Olsen 2001.

Study characteristics
Methods	This was a randomised, parallel‐group trial, which lasted 6 months.
Participants	The study recruited 71 participants (35 in the low‐dose group and 36 in the high‐dose group) with histologically‐proven mycosis fungoides type with ≥ 20% of lymphocytes within the skin biopsy stain positively for CD25 by immunohistochemistry. Further inclusion criteria was as follows: stage Ib‐III CTCL (CTCL Cooperative Group staging) recurred or persisted after ≥ 4 previous treatments for CTCL (excluding topical or systemic corticosteroids) or stage IVa CTCL participants who failed at least 1 previous therapy study consideration, and Eastern Cooperative Oncology Group Performance Status of 0, 1, or 2. Lymph node involvement was no greater than LN₂, and no CTCL involvement of bone marrow. Demographics of the included participants 37 men and 34 women Mean age (range) = 61 years (26 to 90 years) Stages of disease: IB:16, IIA: 10, IIB: 19, III: 11 58% of the participants were lost to follow‐up in total, resulting in 30 participants evaluated Exclusion criteria of the trial Age < 18 years Pregnancy/lactation Disagreement to practice contraception High‐grade large‐cell, poorly‐differentiated tumours, or both Positive test for HIV, HTCLV‐1, or hepatitis B or C Uncontrolled hypertension Any signs of active systemic infection Previous treatment with IL‐2 fusion proteins
Interventions	The intervention group was given 9 µg/kg/day denileukin diftitox intravenous infusion over 15 to 60 minutes for 5 consecutive days every 3 weeks. The control group was given 18 µg/kg/day denileukin diftitox intravenous infusion over 15 to 60 minutes for 5 consecutive days every 3 weeks.
Outcomes	Outcomes of the trial Quality of life measured by the Functional Assessment of Cancer Therapy‐general (FACT‐G) questionnaire Common adverse effects of the treatments Percentage of participants demonstrating complete response Overall survival (assessed 90 days after the end of the intervention) Objective response rate Rare adverse effects
Notes	Only 42% of all randomised participants received 8 courses of treatment as planned. There was no comparison reported between both treatment groups regarding QoL from baseline to the end of the study (only subgroup analyses of responders vs non‐responders). The funding body was Seragen, Inc (a wholly‐owned subsidiary of Ligand Pharmaceuticals Inc, San Diego, CA). Two of the authors had equity interests in Seragen. This study was conducted in 20 centres across the USA.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was stratified by stage of CTCL for multicentre trial. It was likely to have been carried out by a third party and concealed, although this was not formally stated.
Allocation concealment (selection bias)	Unclear risk	Randomisation was stratified by stage of CTCL for multicentre trial. However, it was unclear whether randomisation was concealed.
Blinding of participants and personnel (performance bias) All outcomes	High risk	This was unlikely to be blinded, since the drug was diluted to a certain minimum concentration in both arms and administered by a pump device for 15 to 60 minutes.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "All responses were verified by an independent panel of physicians [the Data End Point Review Committee]." Comment: outcome assessment was blinded
Incomplete outcome data (attrition bias) All outcomes	High risk	41/71 (58%) participants dropped out. Discontinuation was due to adverse events (11/35 (31%) participants in the 9 µg/kg/day group vs 15/36 (42%) in the 18 µg/kg/day group) and treatment failure (6/35 (17%) in the 9 µg/kg/day group vs 2/36 (6%) in the 18 µg/kg/day group).
Selective reporting (reporting bias)	High risk	The quality of life assessment was compared between responders and non‐responders instead of comparing treatment groups. We contacted the corresponding author for additional outcome data, but we received no reply within 4 weeks.
Other bias	Unclear risk	There were insufficient information to permit judgement.