry versus snapshot comparisons in the literature, this paper retrospectively analyzed two case studies in which proprietary small molecules were evaluated with both a gold‐standard telemetry study and a toxicology study that incorporated restraint‐based CV methods. These examples are presented to document the reduced sensitivity of restraint‐based methods to detect both drug‐induced QTc interval prolongation and hypotension. The findings have important safety implications for clinical risk prediction and highlight the need for best practice recommendations to guide the collection and evaluation of CV data in repeat dose toxicology studies.

2. Material & Methods

2.1. Animals

The non‐human primate (NHP; cynomolgus macaque) was used to evaluate AMG 319 in GLP CV safety pharmacology and toxicology studies (Table 1; Covance Research Products Inc., Alice, TX). Non‐naïve males were used in the telemetry study (N = 8; 4–6 years of age; 4–7 kg in weight), and both sexes (naïve) used in the toxicology study (N = 3‐5/group/sex; males: 5–8 yr., 5–8 kg; females: 5–9 year., 3–6 kg). Environmental controls (housing) were set to maintain 18°C–26°C, a relative humidity of 30%–70%, a minimum of 10 room air changes per hour, and a 12‐h light/12‐h dark cycle. NHPs were fed Certified Primate Diet (#5048, PMI Nutrition International) with water provided ad libitum.

TABLE 1.

Outline of studies examined in this paper.

Study #	Species	Compound (year)	Study design	Animal # M/F	Oral dose & frequency	Parameters of interest	Data collection method	Post‐dose data collection
1	NHP	AMG 319 (2010)	LS	8/0	Single dose (0, 10, 30, 100 mg/kg)	QTc interval	Implanted telemetry	24 h
2	NHP	AMG 319 (2010)	RD	5/5	28 doses (0, 10, 30, 100 mg/kg)	QTc interval	Surface ECG (restrained)	2 min (Days 2, 24)
3	Dog	AMG 337 (2013)	ALT RD	0/6	7 doses (vehicle) then 10 doses (30 mg/kg)	Blood pressure and heart rate	Implanted telemetry	24 h (Days 1 b ‐7 vehicle) (Days 8 b ‐17 AMG337)
4	Dog	AMG 337 (2010)	RD	3–5/3–5 a	28 doses (0, 1, 5, 30 mg/kg)	Blood pressure and heart rate	Tail cuff (restrained)	2 min (Days 2, 24)

Abbreviations: ALT RD, alternative design repeat dose; LS, Latin square design; RD, repeat dose.

^a The control and high dose groups had 5 animals/sex.

^b First day of dose administration for each treatment.

The beagle dog was used to evaluate AMG 337 in GLP cardiovascular safety pharmacology and toxicology studies (Table 1; Covance Research Products Inc., Kalamazoo, MI). Non‐naïve female dogs (N = 6, 1–4 years of age, 7–11 kg in weight) were used in the telemetry study. In the toxicology study, naïve male and female dogs (9–10 months of age, 4–7 kg in weight) were used. Dogs were housed individually during the study in stainless steel cages. Dogs were fed Certified Canine Diet (#5007, PMI Nutrition International) and water was provided ad libitum. Environmental controls (housing) were set to maintain 18°C–26°C, a relative humidity of 30%–70%, a minimum of 10 room air changes per hour, and a 12‐h light/12‐h dark cycle.

Animals were cared for in accordance with the Guide for the Care and Use of Laboratory Animals, eighth edition (2011). They were housed individually at an indoor Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) internationally‐accredited facility in species‐specific housing. All studies were reviewed and approved by study site institutional animal care and use committees.

2.1.1. Implanted Telemetry and Instrumentation

All telemetry study animals were surgically implanted with telemetry devices (PCT or PCTP units; Data Sciences International, St. Paul, MN) at least 2 weeks prior to study initiation. Specifically, the units were inserted into the abdomen and sutured to the abdominal wall to enable electrocardiograms (ECGs), blood pressure, and temperature data collection from unrestrained animals. The ECG lead had an approximate Lead II configuration. To measure blood pressure, the pressure catheter was advanced toward the abdominal aorta via the femoral artery. Detailed surgical procedures for these telemetry implantations have been published previously [25, 26].

2.1.2. Surface ECG and Blood Pressure Cuff Instrumentation Under Restraint

Skin (surface) ECG leads were used with NHPs that were physically restrained and lightly sedated with ketamine. Dogs were physically restrained for CV data collection. Electrodes were manually placed to capture multiple lead configurations (I, II, III, aVR, aVL, and aVF) but only lead II data are reported. Arterial blood pressure measurements in dogs were obtained by high‐definition oscillometry (HDO) [18].

2.1.3. In Vitro hERG Assay

AMG 319 was assessed for hERG (human ether‐a‐go‐go‐related gene) block activity in a GLP in vitro hERG channel current assay using a human embryonic kidney (HEK293) cell line expressing the hERG potassium channel. Whole cell patch clamp was performed using glass micropipettes and a micropipette solution containing: (mM) potassium aspartate, 130; MgCl2, 5; EGTA, 5; ATP, 4; HEPES, 10; pH adjusted to 7.2 with KOH. The recording was performed at a temperature of 33°C to 35°C. A commercial patch clamp amplifier was used for whole cell recordings. Cells stably expressing hERG were held at −80 mV. Onset and steady‐state inhibition of hERG potassium current due to AMG 319 were measured using a pulse pattern with fixed amplitudes (conditioning prepulse: +20 mV for 1 s, repolarizing test ramp to −80 mV repeated at 5 s intervals). Each recording ended with a final application of supramaximal concentration of the reference substance E‐4031 at 500 nM to assess the contribution of endogenous currents. The remaining uninhibited current was subtracted off‐line digitally from the data to determine the potency of AMG 319 for hERG inhibition. AMG 319 was tested at 1, 3, and 30 μM.

2.1.4. AMG 319 Clinical ECG Assessment

AMG 319 was evaluated in a Phase 1 study to assess the safety, tolerability, pharmacokinetics, and pharmacodynamics in adult subjects with relapsed or refractory lymphoid malignancies as described previously [27]. Cardiac assessments included 12‐lead ECG (triplicate) collected from supine subjects (approximately 3–5 min) at predose, 1, 4, or 24 h post dose depending on the cycle. The parameters evaluated were QRS, QT, QTcF, RR, and PR intervals. Baseline ECGs were a mean of two sets of triplicate screening ECGs and one set of triplicate ECGs obtained pre‐dose on Day 1. All ECGs were analyzed by a central reader. All clinical studies complied with the International Conference on Harmonization Tripartite Guideline on Good Clinical Practice and applicable FDA regulations/guidelines, and all applicable aspects (protocol, consent form, etc.) were approved by the site's Institutional Review Board.

2.2. Study Design Details

Details of the studies, including study groups, species, group size, treatments, and data collection methods are shown in Table 1.

2.2.1. Test Material and Overview of Studies

AMG 319 is a small molecule kinase inhibitor developed for the treatment of multiple indications including cancer and inflammatory disease [27, 28]. Nonclinical safety studies (GLP) were conducted at contract research laboratories (telemetry: Covance; Madison WI; toxicology: Charles River Labs; Reno NV). The secondary pharmacology profile of AMG 319 indicated hERG blockade (IC₅₀: 8.4 μM; GLP). In the NHP telemetry study, single oral doses were evaluated in a conventional double Latin Square crossover design with a 7‐day washout between doses (Table 1). In the toxicology study, oral doses were given daily to separate dose groups in a parallel design (Table 1). Plasma exposures measured on Days 1 and 28 of the toxicology study were similar between male and female NHP, so combined values are reported. Day 1 values were used to estimate drug exposure in the telemetry study.

AMG 337 is a small molecule kinase inhibitor developed for the treatment of cancer [29]. Nonclinical safety studies (GLP) were conducted at contract research labs (telemetry: MPI; Mattawan MI, 2013; toxicology: Covance; Chandler AZ, 2010). Secondary pharmacology profiling of AMG 337 indicated adenosine uptake inhibition (IC₅₀: 0.28 μM; non‐GLP) as the only off‐target activity [26]. In the dog telemetry study, a repeat dose design in which oral vehicle (7 consecutive days) followed by 30 mg/kg AMG 337 (10 consecutive days) was given so that each animal served as its own control (Table 1). Study design details, including blood sampling for exposure, were published previously [26]. In the toxicology study, oral doses (0, 1, 5, 30 mg/kg/day) were given to separate groups for 28 days (Table 1) in a parallel design. Plasma exposures on Days 1 and 28 were similar between male and female dogs, so combined values are reported.

2.2.2. Data Collection and Analysis

Study 1 (NHP telemetry), arterial pressures and temperature were digitized at a sampling rate of 250 Hz, and ECG at 500 Hz. Quantitative ECG parameters evaluated were RR, PR, QRS, QT, and QTcB (Bazett correction) intervals. Only QTc interval and RR interval data are reported in this analysis. Telemetry data were analyzed in four distinct blocks: baseline data (1 h of data prior to room entry for dosing); block 1 (1 h averages; 1–7 h postdose); block 2 (3 h averages; 8–19 h postdose including night cycle); block 3 (3 h averages 20–25 h postdose including second light cycle). A modified Latin square design analysis of covariance with repeated measures was used, selecting the variance–covariance structure with the smallest Akaike Information Criterion. Group comparisons were made using Dunnett or Dunnett‐Hsu t‐tests, with a significance level of 0.05.

Study 2 (NHP toxicology), ECG waveform measurements were collected from NHPs that were restrained and lightly sedated with ketamine. ECG parameters evaluated were RR, PR, QRS, QT, and QTcB (Bazett correction) intervals. Measurements were taken once prestudy and twice during the drug treatment phase (Days 7 & 28). Data were analyzed separately for males and females. An analysis of variance (ANOVA) followed by a multiple comparisons test, was used, with the Shapiro‐Wilkes and Levene's tests ensuring the assumptions for parametric ANOVA. If these assumptions were not met, a non‐parametric Kruskal‐Wallis ANOVA was applied, followed by Dunn's multiple comparison test.

Study 3 (dog telemetry), arterial blood pressure (systolic, diastolic, and mean) and heart rate (derived from blood pressure) were monitored continuously using implanted telemetry from at least 2 h prior to dosing until at least 24 h postdose on data collection days (Table 1). Blood pressure signals were digitized at a rate of 250 Hz. Data were collected and summarized in 1‐min intervals and reported as 20 or 60 min averages and analyzed using a mixed model analysis using SAS software. Data was reported in summary intervals across various photoperiods—i.e. 20 min bins during the light cycle (0–7 and 19–24 h) and 60 min bins during the dark cycle (7–19 h). Descriptive statistics and repeated measures analysis of covariance were employed. The analysis included treatment and time factors, and the covariate was the average of 2 h of predose data. Pair‐wise group comparisons were adjusted using step‐down simulate adjustment, with overall effects tested at the 0.05 significance level.

Study 4 (dog toxicology), snapshot blood pressure measurements were taken from restrained conscious dogs using HDO (tail cuff). Blood pressure measurements were taken three times during predose (Study Days −2, −1, 1) and twice during drug treatment (Study Days 2 and 24; 2–4 h post‐dosing). Snapshot ECG measurements were also taken but are not reported here. Statistical analysis included descriptive statistics (means, standard error and deviation) and a post‐study one‐way ANOVA analysis (Graphpad Prism 8.4.3).

3. Results

3.1. Unrestrained vs. Restrained ECG Measurements in Cynomolgus Monkeys

AMG 319 produced a dose‐dependent increase in QTc intervals in the dedicated telemetry study that was not observed in the repeat dose toxicology study (Figure 1; Table 2). Specifically, QTc interval prolongation of 9, 27, and 67 ms was observed at doses of 10, 30, and 100 mg/kg respectively in the telemetry study at 2 h post dose, corresponding to Tmax; the mid and high dose effects were statistically significant. This study also showed that the Emax for QTc prolongation occurred 7 h post dose, i.e., after Tmax (Figure 1). Exposure estimates were not determined in the telemetry study; however, C_max (total) values of 2070 ± 559, 7450 ± 1530, and 15,800 ± 7410 ng/mL were obtained from Day 1 of the toxicology study (Table 2). In contrast, in the repeat dose toxicology study utilizing snapshot ECG evaluations in restrained‐sedated NHPs on Days 7 and 28 statistically significant lower QTc interval values were observed compared to vehicle at 30 mg/kg (mid dose) between 2 and 3 h post dose. Moreover, there was no significant effect at the highest dose (100 mg/kg) on Day 7 (ECG data was not collected on Day 28 due to unscheduled termination of the high dose group). The C_max values at 30 mg/kg in the toxicology study were similar on Days 1 and 28 (Table 2).

FIGURE 1.

Effect of AMG 319 on QTc intervals (upper) and heart rate (lower) in telemeterized NHP. Eight male animals were used in cross‐over design. Symbols represent mean ± standard error of the mean. In unrestrained monkeys, telemetry detected dose‐dependent QTc prolongation with AMG 319 (A), while restrained ECG measurements failed to detect this effect, showing QTc shortening instead (B).

TABLE 2.

AMG 319‐induced QTc interval and heart rate effects in Conscious NHP: Telemetry versus Restrained (manual + ketamine) Data Collection.

AMG 319	Monkey telemetry study (Study 1) single dose (N = 8 unrestrained; mean ± SD)			Monkey toxicology study (Study 2) 4 week repeat dose (N = 7–10 restrained; mean ± SD)
Dose (mg/kg)	HR (bpm) pre‐dose	HR (bpm) 2 h post dose	Cmax (ng/mL)	HR (bpm) pre‐dose	HR (bpm) Day 7	HR (bpm) Day 28	Cmax (ng/mL)
0	120 ± 23	131 ± 19	—	169 ± 41	185 ± 28	168 ± 20	—
10	117 ± 17	118 ± 18	2070 ± 559	159 ± 26	178 ± 20	195 ± 41	2970 ± 953
30	118 ± 24	107 ± 30 a	7450 ± 1530	168 ± 20	195 ± 41	207 ± 22	5530 ± 1410
100	114 ± 18	111 ± 19 a	15,800 ± 7410	179 ± 37	181 ± 25	Unec	Unec

AMG 319	Monkey telemetry study (Study 1) single dose (N = 8 unrestrained; mean ± SD)			Monkey toxicology study (Study 2) 4 week repeat dose (N = 7–10 restrained; mean ± SD)
Dose (mg/kg)	QTc (ms)	Δ QTc (ms) 2 h post dose	Cmax (ng/mL)	QTc (ms) pre‐dose	QTc (ms) Day 7	QTc (ms) Day 28	Cmax (ng/mL)
0	310 ± 9	—	—	372 ± 25	382 ± 16	358 ± 24	—
10	319 ± 13	9	2070 ± 559	357 ± 21	347 ± 23	342 ± 21	2970 ± 953
30	337 ± 10 a	27 a	7450 ± 1530	371 ± 28	339 ± 59 b	319 ± 39 b	5530 ± 1410
100	377 ± 20 a	67 a	15,800 ± 7410	361 ± 22	354 ± 17	Unec	Unec

Note: Heart rate and total C_max values in the cardiovascular safety pharmacology study conducted with a cross‐over design, (2 h time point to align with Tmax) and general toxicology study conducted with parallel dose groups (1–3 h timepoint). QTc and C_max values in the cardiovascular safety pharmacology study conducted with a cross‐over design. Total drug values obtained in toxicology animals (Day 1) were used for the telemetry study while total drug values obtained in toxicology animals (Day 28) are presented for the toxicology study. Unec: unscheduled necropsy.

Abbreviation: Unec, unscheduled necropsy.

^a Statistically significant QTc prolongation, p < 0.05.

^b Statistically significant QTc shortening at mid‐dose (vs pre‐dose) but no effect at high dose, p < 0.05.

Also of note, HRs were starkly different between studies during ECG collections. In Study 1, using unrestrained monkeys, HR showed minimal variation across different doses of AMG 319 at pre‐dose and 2 h post dose (around Tmax) time points. The values ranged narrowly from 120 ± 23 bpm to 114 ± 18 bpm pre‐dose and 131 ± 19 bpm to 107 ± 30 bpm at 2 h post dose across all treatment groups, indicating relatively stable, normal, resting heart rates at low dose and significantly lower at the mid and high dose. In contrast, Study 2, using restrained and sedated monkeys, heart rates were tachycardic across groups in pre and post dose collections, with heart rates ranging from 179 ± 37 bpm to 159 ± 26 bpm pre‐dose and 159 ± 26 bpm to 207 ± 22 bpm post dose on Days 7 and 28 (Table 2).

In clinical trials in oncology patients, AMG 319 caused QTc prolongation at oral doses of 200 to 400 mg (Table 3) [27]. Of the 16 patients treated with AMG 319 at dose levels ≥ 200 mg, 11 exhibited significant QTc interval prolongation of 30 to 60 ms (categorical analysis). The estimated free (unbound) concentrations of high‐dose AMG 319 (200–400 mg) achieved plasma drug levels (1.9–2.1 μM free) that were associated with a 20% hERG blockade (IC50 8.4 μM) and QTc interval prolongation in the NHP model (44 ms; Figure 2).

TABLE 3.

Effect of AMG 319 on QTc intervals in patients with lymphoid malignancies.

Cohorts	AMG 319 (oral dose)
	25 mg	50 mg	100 mg	200 mg	300 mg	400 mg
Group size	6	3	3	3	3	10
Cmax (μg/mL, total)	0.77	2.00	3.47	5.53	6.06	5.97
Cmax (μg/mL, free) a	0.10	0.27	0.47	0.75	0.82	0.81
Cmax (μM, free) a	0.30	0.70	1.2	1.9	2.1	2.1
Number of patients with a QTcF interval shift (maximal change from baseline)
≤ 30 ms	6	2	3	2	0	3
> 30–60 ms	0	1	0	1	3	7

Note: Six cohorts were used to assess the activity and cardiac safety of AMG 319. Plasma samples and QTc intervals (maximum change from baseline) were assessed during the control (pre‐treatment period) and specified time points after dosing.

^a Molecular weight (385.4 Da) and unbound fraction (human; 13.5%) were used to calculate free concentrations of AMG 319. C_max value is the group mean per cohort.

FIGURE 2.

Effect of AMG 319 in the in vitro hERG assay and the QTc study in telemeterized NHP. Maximal QTc interval prolongation (left axis) and hERG inhibition (right axis) are shown at the free AMG 319 concentration obtained in each assay. Yellow box: Exposure range measured in oncology patients per Table 3. Red box: The clinical range associated with QTc prolongation > 30–60 ms. Symbols are mean ± standard deviation (QTc peak) or mean ± standard error of the mean (hERG).

3.2. Unrestrained vs. Restrained Blood Pressure Measurements in Beagle Dogs

AMG 337 was evaluated for safety and toxicity in a repeat dose telemetry study (Study 3) and a repeat dose toxicology study (Study 4) utilizing unrestrained and restraint‐based blood pressure data collection methods, respectively. Differences in the CV response to AMG 337 were observed in unrestrained (Study 3) vs. restrained (Study 4) dogs at time‐matched data points coinciding with the expected Tmax at 2 h post dose. Mean blood pressure was similar between restrained and unrestrained animals with vehicle treatment; however, mean heart rate was higher (+21 bpm) in restrained dogs compared to unrestrained dogs at the 2 h time‐matched post vehicle dose time point (Table 4; Figure 3A). Heart rates were also higher during prestudy collections for restrained animals compared to vehicle dosing in unrestrained animals (+29 bpm) (Table 4).

TABLE 4.

AMG 337‐induced hemodynamic effects in conscious dogs: sensitivity of telemetry versus high‐definition oscillometer cuff recording.

Study 3: AMG 337 dog telemetry study, first dose*(unrestrained data collection)				Study 4: AMG 337 dog toxicology study (restraint based data collection), Day 2
Dose (mg/kg)	MAP (mm Hg; mean ± SD)	HR (bpm; mean ± SD)	Cmax (ng/mL; mean ± SD)	MAP (mm Hg; mean ± SD)	HR (bpm; mean ± SD)	Cmax (ng/mL; mean ± SD)
0	98 ± 16	94 ± 22	—	103 ± 9	115 ± 21	—
30	69 ± 11*	133 ± 15*	13,500 ± 2190 (29 μM, 16 μM free)	93 ± 15	148 ± 23	6830 ± 2250 (15 μM, 8 μM free)

Study 4: AMG 337 dog toxicology study with arterial pressure cuff prestudy vs. post dose
Dose (mg/kg)	MAP: pre‐dose (mm Hg; avg. ± SD)	MAP: Day 2 (mm Hg; avg. ± SD)	HR: pre‐dose (bpm; avg. ± SD)	HR: Day 2 (bpm; avg. ± SD)	Cmax (ng/mL; avg. ± SD)
0	97 ± 11	103 ± 9	123 ± 23	115 ± 21	—
1	101 ± 15	103 ± 9	112 ± 24	117 ± 33	193 ± 71
5	93 ± 11	103 ± 10	115 ± 13	119 ± 9	1190 ± 247
30	105 ± 11	93 ± 15	124 ± 21	148 ± 23*	6830 ± 2250

Note: Blood pressure and heart rate data from a cardiovascular study utilizing unrestrained, continuous CV data collection versus a repeat dose toxicology study utilizing restrained snapshot blood pressure collection from dogs administered vehicle (Day 1*) and high dose (30 mg/kg, Day 8*) at the timing of peak effect (2–4 h post‐dose) demonstrate higher heart rates and a lack of sensitivity to blood pressure effects in the restrained animals. Pre‐dose values in the repeat dose toxicology study utilizing restrained snapshot blood pressure collection demonstrate higher heart rates during prestudy and post‐dose collections.

^*Statistically significant at p < 0.05; Molecular weight (463.4 Da) and unbound fraction (dog; 56%) were used to calculate free concentrations of AMG 337.

FIGURE 3.

Effect of AMG 337 on blood pressure and heart rate over 24 h in unrestrained animals. (A) AMG 337 results in a pronounced heart rate (mean ± standard deviation) increase over several hours in unrestrained dogs compared to vehicle. (B) AMG 337 results in a pronounced blood pressure decrease (mean ± standard deviation) compared to vehicle. The snap‐shot data point demonstrates that a similar effect was not observed with restrained data collection. (C) AMG 337 corresponding effects on blood pressure and heart rate demonstrate a clear time dependence indicating the increase in heart rate is a reflexive (baroreceptor) response to decreased blood pressure.

In the unrestrained telemetry study following administration of 30 mg/kg AMG 337, a pronounced decrease in blood pressure (−29 mmHg) and an increase in heart rate (+39 bpm) was observed compared to vehicle. The observed increase in heart rate associated with AMG 337 coincides with the timing of the blood pressure decrease and is interpreted as a baroreflex response to the reduction in blood pressure (Figure 3C). In contrast, the blood pressure response to AMG 337 was of smaller magnitude in restrained dogs on Day 2 (−10 mmHg; not statistically significant) at the 30 mg/kg dose (Study 4) and is interpreted as within normal variability. A trend toward a higher heart rate in restrained animals at 30 mg/kg versus vehicle was observed (148 vs. 133 beats/min, respectively) despite the small magnitude change in blood pressure (Figure 3A,B). The increase in heart rate was largely attributed to increased excitement rather than the baroreceptor reflex, as occurred with unrestrained telemetry dogs. AMG 337 had 2‐fold lower C_max exposures in the repeat dose toxicology study versus the telemetry study at the high dose of 30 mg/kg in both studies (Table 4).

4. Discussion

Our retrospective review of real‐world nonclinical studies in dog and NHP assessed drug‐induced QTc prolongation and hypotension, respectively, and the results clearly demonstrate that snapshot methods were unable to detect even large effects compared to telemetry. Telemetry has been used extensively to detect drug‐induced CV effects with high sensitivity and is the established standard for conducting nonrodent CV safety studies [3, 5, 8]. In addition, jacketed ECG telemetry with external leads has also demonstrated high sensitivity and can be used for optimal CV data collection from unrestrained toxicology study animals [15, 25, 30]. Restraint‐based CV data collection requires manual handling, which is known to cause animal stress and excitement during the short “snapshot” measurement period. This handling stress introduces variability into CV functional measurements; therefore, it is important to understand how restraint impacts the biological sensitivity to detect a drug‐related cardiovascular hazard. However, the common practice of “snapshot” measurement of CV endpoints in nonrodent toxicology studies has not been robustly evaluated for signal detection sensitivity compared to telemetry‐based data collection, which is a critical data gap.

Dosing with AMG 319 and AMG 337 caused QTc interval prolongation and hypotension, respectively, when data were obtained from unrestrained animals with telemetry devices and predicted the clinical CV profile of these agents. However, the peak CV effects were either missed, blunted, or directionally wrong when restraint‐based methods were used. In telemetry‐instrumented NHP, AMG 319 caused a dose‐dependent increase in the QTc interval (Emax: 10 mg/kg–14 ms; 30 mg/kg–44 ms; 100 mg/kg–124 ms) which indicated delayed cardiac repolarization attributed to a known hERG blockade. However, restraint‐based ECG monitoring failed to detect any degree of QTc prolongation and also missed the peak effect time period clearly observed in the unrestrained study; rather, QTc shortening was detected, which was contradictory and misleading. In oncology clinical trials, AMG 319 caused dose‐related QTc interval prolongation, which was accurately predicted by the NHP telemetry study and hERG in vitro patch clamp study during preclinical development. The QTc findings captured under restraint conditions (QTc shortening), however, did not translate to clinical results.

Telemetry studies in dogs treated with AMG 337 detected significant and long‐lasting hypotension and compensatory tachycardia that was attributed to adenosine‐mediated vasorelaxation caused by adenosine transport inhibition [26]. However, restraint‐based monitoring did not clearly detect a blood pressure change, likely due, in part, to the presence of increased sympathetic drive, as evidenced by higher heart rates at the time of snapshot ECG collections. In cancer patients, AMG 337 administration caused a number of adverse events including headache (migraine‐like), hypotension, and heart rate elevation that was mediated by adenosine transport blockade at clinical exposures which ranged from 0.37 μM to 3.4 μM, unbound [26, 29]. AMG 337 unbound exposures in the telemetry (15 μM) and toxicology (8 μM) animals exceeded the IC50 value of 0.28 μM for adenosine transport inhibition by 29×–58×. In addition, AMG 337 tested in an ex vivo arterial response assay at 0.01, 0.1, 0.3, 1, 3, and 10 μM on dog arteries elicited a relaxation response at 3 and 10 μM concentrations, further supporting that the free exposures in the dog telemetry and toxicology studies would be expected to cause vasodilation and a decrease in blood pressure [26]. Therefore, the inability of snapshot to detect a substantial decrease in arterial pressure, yet did detect an increase in heart rate, was largely influenced by the physiologic impact of restraint, rather than a 2× lower drug exposure in the toxicology study.

Heart rates were higher when restraint‐based recordings were used, regardless of treatment, which is attributed to sympathetic activation and behavioral arousal caused by human handling. Such stress‐induced physiological changes can mask or confound the detection of drug‐related CV effects, thereby reducing the pharmacological and statistical sensitivity of the assay and leading to erroneous safety conclusions. With regard to reduced sensitivity for detecting changes in QTc, part of the underlying cause is due to the impact of high heart rates on the ability to heart rate correct the QT interval. High heart rates confound the use of a fixed formula to QT correction (QTcB) which is known to be inferior to an individually derived correction factor. Use of an individually derived correction factor is not an option in the context of “snapshot” ECG collections. The QTcB method is based on the relationship of QT and HR in a small cohort of healthy humans [31] and therefore can undercorrect in the presence of heart rates above 60 bpm [32]. In addition, it is nearly impossible to capture the exact timing of peak effects using the short duration of data collection routinely used for restraint‐based ECG and blood pressure methodologies as part of toxicology studies, contributing to the apparent lack of sensitivity and variability observed in the case studies presented here.

Telemetry‐based methods (implanted or jacket‐based) are best practice for CV de‐risking because they enable monitoring in unrestrained and unstressed animals [8, 25]. Several nonclinical guidance documents (ICH S9, ICH S6, ICH M3R2) indicate that safety pharmacology endpoints can be integrated into toxicology study designs. In this scenario (i.e., no dedicated CV telemetry study is conducted), ICH M3R2 specifically states that when safety pharmacology evaluation is integrated into a toxicology study, the rigor (i.e., sensitivity) should be similar to stand‐alone telemetry studies [33]. Based on our two case studies, snapshot evaluations were inferior and did not accurately detect drug‐induced QTc interval increases or blood pressure decreases and secondary reflexive responses (i.e., baroreceptor). That is, restraint‐based methods are sub‐optimal for CV safety assessment and have a high risk of generating false negative data (due to low quality, misleading or uninformative data) which is the worst‐case scenario for CV safety profiling of new drugs. Building on these insights, our recommendation is to prioritize the use of telemetry‐based methods in toxicology studies, especially if the toxicology study is intended to serve as the definitive or pivotal nonclinical CV safety pharmacology evaluation.

In conclusion, the AMG 319 and AMG 337 case studies underscore the critical importance of study design and data collection methodologies in the accurate detection of drug‐induced CV effects during preclinical drug safety evaluation. These findings highlight the need for the adoption of unrestrained telemetry methods in preclinical toxicology studies, as they offer superior sensitivity and reliability compared to restraint‐based approaches. In addition, these findings indicate an urgent need for updated regulatory guidance and industry best practice recommendations for CV data collection in toxicology study designs to reflect these advancements, ultimately ensuring the safety and efficacy of new drug entities in clinical development.

Author Contributions

K.A.H., N.E., and H.M.V. wrote the manuscript; H.M.V. designed the research; Performance of the research was outsourced to a contract lab, no author included. K.A.H. and N.E. analyzed the data.

Conflicts of Interest

The authors are employees of Amgen Inc. This manuscript discusses best practices in study design using data from drug candidates no longer in development. The study's intent is to provide insights into study design methodology, without promoting any active Amgen products or candidates.