Table 2 Summary of the stably selected variables and their confirmation.

From: Application of regularized regression to identify novel predictors of mortality in a cohort of hemodialysis patients

 

Derivation

Confirmation

n

Stable selection

Sign of EN coefficient

All-cause

Cardiovascular

Infection-associated

AC

CV

INF

AC

CV

INF

HR

CI (95%)

p-value

HR

CI (95%)

p-value

HR

CI (95%)

p-value

Age (per 10 years)

x

x

x

 + 

 + 

 + 

1.64

(1.31–2.06)

 < 0.001*

1.86

(1.23–2.81)

0.003*

1.75

(1.12–2.73)

0.014*

158

Atherosclerosis (non CHD)

x

 

x

 + 

 

 + 

2.14

(1.26–3.61)

0.005*

   

2.70

(1.01–7.21)

0.048*

47

Atrial fibrillation

x

  

 + 

  

2.67

(1.58–4.52)

 < 0.001*

      

42

CVD

 

x

  

 + 

    

2.35

(0.97–5.71)

0.059#

   

62

Dialysis catheter

x

x

x

 + 

 + 

 + 

3.72

(1.59–8.72)

0.002*

3.47

(0.80–15.12)

0.098#

4.09

(0.92–18.12)

0.064#

8

DM

x

x

 

 + 

 + 

 

2.00

(1.19–3.37)

0.009*

1.38

(0.57–3.35)

0.470

   

61

H.o. amputation due to PAD

x

x

x

 + 

 + 

 + 

1.57

(0.67–3.67)

0.296

2.43

(0.71–8.35)

0.158

0.90

(0.12–6.81)

0.918

10

H.o. MI

 

x

  

 + 

    

2.49

(0.97–6.40)

0.057#

   

37

H.o. neoplasia

 

x

  

 + 

    

2.22

(0.90–5.46)

0.082#

   

42

HF

x

  

 + 

  

2.57

(1.49–4.41)

0.001*

      

35

HRD (non AF)

  

x

  

 + 

      

2.07

(0.59–7.31)

0.256

19

Oral anticoagulation

x

  

 + 

  

3.10

(1.71–5.64)

 < 0.001*

      

22

Other ac (dialysis)

  

x

  

 + 

      

0.00

(0.00–Inf)

0.998

7

PAD

x

 

x

 + 

 

 + 

2.13

(1.23–3.69)

0.007*

   

2.56

(0.93–7.05)

0.069#

31

Pulmonary hypertension

 

x

  

 + 

    

3.58

(1.19–10.81)

0.023*

   

13

DBP (24 h) low

  

x

  

 + 

      

3.41

(1.18–9.80)

0.023*

20

MAP (24 h) low

  

x

  

 + 

      

2.22

(0.71–6.87)

0.169

23

Immunosuppression

 

x

  

 + 

    

0.47

(0.06–3.48)

0.457

   

17

Kt/V high

  

x

  

−

      

2.85

(0.98–8.26)

0.054

25

HB low

  

x

  

 + 

      

3.26

(1.18–9.03)

0.023*

29

Serum AST low

x

  

−

  

0.31

(0.10–0.98)

0.046*

      

23

Serum cholesterol high

 

x

  

 + 

    

0.22

(0.03–1.67)

0.144

   

24

Serum creatinine high

x

  

−

  

0.44

(0.18–1.11)

0.083#

      

30

Serum creatinine low

x

x

 

 + 

 + 

 

1.34

(0.77–2.33)

0.295

1.77

(0.72–4.35)

0.210

   

44

Serum hsCRP high

x

  

 + 

  

1.82

(1.00–3.28)

0.048*

      

31

Serum IFN-gamma high

  

x

  

 + 

      

0.95

(0.27–3.33)

0.934

29

Serum IL-12p70 high

  

x

  

 + 

      

3.12

(1.16–8.41)

0.024*

33

Serum IL-13 low

  

x

  

 + 

      

0.26

(0.03–1.95)

0.190

32

Serum IL-6 high

x

x

x

 + 

 + 

 + 

2.96

(1.73–5.07)

 < 0.001*

1.98

(0.75–5.20)

0.165

4.49

(1.67–12.05)

0.003*

36

Serum IP-10 high

 

x

  

 + 

    

1.19

(0.45–3.10)

0.726

   

38

Serum IP-10 low

  

x

  

−

      

0.57

(0.13–2.53)

0.462

30

Serum iron low

  

x

  

 + 

      

1.94

(0.62–6.05)

0.251

26

Serum LDL high

 

x

  

 + 

    

0.49

(0.11–2.13)

0.343

   

24

Serum phosphate low

 

x

  

 + 

    

1.28

(0.49–3.33)

0.615

   

36

Serum triglycerides low

x

x

 

 + 

 + 

 

2.30

(1.16–4.56)

0.018*

1.20

(0.28–5.19)

0.809

   

16

Serum YKL-40 high

x

  

 + 

  

2.51

(1.44–4.35)

0.001*

      

32

  1. For ordinal variables "high" and "low" indicates that the value falls within the top or bottom quintile of the total study population (see “Methods”).
  2. ac, anticoagulation; AC, all-cause mortality; AF, atrial fibrillation; AST, Aspartate transaminase; CHD, coronary heart disease; CI(95%), 95% confidence interval of the hazard ratio; CV, cardiovascular mortality; CVD, cardiovascular disease; D/T, due to; DBT (24 h), average (24 h) diastolic blood pressure; DM, Diabetes mellitus; EN, elastic net; ESRD, end-stage renal disease; H.o., history of; HB, hemoglobin; HF, heart failure; HR, hazard ratio; HRD, heart rhythm disorder; INF, infection-associated mortality; MAP (24 h), average (24 h) mean arterial pressure; MI, myocardial infarction; n, number of non-zero observations of the respective predictor in the confirmation group; PAD, peripheral artery disease.
  3. *p < 0.05; #, two-sided p < 0.1 and same effect direction as in the elastic-net regression derivation dataset for respective all-cause or cause-specific mortality. Effect direction in the derivation dataset is indicated as sign of elastic net coefficient: + , regularized HR > 1; −, regularized HR < 1. Statistics were calculated in the confirmation dataset using univariate Cox regression. Tests were only performed in the confirmation dataset for predictors which passed stability selection for the respective outcome in the derivation dataset.
Back to article page