Table 3—

Multivariate RRs of type 2 diabetes according to quintiles of magnesium intake stratified by major risk factors*

VariableNo casesQuintiles of magnesium intake
P for trend
2345 (high)
Women (NHS)
 BMI
  >27 kg/m23,1470.900.810.740.71 (0.62–0.81)<0.001
  ≤27 kg/m29300.891.080.760.72 (0.56–0.92)0.004
 Physical activity
  High1,3950.830.740.700.64 (0.52–0.78)<0.001
  Low2,0880.950.920.730.78 (0.66–0.92)<0.001
 Family history of diabetes
  Yes1,8130.920.950.790.77 (0.64–0.92)0.001
  No2,2720.910.830.740.69 (0.59–0.81)<0.001
Men (HPFS)
 BMI
  >27 kg/m28370.870.850.750.75 (0.57–0.98)0.018
  ≤27 kg/m24860.791.010.560.71 (0.50–0.99)0.022
 Physical activity
  High4450.730.850.650.68 (0.47–0.99)0.063
  Low7100.890.860.620.71 (0.53–0.94)0.002
 Family history of diabetes
  Yes4720.730.670.680.62 (0.43–0.88)0.013
  No8560.891.040.640.78 (0.60–1.01)0.012
  • *

    * The multivariate RRs are adjusting by all dietary and nondietary variables included in the most adjusted model in Table 2 (see footnote for the explanation of the adjustment). The variables used for stratification were not included in the model. The first quintile was used as the referent in all models;

  • used median values as cutoff points;

  • the number of cases do not match with Table 2 because of the presence of missing values removed in the stratified models.