Randomized controlled trials (RCTs) that examined duloxetine across various dosage regimens in individuals suffering from chronic low back pain (CLBP) were included, compared with placebo. Observational studies, those with duplicated or overlapping populations, and those focusing solely on acute low back pain were excluded. Furthermore, studies that used accompanied treatment, such as exercise, electrotherapy or physiotherapy were excluded from this meta-analysis to maintain the integrity and relevance of the selected literature.

A comprehensive and systematic search strategy was developed based on four prominent electronic bibliographic databases: PubMed (https://pubmed.ncbi.nlm.nih.gov/), Web of Science (https://0710oser8-1106-y-https-www-webofscience-com.mplbci.ekb.eg/wos/woscc/basic-search), Cochrane Central (https://www.cochranelibrary.com/advanced-search) and Scopus (https://07105seqa-1106-y-https-www-scopus-com.mplbci.ekb.eg/search/form.uri?display=basic#basic). The present study aimed to maximize the retrieval of relevant articles published from the inception of each database up to March 2024. The search strategy used was as follows: For PubMed, (duloxetine hydrochloride OR duloxetine OR Cymbalta) AND (low back pain OR back pain OR back pain OR backache); for Web of Science, (duloxetine OR Cymbalta) AND (low back pain OR back pain OR back pain OR backache); for Cochrane, (duloxetine OR Cymbalta) AND (low back pain OR back pain OR back pain OR backache); and for Scopus, (duloxetine OR Cymbalta) AND (low back pain OR back pain OR back pain OR backache).

A total of 2 independent reviewers performed the study selection process. Initially, titles and abstracts were screened to identify potentially eligible studies, followed by full-text screening to confirm eligibility based on the predefined inclusion and exclusion criteria.

Data extraction was conducted using a structured electronic data extraction sheet. A total of 2 independent reviewers extracted the following data from the included studies: Study name, year of publication, study design, interventions, duration of duloxetine treatment, sample size, mean age, male percentage, weight, follow up duration and outcome measures.

The visual analogue scale (VAS) is a subjective, self-reported, pain rating scale. It is used to track pain regression. A handwritten mark placed at one point along the length of a 10 cm line that represents a continuum between the two ends of the scale; ‘no pain’ on the left end (0 cm) of the scale and ‘worst pain’ on the right (10 cm). Measurements from the zero point (left end) of the scale to the patients' marks are recorded in cm and are interpreted as their pain (11).

The Brief Pain Inventory (BPI) is a self-reported assessment. It provides information on the intensity of pain (sensory dimension or BPI-severity) and the degree to which pain interferes with function (reactive dimension or BPI-Interference). BPI-I is rated on a scale of 0 (no interference) to 10 (interferes completely). The average of 7 items or questions that the BPI-I scale addresses (general activity, mood, walking ability, normal work, relations with other people, sleep and enjoyment of life) were reported (12). The Roland-Morris Disability Questionnaire-24 (RMDQ-24) measures physical disability due to CLBP on a scale ranging from 0 (no disability) to 24 (severe disability) (13).

The SF-36 includes 36 questions. In total, 8 different sub-scores (physical and social functioning, physical and emotional role limitations, mental health, energy, pain, and general health perceptions), and a mental and physical summary score can be obtained from these. A maximum score of 100 resembles the best possible health state (14). Patients' global impression of improvement (PGI-I) compares a participant's perception of improvement at the time of assessment with treatment initiation time. Score varies from 1 (very much better) to 7 (very much worse) (15).

Of the included studies, only Raskin et al (16) specifically addressed depression, defining it according to DSM-IV criteria (17). While this definition was referenced in the present study, none of the other 7 studies included a focus on depression. Depression was defined according to the criteria of DSM- IV (17). The included study by Raskin et al (16) also defined depression as in the DSM-IV. The diagnosis was confirmed by the Mini International Neuropsychiatric Interview, a standardized diagnostic interview based on DSM-IV criteria (17). Baseline disease severity was defined by patients' scores on the Hamilton Depression Rating Scale with 17 items (HAM-D) (18). Patients were required to have a HAM-D total score of ≥18 at visits 1 and 2; a Mini-Mental State Examination (MMSE) score of ≥20 with or without mild dementia; and at least one previous episode of major depression. Patients with an MMSE score of 20 to 23 were categorized as having mild dementia, while those with a score of ≥24 were categorized as having no dementia (18).

The risk of bias in the included RCTs was assessed using the well-established Cochrane collaboration tool (19). This evaluation included sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting and other potential sources of bias. Each study's bias risk was categorized as ‘Low’, ‘High’, or ‘Unclear’ ensuring a comprehensive understanding of the methodological quality and robustness of the included studies.

For dichotomous data, risk ratios (RR) with 95% confidence intervals (CIs) were used. For continuous data, the mean differences (MD) with 95% CI were employed. Data synthesis was performed using the R software (meta-package; v.6.5-0; https://www.r-project.org/). Visual inspection of the forest plots and measurement of the Q and I² statistics were used to assess heterogeneity. Significant statistical heterogeneity was indicated by the Q statistic P<0.1 or I²>50%. Random effect model was used in the meta-analysis.

A total of 740 records were identified, 235 redundant entries were removed and 505 unique records were reviewed. After assessing the full-text articles, the pool was narrowed down to 24 relevant studies, and finally a total of 8 studies were included in the meta-analysis. Any discrepancies were resolved through discussion and consensus to ensure the robustness and integrity of the study selection process. The study selection process is shown in Fig. 1.

The included studies collectively enrolled a total of 1,985 patients. These studies have exhibited diverse characteristics, as summarized in Table I. The methodological quality of each study was evaluated using the Cochrane collaboration tool, illuminating a spectrum ranging from moderate to high quality (Fig. 2).

The results revealed that duloxetine was superior to placebo in reducing pain intensity, as demonstrated by significant improvements on both the VAS (20,21) [MD, -2.82; 95% CI, (-3.67 to -1.97); P<0.0001]. A subgroup analysis was conducted to differentiate patients with concomitant depression from those with only CLBP. The results of both groups (CLBP with and without depression) showed that duloxetine had significant advantage over control [MD, -2.71; 95% CI, (-3.57 to -1.85) and MD, -7.20; 95% CI, (-12.62 to -1.78), respectively] (Fig. 3).

Furthermore, the overall mean difference favoured duloxetine over control in terms of the BPI score (7,22-24) [MD, -0.78; 95% CI, (-1.42 to -0.13); P<0.0001] (Fig. 4). In addition, the overall mean difference between favoured duloxetine over control regarding the RMDQ-24 scale [MD, -0.87; 95% CI, (-1.36 to -0.39), P<0.0001] (Fig. 5), BPI-S average pain [MD, -0.57; 95% CI, (-0.78 to -0.36); P<0.00001] (Fig. 6) and worst pain [MD, -0.53; 95% CI, (-0.80 to -0.26); P<0.0001] (Fig. 7).

Quality of life. One study (21) reported on impact of duloxetine on overall SF-36 scores, and showed no significant difference between the compared groups [MD, 4.30; 95% CI, (-3.59 to 12.19); P>0.05].

PGI-I. Assessment of PGI-I across the 4 studies (7,22-24) revealed significant advantages of duloxetine over placebo [MD, -0.33; 95% CI, (-0.46 to -0.21); P<0.0001)], with no significant heterogeneity (P=0.92 and I²=0%) (Fig. 8).

Safety and tolerability. In total, 7 studies reported data on safety and tolerability while the study conducted by Samadi et al (21) did not report any data on side effects. The analysis revealed no significant difference in the occurrence of serious adverse events (chest pain, vertigo, dyspnoea, perioral numbness, perioral numbness, myocardial infarction, toxic myopathy, asthma and alcohol poisoning) between the duloxetine and control groups [RR, 1.92; 95% CI, (0.80 to 4.60); P=0.054], with studies (7,22-24) demonstrating homogeneity (P=0.36 and I²=6%) (Fig. 9).

Compared with control, duloxetine exhibited more constipation [RR, 3.00; 95% CI, (1.86 to 4.84); P<0.00001] (Fig. 10), diarrhoea [RR, 1.82; 95% CI, (1.08 to 3.08); P<0.0001] (Fig. 11), nausea [RR, 4.37; 95% CI, (2.86 to 6.67); P<0.00001] (Fig. 12) and dizziness [RR, 2.94; 95% CI, (1.70 to 5.09); P=0.0001] (Fig. 13). In addition, no significant difference was found between both groups regarding insomnia [RR, 1.17; 95% CI, (0.65 to 2.10); P=0.54] (Fig. 14) and headache [RR, 1.08; 95% CI, (0.42 to 2.77); P=0.82] (Fig. 15).

Sensitivity analysis. A sensitivity analysis was performed to evaluate the impact of excluding the study by Samadi et al (21) that included post-surgical patients. The aforementioned study only reported on the VAS outcome. After removing it, the analysis did not show significant difference between duloxetine and placebo in terms of VAS outcome, as revealed in Fig. 16.