Table 2. Summary of case-control study quality assessment system.

	Risk of Bias (score)
Criterion	High (−1)	Medium (0)	Low (+1)
1. Case definition explicit and appropriate?	Definition and/or incl/excl criteria not given, ambiguous, or clearly unsuitable	Basic definition given; enough to satisfy that chosen cases (and the criteria used to select them) are suitable	Detailed definition and explanation; all suitable cases included
2. How is maternal vitamin D status measured?	Dietary intake only or insufficient information	Blood levels of 25(OH)-vitamin D	Blood levels of circulating 25(OH)-vitamin D, with details of precision, pick up of D2 and D3 and assay used
3. Participants grouped according to Vitamin D status?	Subjects divided and analysed in groups based on pre-existing vitamin D thresholds	Subjects divided and analysed in groups according to Vitamin D level based on group characteristics	Subjects not divided into groups according to Vitamin D level/ or grouped according to at threshold generated from the study
4. Measurements of outcomes reliably ascertained?	Inadequately explained or obviously unsuitable	Adequate description and reliability/suitability of at least one of the following: instruments, technique/ definition/protocol, people, place	Detailed description and reliability of one and at least adequate description of the others
5. Measurements of later outcomes objective?	Subjective measure, eg bone or muscle pain, wheezing	Ascertained from researcher examination	Objective measure e.g. DXA, bone biopsy, lung function tests
6. Control selection appropriate?	No information at all, ambiguous, or not selected from population of cases or otherwise clearly inappropriate to the study objectives	Selection is from population of cases, and is basically appropriate and similar to cases for all factors other than the outcome of interest, but not optimally, or with incomplete information	Selection is from population of cases in a manner wholly appropriate to the study objectives, and in such a way as to make them as similar as possible to cases in all respects except the outcome of interest
7. Measures of vitamin D intake/ 25(OH)-vitamin D level, bone outcomes rounded?	Categorisation or very rough rounding, or if any clear evidence of rounding exists without explanation in the text	Measures are rounded, but not by much	No information given, and no obvious reason to suspect rounding has occurred. Or: explicitly stated that measurements were not rounded.
8. Setting and population appropriate?	Ambiguously described, obviously bias inducing or unsuitable for the objectives and stated conclusions	Possibly restricting but reflected in the scope of the objectives and the stated conclusions	Planned to minimise bias and allow generalisability beyond the immediate scope of the objectives
9. Outcome assessment blind to vitamin D status?	N/A	No details given	Some details or statement given
10. Analysis rigorous and appropriate?	No statistical analyses carried out (just tables or description), or analysis badly carried out	Tables of means and differences given with statistical tests (e.g. t-tests), or some regression but without clear/valid measure of association	Regression (or similar technique) is used which gives a valid measure of association (e.g. odds ratios, hazard ratios, relative risks)
11. Response rates for: e.cases f.controls (a separate score for each should be given)	Low (<70%)	Medium (70-90%) or not given	High (>90%)
12. Info on representativeness and non-participants	Cases obviously unrepresentative of wider population alluded to in text	Some information on cases and controls lost or excluded, or no information but with no reason to suspect a detrimental lack of representativeness	Detailed information on cases and controls lost or excluded, with numbers and reasons.
13. Sample sizes for: e.cases f.controls (a separate score for each should be given)	Extremely ambiguous, not given, or small (under 100)	Average (100 to 1000)	Large (over 1000)
14. Adequate consideration of important confounding factors? (e.g. season, sunlight exposure, calcium intake, maternal compliance, infant feeding)	One factor matched on or controlled for in tables; nothing for the others (NB whether they were measured or not is irrelevant)	Most factors matched on or controlled for in tables, or fewer if one or more is adjusted for in regression	Most factors adjusted for in regression