Baseline characteristics of 45,436 participants overall and stratified by baseline disease number categories are shown in Table 1. Participants had a mean age of 55.67 ± 8.21 (SD, standard deviation) years, and 55.35% were female. There were significant differences between disease number categories for all covariates (all Ps < 0.001). Age-related disease numbers identified at baseline ranged from 0 to 11, with 20,857 reported none of the disease and 13,372 participants reported one and 11,207 reported two or more. Compared with participants with zero disease, participants with multimorbidity at baseline tended to be older, male, of white ethnicity, to have higher TDI and lower education level, less active physically, more smokers, and less alcohol consumers, and to have higher body mass index, a lower longevity GRS.

Retinal age gap in the study population followed a nearly normal distribution (Supplementary Figure 1). The mean (SD) of retinal age gap is 0.54 (4.06) years. Age- and sex-adjusted models showed that older age, male sex, non-White ethnicity, meeting exercise recommendations, and higher longevity genetic risk scores were associated with lower retinal age gaps while higher Townsend deprivation, smoking, alcohol consumption, and BMI was significantly associated with larger retinal age gaps (all P value < 0.05, Supplementary Table 3).

The associations between disease numbers and retinal age gap are shown in Table 2. After adjusting for age and gender, disease numbers had significantly associations with retinal age gaps (model 1: β = 0.093, 95% CI 0.064, 0.122, P < 0.001). When participants were categorized into different disease number groups, compared to participants with zero disease number, those with multimorbidity and one disease both showed significant increases in retinal age gaps (model 1: β = 0.306, 95% CI 0.219, 0.392; P < 0.001; β = 0.249, 95% CI = 0.171, 0.327; P < 0.001; respectively). These findings remained significant after comprehensive adjustments for covariates (model 3: β = 0.254, 95% CI 0.154, 0.354; P < 0.001; β = 0.203, 95% CI 0.116, 0.291; P < 0.001; respectively).

After a median follow-up period of 11.38 (IQR, 11.26–11.53; range, 0.02-–11.81) years, a total of 3607 (17.29%) participants had incident multimorbidity. As shown in Table 3, compared with those who did not experience the outcome, participants with incident multimorbidity tended to be older, male, of white ethnicity, and have lower TDI and education level, smokers, to have higher body mass index, and with lower longevity GRS (all Ps < 0.001).

After adjusting for age and sex, each 5-year increase in retinal age gap was independently associated with a 12% increase in the risk of multimorbidity (model 1: HR = 1.12, 95% CI =1.07, 1.18; P < 0.001), as shown in Table 4. This association remained significant after further adjustments (model 2: HR = 1.09, 95% CI 1.03, 1.15, P = 0.003; model 3: HR = 1.08, 95% CI 1.02, 1.14, P = 0.008). Compared with groups of retinal age gap within ± 3 years, retinal age gap less than minus 3 years was associated with a 9% decreased multimorbidity risk (model 1: HR = 0.91, CI 0.83, 0.99, P = 0.038) while retinal age gap more than 3 years showed an 15% increased risk of multimorbidity incidence (model 1: HR = 1.15, CI 1.06, 1.25, P = 0.001). Individuals with a retinal age gap of more than 3 years showed 12% increased risk of multimorbidity incidence in fully adjusted models. In addition, the Kaplan-Meier survival curves for each retinal age gap group did not cross, supporting the proportional hazards assumption (Supplementary Figure 2).

Restricted cubic spline analyses showed that the risk of incident multimorbidity increased significantly when the retinal age gap reached −1.78 years (P-overall = 0.002; P-nonlinear = 0.028, Fig. 1). After adding interactive terms in the cox model, we identified significant interaction effects between retinal age gap and smoking status (p = 0.017) as well as education level (p < 0.001).

Nonlinear association between retinal age gap and incident multimorbidity. The model was fitted with a restricted cubic spline for retinal age gap adjusted for age, sex, ethnicity, Townsend, education, physical activity, smoking status, alcohol drinking status, body mass index, and longevity genetic risk scores (GRS). Evidence of an overall and nonlinear association between retinal age gap and multimorbidity risk was observed (Poverall = 0.002; Pnonlinear = 0.028). The association between retinal age gaps and multimorbidity is depicted as a J-shaped curve, where positive retinal age gaps were associated with substantially increased risks of multimorbidity