Skip to main content

Table 3 Results of the multivariate linear regression model obtained using twelve predictors of the increase of BCVA from baseline as independent variables

From: Evaluation of markers of outcome in real-world treatment of diabetic macular edema

 

Unstandardized Coefficients

Standardized Coefficients

  

95% Confidence Interval for B

Model

B

SE

Beta

t

p

Lower Bound

Upper Bound

Constant

38.773

4.219

 

9.190

0.000

30.414

47.131

BCVAi (L)

− 0.386

0.056

− 0.611

−6.862

0.000

−0.497

− 0.274

Baseline EZ

4.291

1.407

0.257

3.049

0.003

1.502

7.079

Laser

−2.457

1.212

−0.155

−2.027

0.045

−4.858

−0.055

HbA1c_bin

−4.188

1.350

−0.232

−3.102

0.002

−6.862

−1.513

DM_bin

−0.960

1.193

−0.061

−0.805

0.423

−3.323

1.403

RNZ

−1.523

2.469

−0.062

−0.617

0.538

−6.415

3.368

AFL

−1.588

1.811

−0.086

−0.877

0.382

−5.176

1.999

  1. Abbreviations: BCVAi (L)=baseline best corrected visual acuity in ETDRS letters; EZ=ellipsoid zone; HbA1c_bin=glycated hemoglobin level entered as a dichotomous variable (≤7 versus > 7); DM_bin=diabetes duration entered as a dichotomous variable (≤15 years versus >15 years); RNZ=ranibizumab; AFL=aflibercept; B=regression coefficient; SE=standard error for B
  2. aBaseline BCVA: for each unit of increase in the baseline BCVA, there is an average decrease of 0.386 letters in the dependent variable (increase of BCVA after 6 months)
  3. bBaseline EZ: eyes with intact EZ have an average increase of 4.291 letters in the BCVA after 6 months;
  4. cLaser: eyes with history of macular photocoagulation have an average decrease of 2.457 letters in the dependent variable (increase of BCVA after 6 months);
  5. dHbA1c: eyes with HbA1c > 7 have an average decrease of 4.188 letters in the dependent variable. In this model, diabetes duration and the use of ranibizumab or aflibercept were not statistically significant. The regression model obtained was statistically significant (F(4,116) = 14.791, p < 0.001) and the variables explained about 32% of the variance (R_adj = 0.315). The assumptions of the model regarding residuals were observed as well as collinearity