USA 1992.

Methods	Consecutive numbers were obtained from a random number table. No losses to follow‐up were reported, although 10 women did not complete the questionnaire on side effects. O'Sullivan 1964 criteria were used to diagnose GDM.
Participants	76 Native Alaskan women at 24 to 28 weeks' gestation without a history of diabetes. Settings: Ketchikan Native Health Clinic and Mt. Edgecumbe Hospital, Alaska, USA.
Interventions	Experimental: 35 women; 50 g glucose polymer drink. Control: 41 women; 50 g glucose monomer drink.
Outcomes	Women: mean 1 hour venous plasma glucose, side effects. Babies: none reported.
Notes	Study dates: January 1988 to May 1990. Funding sources: grant from the Diabetes Research and Education Foundation, Bridgewater, NJ; declarations of interest: not reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Consecutive numbers from a random number table.
Allocation concealment (selection bias)	Unclear risk	Allocation concealment was not reported, although study was described as "double blind".
Incomplete outcome data (attrition bias) All outcomes	Low risk	No losses to follow‐up were reported, although 10 women did not complete the questionnaire on side effects.
Selective reporting (reporting bias)	Low risk	Specified outcomes of side effects and plasma glucose values were reported.
Other bias	Unclear risk	No table of participants' baseline characteristics was provided to permit judgement of the effectiveness of randomisation.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Study was described as double blind, which is possible as test drinks may have appeared identical. Outcomes of side effects could have been affected by knowledge of treatment groups, but the plasma glucose value would not have been affected.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Study was described as double blind. Side effects were self‐reported on a questionnaire and thus were not subject to assessor bias. Plasma glucose values would not have been affected by lack of blinding on the part of assessors.