Exercise 9:    Smoothing plot and Threshold analysis

·         This exercise will use demo2 project. If you have not started demo2 project, click here to follow exercise 1: start the new project demo2

·         This exercise will use new variables. Click here to follow through Exercise 2:  Create new variables, code and label variables, if you have not done so.

Objectives:

·         Smoothing plot of SBP and DBP versus BMI by SEX

·         Threshold analysis of BMI with SBP and DBP

================================================================================================

 

Objective 1: To create following smoothing plot:

demo2_15_gam

Follow following steps to create the above table:

·         Click “Advanced analysis” -> “Spline smoothing plot”

·         Drag and drop variable SBP and DBP to “Select outcome variables (Y)”

·         Drag and drop BMI to “Select risk factors(X)”

·         Drag and drop AGE, EDUC, SMK, OCCU and ALH to “Select adjustment variables”.

·         Drag and drop SEX to “Select stratified variable(s)”

 

  

 

·         Click “View Graph” button, wait for a few seconds, you will get the above graphs

 

Done!

Objective 2:

According to the above smoothing plot, there is a possible threshold effect of BMI on SBP and DBP.  Follow following steps to conduct “Threshold analysis”:

·         Click “Advanced analysis” -> “Interactions test”

·         Drag and drop variable SBP and DBP to “Select outcome variables (Y)”

·         Drag and drop BMI to “Select risk factors(X)”

·         Drag and drop AGE, EDUC, OCCU and ALH to “Adjusting variables”

·         Drag and drop SEX to “Select stratified variable(s)”

             

·         Click “View Table” button, wait for a few seconds, you will get following table:

Threshold effect analysis

SEX = Male

SEX = Female

Total

SBP

Model I

1-line(ß)

-0.1 (-1.1, 0.8) 0.772

1.4 ( 0.6, 2.1) <0.001

0.8 ( 0.2, 1.4) 0.012

Model II

Turning point (K)

18.6 (10% percentile)

22.0 (66% percentile)

23.6 (86% percentile)

< K, ß(1)

4.3 ( -1.2, 9.9) 0.125

0.3 ( -1.0, 1.6) 0.639

0.3 ( -0.6, 1.1) 0.538

> K, ß(2)

-0.6 ( -1.7, 0.5) 0.309

2.6 ( 1.1, 4.1) <0.001

2.8 ( 0.7, 5.0) 0.008

ß(2)-ß(1)

-4.9 (-10.9, 1.1) 0.107

2.3 ( -0.1, 4.7) 0.062

2.6 ( 0.1, 5.1) 0.045

LLR test

0.103

0.059

0.044

DBP

Model I

1-line(ß)

-0.2 (-0.8, 0.3) 0.416

0.6 ( 0.2, 1.0) 0.005

0.3 (-0.1, 0.6) 0.117

Model II

Turning point (K)

18.6 (10% percentile)

21.9 (64% percentile)

24.3 (90% percentile)

< K, ß(1)

2.6 (-0.5, 5.7) 0.096

-0.1 (-0.8, 0.6) 0.803

0.0 (-0.4, 0.4) 0.928

> K, ß(2)

-0.5 (-1.1, 0.1) 0.113

1.3 ( 0.5, 2.1) 0.001

1.7 ( 0.3, 3.2) 0.018

ß(2)-ß(1)

-3.1 ( -6.5, 0.2) 0.066

1.4 ( 0.1, 2.7) 0.032

1.7 ( 0.1, 3.3) 0.038

LLR test

0.063

0.030

0.037

ß (95%CI) p-value / OR/RR (95%CI) p-value
Outcome: SBP and DBP
Risk factor: Body mass index
Model each outcome versus each risk factor respectively
Adjust for: AGE, Education, Occupation, Cigarette smoking and Alcohol consumption
Stratified by: SEX
Total group analysis also adjusted for SEX

 

Interpretation of the table:

A threshold of BMI=18.6 for male and BMI=22 for female were detected.  The effect of BMI on SBP and DBP is different at BMI less than turning point (18.6 for male, 22.0 for female) comparing to BMI greater the turning point.

 

Done!