Coupling Data Mining and Laboratory Experiments to Discover Drug Interactions Causing QT Prolongation

Orthogonal Learning

As a co-author of the recent publication by Lorberbaum et al., I acknowledge my obvious bias in this Commentary. However, I welcome the opportunity to discuss the paper and its scientific approach, orthogonal learning, that is so aptly described in the accompanying editorial by Roden et al.

Having read many ECGs, cardiologists are familiar with the concept of using the “orthogonal” approach to learn about the heart by examining different leads. The article and the accompanying editorial discuss an orthogonal approach to “learn and confirm,” one that we should consider in this era of “big data” and its promise of “big learning.” We have seen examples where research that only examined a question from one perspective has been misleading, eg, the value of PVC suppression in acute MI versus in post-MI patients. Lorberbaum et al. used multiple scientific tools and innovative research methods to search for potential adverse drug-drug interactions (DDIs). Through incremental learning and confirmation, they were able to identify evidence for a novel DDI that would not have been discovered using conventional means. The next step, proof of clinical validity, will be the final test of this research, but nevertheless, the approach has attractive features.

The researchers began with clinical data (adverse events and QT prolongation) and then tested the plausibility of their findings in a lab model (hERG). This approach may supplement our conventional efforts to detect DDIs, which today begin in the lab and then move to the clinic. Drug developers determine whether their drug of interest is metabolized in vitro and ask if other drugs can inhibit or induce its metabolism. When in vitro interactions are found, the developer is expected to conduct confirmatory pharmacokinetic studies in normal volunteers. This very often leaves the question of clinical relevance unanswered. Using clinical data as a starting point seems attractive because we have previously seen how human biology can surprise us with unanticipated pharmacologic or toxic drug effects. Few, if any, would have predicted that a proton pump inhibitor could so dramatically augment an antibiotic’s ability to block the cardiac hERG channel. 

Patients today take so many medications under so many different clinical conditions that we can never expect sponsors to conduct clinical studies to rule out or confirm every potential interaction. Therefore, a series of logical orthogonal experiments such as those applied in this research could serve as the basis for selecting interactions with high likelihood of clinical relevance. This will allow us to invest our clinical resources into the study of interactions with a greater likelihood of improving patient outcomes.

As noted in the editorial by Roden et al., these results alone do not at this time support a change in the use of these drugs. However, the findings do point to the need for awareness and a closer examination of the safety of their concomitant use. Perhaps of even greater importance, the data suggest that there may be a biological connection between the functioning of the hERG channel and the myocardial proton pumps that deserves exploration. The real impact of this research may have less to do with drug-drug interactions and more to do with cardiac physiology.

This study shows that the combination of ceftriaxone and lansoprazole prolongs QTc interval and relates this observation to the blockade of the human Ether-à-go-go-Related Gene (hERG) potassium channel blockade. The pertinent aspect is that this study points into the direction of the arrhythmia risks associated with proton pump inhibitors (PPIs). These drugs are widely used in medical practice and a growing number of cases of hypomagnesemia with chronic use of PPIs have been described, attributed to impaired intestinal absorption. This is even more pronounced in patients who are concomitantly treated with diuretics, eg, those with systemic hypertension. Hypokalemia is not generally caused by PPIs alone. However, in extreme alkalosis or with an impaired potassium recycling system, PPIs may cause hypokalemia even unrelated to hypomagnesemia. The hERG encodes for a potassium channel protein known as Kv11.1, which colocalizes with the magnesium channel TRPM6 in the distal collecting tubules, and interference with Kv11.1 may interfere with magnesium reabsorption. These dynamics may thus set up the perfect storm for torsades to develop in those on chronic PPI therapy and especially in combination with diuretic therapy. Practitioner needs to take these dynamics into consideration in the care for their patients, particularly given that diseases such as GERD, peptic ulcer disease, and hypertension are so common as is the prescribing pattern of these drugs often thought to be so harmless. H2 blockers may be considered as an alternative to PPIs as suitable, and if not, magnesium levels should be followed and replaced. Recovery from hypomagnesemia is relatively quick after stopping PPIs, usually within a few days. While not mentioned in this study directly, these are important aspects for daily patient management.