Simulation study

We present a simulation study in Stata (script name "Simulation study")

Simulated Data

we simulated data for 10,000 individuals. The simulation process is described below

Smoking status at time zero

variable name: sm_status0

32% never smokers 20% former smokers 48% current smokers

Sex

variable name: sex

50% men 50% women

Age (in years) at time zero

variable name: age

Drawn from a normal distribution with: Mean: 50, sd: 2.5

Family history of CVD

variable name: fh_CVD

It is drawn from a Bernoulli distribution with Pr(fh_CVD=1) = 0.12+ sd 0.05

Diuretics at time zero

variable name: diuretics0

It is drawn from a Bernoulli distribution with Pr(diuretics0 =1) = -0.14+0.03age+0.1sex + sd 0.04

BMI at time zero

variable name: BMI0

Conditional on age, sex, family history of CVD, smoking status (at time zero) and diuretics (at time zero), BMI at time zero was drawn from a normal distribution with mean 25.8+0.06*(age-18)+0.1sex+0.5fh_CVD-0.1sm_status0-0.2diuretics0 and sd 0.56

Height

variable name: height

Drawn from a normal distribution with: Mean: 1.63m, sd: 0.03m --> for women Mean: 1.80m, sd: 0.03m --> for men

Smoking cessation between time zero and the 1st follow-up

variable name: sm_cess1

It is equal to 0 for never and former smokers at time zero and for those who were current smokers at time zero, it is drawn from a Bernoulli distribution with Pr(sm_cess=1) = -1.8+0.041age-0.01sex+0.2*diuretics0+ sd 0.04

Diuretics at the 1st follow-up

variable name: diuretics1

It is drawn from a Bernoulli distribution with Pr(diuretics1 =1) = 1.4+0.03age+0.1sex+0.1*diuretics0 + sd 0.04

CVD at the 1st follow-up

variable name: CVD1

It is drawn from a Bernoulli distribution with Pr(CVD1 =1) = -1.9+0.035age+0.03sex+0.3fh_CVD+0.1diuretics0 + sd 0.03

BMI at the 1st follow-up

variable name: BMI1

Conditional on BMI at time zero, age, family history of CVD, smoking status (at time zero) and diuretics (at time zero), BMI at time zero was drawn from a normal distribution with mean BMI0-0.9CVD1+ 0.1 sm_status0 +2.5sm_cess1-1.8diuretics0-0.9diuretics1+0.005age and sd 1.4

Weight

variable name: : weight0 (at time zero) and weight1 (at 1st follow-up)

Weight0=BMI0height^2
Weight1=BMI1height^2

Weight change

Combining information on weight1 and weight0, we created the 3 categories of weight change, i.e.

weight loss:

(weight1- weight0)/ weight0 < -5%

weight maintenance:

(weight1- weight0)/ weight0 ≥ -5% & (weight1- weight0)/ weight0 ≤ 5%

weight gain:

(weight1- weight0)/ weight0 >5%

Time to event data

We then simulated time to event data for the next 18 years from a Weibull distribution with lambda 10^(-4) and gamma 1.1 and the following log hazard ratio for the covariates (Table S4), after deleting the participants that developed CVD between time zero and the 1st follow-up

Weighted Kaplan-Meier curves

We run an example in Stata (script "Weighted Kaplan Meier curves").

We present how we can estimate the Weighted Kaplan-Meier curves using IPW from the 1st simulated dataset "Final dataset_1" from the "simulation study" script

Risk Curves using the g-formula

We run an example in Stata (script "Risk Curves using the g-formula").

We present how we can estimate the risk curves using the g-formula from the 1st simulated dataset (in a long format) "Final dataset (pooled)_1" from the "simulation study" script