Background Cancer cachexia is a multifactorial syndrome characterised by an ongoing loss of skeletal muscle mass, with or without a loss of fat mass, leading to progressive functional impairment. Physical exercise may attenuate cancer cachexia and its impact on patient function. This is the first update of an original Cochrane Review published in Issue 11, 2014, which found no studies to include. Objectives To determine the effectiveness, acceptability and safety of exercise, compared with usual care, no treatment or active control, for cancer cachexia in adults. Search methods We searched CENTRAL, MEDLINE, Embase, and eight other databases to March 2020. We searched for ongoing studies in trial registries, checked reference lists and contacted experts to seek relevant studies. Selection criteria We sought randomised controlled trials in adults with cancer cachexia, that compared a programme of exercise alone or in combination with another intervention, with usual care, no treatment or an active control group. Data collection and analysis Two review authors independently assessed titles and abstracts for relevance and extracted data on study design, participants, interventions and outcomes from potentially relevant articles. We used standard methodological procedures expected by Cochrane. Our primary outcome was lean body mass and secondary outcomes were adherence to exercise programme, adverse events, muscle strength and endurance, exercise capacity, fatigue and health-related quality of life. We assessed the certainty of evidence using GRADE and included two Summary of findings tables. Main results We included four new studies in this update which overall randomised 178 adults with a mean age of 58 (standard deviation (SD) 8.2) years. Study sample size ranged from 20 to 60 participants and in three studies the proportion of men ranged from 52% to 82% (the fourth study was only available in abstract form). Three studies were from Europe: one in the UK and Norway; one in Belgium and one in Germany. The remaining study was in Canada. The types of primary cancer were head and neck (two studies), lung and pancreas (one study), and mixed (one study). We found two comparisons: exercise alone (strength-based exercise) compared to usual care (one study; 20 participants); and exercise (strength-based exercise/endurance exercise) as a component of a multimodal intervention (pharmacological, nutritional or educational (or a combination) interventions) compared with usual care (three studies, 158 participants). Studies had unclear and high risk of bias for most domains. Exercise plus usual care compared with usual care We found one study (20 participants). There was no clear evidence of a difference for lean body mass (8 weeks: MD 6.40 kg, 95% CI -2.30 to 15.10; very low-certainty evidence). For our secondary outcomes, all participants adhered to the exercise programme and no participant reported any adverse event during the study. There were no data for muscle strength and endurance, or maximal and submaximal exercise capacity. There was no clear evidence of a difference for either fatigue (4 to 20 scale, lower score was better) (8 weeks: MD -0.10, 95% CI -4.00 to 3.80; very low-certainty evidence) or health-related quality of life (0 to 104 scale, higher score was better) (8 weeks: MD 4.90, 95% CI -15.10 to 24.90; very low-certainty evidence). Multimodal intervention (exercise plus other interventions) plus usual care compared with usual care We found three studies but outcome data were only available for two studies. There was no clear evidence of a difference for lean body mass (6 weeks: MD 7.89 kg, 95% CI -9.57 to 25.35; 1 study, 44 participants; very low-certainty evidence; 12 weeks: MD -2.00, 95% CI -8.00 to 4.00; one study, 60 participants; very low-certainty evidence). For our secondary outcomes, there were no data reported on adherence to the exercise programme, endurance, or maximal exercise capacity. In one study (44 participants) there was no clear evidence of a difference for adverse events (patient episode report) (6 weeks: risk ratio (RR) 1.18, 95% CI 0.67 to 2.07; very low-certainty evidence). Another study assessed adverse events but reported no data and the third study did not assess this outcome. There was no clear evidence of a difference in muscle strength (6 weeks: MD 3.80 kg, 95% CI -2.87 to 10.47; 1 study, 44 participants; very low-certainty evidence; 12 weeks MD -5.00 kg, 95% CI -14.00 to 4.00; 1 study, 60 participants; very low-certainty evidence), submaximal exercise capacity (6 weeks: MD -16.10 m walked, 95% CI -76.53 to 44.33; 1 study, 44 participants; very low-certainty evidence; 12 weeks: MD -62.60 m walked, 95% CI -145.87 to 20.67; 1 study, 60 participants; very low-certainty evidence), fatigue (0 to 10 scale, lower score better) (6 weeks: MD 0.12, 95% CI -1.00 to 1.24; 1 study, 44 participants; very low-certainty evidence) or health-related quality of life (0 to 104 scale, higher score better) (12 weeks: MD -2.20, 95% CI -13.99 to 9.59; 1 study, 60 participants; very low-certainty evidence). Authors' conclusions The previous review identified no studies. For this update, our conclusions have changed with the inclusion of four studies. However, we are uncertain of the effectiveness, acceptability and safety of exercise for adults with cancer cachexia. Further high-quality randomised controlled trials are still required to test exercise alone or as part of a multimodal intervention to improve people's well-being throughout all phases of cancer care. We assessed the certainty of the body of evidence as very low, downgraded due to serious study limitations, imprecision and indirectness. We have very little confidence in the results and the true effect is likely to be substantially different from these. The findings of at least three more studies (one awaiting classification and two ongoing) are expected in the next review update.