Cochrane, PubMed, Web of Science (WOS), Embase and EBSCOhost, were accessed for searching the related articles with the keywords of ‘chronic non-specific low back pain’, ‘NSCLBP’, ‘chronic low back pain’, ‘CLBP’, ‘low back pain’, ‘LBP’, ‘back pain’, ‘lumbar pain’, ‘spinal pain’, ‘lumbar pelvic pain’, ‘lumbar-pelvic pain’, ‘healthy participant’, ‘healthy control’, ‘asymptomatic individual’, ‘asymptomatic subject’, ‘healthy subject’, ‘gait pattern’, ‘gait’, ‘gait parameter’, ‘walking’, ‘gait characteristic’, ‘gait analysis’, ‘gait behavior’, ‘gait kinetics’. Boolean operators ‘AND’ and ‘OR’ were applied to combine the keywords in a search, and a valid wildcard of an asterisk (*) was added to some of the keywords to broaden the search. The concrete searching strategies of each database can be referred to in Table 1. Only document types (full-text or trials), language (English) and human studies were limited during search in each database. The latest search was updated on 22nd July 2024.

Two researchers (JG and JST) independently evaluated the eligibility of articles according to predefined inclusion and exclusion criteria to minimize bias, with any disagreements resolved through consultation with a third reviewer (SW).

Full-text, peer-reviewed original research articles since 2000 were included if the following inclusion criteria were met: 1) The primary study population consisted of patients with NSCLBP who had experienced pain for a minimum of three months; 2) Gait or walking pattern is captured by the motion capture system, treadmill and gait mat; 3) The study reported spatiotemporal variables such as walking speed (m/s), step length (m), cadence (steps/min), step width (m), stance phase (%), single support phase (%), swing phase (%), and stride length (m).

Articles were excluded if those met the following exclusion criteria: 1) The full text of the article was unavailable; 2) Mean values and standard deviations of relevant parameters were not reported; 3) The parameters were not related to spatiotemporal parameters; 4) The study population included other types of low back pain, such as acute low back pain, sub-acute low back pain, or low back pain with specific causes.

To minimize bias, two researchers (JG and JST) independently assessed the methodology quality and referred to the modified version of the Downs and Black checklist [18, 19]. Four categories were evaluated, including the reporting (questions 1, 2, 3, 5, 6, 7, and 10), external validity (questions 11 and 12), internal validity (questions 16, 18, and 21), and confounding (questions 21, 22, and 25). Each question was scored as either 1 (Yes) or 0 (No or unable to determine), except for question 5, which was scored as 2 (Yes), 1 (Partial), or 0 (No), giving a maximum total score of 16. Discrepancies were resolved through discussion, with a third researcher consulted when necessary. Inter-rater reliability was assessed using Kappa correlation.

Review Manager 5.4 (Copenhagen, the Nordic Cochrane Centre, Cochrane Collaboration) software was used to do a meta-analysis. The forest plot with the random effects model was compared to the kinematic gait data between the NSCLBP patients and the healthy controls. I² was used to check the heterogeneity among studies, and it could be classified into three levels, which included low heterogeneity (0–50%), moderate heterogeneity (50–75%), and high heterogeneity (> 75%) [20]. The total standardized mean difference (SMD, effect size [ES]), 95% confidence interval (CI), and p-value were calculated. ES was categorized into small effect (0.2–0.5), medium effect (0.5–0.8), and large effect (> 0.8) [21]. The percentage would be calculated with the formula (percentage = scored/total score (16) × 100%). Kappa’s correlation was analyzed by SPSS (v26.0; CED, Cambridge, United Kingdom), and no more than 60%, 60% to 75%, and above 75% were considered lower quality, moderate quality, and high quality, respectively [18].

822 papers were found from the databases of Cochrane (128/822), PubMed (103/822), WOS (221/822), Embase (205/822), and EBSCOhost (165/822). 359 duplicated papers were removed. Non-English (1/463), non-related title or abstract (359/463), case report (2/463), conference (1/463), review (4/463), and research protocols (74/463) were excluded. Meanwhile, nine studies out of 22 were not retrieved due to violating the eligibility assessment of the unrelated parameters or targeted participants. Eventually, thirteen studies were included in the systematic review [15, 16, 22–32]. One [26] out of 13 was not entered into the meta-analysis due to missing reporting of the mean and standard deviation. Therefore, twelve studies [15, 16, 22–25, 27–32] were included in the meta-analysis (Figure 1).

Thirteen studies [15, 16, 22–32] were assessed for quality. Two studies [26, 27] scored 50%, one study [32] scored 62.5%, three studies [22, 24, 31] scored 68.25%, four studies [15, 16, 25, 28] scored 75%, and three studies [23, 29, 30] scored 81.25%. Based on these scores, seven studies [15, 16, 23, 25, 28–30] were classified as high quality, six studies [22, 24, 26, 27, 31, 32] as moderate quality, and only two studies [26, 27] as low quality. All studies [15, 16, 22–32] reported poor external validity. Regarding internal validity and confounding, four studies [22, 24, 31, 32] showed low validity in items 21 and 22, one study [27] in items 21, 22, and 25, and one study [24] in item 21. Inter-rater reliability was high, with a Kappa of 0.836 (p < 0.001) (Table 2).

A total of 678 participants were recruited from the thirteen studies [15, 16, 22–32] for the systematic review. 378 NSCLBP were compared to different gait parameters in the 300 healthy controls, and 168 males (NSCLBP: 98, control group [CON]: 70) and 278 females (NSCLBP: 135, CON: 143) were reported among nine studies [16, 22–26, 29–31] out of 13. Twelve studies [15, 16, 22–26, 28–32] described participants’ age, with mean ages of 46.82 years for the NSCLBP group and 44.76 years for the healthy group. Nine studies [22–26, 28–30, 32] reported pain intensity in NSCLBP patients, with a combined mean pain score of 3.85 ± 1.29. The spatiotemporal parameters of walking speed, step length, cadence, step width, stance phase, single support phase, swing phase, and stride length were selected to compare the NSCLBP and healthy controls. Six studies [15, 16, 24, 25, 28, 30] reported that patients with NSCLBP walked significantly more slowly than healthy participants. Two of these studies [15, 30] specifically found a statistically significant reduction in cadence in the NSCLBP group compared to healthy controls, with values of 107.2 ± 8.4 vs. 111.0 ± 7.1 steps/min and 107.03 ± 8.7 vs. 110.9 ± 8.8 steps/min, respectively (p < 0.05). Additionally, three studies [15, 16, 28] observed shorter step lengths in NSCLBP patients, while one study [28] reported a smaller stride length compared to healthy controls. No studies reported significant differences between NSCLBP patients and healthy participants in step width, single support phase, stance phase, or swing phase [15, 16, 22–32] (Table 3).

Walking speed, an essential spatiotemporal parameter, is calculated via distance divided by displacement time. In this review, ten out of twelve studies investigated walking speed. Five studies [15, 16, 24, 25, 28] reported that NSCLBP patients walked significantly slower (1.03 m/s to 1.84 m/s) compared to healthy participants (1.17 m/s to 1.97 m/s). However, another five studies [23, 27, 30–32] found no significant difference in walking speed between the two groups. The potential causes of these differences can be attributed to multiple factors, including pain intensity, gender, age, body weight, muscle inhibition, restricted trunk-lower limb flexibility, and other variables.

Pain severity is a factor that may influence walking speed. Cross-sectional studies comparing NSCLBP patients with age- and gender-matched healthy controls normalized gait velocity by accounting for stride length, body weight, and stride time, and the findings revealed that NSCLBP patients with higher pain intensity tend to walk at a slower pace [33]. Therefore, pain intensity may be a potential factor contributing to the variability in walking speed. This evidence is further supported by an observational study that compared muscle activation between NSCLBP patients and healthy controls [34]. The study found that NSCLBP patients tend to exhibit reduced muscle activation during voluntary contractions, particularly in muscles such as the lumbar multifidus, knee extensors, and hip extensor muscles. The decreased activation of these key muscle groups may contribute to limited knee and hip extension, which in turn could impair overall gait mechanics. This reduction in muscle function may restrict the range of motion and efficiency of movement, ultimately leading to a slower walking speed in patients with NSCLBP. Thus, the compromised muscle activation observed in NSCLBP patients may play a significant role in the altered gait patterns and reduced mobility seen in this population.

Gender is a recognized factor influencing walking speed, with females typically walking slower than males. A descriptive meta-analysis highlighted differences in walking speed across sexes and age groups, reporting preferred walking speeds ranging from 0.97 to 1.43 m/s for males and 0.94 to 1.39 m/s for females [35]. Ayis et al. [36] observed that healthy women aged 35 to 44 years had an average walking speed of 1.26 m/s (95% CI: 1.19–1.34), while men in the same age group averaged 1.35 m/s (95% CI: 1.28–1.42). Therefore, gender should be considered when evaluating walking speed in NSCLBP patients. In this review, all studies included male and female participants, which may influence the consistency of findings. Additionally, walking speed may be affected by age. One study categorized participants into young (below 30 years), middle-aged (31–45 years), and older adults (above 45 years) and found that walking speed decreases with age [37]. Similarly, Hageman [38] also reported a declining trend in walking speed with ageing, with averages of 1.76 m/s (age 39.2 ± 12.6), 1.60 m/s (age 61.5 ± 3.4), and 1.39 m/s (age 74.7 ± 6.6).

This review reported walking speeds ranging from 1.03 m/s to 1.84 m/s in NSCLBP patients aged 20 to 60 and from 1.17 m/s to 1.97 m/s in healthy individuals. Notably, the fastest walking speed was 1.97 m/s in healthy participants and 1.84 m/s in NSCLBP patients aged 20 to 40 years [25], exceeding previously reported values for females (1.34 m/s) and males (1.42 m/s). In Müller et al.’s [25] study, participants were allowed to select their walking and running speeds, which may account for differences in walking speed on level ground compared to earlier findings. It is also important to consider that walking speed variability could be influenced by factors such as measurement devices, walking conditions, individual muscle strength, and other variables [25]. Therefore, while evidence suggests that walking speed is significantly reduced in NSCLBP patients compared to healthy controls, these findings should be interpreted with caution. Nevertheless, this finding is clinically significantly important for clinicians to address the issue during rehabilitation.

The linear distance from the heel of one foot to the heel of the opposite foot is a sensitive parameter commonly used to evaluate gait patterns in NSCLBP patients. In this review, step length ranged from 0.44 m to 0.66 m in NSCLBP patients and from 0.44 m to 0.74 m in healthy controls [15, 16, 27–29, 32]. NSCLBP patients typically exhibit shorter step lengths compared to healthy controls [15, 16, 27, 28], which may be attributed to reduced pelvic rotation in the horizontal plane. This restriction is often associated with shortened hip flexors, causing anterior pelvic tilt [39], which limits the pelvic rotation required for achieving larger step lengths [40]. Additionally, pain in the lower back may contribute to reduced step length. NSCLBP patients often decrease lumbar and pelvic movement to minimize pain, further restricting step length [41]. However, two studies [29, 31] found similar step lengths in NSCLBP patients and healthy controls. It is worth mentioning that these studies primarily involved younger participants aged 20 to 30 years, which may explain the less pronounced changes in step length despite back pain persisting for over three months.

This meta-analysis suggests no significant differences in cadence between NSCLBP patients and healthy controls, although a small ES indicates a decreasing trend. Cadence, defined as the number of steps per minute, is influenced by walking speed and stride length. Most evidence indicates that NSCLBP patients tend to decrease cadence during walking tasks [15, 16, 24, 27, 31], likely as a strategy to minimize pain irritation. For instance, NSCLBP patients often take longer to complete walking tasks with reduced step frequency, reflecting an adaptive strategy to avoid discomfort. A cross-sectional study observed that NSCLBP patients had a longer cadence duration (132.90 ± 39 s) compared to healthy controls (122.18 ± 21.91 s), suggesting fewer steps per minute [28]. This trend toward a slower cadence highlights an effort by NSCLBP patients to reduce pain while walking. The reduction in cadence may serve as a compensatory mechanism, where patients prioritize minimizing discomfort over maintaining normal gait patterns. Overall, while cadence differences between NSCLBP patients and healthy controls are subtle, the observed trends underscore the importance of considering cadence as a key factor when assessing gait adaptations and pain management strategies in NSCLBP patients.

In conclusion, this systematic review and meta-analysis found moderate evidence that patients with NSCLBP exhibit significantly slower walking speed and shorter step length compared to healthy controls. These findings suggest a distinct clinical alteration in walking pattern that should be targeted during rehabilitation to improve functional outcomes. However, other spatiotemporal parameters—including cadence, step width, stride length, phase timings-showed high variability across studies due to clinical and methodological differences. As these findings are specific to NSCLBP, their generalizability to other types of low back pain remains uncertain. Future high-quality studies are warranted to explore multidimensional gait variability, assess the impact of pain intensity, and examine gait characteristics across diverse low back pain populations.

This review employed a comprehensive search strategy and a two-reviewer screening process, adhered to PRISMA guidelines, and included a quantitative synthesis using meta-analysis. Therefore, the results provide robust evidence supporting and explaining the differences in walking patterns between individuals with NSCLBP and healthy controls.

However, this review has several limitations that warrant attention. First, the study focuses solely on spatiotemporal parameter changes to summarize gait characteristics and does not incorporate other kinematic parameters, such as joint angles in the sagittal, frontal, and transverse planes, or kinetic data on force generation during different gait phases. Consequently, the high heterogeneity observed in walking speed and step length may reflect unmeasured variability in these additional gait dimensions, limiting the generalizability of the pooled results. Second, the severity of pain, a critical factor influencing spatiotemporal gait parameters, was not addressed. Differences in pain intensity may further contribute to the observed variability and warrant further investigation. Finally, this systematic review and meta-analysis exclusively examined gait variability in patients with NSCLBP, so the applicability of these findings to other types of low back pain remains uncertain.