Hill 1991

Methods	Location: Seattle, USA. Number of centres: One. Funding source: National Cancer Instiute grant (CA38552) and National Institute on Drug Abuse grant (DA 05513). Abbott Laboratories supplied the PCA infusers and drug supplies. Recruitment period: Not stated. Randomised but not blind. Eligible patients had severe oral mucositis pain. Clear information on withdrawals. 35% lost to follow‐up.
Participants	Inclusion: Adults with acute lymphocytic leukemia, acute non‐lymphocytic leukemia, chronic myelogenous leukemia, Hogkin's and non‐Hodgkin's lymphoma or preleukemic syndrome planned to undergo bone marrow transplant. 54 eligible, 54 enrolled, 35 completed.
Interventions	Gr A (n = 31) PCA Morphine bolus 1 to 2 mg morphine sulphate as bolus, with 10 min lockout. Gr B (n = 23) PKPCA (pharmacokinetically based patient controlled analgesic infusion system). Individual pharmacokinetic profiles for morphine were determined on the day prior to bone marrow transplant. 75 μg/kg dose was administered and venous blood samples taken over following 8 hours to determine PK profile, and this information was used to tailor the individual dose for the PKPCA intervention.
Outcomes	Average pain VAS score (0 to 100) for each study day, day 7 used. Mean opiate use mg/hour for day 7. Severity of mucositis assessed by dental hygienist.
Notes	Data obtained from graphs. No power calculation reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Quote: "randomly assigned". Author clarified this.
Allocation concealment?	Low risk	No information given but clarified by correspondence with author.
Blinding of participants/carers?	High risk	Comment: Unblinded patients self assessed pain; nurse assessment of amount morphine, hygienist completed mucositis ratings unclear if blinded.
Blinding of outcome assessors?	Unclear risk	Comment: Unblinded patients self assessed pain; nurse assessment of amount morphine, hygienist completed mucositis ratings unclear if blinded.
Incomplete outcome data addressed?	Low risk	19/54 (35%) of those randomised were excluded, 11/31 (35%) from Gr A and 8/23 (35%) from Gr B. Reasons for each withdrawal given in table 2.
Free of selective reporting?	Low risk	Planned outcomes described and reported.
Free of other bias?	Unclear risk	Unclear.