Genetic test positivity prediction model
The genetic test positivity prediction model developed in this study is useful to analyse a priori a patient’s chance of finding a mutation when ordering genetic testing. In the clinical practice scenario, cardiologists can give a more accurate estimate to patients regarding the molecular test. In the large scale screening scenario, such as in national cascade screening projects, only patients with a greater chance of having a positive genetic test would be analysed, which could optimize the use of reagents and analysis time. This model is particularly important in centres with fewer resources, because the genetic test is still very expensive. Especially in the Brazilian scenario, where few academic centres have the structure and budget to perform the test, this method can serve as a cost-effective tool. Our main focus with this model is the cascade screening. Patients with borderline hypertrophy or with uncertain diagnosis were not included in this study, so such predictor may not be accurate for this group.
The sensitivity and specificity values can be adjusted according to a particular interest. For example, if one wants to maximize the number of positive patients included, the sensitivity and specificity can be modulated by changing cut-off values. In the simulations for the studied population, adjusting the predicted probability cutoff for 0.34, which represents an approximate 90% sensitivity, from the 122 included in the analysis, 91 would be tested. Of those, 60 patients would be positive. From the 31 patients not tested, only 7 would have a positive result, so we would not be testing 24 patients predicted as negative. These savings, in a national screening program, can signify an important economic resource.
To applied this model, one can use the values from Table
3 to calculate the predicted probability of a positive genetic test. P is equal the predicted probability, βo is the constant value, β1 is the variable constant and x is the variable value if continuous or 0 and 1 if the variable is categorical.
In this work, we are using predicted probability, thus what we can conclude is regarding higher or lower probabilities, not certainties. For example, we saw in the logistic regression model that each patient’s year addition decreases the chance of finding a mutation in the genetic test. We did not define a cutoff for this measure, meaning that one can find the mutation in any age, but the older the patient is, the lower is the probability of finding a mutation. On the other hand, the presence of a confirmed HCM family history increases the chance of finding a mutation almost 6 times when compared to patients who don’t have a positive HCM family history.
Ingles et al. [
5] also used this approach to identify positivity predictors for genetic testing in an Australian HCM population. The multivariate analysis of this population identified female sex, LV thickness, HCM familial history, and SD familial history as associated with a higher chance of mutation identification. The authors considered familial history as a key predictor of a positive genetic test in their population, with a 3 times higher chance of a positive result compared to patients without a familial history, which was similar to what we found in this study. In their study, this detection rate was even higher when the patient also had an SD familial history. Differently from what was found in this study, age at diagnosis was not significant in the Australian population, although the p value was 0.052.
Another study performed by Gruner et al. [
6] in a Toronto HCM population showed that in these patients age at diagnosis, female sex, HCM familial history, and SD familial history were also correlated with a higher probability of a mutation identification. In addition, the study correlated hypertension and dyslipidaemia as negative predictive factors, with a higher frequency of both in genetically negative patients. Other strong predictors identified were morphology subtype, as previously described [
13] and maximal wall LV thickness and LW thickness.
The identification of NSVT as one of the predictors for a positive genetic test is a very interesting observation, since this is one risk factor for sudden death in HCM patients. In a previous work, Olivotto et al. [
14] compared patients with and without an identified sarcomeric mutation and found that there was a difference regarding those, with patients with a positive genetic test being related to a less favourable clinical outcome, especially regarding end stage heart failure, but not sudden death. This finding in our study may show that the presence of a mutation can indeed be related with a higher risk of sudden death related risk factors.
Study limitations
The limitations of this study are the lack of patients with hypertension, since it was an exclusion criterion for HCM diagnosis in the participating centres; therefore, we could not test it as a negative predictor. Also, we only studied the three most important genes (MYH7, MYBPC3, and TNNT2), thus patients with mutations in other sarcomeric genes may be wrongly labelled as mutation negative. But the frequency of these genes in HCM is very low, so we believe that the lack of these data does not change the results found.
These finding may be limited to the Brazilian population. Replication studies in other populations should be made to confirm these results.
As new technology becomes available, such as next generation sequencing screening techniques, this reality may change and screening programs will have to constantly adapt to new molecular data.