Kalnins 1996.

Methods	Quasi‐randomised study. Duration: 3 months. Parallel design. Single centre. Location: Canada.
Participants	Adolescents and adults (over 10 years of age) with CF. Most recent published report states 13 completed the trial (2 drop‐outs). Mean (SD) age: treatment group 19.5 (11.3) years; control group 16.4 (6.7) years. Gender split: 3 males, 10 females. Less than 90% ideal WFH or 5% reduction in ideal WFH over 3 months. All participants pancreatic insufficient, except 1 in treatment group.
Interventions	Treatment: high‐calorie drink to increase energy intake by 20% of predicted energy needs. Control: nutritional counselling to increase energy intake by 20% of predicted energy needs by eating high calorie foods.
Outcomes	Z scores for weight* and height*, WFH* Anthropometric measures* Pulmonary function (FEV1 % predicted)* Energy* and nutrient* intake Faecal balance studies.
Notes	Information from lead author: A modified randomization process was used: 1. males and females were segregated in 2 groups 2. within each gender, participants were separated into groups by age: 10 ‐ 14 years, 15 ‐ 18 years and > 18 years 3. if a participant within his/her age group were to select a sealed envelope card containing supplement, as the arm, then the next patient of the same group would be automatically assigned to diet counselling. If there was a drop out, then the next participant of that gender and age would replace the drop out in that group. The above was done because of the relatively small number of participants in each group.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Paper states "randomized" but gives no further detail on method of randomisation. Personal communication confirmed that participants split into groups according to age and gender and first participant in each group chose a sealed envelope that contained a card allocating to either supplement or dietary advice group.
Allocation concealment (selection bias)	High risk	Unclear from paper. However communication with lead author confirmed that allocation was concealed from initial participant as the cards were in sealed envelopes ‐ low risk of bias. But other participants were allocated alternately ‐ high risk of bias as clinician can foresee which treatment next participant will receive.
Blinding (performance bias and detection bias) All outcomes	Low risk	Participants could not be blinded as treatment options were counselling or dietary supplements. No discussion in paper of whether outcome assessors were blinded to participants' treatment group. However, lead author confirmed that different physicians were used, but she was not sure if patients had discussed their treatment arm during clinic visits therefore we judge that blinding was not present.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Two participants dropped out (one from each group) after baseline measurements. Reasons given (feeling unwell, change of mind).
Selective reporting (reporting bias)	Unclear risk	Paper appears to address all the outcomes measured in the 'Resuts' section. No access to study protocol to double check, however the lead author has confirmed that all planned outcome measures were reported.