Table A1.
Comparison of epidemiological approaches to determine relationships between vitamin D and health outcomes.
| Approach | Method | Strengths | Weaknesses |
|---|---|---|---|
| Ecological, geographical |
Compare health outcomes with indices of solar UVB doses and other risk-modifying factors | Can include large numbers of participants | Subject to confounding factors; indices used may not apply to those with health outcomes |
| Observational, prospective |
Enroll participants, draw blood, obtain information, follow for a long period. | 25(OH)D has inputs from solar UVB, diet, and supplements. | 25(OH)D changes with time including season so effect decreases as follow-up time increases; control of confounding may not be complete. |
| Observational, case-control |
Measure 25(OH)D near time of diagnosis, match with controls. | Appropriate when health outcome is affected by recent 25(OH)D or 25(OH)D changes little over time. | Disease status may affect 25(OH)D; control choice may be biased. |
| Observational, cross-sectional |
Measure 25(OH)D and health status of a representative number of a population | Many health outcomes can be studied. | 25(OH)D may not be similar to that prior to health outcome; health status may affect 25(OH)D. |
| Non-randomized vitamin D supplementation | Enroll participants, measure parameters, instruct. | High vitamin D doses can be used; | Confounding by self-selected use of supplements |
| RCT | Enroll participants, randomize to vitamin D or placebo supplementation, follow. | Effects found are likely due to vitamin D. | False negatives are possible because enrollees often have high 25(OH)D or may be given low vitamin D doses; compliance issues; other sources of vitamin D occur. Initiation may be after onset of disease and duration may be too short to have an effect. |
| MR | Measure alleles of genes that affect 25(OH)D levels. | Independent of many factors including actual 25(OH)D. | The alleles may not reflection biological activity, and confounding is still possible. Can be misleading when relationship of vitamin D is nonlinear, and statistical power can be low. |
MR—Mendelian randomization; RCT—Randomized controlled trial.