Table 3.
Risk factor | Univariate analysis | Multivariate analysis | Multivariate excluding mediators* | ||||||
---|---|---|---|---|---|---|---|---|---|
HR | 95% CI | P | HR | 95% CI | P | HR | 95% CI | P | |
Iron deficiency | |||||||||
Age, y | 0.98 | 0.96-0.99 | .005† | 0.99 | 0.92-1.06 | .735 | 0.98‡ | 0.97-1.00 | .028† |
Female sex | 2.85 | 1.58-5.13 | <.001† | 0.80§ | 0.01-44.3 | .914 | 2.30§ | 1.27-4.17 | .006† |
Malabsorptive procedure | 2.46 | 1.59-3.80 | <.001† | 1.92 | 1.20-3.06 | .006† | 2.36 | 1.50-3.71 | <.001† |
Baseline anemia | 2.94¶ | 1.88-4.60 | <.001† | 1.49¶ | 0.92-2.44 | .107 | — | — | — |
Baseline ferritin, μg/L | 0.96¶ | 0.95-0.97 | <.001† | 0.96¶ | 0.95-0.97 | <.001† | — | — | — |
Weight loss at 6 mo, %TBW | 1.01 | 0.99-1.03 | .480 | 0.98 | 0.96-1.00 | .066 | 0.99 | 0.97-1.01 | .492 |
Iron deficiency anemia | |||||||||
Age, y | 0.98 | 0.96-1.00 | .072 | 0.90 | 0.82-0.99 | .028† | 0.98 | 0.96-1.00 | .088 |
Female sex | 1.64 | 0.70-3.81 | .252 | 0.02§ | 0.00-2.62 | .114 | 1.48§ | 0.62-3.52 | .376 |
Malabsorptive procedure | 1.01 | 0.58-1.77 | .973 | 0.84 | 0.45-1.59 | .602 | 1.01 | 0.55-1.85 | .969 |
Baseline anemia | 24.6¶ | 11.0-55.1 | <.001† | 19.6¶ | 7.85-48.9 | <.001† | — | — | — |
Baseline ferritin, μg/L | 0.96¶ | 0.94-0.97 | <.001† | 0.96¶ | 0.95-0.98 | <.001† | — | — | — |
Weight loss at 6 mo, %TBW | 0.99 | 0.95-1.02 | .399 | 0.98 | 0.94-1.02 | .335 | 0.98 | 0.95-1.02 | .278 |
Administration of IV iron|| | |||||||||
Age, y | 0.98 | 0.94-1.01 | .203 | — | — | — | 0.98 | 0.94-1.02 | .262 |
Female sex | 2.45 | 0.33-18.3 | .382 | — | — | — | 2.12§ | 0.28-16.1 | .469 |
Malabsorptive procedure | 1.62 | 0.60-4.36 | .342 | — | — | — | — | — | — |
Baseline anemia | 1.39 | 0.47-4.14 | .550 | 1.21 | 0.40-3.63 | .736 | — | — | — |
Baseline ferritin, μg/L | 0.99 | 0.98-1.00 | .071 | 0.99 | 0.98-1.00 | .080 | — | — | — |
Weight loss at 6 mo, %TBW | 0.98 | 0.92-1.04 | .450 | — | — | — | — | — | — |
—, not included in the analysis.
This multivariate analysis excludes baseline anemia and baseline ferritin.
P < .05, statistically significant.
The effect of age was not statistically significant with the inclusion of an interaction term for age and sex.
There was no significant interaction between age and sex.
Time-dependent covariates were used because proportional hazards assumption was not satisfied in univariate analysis. Time measured in months.
The multivariate analysis included only 2 covariates to prevent overfit.