Skip to main content

Table 3 Contribution of MBDA-based CVD risk score and other models to prediction of 3-year CVD risk

From: Derivation and internal validation of a multi-biomarker-based cardiovascular disease risk prediction score for rheumatoid arthritis patients

CVD risk prediction score

Single-score analyses (univariable)

Sequential analyses of paired scores (bivariable)

A. Non-MBDA-based CVD Risk Score added to a base model of MBDA-based CVD Risk Score

B. MBDA-based CVD Risk Score added to a non-MBDA-based CVD Risk Score base model

HR (95% CI)

LRT statistic

p value

HR (95% CI)

Increment in LRT statistic

p value

HR (95% CI)

Increment in LRT statistic

p value

Age + sex

3.44 (2.62–4.53)

82.2

1.22 × 10−19

1.34 (0.96–1.86)

3.0

0.084

2.62 (2.14–3.20)

83.3

7.15 × 10−20

Age + sex + CRP

2.97 (2.38–3.70)

95.8

1.29 × 10−22

1.13 (0.84–1.54)

0.7

0.412

2.71 (2.15–3.41)

67.4

2.20 × 10− 16

Clinical

3.08 (2.50–3.80)

115.1

7.34 × 10−27

1.24 (0.89–1.71)

1.7

0.197

2.56 (1.98–3.29)

49.0

2.51 × 10−12

Clinical + CRP

2.94 (2.43–3.55)

127.5

1.44 × 10−29

1.12 (0.79–1.60)

0.4

0.526

2.67 (1.96–3.63)

35.4

2.67 × 10−9

MBDA-based CVD risk score

2.89 (2.46–3.41)

162.5

4.67 × 10−38

–

–

–

–

–

–

  1. In single-score (univariable) analyses, each of the five risk scores derived from training was analyzed as a single independent variable for predicting risk for a CVD event in the validation dataset (N = 10,275). In sequential (bivariable) analyses of paired scores, the risk scores of the MBDA-based model and each non-MBDA-based risk model were evaluated as the only two variables used to predict CVD risk: (A) with the MBDA-based risk score as the base model and (B) with the non-MBDA-based CVD risk score as the base model. The increment in LRT statistic represents the extent to which a second CVD risk score adds to the CVD risk prediction ability of a first CVD risk score (i.e., the base model). Non-MBDA-based CVD risk score refers to the first four scores in the first column. P values by the likelihood ratio test. Statistically significant HR values are bolded. Clinical model includes age + sex + tobacco use + diabetes + hypertension + history of CVD (excluding MI and stroke)
  2. CRP C-reactive protein, CVD cardiovascular disease, HR hazard ratio, LRT likelihood ratio test, MBDA multi-biomarker disease activity