Griffiths 1998.

Study characteristics
Methods	Randomised, triple‐blind, cross‐over study. The study was described as being triple‐blind as neither the children, their parents nor the researchers knew which intervention the children received. Duration 6 months.
Participants	16 participants with PKU (10 males, 6 females, mean age 12.6 years). All diagnosed by neonatal period and were commenced on a low‐phenylalanine diet within 3 weeks of birth. Diet continued from diagnosis until study entry.
Interventions	Continuation of the low phenylalanine diet in addition to a phenylalanine‐free amino acid supplement for 3 months followed by a low phenylalanine diet plus an amino acid supplement containing phenylalanine for 3 months, or vice versa.
Outcomes	Blood phenylalanine levels measured at baseline and at the end of the first and second 3‐month period. The following neuropsychological tests were carried out at the same time points: matching familiar figures; Rey verbal learning; digits forwards; paired‐associate learning; Corsi block‐tapping; Rey‐Davis manual labyrinth; Purdue pegboard and hole‐type steadiness tester.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Method of randomisation not given.
Allocation concealment (selection bias)	Low risk	Usual formula and experimental product prepared and coded by pharmaceutical company. Key to the code kept by a senior member of the hospital medical staff.
Blinding (performance bias and detection bias) Participants	Low risk	Triple‐blinded, neither the children, their parents nor the researchers knew which intervention the children received. Participants continued on their usual diet but with the phenylalanine content of the diet manipulated by using additional supplements.
Blinding (performance bias and detection bias) Clinicians	Low risk	Triple‐blinded, neither the children, their parents nor the researchers knew which intervention the children received.
Blinding (performance bias and detection bias) Outcome assessors	Low risk	Triple‐blinded, neither the children, their parents nor the researchers knew which intervention the children received.
Incomplete outcome data (attrition bias) All outcomes	Low risk	An intention‐to‐treat analysis was employed.