A new logistic regression model for the detection of chronic low back pain, based on a case-control study in the Spanish population

Article information

Asian Spine J. 2025;.asj.2025.0336

Publication date (electronic) : 2025 September 23

doi : https://doi.org/10.31616/asj.2025.0336

Pilar Alberola-Zorrilla¹, Ana Queralt², Maria Ángeles Pamblanco-Valero^1,3, Daniel Sánchez-Zuriaga¹

¹Research Group on the Clinical Anatomy of the Musculoskeletal System (GIACAL), Department of Anatomy and Human Embryology. Universitat de València, València, Spain

²AFIPS Research group, Department of Nursing, Universitat de València, València, Spain

³Conservatori Superior de Dansa de València, València, Spain

Corresponding author: Daniel Sánchez-Zuriaga, Department of Anatomy and Human Embryology, Facultat de Medicina i Odontologia, Universitat de València, Av. Blasco Ibáñez,15. 46010 València, Spain, Tel: +34-963983522, Fax: +34-963864159, E-mail: daniel.sanchez@uv.es

Received 2025 June 8; Revised 2025 July 4; Accepted 2025 July 6.

Abstract

Introduction

Low back pain (LBP) is considered chronic when it persists for more than 12 weeks. In 85%–90% of cases, the exact cause of LBP cannot be identified, and such patients are classified as having non-specific low back pain (NSLBP) [1].

Chronic NSLBP is a complex condition involving interactions between psychosocial and musculoskeletal factors [2,3]. Among these, muscular activity and lumbopelvic motion kinematics are often evaluated in patients with LBP [4]. Electromyography (EMG) and motion pattern analysis are relatively affordable techniques. However, despite their hypothetical clinical utility [5], methodological heterogeneity across studies has led to inconsistent results [2].

While some studies have solely focused on kinematic variables [6] or combined EMG with trunk kinematics [7] during complex motor tasks, trunk flexion-extension remains the most commonly investigated movement in NSLBP patients. This task is easy to standardize and reliably reproduces symptoms in this population [8]. Moreover, it elicits a distinct EMG activation pattern known as the flexion-relaxation phenomenon (FRP), characterized by relaxation of the erector spinae (ES) muscles toward the end of lumbar flexion. The FRP is considered a key electromyographic biomarker of muscular activity in LBP [3,8], and its absence is frequently observed in NSLBP patients [9].

EMG measures of the FRP demonstrate greater sensitivity and specificity for identifying LBP patients when combined with motion analyses [4]. However, findings remain inconsistent, often derived from laboratory settings that do not reflect real-world clinical practice [2].

The utility of NSLBP research is further limited by substantial heterogeneity in study populations, which often vary in functional impairment and symptom duration [10]. Additionally, participants are frequently recruited from the general population without prior professional assessment, leading to inconsistent characterization of the condition.

This study aimed to conduct a detailed biomechanical analysis of trunk flexion-extension in healthy individuals and a homogeneous clinically validated sample of patients with chronic NSLBP. By combining spinal motion curves with muscular activity biomarkers—such as flexion-extension and relaxation EMG activity ratios—we sought to identify kinematic and electromyographic variables that are both clinically accessible and effective in distinguishing healthy subjects from chronic NSLBP patients.

Materials and Methods

Study design and participants

This was a case-control study comparing healthy participants and patients with chronic NSLBP. An a priori analysis was conducted to determine the required sample size, targeting a statistical power of 90%. Effect size was estimated using Cohen’s d, based on data from previously published studies that examined similar dependent variables (lumbar muscle relaxation ratios, time to maximum lumbar flexion) in LBP patients and pain-free subjects [11]. Sample size calculations were performed using the G*Power 3 software (Heinrich-Heine-Universität Düsseldorf, Düsseldorf, Germany; http://www.gpower.hhu.de/) [12], yielding an estimated requirement of 43 pain-free controls and 43 NSLBP patients.

Patients were recruited from the Orthopedic Surgery and Rheumatology departments of a tertiary referral hospital. The inclusion criteria for the 43 chronic NSLBP patients were as follows: (1) a history of NSLBP, with recurrent episodes of lumbar pain over the last month and a total symptom duration of ≥12 weeks; (2) functional disability directly attributable to LBP, including time off work within the past 6 months; (3) no current or prior history of low back surgery and no evidence of specific pathologies that could account for the symptoms (e.g., spinal infection, tumor, osteoporosis, fracture, structural deformity, inflammatory disorders, radiculopathy, or cauda equina syndrome); and (4) no involvement in workers’ compensation, litigation, or disability insurance claims. The control group consisted of 43 individuals with no history of LBP or other low back disorders. An independent t-test revealed no significant differences between the two groups in terms of age, sex, weight, or height (Table 1). In the NSLBP group, pain intensity and pain-related disability were assessed using the Million Visual Analog Scale [13].

Table 1

Descriptive characteristics of the participants

Instrumentation, procedure, and outcome data

Instrumentation, testing procedures, and data processing methods are described in detail elsewhere [14]. Sagittal plane angular displacements of the lumbar spine and the pelvis were recorded using a two-sensor Liberty 240/16 electromagnetic motion capture system (Polhemus Inc., Colchester, CT, USA). One sensor was affixed to the skin over the L1 spinous process, while the second sensor was placed at the level of S1 to capture sacral inclination at the coxofemoral joint (pelvic flexion) (Fig. 1). True lumbar spine motion was calculated by subtracting the S1 data from the L1 data [15].

Fig. 1

Location of the erector spinae (ES) electromyography electrodes and the lumbopelvic position sensors. L1 and S1: sensors on first lumbar and sacral vertebrae.

Surface EMG activity of the ES was recorded using an EMG100C Biopac module (Biopac Systems Inc., Goleta, CA, USA). Registration points of ES activity were located 3 cm to the right of the L3 spinous process (Fig. 1).

Participants performed five repetitions of a standardized trunk flexion-extension movement, with each flexion and extension phase lasting 4 seconds and a 1-second pause at maximum flexion.

ES EMG values were expressed as a percentage of the maximum ES EMG amplitude recorded during each flexion-extension cycle. Similarly, spinal flexion angles were normalized to the percentage of maximum spine flexion achieved within each cycle. Summary measurements were calculated for each of the three central flexion-extension cycles [16] and subsequently averaged. ES muscle relaxation was quantified using a relaxation ratio [17], defined as the average EMG activity during 85%–100% of the flexion phase—when the FRP is typically observed—divided by the average activity during 45%–60% of the flexion phase, corresponding to peak eccentric activity. Additional summary measurements included: (1) the maximum range of lumbar flexion; (2) percentage of flexion-extension cycle during which the lumbar spine remained flexed beyond 90% of its maximum (i.e., time of maximum lumbar flexion) [11]; (3) spinal postures at the start and end of ES relaxation; and (4) average ES EMG activity during maximum lumbar flexion and the FRP. The initiation and termination points of flexion, extension, and EMG activation peaks were identified using a threshold-based method [18] (Fig. 2). Data analyses were performed using MATLAB 2020b (MathWorks Inc., Natick, MA, USA). To assess inter-observer reliability, the analyses of recordings from the first 12 pain-free subjects were independently repeated by two experienced researchers.

Fig. 2

Representative curves of lumbar spine motion (grey) and erector spinae activation (black) during trunk flexion-extension from one pain-free participant. % max electromyography (EMG) and range of motion (ROM): percentages of maximal erector spinae activation and lumbar ROM. ^a)Start of eccentric activation. ^b)End of eccentric activation and start of flexion-relaxation phenomenon. ^c)End of flexion-relaxation phenomenon and start of concentric activation. ^d)End of concentric activation. ^e)Time of lumbar flexion over 90%.

Results

Inter-observer ICC values were ≥0.8 for all variables, indicating very good reliability. NSLBP patients showed a mean score of 72±30 out of 150 on the Million Visual Analog Scale, corresponding to a severe level of pain-related disability [13].

Three lumbopelvic motion variables showed significant differences between groups: lumbar flexion angle in the standing position, maximum range of lumbar flexion, and the percentage of the flexion-extension cycle during which the lumbar spine remained flexed beyond 90% of its maximum (time of maximum flexion). Regarding the EMG activity of the ES, significant differences were found in the ES relaxation ratio and the average ES activity during maximum flexion and the myoelectrical silence (Table 2). All three variables reflect the incomplete relaxation of the ES muscles during maximum flexion and were therefore highly correlated. To avoid multicollinearity in the logistic regression model, only the relaxation ratio was included, as it is more methodologically robust and does not require normalization of the EMG signal.

Table 2

Comparisons between pain-free and chronic non-specific LBP participants

Binary logistic regression identified two variables with significant odds ratios for the presence of LBP: time of maximum flexion (odds ratio, 0.92; 95% confidence interval, 0.86–0.99) and the relaxation ratio (odds ratio, 1.08; 95% confidence interval, 1.04–1.13).

Discussion

In the present study, patients with chronic NSLBP were recruited from a tertiary care hospital and had been previously evaluated by clinicians with expertise in LBP management. This group reported severe pain and functional impairment, suggesting they are representative of the chronic NSLBP population typically seen in clinical practice [20].

Most studies that incorporate lumbopelvic kinematic measurements alongside EMG recordings focus primarily on maximal ranges of motion [8]. Lumbopelvic range of motion measurements depend on the flexibility of the subject, and their measurement is affected not only by the limitations of the specific techniques used but also by subject-related factors such as motivation, kinesiophobic behaviors, and asymptomatic age-related structural changes [2,21]. Given these sources of variability, several authors have recommended exploring alternative biomechanical indicators beyond the maximum range of motion when assessing patients with LBP [22]. Paquet et al. [17] found no significant differences in maximum ranges of motion between NSLBP patients and pain-free controls, but reported alterations in ES flexion-relaxation ratios. They also noted differences in the shape of lumbar displacement curves, with healthy subjects reaching their maximum amplitude of flexion earlier than patients. In an effort to quantify these curve alterations, Sánchez-Zuriaga et al. [11] measured the relative time during which the lumbar spine remained flexed beyond 90% of its maximum. However, their study focused exclusively on patients with lumbar disc herniation. Moreover, both studies had limited sample sizes—10 subjects per group in Paquet et al. [17] and 15 per group in Sánchez-Zuriaga et al. [11], potentially limiting the generalizability of their findings.

The present study found reduced maximum ranges of lumbar flexion in participants with chronic NSLBP, consistent with several previous studies [9]. In addition, patients with pain demonstrated a shorter duration of time spent in lumbar flexion, quantitatively supporting the descriptive observations of Paquet et al. [17] regarding altered lumbar movement curve profiles in this population. Among the kinematic variables analyzed, the time spent in >90% of maximum lumbar flexion was the only variable to yield a statistically significant odds ratio in distinguishing NSLBP patients from healthy controls. While the maximum range of lumbar flexion also differed significantly between groups in the t-test, the confidence interval for its odds ratio in logistic regression included the null value, suggesting that it may be less robust as a predictive marker when accounting for other variables.

As the odds ratio for the time of maximum flexion was less than one, its increase would imply a lower risk of LBP. In contrast, the odds ratio for the ES relaxation ratio was greater than one, indicating that lower levels of relaxation at maximum flexion would indicate a higher risk of LBP. Although the magnitude of these odds ratios may appear modest, it is important to note that both variables are continuous. Therefore, the reported odds ratios represent the change in risk per one-unit (i.e., 1%) increase in each variable [23]. These effects are cumulative, meaning that larger deviations in either variable could result in substantially greater differences in risk.

The observed reduction in maximum flexion time suggests that individuals with NSLBP reach 90% of their maximum lumbar flexion later during the flexion phase and sooner upon beginning to extend the spine from the maximally flexed position. Previous studies have similarly suggested that patients with LBP tend to reach maximum trunk flexion later than pain-free individuals [24]. At end-range flexion, small variations in flexion angles can produce substantial alterations in bending moments [25]. Therefore, the reduced time spent at maximum flexion may reflect a protective strategy aimed at minimizing mechanical stress on the damaged spinal structures. Comparable alterations in lumbar motion patterns have been reported in patients with discogenic LBP [11], as well as in pregnant women during the last trimester, when spinal loading due to fetal weight is at its peak [14].

In assessing the electromyographic activity of the spinal extensor muscles, visual inspection by trained observers to identify a distinct and abrupt reduction in muscle activity near end-range flexion—indicative of the FRP—is considered the benchmark method. However, visual inspection is inherently subjective and may be influenced by both inter- and intra-observer variability. Therefore, it is recommended that at least two assessors independently evaluate the data [18]. In the present study, high inter-observer reliability was demonstrated for all EMG-derived variables, supporting the consistency and robustness of these measures.

An alternative and more straightforward method for assessing the presence of an altered FRP is the calculation of flexion-relaxation ratios, comparing EMG activities during periods of minimum and maximum activation throughout trunk flexion [17]. This approach is objective, requires no specialized training, and can be performed by a single assessor. Moreover, it enables quantification of the extent to which FRP occurs, offering greater informational value than methods that simply identify whether FRP is present. Owens et al. [26] found that higher flexion-relaxation ratios were associated with greater disability, more pronounced clinical findings, and reduced spinal mobility. Similarly, a recent meta-analysis by Gouteron et al. [3] reported significant differences in flexion-relaxation ratios between asymptomatic subjects and NSLBP patients. However, they also noted considerable heterogeneity across studies and emphasized the need for further research to validate the clinical utility of these ratios. The present study supports the usefulness of the relaxation ratio in assessing NSLBP, as it was the only EMG-related variable that showed a statistically significant odds ratio for distinguishing healthy individuals from those with chronic NSLBP.

This study has several limitations. While the sample was homogeneous in terms of functional impairment and recruitment from a specialized clinical setting, it may limit the generalizability of the findings. Future research should investigate the behavior of these variables in individuals with varying degrees of pain intensity and functional impairment, as well as in populations with acute LBP. Additionally, further studies are needed to validate these markers against potential confounding factors such as age, physical activity level, body composition, body mass index, and other comorbidities. Implementing these studies in routine clinical practice settings would help evaluate the usability and clinical utility of this type of biomechanical and EMG analysis.

Conclusions

In this study, we identified two variables derived from a simple trunk flexion-extension test that may help identify patients with chronic NSLBP: the duration of maximum lumbar flexion and the ES relaxation ratio. Both variables were independently associated with the presence of NSLBP in a logistic regression model. Importantly, these variables can be measured using affordable, clinically accessible methods involving lumbar motion analysis and surface EMG recording, supporting their potential applicability in routine clinical assessment.

Article information Continued