Background Pelvic floor muscle training can reduce prolapse severity and symptoms in women seeking treatment. We aimed to assess whether this intervention could also be effective in secondary prevention of prolapse and the need for future treatment. Methods We did this multicentre, parallel-group, randomised controlled trial at three centres in New Zealand and the UK. Women from a longitudinal study of pelvic floor function after childbirth were potentially eligible for inclusion. Women of any age who had stage 1–3 prolapse, but had not sought treatment, were randomly assigned (1:1), via remote computer allocation, to receive either one-to-one pelvic floor muscle training (five physiotherapy appointments over 16 weeks, and annual review) plus Pilates-based pelvic floor muscle training classes and a DVD for home use (intervention group), or a prolapse lifestyle advice leaflet (control group). Randomisation was minimised by centre, parity (three or less vs more than three deliveries), prolapse stage (above the hymen vs at or beyond the hymen), and delivery method (any vaginal vs all caesarean sections). Women and intervention physiotherapists could not be masked to group allocation, but allocation was masked from data entry researchers and from the trial statistician until after database lock. The primary outcome was self-reported prolapse symptoms (Pelvic Organ Prolapse Symptom Score [POP-SS]) at 2 years. Analysis was by intention to treat. This trial is registered with ClinicalTrials.gov, number NCT01171846. Findings Between Dec 21, 2008, and Feb 24, 2010, in New Zealand, and Oct 27, 2010, and Sept 5, 2011, in the UK, we randomly assigned 414 women to the intervention group (n=207) or the control group (n=207). One participant in each group was excluded after randomisation, leaving 412 women for analysis. At baseline, 399 (97%) women had prolapse above or at the level of the hymen. The mean POP-SS score at 2 years was 3·2 (SD 3·4) in the intervention group versus 4·2 (SD 4·4) in the control group (adjusted mean difference −1·01, 95% CI −1·70 to −0·33; p=0·004). The mean symptom score stayed similar across time points in the control group, but decreased in the intervention group. Three adverse events were reported, all of which were in the intervention group (one women had a fall, one woman had a pain in her tail bone, and one woman had chest pain and shortness of breath). Interpretation Our study shows that pelvic floor muscle training leads to a small, but probably important, reduction in prolapse symptoms. This finding will be important for women and caregivers considering preventive strategies. Funding Wellbeing of Women charity, the New Zealand Continence Association, and the Dean's Bequest Fund of Dunedin School of Medicine.