This randomized clinical trial was conducted at Centro Universitário São Camilo (CUSC), São Paulo, Brazil. The study was approved by the CUSC Research Ethics Committee (protocol 53028116.0.0000.0062) and reported according to CONSORT guidelines.15 This project was funded by the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) and prospectively registered at Clinicaltrials.gov (NCT02517606).

Seventy patients with nonspecific LBP were enrolled in this randomized clinical trial. Patients were randomly assigned to 1 of 2 groups: a manual therapy and lumbar segmental stabilization group (MTLS group, n = 35) or specific hip strengthening exercises plus manual therapy and lumbar segmental stabilization group (MTLSHS group, n = 35).

All volunteers were informed of the study procedures and signed the informed consent forms prior to data collection. To be included in this study participants had to report chronic nonspecific LBP for at least 3 months without radicular symptoms. Patients were eligible if they presented with a strength deficit greater than 20% using a hand-held dynamometer for at least one of the hip muscle groups (abductors or extensors) compared to our database of asymptomatic individuals collected in a previous study.16 Participants were recruited from the Rehabilitation Service at CUSC by a single physical therapist with more than 15 years of clinical experience and who did not participate in treatment sessions. Participants were referred to the study from waiting list and local and online advertisement. Participants were excluded if they had any contraindication to exercise, neurological disorders, severe spinal diseases (e.g. fractures, tumors, or inflammatory conditions such as ankylosing spondylitis), lumbar radiculopathy (i.e. pain radiating below the knee with neurological deficits such as muscle weakness, reflex changes, or sensory deficits), spondylolisthesis with neurological impairment or stenosis, or severe cardiorespiratory diseases. We also excluded pregnant women or those patients who were taking corticosteroids or anti-inflammatory medication.17 A standard spine and lower extremities clinical examination was performed to rule out concomitant pathologies.

Randomization was conducted using a computer-generated randomized table of numbers created by an independent statistician. The allocation was conducted by an independent researcher who was not involved with other experimental procedures. Simple randomization results were concealed in sealed envelopes with consecutive numbers.15

All patients received a course of 10 treatment sessions, twice a week for 5 weeks. In the MTLS group, treatment sessions lasted approximately 30 min. In the MTLSHS group, sessions lasted approximately 45 min. Three physical therapists were trained in the study's intervention protocols and provided all treatment.

The MTLS group received a treatment consisting of a combination of manual therapy and lumbar segmental stabilization exercises. The manual therapy intervention delivered included joint mobilization in accordance with the Maitland method18 and myofascial release following the tender point principles.19 These manual techniques were intended to reduce muscle hyperactivity and joint stiffness as well as to improve lumbar spine range of motion. Posterior-anterior central mobilization was performed at selected segments between L1 to L5 that were considered stiff upon manual examination. The lumbar segmental stabilization focused on training the recruitment of the deep stabilizer muscles, such as transverse abdominal muscle, multifidus muscle, pelvic floor muscles, and diaphragm.20,21

Participants allocated to the MTLSHS group received the same treatment as described above with the addition of four exercises specifically designed to strengthen the hip muscles, performed at the end of the treatment session (Table 1).22,23 The load during training was standardized to 70% of the estimated 1-repetition maximum test, defined as the maximum load that can be lifted once while maintaining correct technique and without increasing pain. Exercises that used elastic resistance were standardized to the maximum resistance at which the patient was able to perform 10 repetitions. Patients performed these exercises solely during the physical therapy session and were not instructed to perform exercises at home.

All clinical tests and questionnaires were conducted by a single examiner at baseline, posttreatment, and 6- and 12-months posttreatment. This examiner was blinded to the patients’ group assignment and did not participate in the intervention sessions.

A 10 cm visual analog scale was used to quantify average pain intensity in the last 7 days,24 where 0 corresponded to no pain and 10 to the worst imaginable pain. A change of 2.0 cm or greater in the visual analog scale was considered the minimum clinically important difference.25 The Roland-Morris Questionnaire24,26 has been used to measure function in patients with LBP. The Roland-Morris is a 24-item functional assessment, with higher scores indicating greater disability. The minimal clinically important difference is 3.5 points in absolute values.25 These measures were administered before treatment, posttreatment, and at 6- and 12-months follow-up.

A previously calibrated hand-held dynamometer (Lafayette Instrument Company, Lafayette IN, USA) was used to measure hip muscle strength.16 Two submaximal trials were used to familiarize the patient with each test position followed by two maximal isometric contractions for each muscle group on both sides. The tested side (right or left) and the order of muscle testing were randomized. To stabilize the lower limb, a rigid band was employed around the treatment table. Data from the two maximum effort trials were averaged and used for statistical analyses. Each contraction was held for 5 s, with a 30-second rest period between trials. Strength values were measured in kilograms and were normalized to body mass index. The position of the patient and resistance application were based on a previous study by Magalhães et al.16

Two-dimensional kinematic analysis of the lower limb, pelvis, and trunk were assessed using the Myovideo system (Noraxon, Scottsdale, AZ, USA) while the patient walked on a treadmill and while performing a single-limb stepdown, a detailed description of these procedures is in Supplementary Material. The measures of strength and the kinematic parameters were analyzed at pretreatment and posttreatment. Although the reliability of strength assessment and kinematics has already been tested in previous studies,27,28 a pilot study was conducted 1 month prior to data collection with 10 asymptomatic individuals to evaluate the test-retest reliability in the laboratory. The individuals were tested according to the protocol described in the supplementary material, with a 1-week interval between the 2 testing sessions. We found in the pilot study that all comparisons of strength and kinematic data showed excellent reliability with intraclass correlation coefficients between 0.83 and 0.98.

Descriptive statistics for demographic data and all outcome measures were expressed as mean ± standard deviation (SD). The normality of distribution was verified through the visual inspection of histograms and assessed with the Shapiro-Wilk test. Data from the visual analogue scale and the Rolland-Morris Questionnaire were analyzed using separate 2-by-4 (group-by-time) mixed model analyses of variance. The factor of group had 2 levels (MTLS and MTLSHS) and the repeated factor of time had 4 levels (pretreatment, posttreatment, and 6- and 12-months posttreatment). Between-group difference in change scores were expressed as mean difference and 95% confidence interval (CI). Measures of strength and kinematic data at baseline and posttreatment were evaluated using pairwise comparisons. All participants were analyzed in the group they were randomized, following the intention-to-treat approach. Missing data were imputed using the last observation carried forward method. Statistical significance was set at 0.05. Data were analyzed with SPSS Version 20.0 (SPSS Inc, Chicago, IL, EUA). The sample size calculation was based on the pain visual analog scale (average pain in the last 7 days) with 80% power to detect an intervention effect that would generate a difference between groups of 1.5 ± 2.1 cm29,30 with an alpha value set at 0.05. A total of 31 participants per group was estimated. To allow for potential dropouts, 70 participants were recruited for this study.

Both groups showed improvements in pain and disability relative to baseline, however, there was no meaningful difference between groups at posttreatment and follow-ups (Table 3). The mean difference in changes in pain level between groups at posttreatment, and at 6- and 12-month follow-up were 0.5 points (95%: −0.5, 1.5), 0.3 points (95% CI: −0.9, 1.5), and 0.0 points (95% CI: −1.1, 1.1), respectively. The mean differences in disability were 0.8 points (95% CI: −1.3, 2.7), 0.0 points (95% CI: −2.4, 2.4), and 0.4 points (95% CI: −2.0, 2.8), respectively (Table 3).

Although normalized hip extensor strength increased at posttreatment in both groups, we observed no difference in either hip extensor or abductor strength between the two groups. The mean difference in hip extensors strength posttreatment was 0.8 (95% CI: −13.9, 15.5) (Table 4). There was no significant intra and between-group interaction observed for any of the kinematic variables (p > .05) (Table 4).