Background A better understanding of the influence of restricted dietary intake on circulating ketone concentrations might help to explain heterogeneity in the amount of weight lost in dietary interventions. We sought to determine the association between the magnitude of weight loss and changes in fasting serum ketone concentrations in adults with severe obesity completing a meal replacement programme. Methods In this single-centre prospective cohort study, adults with severe and complicated obesity attending our regional bariatric service underwent an eight-week, milk-based, low-energy meal replacement programme. Fasting serum beta-hydroxybutyrate levels were measured with a spectrophotometric assay at baseline and after two- and eight weeks of follow-up. Results Of 120 patients who started the programme, 74 (62%) completed eight weeks. Of these, 60% were female, mean age was 49.2 ± 11.9 (range 18–75.1) years, mean body mass index was 51.2 ± 10.5 kg−2. The median [IQR] total weight loss percentage after eight weeks was 10.2 [8.6, 12.2] %. Fasting serum ketones increased between zero and eight weeks from a median of 95 [70, 140] to 185 [130, 320] μmoL/l (p < 0.0001). For each one percent reduction in body weight at eight weeks, there was a 112.6 [77.4, 147.8] μmoL/l increase in fasting beta-hydroxybutyrate concentrations, (p < 0.0001). We found similar associations between weight loss at eight weeks and increases in ketones at two weeks. However, with receiver operating characteristics (ROC) curve analyses, changes in ketones at two weeks were not a strong predictor of attaining 5% or 10% weight loss at eight weeks. Discussion In a specialist bariatric clinic-based cohort of Irish adults with severe and complicated obesity, the amount of weight lost at eight weeks was proportional to increases in fasting beta-hydroxybutyrate concentrations after two and eight weeks of milk-based, low-energy meal replacement, but early changes in fasting ketones were not a good predictor of subsequent intervention success.