Table 2 Results of the two-piecewise linear-regression model of mean value of HbA1c and outcomes

Primary Outcome

K-means Variability Group

Low Variability

Medium Variability

High Variability

Total

Mean Value of HbA1c

P for interaction: 0.05

One linear-regression model

1.33 (1.14, 1.56) P < 0.01

1.14 (0.99, 1.30) P = 0.06

1.06 (0.90, 1.23) P = 0.49

1.16 (1.07, 1.27) P < 0.01

 Two-piecewise linear-regression model

  Inflection point (K and 95%CI) (%)

6.74 (6.57, 6.95)

7.49 (7.34, 7.65)

7.88 (7.68, 8.09)

8.82 (8.17, 8.87)

  <K Effect size β (95%CI)

1.12 (0.70, 1.79) P = 0.63

0.88 (0.67, 1.15) P = 0.36

0.80 (0.54, 1.18) P = 0.27

1.18 (1.07, 1.31) P < 0.01

  >K Effect size β (95%CI)

1.42 (1.13, 1.78) P < 0.01

1.38 (1.11, 1.71) P < 0.01

1.19 (0.96, 1.47) P = 0.11

1.08 (0.79, 1.47) P = 0.61

  P for Log-likelihood ratio test

0.46

0.04

0.15

0.60

All-cause Mortality

K-means Variability Group

Low Variability

Medium Variability

High Variability

Total

Mean Value of HbA1c

P for interaction: 0.02

One linear-regression model

1.23 (1.01, 1.51) P = 0.04

0.92 (0.76, 1.11) P = 0.38

0.86 (0.72, 1.03) P = 0.10

0.98 (0.88, 1.09) P = 0.65

 Two-piecewise linear-regression model

  Inflection point (K and 95%CI) (%)

5.99 (5.97, 6.21)

7.44 (7.30, 7.59)

7.86 (7.66, 8.06)

6.11 (6.10, 6.43)

  <K Effect size β (95%CI)

None

0.49 (0.36, 0.69) P < 0.01

0.63 (0.42, 0.95) P < 0.01

None

  >K Effect size β (95%CI)

1.19 (0.96, 1.47) P = 0.11

1.57 (1.19, 2.08) P < 0.01

1.01 (0.78, 1.31) P = 0.11

0.95 (0.85, 1.07) P = 0.14

  P for Log-likelihood ratio test

0.29

< 0.01

0.12

0.10