P3-056: An interactive frailty index predicts progression of cognitive impairment and mortality

Frailty is the concept that multiple minor physiological abnormalities interact to non-specifically increase the risk of adverse outcomes, including cognitive impairment and death. However, existing frailty scales are all additive. We tested if an interactive frailty index (IFI) predicts cognitive impairment and death better than an additive frailty index (AFI). Inclusion criteria were age over 60; CAMCOG scores over 70; blood data at initial assessment; subsequent survival of at least 18 months. We assessed patients 6 monthly and cognitively healthy controls annually using CAMCOG. We constructed frailty indices from 24 routine biochemical and haematological variables. We first transformed each variable to a standard normal distribution (mean =0, standard deviation=1). The AFI was the number of standardised variables lying outside the 2.5th–97.5th centiles; the IFI was the sum of the 2-way cross-products of the absolute raw standardised values. Finally, we analysed the relations of these frailty indices with CAMCOG scores and survival using generalised linear mixed models and Weibull regression. Of 545 participants, 350 were non-demented, 120 had Alzheimer's disease and 57 had other dementias. 34% died (median survival 5.7 years), with more deaths among dementia patients (OR=3.8, p<0.001). Higher AFI scores related to worse initial CAMCOG scores (z=-2.98, p=0.003) and faster cognitive decline (z=-2.20, p=0.028). IFI scores showed similar but stronger relationships (CAMCOG: z=-3.64, p=0.0003; decline: z=-2.40, p=0.017). When we included both the AFI and IFI in one model, the IFI still related to lower CAMCOG scores overall (z=-2.16, p=0.03), but the AFI did not (z=0.67, p=0.5). Higher AFI scores predicted shorter survival (z=-2.13, p=0.033). Higher IFI also predicted shorter survival in a separate model (z=-3.03, p=0.0025). Again, when we included the AFI and IFI indices in one model, the effect of IFI was still significant (z=-2.09, p=0.037) but that of AFI was not (z≈0.5, p≈0.6). Our findings support the previously untested hypothesis that frailty may consist in multiple interactions between minor physiological abnormalities. Such frailty relates to cognitive impairment and death, especially in dementia. Additive frailty indices used clinically may obtain their relationship with adverse outcomes vicariously, via their positive relations to interactive effects.