Effectiveness of high-intensity laser therapy in the management of patients with knee osteoarthritis: A systematic review and meta-analysis of randomized controlled trials

2.1Literature search

We searched for articles in the MEDLINE, EMBASE, Cochrane CENTRAL, and Web of Science databases from inception to March 14, 2020. Only HILT-related terms were used and disease-related terms were not included for a higher sensitivity of the search strategy (Supplementary Table 1). In Web of Science, the word ‘pain’ was added to the search. MeSH terms and text words such as “high-intensity laser” and “HILT” were used as part of the search strategy.

2.2Study selection

Two independent reviewers selected the studies. In the scenario of a difference of opinion between the two reviewers, the selection was discussed with a third reviewer. The inclusion criteria were adult patients with knee osteoarthritis, HILT intervention (power 500 mW or more) [14], placebo or active intervention such as LLLT as comparator groups, those who reported pain relief as the primary outcome, randomized controlled trials (RCT) regardless of double-blindness, and peer-reviewed original articles. Studies were excluded if they did not include humans, pain associated with disease, or were reviews or abstracts.

2.3Quality assessment

The risk of bias (ROB) was used for the quality assessment of the RCT study design [15]. We estimated seven criteria including random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and funding source. Low, unclear, and high levels were assessed for each criterion. Two reviewers conducted the quality assessment and another person joined the discussion if there was a difference in opinion.

Figure 1.

PRISMA flow diagram of study selections.

2.4Data extraction

Two independent reviewers conducted the data extraction using a pre-specified data extraction form. Data on diagnosis, sample size, interventions, co-interventions, mean age and standard deviation (SD), the percentage of men, Kellgren-Lawrence (K-L) radiological stage [16], evaluation point, country, and outcomes were extracted. The primary outcome of interest was pain and secondary outcomes were stiffness and function using a disease-specific instrument. For the HILT intervention, we collected data on the laser type, wavelength, average output, energy density, J per point, total dose, application time, frequency of treatment, and number of repetitions.

2.5Data analyses

The primary outcome of our study was pain; additionally, knee stiffness and function were secondary outcomes. The pooled estimate was analyzed using an inverse-variance random effect model for continuous variables considering high heterogeneity. Mean difference (MD) for visual analogue scale (VAS) pain was conducted due to the same unit, and standardized mean difference (SMD) was utilized for the measurement of different units. To combine outcomes, we used mean, SD, and the number of patients. If standard error (SE) or p-value was reported, SD was calculated using an equation by the Cochrane group [17]. In the base-case analysis, we prioritized the placebo comparison group and subsequently the active comparison group (e.g., LLLT). If the study reported outcomes of more than one group of HILT, we used the combined mean and SD for meta-analysis by the Cochrane group [18, 19]. If the study reported pain scores with several measurements except for VAS, we chose the results of after-treatment when the baseline scores between the HILT control groups were similar.

Heterogeneity was assessed visually using a forest plot and Higgins I2 with the p-value. A funnel plot and Egger’s and Begg’s tests were conducted to identify publication bias. Based on the Cochrane guidelines [18], Egger’s and Begg’s tests could not be applied if fewer than 10 studies were included in the meta-analysis. Subgroup analyses were performed for HILT with or without exercise, follow-up period, and active comparator groups. RevMan version 5.3 was used for the meta-analysis.

3.1Literature search

A total of 1,182 articles were identified during the initial search (Fig. 1). After excluding duplicate articles, the title or abstract was screened for 1,043 articles. We reviewed 85 full-text articles and 80 articles were excluded because they did not include patients with knee osteoarthritis or a HILT therapy group, contained ineligible study designs, no interest outcomes reported, or were not original peer-reviewed studies. Finally, six articles were included in this meta- analysis [13, 20, 21, 22, 23, 24].

Table 3

Quality assessment of included studies using Risk of Bias Assessment tool for Non-randomized Studies (ROBANS). (a) ROBANS graph, (b) ROBANS summary +: low risk of bias; -: high risk of bias.

3.2General characteristics of the included studies

Figure 2.

Mean difference in visual analogue scale (VAS) pain between HILT and control.

The mean age of the study populations ranged from 54 to 65 years (Table 1). Two studies included only men and one study included only women. The percentage of men in the remaining studies ranged from 31.4% to 45.6%. The Kellgren-Lawrence (KL) radiological grade was mostly grades II and III. The countries where the studies were conducted were Saudi Arabia, Bulgaria, South Korea, Iran, and Poland. The final evaluation time points ranged from 2 weeks to 6 months. Pain was assessed using the VAS and/or the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), while stiffness and function were assessed using the WOMAC and the Knee Society Clinical Rating Score (KSCRS). VAS pain with a score ranging from 0 (no pain) to 10 (unbearable pain) was the primary outcome [25]. WOMAC is an arthritis-specific instrument with four levels (none, low, moderate, and severe) and three subscales (pain, stiffness, and physical function) [26, 27]; wherein higher scores indicate higher disability. The Knee Society developed the KSCRS [28], which includes pain, range-of-motion, and function. High scores were 50, 25, and 50, respectively, with high scores indicating severe problems. None of the studies reported adverse events. The modified Latinen pain index (mLPI) includes four dimensions of intensity, frequency, analgesics, and social/professional activities and graded the pain into three groups: slight pain (0–4 points), considerable pain (5–8 points), and severe pain (9–12 points) [29]. Table 2 shows the characteristics of the HILT.

3.3Quality assessment

The ROB results are presented in Fig. 3. Among the included studies, 50% or more studies were estimated to have a low risk for bias regarding the blinding of participants or assessment, selective reporting, and funding source. The method of random sequence generation in four studies was estimated to potentially be at risk for bias. The proportion of low risk for bias in allocation concealment and incomplete outcome data was approximately 30%. The overall quality across studies was fair, despite being vulnerable to selection bias and detection bias of the included studies in this review.

Figure 3.

Mean difference in disease-specific measurement between HILT and control. (a) Stiffness, (b) function.

3.4Clinical outcome: Pain

A total of six studies that evaluated pain using the VAS, including 334 patients, were included in the meta-analysis. We used the combined mean and SD from two HILT groups in Gwory et al.’s study. HILT was observed to significantly decrease the knee pain compared to the control (MD, -1.18; 95% CI, -1.68 to -0.69) (Fig. 2). There was significant heterogeneity (I2= 90%, p< 0.001). The funnel plot showed that publication bias could not be determined (Supplementary Fig. 1). We performed a subgroup analysis of HILT with and without exercise in both groups; three studies were included in each subgroup (Table 4). The MD of HILT with exercise was higher than that of HILT without exercise (MD, -1.51; 95% CI, -1.74 to -1.28; versus MD, -0.68; 95% CI, -1.14 to -0.23), when compared to the control. The longer the follow-up, the more decrease in the MD of HILT for pain was observed. In three studies, the HILT for pain was improved compared with active controls (MD, -0.83; 95% CI, -1.23 to -0.42).

No significant difference was observed in the random effect model between the HILT and placebo for disease-specific pain measures such as WOMAC, KSCRS, or mLPI in five studies including 262 patients (SMD, -1.31; 95% CI, -2.66 to 0.05), but there was significant heterogeneity (I2= 95%, p< 0.001).

3.5Clinical outcome: Stiffness

In four studies, 168 patients were included to determine the pooling effectiveness of stiffness using disease-specific measures. The SMD of HILT versus placebo was -1.00 (95% CI -1.32, -0.68). Therefore, HILT significantly decreased stiffness when compared with the placebo (Fig. 3a). Heterogeneity of the pooling effect was not significant (I2= 0%, p= 0.44).

3.6Clinical outcome: Function

Four studies, including 87 patients who received HILT and 81 placebos, had pooled results using the disease-specific function. Upon statistical analysis, it was observed that HILT significantly improved function, compared with placebo (SMD, -5.36; 95% CI, -7.39 to -3.34) (Fig. 3b). It should be noted that heterogeneity was significant (I2= 89%, p⩽ 0.001).