Summary Physical exercises help reduce pain and disabilities in people with non-acute low back pain (> 3-4 weeks), but the effects obtained are limited. To increase the efficacy of this type of intervention, it is important to identify which patients respond most favourably to each exercise method (matching of patient/intervention) and for which reasons (underlying mechanisms). The focus of this study was a lumbar stabilization exercise program (LSEP), an active exercise method that is gaining in credibility and popularity. The study had two objectives: (1) to finalize the phase of deriving clinical prediction rules (CPRs) for a successful treatment outcome1 in order to identify, during clinical examination, those patients likely to respond most favourably to these exercises; (2) to examine the mechanisms (physical, psychological and neuromuscular) activated by these exercises by using more specific measures in order to describe the effects of the treatment. A preliminary study involving 48 patients had produced sufficient results to warrant recruiting additional patients and thus attain the number of patients needed to meet these objectives. Based on the results of the preliminary study, 64 additional patients (experiencing pain for more than four weeks) had to be recruited to attain a sample of 107 patients, i.e. the total sample size needed for statistical purposes. The exercise program was carried out over eight weeks (2 sessions/week) in physiotherapy clinics. The main measures of the results [pain intensity; perceptions of disability (Oswestry Disability Index)], as well as several measures obtained via questionnaires (psychological measures [PSY] associated with pain and treatment adherence), were collected at baseline (T0), at weeks 4 (T4) and 8 (T8 - end) of the exercise program, and at six months post-treatment (T34). The other measures – obtainable in clinical settings and therefore likely to be retained to develop the CPRs (objective 1), i.e. physical tests (measures from the clinical physiotherapy examination [PHT]) – were taken at T0 and T8. These physical tests included tests for joint instability (n = 4), flexibility (n = 6), sensorimotor control deficiencies. (n = 8), physical performance (n = 4) and muscular endurance (n = 3). Six laboratory tests were also performed at T0 and T8 to study the neuromuscular action mechanisms (neuromuscular measures [NRMs] for objective 2) in a sub-sample of 77 patients. The analyses for both objectives involved, first, forming three sub-groups of patients according to their level of success following the LSEP (1. success; 2. clinically significant improvement; 3. failure), as assessed using the Oswestry Low Back Pain Disability Questionnaire, which measures the perception of disability related to back pain. The groups at each end of the spectrum were then compared (successful sub-group (SG) vs failed sub-group (FG)) to ensure a clear differentiation. Moreover, at both T8 and T34, the effect sizes of the LSEP corresponding to the SG were much higher than those for the FG. This result was anticipated for the measure of disability perception, but was also observed (to a lesser degree) for that of pain intensity. Results and discussion related to objective 1 (development of the CPRs): Our two CPRs for success (at T8 and T34) – both comprising tests with a direct theoretical link to the concept of low back instability – showed a very strong predictive value. The CPR for success at T8, derived from 99 patients (SG + FG) to predict a successful outcome at the end of the eight-week clinical treatment, retained four variables from the physical examination. When at least two of these variables were positive, the positive likelihood ratio was very high (LR+ = 17.9), raising the probability of success in the LSEP from 49% (without applying the CPR) to 96% at T8, i.e. a 47% improvement in the probability of success. These diagnostic performance indicators were higher than those of the preliminary CPR derived by Hicks et al., (2005) for a LSEP, a rule comprising four variables and offering a LR+ of 4.0 and a probability of success that rose from 33% (without applying the CPR) to 67% at T8 (a 34% improvement). Our CPR for success at T34, for which the clinical derivation was based on a sample of 89 patients to predict a successful outcome after a six-month follow-up period, retained three predictive age-adjusted variables. When at least three of these variables were positive, the positive likelihood ratio was also very high (LR+ = 17.0), raising the probability of success in the LSEP from 53% (without applying the CPR) to 85% at T34, i.e. a 32% improvement in the probability of success. Compared with the preliminary CPR of Hicks et al. (2005), derived to predict success at T8 (the end of the clinical program), the CPR at T34 offered a much higher LR+ (17.0 vs 4.0), but a comparable change in terms of predicting success (improvement of 32% vs 34%). It is important to note that this is the first CPR in the field of physiotherapy for predicting success after a follow-up period, in this case, of six months. Results and discussion related to objective 2 (study of action mechanisms): Comparisons between the SG and FG over time revealed that several physical mechanisms (measures obtained in the clinical examination) and psychological mechanisms (measures obtained through questionnaires) may potentially explain success in the LSEP, which was not the case for neuromuscular mechanisms (measures obtained in the laboratory). The LSEP thus made greater improvements in several clinical indicators (e.g. physical endurance tests) and psychological indicators (e.g. fears and beliefs regarding physical activity) in the SG than in the FG. This finding supports the idea that the LSEP provides a gradual exposure to activity that is conducive to reducing pain- and movement-related fears, as proposed in the Fear-Avoidance Model. Lastly, closer examination of adherence to the at-home exercise program suggests that some of the psychological mechanisms involved might play a role in this adherence, in turn impacting the success of the LSEP. In conclusion, this study successfully completed the phase of deriving two CPRs for success in a LSEP, i.e. for predicting a successful outcome at the end of eight weeks of treatment (CRP at T8) or after a six-month follow-up (CPR at T34). This finding justifies continuing efforts to develop this type of tool, which allows for better matching of patients with this particular type of exercise program. The study of the action mechanisms also suggests that this type of exercise program includes action mechanisms that are beneficial for patients on both physical and psychological levels.  The goal was initially to produce a CPR for a successful outcome and a CPR for a failed outcome, but during this study, it was deemed more useful to concentrate on the CPR for a successful outcome. However, rather than considering only the immediate success at the end of the clinical program, a CPR for a successful outcome was also derived for predicting success at the six-month follow-up.