Predicting Drug Safety and Communicating Risk: Benefits of a Bayesian Approach.

Drug toxicity is a major source of attrition in drug discovery and development. Pharmaceutical companies routinely use preclinical data to predict clinical outcomes and continue to invest in new assays to improve predictions. However, there are many open questions about how to make the best use of available data, combine diverse data, quantify risk, and communicate risk and uncertainty to enable good decisions. The costs of suboptimal decisions are clear: resources are wasted and patients may be put at risk. We argue that Bayesian methods provide answers to all of these problems and use hERG-mediated QT prolongation as a case study. Benefits of Bayesian machine learning models include intuitive probabilistic statements of risk that incorporate all sources of uncertainty, the option to include diverse data and external information, and visualizations that have a clear link between the output from a statistical model and what this means for risk. Furthermore, Bayesian methods are easy to use with modern software, making their adoption for safety screening straightforward. We include R and Python code to encourage the adoption of these methods.

An Analysis of the Relationship Between Preclinical and Clinical QT Interval-Related Data.

There has been significant focus on drug-induced QT interval prolongation caused by block of the human ether-a-go-go-related gene (hERG)-encoded potassium channel. Regulatory guidance has been implemented to assess QT interval prolongation risk: preclinical guidance requires a candidate drug's potency as a hERG channel blocker to be defined and also its effect on QT interval in a non-rodent species; clinical guidance requires a "Thorough QT Study" during development, although some QT prolonging compounds are identified earlier via a Phase I study. Clinical, heart rate-corrected QT interval (QTc) data on 24 compounds (13 positives; 11 negatives) were compared with their effect on dog QTc and the concentration of compound causing 50% inhibition (IC50) of hERG current. Concordance was assessed by calculating sensitivity and specificity across a range of decision thresholds, thus yielding receiver operating characteristic curves of sensitivity versus (1-specificity). The area under the curve of ROC curves (for which 0.5 and 1 indicate chance and perfect concordance, respectively) was used to summarize concordance. Three aspects of preclinical data were compared with the clinical outcome (receiver operating characteristic area under the curve values shown in brackets): absolute hERG IC50 (0.78); safety margin between hERG IC50 and clinical peak free plasma exposure (0.80); safety margin between QTc effects in dogs and clinical peak free plasma exposure (0.81). Positive and negative predictive values of absolute hERG IC50 indicated that from an early drug discovery perspective, low potency compounds can be progressed on the basis of a low risk of causing a QTc increase.

Assessment of extracellular field potential and Ca2+ transient signals for early QT/pro-arrhythmia detection using human induced pluripotent stem cell-derived cardiomyocytes.

Cardiovascular toxicity is a prominent reason for failures in drug development, resulting in the demand for assays that can predict this liability in early drug discovery. We investigated whether iCell® cardiomyocytes have utility as an early QT/TdP screen. Thirty clinical drugs with known QT/TdP outcomes were evaluated blind using label-free microelectrode array (parameters measured were beating period (BP), field potential duration (FPD), fast Na+ amplitude and slope) and live cell, fast kinetic fluorescent Ca2+ transient FLIPR® Tetra (parameters measured were peak count, width, amplitude) systems. Many FPD-altering drugs also altered BP. Correction for BP, using a Log-Log (LL) model, was required to appropriately interpret direct drug effects on FPD. In comparison with human QT effects and when drug activity was to be predicted at top test concentration (TTC), LL-corrected FPD and peak count had poor assay sensitivity and specificity values: 13%/64% and 65%/11%, respectively. If effective free therapeutic plasma concentration (EFTPC) was used instead of TTC, the values were 0%/100% and 6%/100%, respectively. When compared to LL-corrected FPD and peak count, predictive values of uncorrected FPD, BP, width and amplitude were not much different. If pro-arrhythmic risk was to be predicted using Ca2+ transient data, the values were 67%/100% and 78%/53% at EFTPC and TTC, respectively. Thus, iCell® cardiomyocytes have limited value as an integrated QT/TdP assay, highlighting the urgent need for improved experimental alternatives that may offer an accurate integrated cardiomyocyte safety model for supporting the development of new drugs without QT/TdP effects.

Enhanced characterization of contractility in cardiomyocytes during early drug safety assessment.

We sought to investigate whether drug-induced changes in contractility were affected by pacing rates that represent the range of heart rates encountered in vivo. Using the cell geometry measurement system (IonOptix), we paced dog cardiomyocytes at different cycle lengths (CLs) of 2000, 1000, 500, and 333.3 ms, before and after exposure to 13 inotropic drugs. Time course data using vehicle control (0.1% dimethyl sulfoxide (DMSO)) demonstrated stability of the system at all CLs tested. Seven positive inotropes (eg isoproterenol) exerted rate-dependent increases in sarcomere shortening (Sarc. short.; maximal effect at a CL of 333.3 ms [0.1 µM isoproterenol increased Sarc. short. by 41.1% and 145.9% at 2000 and 333.3 ms, respectively]). Omecamtiv mecarbil showed an atypical profile (increased Sarc. short. at 2000 ms [106.9%] and decreased at 333.3 ms [IC(50) = 0.64 µM]). Four negative inotropes (eg flecainide) showed rate-independent inhibition of Sarc. short. (IC(50)s: 3.3 µM [2000 ms] versus 2.3 µM [333.3 ms]). The remaining negative inotropes, verapamil, and BTS (N-benzyl-p-toluene sulphonamide) produced an increase (IC(50)s: 3.9 µM [2000 ms] versus 0.043 µM [333.3ms]) and decrease (IC(50)s: 18.3 µM [2000 ms] versus 34.0 µM [333.3 ms]) in potency, respectively. Negative inotropes (eg flecainide, BTS, and verapamil) decreased the area of the Ca(2+) transient versus Sarc. short. hysteresis loop, although rate dependency was seen with verapamil only. Positive inotropes (eg isoproterenol and levosimendan) induced a rate-dependent increase in the area, however Omecamtiv mecarbil increased and decreased the area at CLs of 2000 and 333.3 ms, respectively. Thus, the use of different pacing rates may improve the detection of inotropes in early drug discovery and illustrate the potential for finger-printing different mechanisms of action.

Reducing attrition in drug development: smart loading preclinical safety assessment.

Entry into the crucial preclinical good laboratory practice (GLP) stage of toxicology testing triggers significant R&D investment yet >20% of AstraZeneca's potential new medicines have been stopped for safety reasons in this GLP phase alone. How could we avoid at least some of these costly failures? An analysis of historical toxicities that caused stopping ('stopping toxicities') showed that >50% were attributable to target organ toxicities emerging within 2 weeks of repeat dosing or to acute cardiovascular risks. By frontloading 2-week repeat-dose toxicity studies and a comprehensive assessment of cardiovascular safety, we anticipate a potential 50% reduction in attrition in the GLP phase. This will reduce animal use overall, save significant R&D costs and improve drug pipeline quality.

Preservation of cardiomyocytes from the adult heart.

Cardiomyocytes represent one of the most useful models to conduct cardiac research. A single adult heart yields millions of cardiomyocytes, but these cells do not survive for long after isolation. We aimed to determine whether inhibition of myosin II ATPase that is essential for muscle contraction may preserve fully differentiated adult cardiomyocytes. Using inhibitors of the myosin II ATPase, blebbistatin and N-benzyl-p-toluene sulphonamide (BTS), we preserved freshly isolated fully differentiated adult primary cardiomyocytes that were stored at a refrigerated temperature. Specifically, preserved cardiomyocytes stayed viable for a 2-week period with a stable expression of cardiac genes and retained the expression of key markers characteristic of cardiomyocytes. Furthermore, voltage-clamp, action potential, calcium transient and contractility studies confirmed that the preserved cardiomyocytes are comparable to freshly isolated cells. Long-term exposure of preserved cardiomyocytes to four tyrosine kinase inhibitors, sunitinib malate, dasatinib, sorafenib tosylate and imatinib mesylate, revealed their potential to induce cardiac toxicity that was manifested with a decrease in contractility and induction of cell death, but this toxicity was not observed in acute experiments conducted over the time course amenable to freshly prepared cardiomyocytes. This study introduces the concept that the inhibition of myosin II ATPase safeguards the structure and function of fully differentiated adult cardiomyocytes. The fact that these preserved cardiomyocytes can be used for numerous days after preparation makes them a robust and versatile tool in cardiac research and allows the investigation of long-term exposure to novel drugs on cardiomyocyte function.

Safety pharmacology--a progressive approach.

Although deaths and life-threatening adverse drug reactions (ADRs) in Phase I clinical trials are extremely rare, less severe ADRs occur with an incidence of over 13%. Of the candidate drugs (CDs) that fail prior to marketing, it is generally acknowledged that about 1 in 5 do so because of ADRs in the clinic. Once new chemical entities (NCEs) are on the market, ADRs are estimated to be the fourth leading cause of death in the USA. These various statistics indicate that there is room for improvement in preclinical safety assessment, and a smarter approach to safety pharmacology (SP) can contribute to this. Rather than 'bundling' the SP studies together just prior to Phase I trials, a step-wise, streamlined approach can be adopted throughout the drug discovery process. In this way, the SP information can contribute to making informed judgements at each milestone throughout the preclinical drug discovery process: (i) to assist in series and compound selection; (ii) to assess potential risk of failure in the clinic due to ADRs; (iii) to predict potential ADRs that the clinical pharmacologists can focus on; (iv) to define a therapeutic window for acute dosing in humans. To achieve these objectives, the SP tests need to be carefully selected, adequately validated in-house, and be robust and reliable. To achieve (ii) above, outcome criteria have to be set which, for each test (in vitro and in vivo), take into account acceptable safety margins for the particular therapeutic target. Thus, highly sensitive and predictive SP tests positioned strategically and as early as possible should contribute to reducing attrition during clinical development and ultimately to marketing safer medicines more rapidly.