Multidomain Hypertension-Mediated Organ Damage in Ghanaian Adults: Prevalence, Correlates, and Brachial-Ankle Pulse Wave Velocity Performance

Background: Comprehensive assessment of hypertension-mediated organ damage (HMOD) across multiple organ systems in sub-Saharan Africa is limited. We assessed the prevalence and correlates of multidomain HMOD in a geographically diverse population in Ghanaian adult. Methods: This cross-sectional secondary analysis of the Ghana Heart Study, which included 1,106 adults aged [&ge;]18 years from four Ghanaian regions between September 2016 and March 2017. Multidomain HMOD was determined using a pre-specified 9-domain composite score [&ge;]2, using an ESH/ESC 2018 guideline-informed selection of HMOD domain with baPWV instead of carotid-femoral PWV (cfPWV), due to device unavailability, and a threshold of [&ge;]14 m/s which was derived from analysis within the cohort. LODO sensitivity analyses were used to address issues of predictor-outcome circularity. We used logistic regression models to examine association between each predictor and multidomain HMOD, adjusted for age, systolic blood pressure, body mass index, presence of dyslipidaemia and smoking status. We also performed receiver operating characteristic (ROC) analysis to determine correlates of multidomain HMOD and compare the discriminative ability of each predictor against the others. Results: The mean age of participants was 46.9{+/-}17.2 years of which 58% were females. Multidomain HMOD was observed in 21.3% (235/1,106; zero-imputation lower bound 21.2%) of participants studied. There was a marked increase in the prevalence of multidomain HMOD with advancing age. Thus, while 8.6% (44/ 511) of adults<45years had multidomain HMOD, 20.6% (63/306) of 45- to 59-yr-olds and 44.4% (128/ 288) of individuals [&ge;]60 years had multidomain HMOD. HMOD-positive adults were older (59.1{+/-}8.4 vs 43.6{+/-}13.4y, p<0.001), had higher systolic BP (147{+/-}22 vs 123{+/-}21 mmHg, p<0.001), and had higher prevalence of hypertension (73% vs 28%, p<0.001) than their HMOD-negative counterparts. Using the primary (circular) specification, the strongest co-occurrence among all domains of HMOD was observed between peripheral artery disease and other HMOD (OR 41.2, 95% CI 20.7-81.6; p<0.001) followed by valvular burden and other HMOD (OR 14.4, 95% CI 4.8-43.8; p<0.001) and between ECG-LVH and other HMOD (OR 9.0, 95% CI 5.9-13.8; p<0.001) (S2 Table). After LODO correction to remove the self-inclusive co-occurrence between each predictor domain and the outcome (all p-values calculated in S2 Table), there was no significant association between the remaining 8 HMOD domains and the prevalence of multidomain HMOD (all p-values>0.05; S2 Table). This was not the case for baPWV, however. Thus, whereas the AUC of the best performing non-self-inclusive HMOD domain (ECG-CMD) only reached 0.688{+/-}0.016 (vs 0.827{+/-}0.008 for self-inclusive AUC calculated for the sake of interest only and provided as supplementary material), baPWV demonstrated good discriminative capacity (LODO-adjusted AUC = 0.702, 95% CI 0.654-0.751; S3 Fig). However, this AUC did not significantly exceed that for age alone (AUC = 0.752; {Delta}AUC = -0.050, 95% CI ?0.103 to 0.03; p=0.106; S3 Fig). Most importantly, after adjustment for SBP (a direct mediator in this pathway), the LODO AUC for baPWV did not exceed that for the single variable age (S3 Fig), indicating that baPWV does not possess independent discriminative power for multidomain HMOD above and beyond the information provided by SBP and age. Importantly, however, the adjusted OR for baPWV did not reach statistical significance (OR 1.094, 95% CI 0.986-1.213; p=0.091), suggesting that while circularity prevented validation of biological association, it did not prove the absence of association altogether. Sensitivity analysis (estimating total as opposed to direct effect) in which SBP was excluded from the regression model to estimate the total effect of baPWV on the prevalence of HMOD showed that, indeed, the OR for baPWV was significantly elevated (OR 1.261; 95% CI 1.150-1.382; p<0.001) in this specification. The effect of SBP, a direct mediator in this pathway, therefore apparently accounted for the non-significance in the original model entirely. Formal mediation analysis using the aforementioned specification yielded that SBP indeed mediated 69.9% (95% CI 41.3-128.8%) of the effect of baPWV on the prevalence of HMOD. Conclusions: One in five Ghanaian adults has hypertension-mediated organ damage in multiple HMOD domains. baPWV has good discriminative power for HMOD risk prediction in a Ghanaian adult population under the non-circular LODO estimand (LODO- adjusted AUC = 0.702; 95% CI: 0.654, 0.751) than the PCE (AUC = 0.496; 95% CI: 0.438, 0.555; {Delta}AUC = +0.206; p < 0.001). However, baPWV LODO AUC (0.702) was not statistically significantly greater than age alone (AUC = 0.752; 95% CI: 0.730, 0.774; {Delta}AUC = -0.050, p = 0.106). AUC for self- inclusive model was provided in supplementary materials for the reader's perusal, and that AUC (0.827; 95% CI: 0.794, 0.860) is circular. The prevalence of ECG-LVH was substantially higher (42%) than that of echocardiographic- LVH (5.9%) in this Black African population. These findings support further research on the role of baPWV for HMOD risk prediction in a Ghanaian adult population. Prospective validation of baPWV would be needed before clinical use.