Table 2. Association of Eye Surgery as a Time-Varying Exposure and Subsequent All-Cause Dementia as the Outcome Among People With Cataract Diagnosis.

Modela	Model description	Hazard ratio (95% CI)
		Surgery exposure (time varying)	Time since surgeryb
			>0 to 5 y	>5 y
Model 1c	Primary model	0.71 (0.62-0.83)	0.68 (0.56-0.81)	0.76 (0.63-0.92)
Sensitivity analysesd
1a	Omit 1994-1996 enrollment cohort	0.52 (0.39-0.69)	0.47 (0.34-0.66)	0.63 (0.59-0.95)
1b	Exclude surgery 2 y prior to censoring	0.57 (0.48-0.66)	0.44 (0.35-0.55)	0.70 (0.58-0.85)
1c	Adjust for additional covariates	0.75 (0.65-0.88)	0.72 (0.59-0.86)	0.80 (0.66-0.97)
1d	Consider only incident cataract cases	0.70 (0.56-0.87)	0.69 (0.53-0.89)	0.72 (0.54-0.95)
1e	Incident cataract cases, controlling for CASI at time of cataract diagnosis	0.70 (0.57-0.87)	0.69 (0.53-0.89)	0.72 (0.55-0.96)
1f	Adjust recent vs long-term threshold to 2-y window	NA	0.60 (0.46-0.79)	0.75 (0.64-0.88)
1g	Adjust recent vs long-term threshold to 10-y window	NA	0.71 (0.61-0.83)	0.72 (0.54-0.97)
Model 2e	Marginal structural model with weights for surgery, death, and dropout to account for healthy patient bias	0.71 (0.60-0.85)	0.73 (0.61-0.88)	0.66 (0.51-0.86)
Sensitivity analyses
2a	Marginal structural model with weights for surgery only	0.73 (0.62-0.87)	0.75 (0.62-0.90)	0.70 (0.54-0.90)
2b	Adjust for additional covariates	0.72 (0.61-0.86)	0.73 (0.60-0.88)	0.70 (0.54-0.91)
Model 3f	Glaucoma surgery	1.08 (0.75-1.56)	1.15 (0.72-1.83)	1.00 (0.59-1.70)

Abbreviations: ACT, Adult Changes in Thought; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); CASI, Cognitive Abilities Screening Instrument; IADL, instrumental activity of daily living; NA, not applicable.

^a All models use age as the time axis and adjust for years of education, self-reported White race, and past or current smoking and are stratified by any apolipoprotein E ε4 alleles, sex (to meet proportional hazards assumptions), and age at first cataract diagnosis (<68, 68-71, 72-76, and ≥77 years).

^b A time-dependent covariate that is set at the first category (0 to 5 years) for the first 5 years after surgery and then to the second category (>5 years) after that.

^c During follow-up of 23 554 person-years, there were 853 cases of incident dementia; 504 cases occurred during the 15 941 person-years before or without cataract surgery (0.033 per person-year), and 320 occurred during the 7603 person-years after surgery (0.042 per person-year).

^d Sensitivity analyses include results after excluding the original ACT cohort recruited between 1994 and 1996 (model 1a; n = 2868); ignoring any cataract surgical procedure occurring within 2 years of dementia diagnosis or censoring (model 1b); adding these additional covariates to model 1: diabetes, systolic blood pressure, hypertension, heart disease, cardiovascular disease, BMI, self-rated health, Charlson Comorbidity Index, number of activities of daily living and IADL limitations, at least 15 minutes of physical activities 3 times a week, performance-based physical function scores, Centers for Epidemiologic Studies Depression Scale scores, retirement status, and difficulty with near and distance vision (model 1c); excluding data from people with prevalent cataract diagnosis at time of ACT study entry (model 1 day; n = 1556); including baseline CASI score at the time of diagnosis (model 1e; n = 1556); limiting the recent cataract category to a 2-year window (model 1f); and limiting the recent cataract category to a 10-year window (model 1g).

^e Model 2 used stabilized time-varying weights to adjust for the probability of surgery, death, and dropout (eMethods 4 and eTables 2, 3, and 4 in the Supplement). Model 2a used stabilized time-varying weights in a marginal structural model adjusting only for the probability of surgery (eMethods 4 and eTables 2-4 in the Supplement). Model 2b is model 2 additionally controlled for diabetes, systolic blood pressure, hypertension, heart disease, cardiovascular disease, BMI, self-rated health, Charlson Comorbidity Index, number of activities of daily living and IADL limitations, at least 15 minutes of physical activities 3 times a week, performance-based physical function scores, Center for Epidemiologic Studies Depression Scale scores, retirement status, and self-reported difficulty with near and distance vision.

^f Model 3 is survival analysis with the same covariates and dementia outcome as in model 1 but with the exposure of interest as history of glaucoma surgery instead of cataract surgery (n = 728) and risk starting with first glaucoma diagnosis. During 5029 person-years of follow-up, there were 230 cases of incident dementia; 194 cases occurred during the 4497 person-years before or without glaucoma surgery (0.043 per person-year), and 36 cases occurred during the 553 person-years after surgery (0.062 per person-year).