Frailty trajectory, not blood pressure direction, predicts mortality and cardiovascular disease in adults aged 65 and older: harmonized joint- trajectory analyses of CHARLS and HRS Running title: Frailty trajectory versus blood pressure direction aft

Abstract Background A fall in systolic blood pressure (SBP) after age 65 is ambiguous: it may mark successful treatment or deteriorating health. Whether joint longitudinal phenotypes of SBP and frailty add prognostic meaning beyond frailty itself is unclear. Methods Parallel landmark analyses in CHARLS (China) and HRS (USA). Participants were aged 65 years or older at entry with three paired SBP and harmonized 30-item frailty index (FI) measurements (CHARLS n = 1,902; HRS n = 3,839). We independently fitted linear group-based multi-trajectory models, then estimated mortality (attained-age Cox models) and self-reported incident cardiovascular disease (CVD; complementary log-log models) under retention and selection weighting and 10 covariate imputations, decomposing the information contributed by the FI slope, the SBP slope and joint classes. Results Four classes emerged per cohort, ordered by frailty severity rather than SBP height. The highest-FI class—ordinary-to-moderate SBP with high, rapidly rising frailty—had mortality hazard ratios (HRs) versus the lowest-FI class of 1.65 (95% CI 1.18–2.31) in CHARLS and 2.21 (1.64–2.97) in HRS, and CVD HRs of 3.21 (1.88–5.49) and 2.51 (1.32–4.77). Rapid FI increase was associated with mortality within every SBP direction (HRs 1.43–2.15); no multiplicative interaction was detected, and the SBP slope itself carried no signal (HR per SD 1.06 and 0.94). Joint classes never outperformed a simpler model of initial status plus the FI slope (change in apparent C-index 0.004–0.010). Only about half of baseline-eligible participants—younger, less frail survivors—completed the three-measurement requirement; weighting left estimates essentially unchanged. Conclusions After age 65, the frailty trajectory, not the direction of blood pressure, carries the prognostic signal. A falling SBP warrants concurrent assessment of frailty, but joint trajectory classes are descriptive phenotypes, not a validated prediction tool or a treatment target.