Are Genome-wide Association Studies Worth the Trouble?

Practicing clinicians spend their days managing the variability in each patient’s response to anesthetic drugs. The variability in sensitivity to propofol, for example, is greater than at least a factor of two. Underdosing leads to risk of intraoperative awareness, and overdosing leads to unnecessary cardiovascular and central nervous system inhibition. For both clinical and scientific reasons, it would be very useful to systematically explain this between-patient variation in the concentration–response curves. However, the incorporation of the usual demographic and disease factors will typically explain only a small proportion, estimated to be about 10%, of the total variation. In this age of relatively inexpensive genome sequencing, it would seem reasonable to examine genetic variation as another contributing factor for patients’ differing responses to propofol. The role of variation in genes and their expression as drivers of propofol sensitivity would not only assist clinicians in planning patient care but also aid in the understanding of the mechanisms underlying the effects of propofol on consciousness.

Surprisingly, there are not a large number of publications looking at genetic influences on propofol potency in clinical data, and most have concentrated on the pharmacokinetic aspects of propofol dosage.  In this month’s Anesthesiology, Ahlström et al. report a genetic study of propofol sensitivity in a relatively large homogeneous patient group, with almost all having one type of surgery. By narrowing down onto a standardized clinical scenario, the authors hoped to minimize environmental variation and hence maximize the chance of discovering genetic influences on propofol sensitivity. They then performed a genome-wide association study (GWAS) that incorporated the usual demographic variables that might be related to propofol potency as covariates in their model of propofol sensitivity. A GWAS study takes advantage of a number of relatively common natural variations among individuals in their DNA sequences. These variations (also termed “single-nucleotide polymorphisms”) may or may not be in genes (they can also be in the “silent” or noncoding part of DNA). The authors then looked for cosegregation of some of these variations (single-nucleotide polymorphisms) with changes in propofol sensitivity as determined by electroencephalography responses to the drug.

Sequences of DNA are broadly divided into those that directly code for proteins (exons) and the rest (noncoding), which includes those sequences required for gene regulation, noncoding RNA, and other functions. In this study, Ahlström et al.  did not find any exon variants that were significantly related to propofol sensitivity, i.e., no mutant genes. They did find cosegregation of some noncoding DNA regions with sensitivity. These variations may be associated with regulation of proteins involved in the action of propofol or may simply be close to such important regions. However, none of these were directly related, either by function or location, to the classically described molecular targets of propofol, such as the γ-aminobutyric acid type A receptor. Importantly, their data did not replicate those from previous studies that had identified altered propofol sensitivity being dependent on single-nucleotide polymorphisms that may alter the rate of propofol metabolism and some γ-aminobutyric acid receptor polymorphisms.  Although the targets identified do not fit our preconceived notion of the “right” answer, the entire point of a genetic screen is to identify something one did not know. Therefore, although the significance of these findings remain to be established, the results are thought provoking and suggest new directions for more directed studies.

These somewhat inconclusive and confusing results are the typical outcome for many (perhaps the majority) of genome-wide association studies into a wide range of other diseases. Often the limiting problem is the size of the population being used for the study. For example, many hundreds of gene variants (single-nucleotide polymorphisms) have been found to be associated with type 2 diabetes in huge studies totaling over a million patients, but this information does not seem to materially improve clinical management or risk prediction.  In this aspect, the current study, when viewed from statistical considerations, has a rather small number of participants. However, it represents a very significant effort when considering that the participants are humans who must be anesthetized, and it more than doubles the previous number of individuals used for such studies with propofol.

At present, there is a bit of a crisis in the interpretation of GWAS studies. One possible explanation is that genetics has very little to do with anesthetic sensitivity. This would violate one of our strongest underlying tenets in medicine—that responses to external factors are controlled largely by our genes. However, the confusing results of GWAS do suggest that the mechanisms of how the genome manifests expression in a living human may be more subtle and complex than the traditionally taught paradigm of DNA–mRNA–tRNA–protein. Seldom do our interpretations of GWAS involve microRNAs, silencing RNAs, epigenetics, or other “(not) silent” DNA effects. The sincere hope is that as studies progress, we will improve in our abilities to include these other factors and apply them retrospectively to these current studies.

Of course, an alternative explanation for the GWAS results is that the sheer numbers overwhelm the statistical methods. There are several million single-nucleotide polymorphisms in each patient (out of 3 billion base pairs) of which this study tested 653,034 single-nucleotide polymorphisms. Even if only a tiny percentage cosegregate with sensitivity, there is still an enormous problem of false-positive and -negative results appearing because of huge numbers of multiple comparisons. The problem can often be addressed by performing meta-analyses on multiple such studies, done in as close to identical protocols as is possible, or by comparing the results of the GWAS studies to related genetic studies in other animals. With that final consideration in mind, it is very interesting that this study implicates γ-aminobutyric acid neurogenesis and synaptic neurotransmitter release (FEZ1), a sodium leak channel (NALCN), and has suggestive data implicating a sulfotransferase. Each of these has been implicated in other studies of anesthetic sensitivity.

So, how does one proceed in the light of the confusing data. First, it is important, in our opinion, to keep looking. Meta-analyses can only be done if there are many data sets. Each may disappoint on their own but converge when grouped into a clarified understanding. Second, those sites that show cosegregation need to be followed up in directed studies, including designs that can establish causation by experimental manipulation of expression of these sites. The pathways involving neurotransmitter release, sodium leak channel, and others that were close to significance need to be studied on their own. One or more of them will eventually contribute to our improved understand.

Why pursue this path of inquiry in the first place? First, and foremost, we use these medications for millions of patients a year. In our opinion, we are obligated to know as much as possible about how they work, who might not respond like the majority, and what the long-term effects might be. We owe that to our patients. Second, as anesthesiologists, in many ways we perform the most neuroscience “experiments” of any one group in science or medicine. If we do not figure out these mechanistic details . . . who will?