The Value of Genetic Risk Scores to Predict Hypertension

For most hypertension researchers and clinicians, the genetic revolution, which has consumed so much of our scientific resources and funding in the past few decades, might be thought of as somewhat disappointing. Partly, this has come from the realization that for most patients with hypertension, the genetic disposition is highly polygenetic, which means that each of the more common variants contributes very little on their own and as such they may be difficult to find. On the plus side, a number of monogenetic causes of hypertension, such as Liddle syndrome,1 have been discovered as well as single genes that are associated with lower blood pressure (BP), such as Gitelman syndrome, in which a mutation in the thiazide-sensitive sodium chloride cotransporter in the distal tubule of the kidney leads to a loss of function.2 A major breakthrough was the discovery of mutations in WNK1 and WNK4 kinase, which cause pseudohypoaldosteronism type II.3 However, such mutations account for a very small number of cases, and finding the genetic contribution to hypertension has proved to be extremely elusive. It is not that the genetic contribution is so small, however. The estimates indicate a range somewhere between 30% and 70%4 but a more reasonable estimate from a number of combined studies suggests that it is more likely closer to 30%.5 There are two main approaches to tackle the problem of finding the causative genes in a polygenetic disease. Firstly, a hypothesis-based approach can be used where known disease mechanisms generate a range of genes to be tested for their association with the disease. However, this approach does not necessarily discover new or novel mechanisms. The second approach is to look for genetic associations that are not hypothesis driven. For a minimal cost of analysis of a sample from an individual patient, one can reveal a myriad of disease risks, thus providing a wealth of information to research scientists through genome-wide associations study (GWAS). However, the issue is to link this large amount of data into a framework that provides useful clinical information. It must be remembered that the identified risk of single nucleotide polymorphisms (SNPs) may not cause the disease but may simply be associated or linked with other unmeasured genetic variants that do cause the condition. Also, it has been suggested that the number of genes contributing to the condition has been underestimated and the power of linkage studies to find genetic variants that make only a modest contribution to hypertension is low because of the very high statistical threshold set for eliminating type 1 error.6 A major advance in the field has been to use genetic risk scores (GRSs), which combine the risk-associated variation across the genome from multiple SNPs from GWAS findings. The advantage of this approach is that each individual SNP is less important and the score is less influenced by imperfect linkages. There have been some very large and important GRSs linking loci and hypertension7, 8 and the field has been nicely reviewed by Ehret.4 However, as exemplified by Taal and colleagues9 in a 2012 study, the variance of adult systolic BP explained by GRS, while four- to five-fold greater than single SNPs, is still very small at about <1.2% of the total and very much short of the expected level of inheritability. The manuscript by Niiranen and colleagues10 in this issue of The Journal of Clinical Hypertension examines the predictability of BP change over time and incident hypertension at baseline using a GRS for hypertension constructed from 32 SNPs. This large study included more than 5000 patients at baseline and more than 3000 at follow-up after 11 years. The important features of this study that make it stand out among previous studies is the random population selection, a long follow-up period, and the use of identical techniques used to assess patients at entry and follow-up.10 Trained nurses measured BP with duplicate measurements 5 minutes apart on both occasions, which reduces the white-coat effect considerably.11 The analysis included multivariate models of important nonmodifiable and modifiable risk factors for hypertension, with the most sophisticated accounting for cofounders of age, sex, smoking status, diabetes, education, hypercholesterolemia, exercise, and body mass index. The analysis showed that the GRS was associated with higher systolic and diastolic pressure values at baseline but was not particularly suitable for prediction of the increase in BP over time. Being such a large study, it was highly powered to detect relatively small contributions, but it appears that the GRS yielded no more than the simpler approach of using family history of hypertension. Despite the highly significant association at baseline, the model explained only 1% of the variance in systolic BP, which is in line with all previous studies of this type. Niiranen and colleagues suggest that better BP phenotyping could be used as well as next-generation sequencing to fully explore the whole genome and exome.10 Both of these are considerable undertakings in studies that require large data sets, but there are a number of international collaborations that have collated quality ambulatory BP measurements as well as home and clinic measurements, which would markedly improve the characterization of the true level of BP in each patient. Examples are the International Database of Ambulatory Blood Pressure in Relation to Cardiovascular Outcome (IDACO)12 the Dublin Outcome study13 the Ohasama study14 and Australian studies.11 However, whether they have samples for genetic analysis is the issue. A smaller but well-characterized BP study using multiple measurements according to well-defined guidelines may be able to determine how GWS can better explain the genetic variance associated with BP and hence hypertension. The surprising aspect of the study by Niiranen and colleagues was that they could find only very weak associations of the GRS with BP increase over the 11-year follow-up.10 Participants in the highest tertile had only a one-third greater likelihood to develop hypertension compared with those in the lowest tertile. Over this period there was a 5-mm Hg increase in systolic BP and no change in diastolic BP, which is consistent with known patterns of aging and BP. In addition, nearly 9% of patients commenced antihypertensive therapy during this time. One aspect of the population studied was that the average participant was only just slightly overweight and the increase in body mass index over 11 years was quite modest. However, the values for body mass index are almost identical to those reported between 2000 and 2005 in an Australian population study15 suggesting that this may be typical of such population studies. The study by Niiranen and colleagues indicates that the gene linkages associated with the development of hypertension at different ages may be quite different.10 As suggested by Korner, genetic studies do not differentiate between those genes that initiate hypertension and those that are regulating the structural changes in the heart and vasculature, which determine the degree of rise in BP over the long term.5 One of Pickering's legacies has been the idea that the number of genes contributing to essential hypertension is large. The unanswered acceptance of this could be misplaced at least with the initiation of essential hypertension. In any event the question is unanswerable in the absence of a hypothesis about the pathogenesis of hypertension. Thus, we may need to combine the hypothesis-driven approach with GRS to increase the likelihood of success. The ability to sequence the entire genome in itself is a major step forward but now the speed and low cost of next-generation sequencing may help considerably in our ability to associate multiple genes with chronic diseases such as hypertension. However, we are a long way off from the day when genetic profiling will reveal individuals at high risk for developing hypertension over and above what can be predicted from family history. Clearly, more effort needs to be taken in correctly assessing each patient's true BP level during their normal day and night with ambulatory BP measurement, the gold standard, as recommended by many hypertension societies and experts.16-20 The authors declare no funding for this work. GAH is supported by NHMRC Principal Research Fellowship (1002186).