Background: Low back pain is highly prevalent and a major source of disability worldwide. Spa therapy is frequently used to treat low back pain, but the associated level of evidence for efficacy is insufficient. To fill this knowledge gap, this protocol proposes an appropriately powered, prospective, evaluator-blinded, multi-centre, two-parallel-arm, randomised (1:1), controlled trial that will compare spa therapy in addition to usual care including home exercise (UCHE) versus UCHE alone for the treatment of chronic low back pain. Methods: Eligible patients (anticipated sample size of 358) will have had low back pain for more than 3 months and scores for pain greater than 40 mm on a visual analogue scale (VAS). Following initial consent for UCHE and baseline evaluations, patients are randomised (1:1) to UCHE alone, or UCHE plus spa therapy (18 days of mud packs, underwater massages, showers and water exercises under medical supervision). Patients in the latter arm will be requested to sign an additional consent form as per Zelen randomisation. Follow-up visits will occur at approximately months 1, 6 and 12 and (along with baseline assessments) will cover changes over time in VAS pain scores, the impact of lower back pain on daily life (the Rolland and Morris Disability Questionnaire (RMDQ)), inappropriate fears and beliefs about lower back pain (the fear, avoidance, belief questionnaire (FABQ)), general quality of life (the Euroqol Group 5 dimension, 5 level questionnaire (EQ-5D-5 L)), Patient Acceptable Symptom State (PASS), consumption of analgesic drugs and nonsteroidal anti-inflammatory drugs (NSAIDs), and overall state of health. Health resource use and days of sick leave (and subsequently the associated costs) will also be recorded. The primary outcome is the presence/absence of a clinically relevant change (improvement of at least 30%) in the VAS score for pain at 6 months. Discussion: Despite the fact that previous, rather dated recommendations encourage spa therapy for the treatment of low back pain, the current literary corpus is methodologically poor. This protocol has been designed to provide results spanning a thorough range of outcomes at the highest evidence level possible.