Sarah Barton1, Clare Peckitt2, Francesco Sclafani1, David Cunningham1, Ian Chau3. 1. Department of Medicine, Royal Marsden Hospital, London & Surrey, UK. 2. Department of Clinical Research and Development, Royal Marsden Hospital, Surrey, UK. 3. Department of Medicine, Royal Marsden Hospital, London & Surrey, UK. Electronic address: ian.chau@rmh.nhs.uk.
Abstract
PURPOSE: Correct interpretation of subgroup analyses (SGA) is important as it influences selection of therapeutic interventions for patient subsets. The primary aim of our study was to compare reporting of SGA between industry and non-industry sponsored trials. METHODS: We performed a systematic literature review and extracted data from journal articles (JA) and conference abstracts (CA) published over a decade reporting SGA results of phase III randomised controlled gastrointestinal (GI) oncology trials with patient participants of ≥150. RESULTS: In JA, SGA was reported in 100/145 (69%) trials: 41/54 industry sponsored (76%; 95% confidence interval [CI]: 63-86%) and 59/91 non-industry sponsored (65%; 95% CI: 55-74%) trials (p = 0.16). In CA, SGA was reported in 86/204 (42%) trials: 43/83 industry sponsored (52%; 95% CI: 41-62%) and 43/121 non-industry sponsored (36%; 95% CI: 28-44%) trials (p = 0.02). Number of SGA performed per trial was significantly larger for industry compared to non-industry sponsored trials in both JA (median 6 versus 2, p = 0.003) and CA (median 1 versus 0, p = 0.023). Claims of subgroup effect were made in 52% of trials in JA and 50% in CA, with significant test of interaction evident in only 25% of JA and 16% of CA, with no difference between industry and non-industry trials. Industry sponsored trials with a significant primary end-point reported more SGA (p < 0.001 JA; p = 0.046 CA). CONCLUSIONS: Industry sponsored trials reported more SGA. Claimed subgroup effects were often not accompanied by significant interaction test; thus circumspection should be adopted when using SGA to deviate from standard therapeutic decision-making in GI oncology.
PURPOSE: Correct interpretation of subgroup analyses (SGA) is important as it influences selection of therapeutic interventions for patient subsets. The primary aim of our study was to compare reporting of SGA between industry and non-industry sponsored trials. METHODS: We performed a systematic literature review and extracted data from journal articles (JA) and conference abstracts (CA) published over a decade reporting SGA results of phase III randomised controlled gastrointestinal (GI) oncology trials with patientparticipants of ≥150. RESULTS: In JA, SGA was reported in 100/145 (69%) trials: 41/54 industry sponsored (76%; 95% confidence interval [CI]: 63-86%) and 59/91 non-industry sponsored (65%; 95% CI: 55-74%) trials (p = 0.16). In CA, SGA was reported in 86/204 (42%) trials: 43/83 industry sponsored (52%; 95% CI: 41-62%) and 43/121 non-industry sponsored (36%; 95% CI: 28-44%) trials (p = 0.02). Number of SGA performed per trial was significantly larger for industry compared to non-industry sponsored trials in both JA (median 6 versus 2, p = 0.003) and CA (median 1 versus 0, p = 0.023). Claims of subgroup effect were made in 52% of trials in JA and 50% in CA, with significant test of interaction evident in only 25% of JA and 16% of CA, with no difference between industry and non-industry trials. Industry sponsored trials with a significant primary end-point reported more SGA (p < 0.001 JA; p = 0.046 CA). CONCLUSIONS: Industry sponsored trials reported more SGA. Claimed subgroup effects were often not accompanied by significant interaction test; thus circumspection should be adopted when using SGA to deviate from standard therapeutic decision-making in GI oncology.
Authors: Stefan Schandelmaier; Matthias Briel; Ravi Varadhan; Christopher H Schmid; Niveditha Devasenapathy; Rodney A Hayward; Joel Gagnier; Michael Borenstein; Geert J M G van der Heijden; Issa J Dahabreh; Xin Sun; Willi Sauerbrei; Michael Walsh; John P A Ioannidis; Lehana Thabane; Gordon H Guyatt Journal: CMAJ Date: 2020-08-10 Impact factor: 8.262
Authors: Jacob J Mandel; Shlomit Yust-Katz; Akash J Patel; David Cachia; Diane Liu; Minjeong Park; Ying Yuan; Thomas A Kent; John F de Groot Journal: Neuro Oncol Date: 2018-01-10 Impact factor: 12.300
Authors: Mahmood AminiLari; Vahid Ashoorian; Alexa Caldwell; Yasir Rahman; Robby Nieuwlaat; Jason W Busse; Lawrence Mbuagbaw Journal: Korean J Pain Date: 2021-04-01
Authors: Lawrence Mbuagbaw; Daeria O Lawson; Livia Puljak; David B Allison; Lehana Thabane Journal: BMC Med Res Methodol Date: 2020-09-07 Impact factor: 4.615