Criteria | |

Design | Assessment |

1. Is the subgroup variable a characteristic measured at baseline rather than after randomisation? | Yes, measured at baseline. |

2. Is the effect suggested by comparisons within rather than between studies? | Yes, within study comparison. |

3. Was the hypothesis specified a priori? | Yes, predefined before unblinding and data analysis. |

4. Was the direction of the subgroup effect specified a priori | Yes, authors stated that they expected a ‘greater effect with more severe deficiency’. |

5. Was the subgroup effect one of a small number of hypothesised effects tested? | Yes, only one tested. |

Analysis | |

6. Does the interaction test suggest a low likelihood that chance explains the apparent subgroup effect? | Yes, p value for interaction test=0.04. |

7. Is the significant subgroup effect independent? | Yes, only one interaction tested. |

Context | |

8. Is the size of the subgroup effect large? | Yes, in patients with severe deficiency (HR 0.56; 95% CI 0.35 to 0.90) vs those with less severe deficiency (HR 1.12; 95% CI 0.72 to 1.77). |

9. Is the interaction consistent across studies? | No, not confirmed in at least one other larger RCT (n=1078) in an ICU population with higher risk conditions; there are ongoing studies.38 |

10. Is the interaction consistent across closely-related outcomes within the study? | No, while close (p=0.10, 0.06, 0.12), no statistically significant interactions for all other mortality outcomes (ICU, 28-day and 6-month mortality, respectively). |

11. Is there indirect evidence that supports the hypothesised interaction (biological rationale)? | Yes, based on mechanistic data39 and meta-analysis of observational studies, vitamin D deficiency is associated with increased hospital mortality in critically ill patients.40 |

ICU, intensive care unit; RCT, randomised controlled trial.