We performed a comprehensive systematic search across four online databases: PubMed, Scopus, Web of Science, and Embase, covering publications from their inception through April 30, 2024. Additionally, we manually searched Google Scholar to identify relevant gray literature. Our search strategy involved an exhaustive examination of Medical Subject Headings (MeSH) terms in the title and abstract ([title/abstract]) for Boswellia, glycemic markers, and lipid profile.

Our research question was formulated using the Patient, Intervention, Comparison, and Outcome (PICO) framework, which guided the selection of studies and ensured the relevance and specificity of the search strategy ( Table 1 ) (22).

Studies were included if they involved individuals with T2DM of any age and health status, with interventions consisting of boswellia or its extract and comparisons to placebo, no treatment, or alternative interventions, specifically examining changes in lipid profile and glycemic markers. Exclusion criteria encompassed studies involving participants with other types of diabetes (e.g., type 1 diabetes mellitus, gestational diabetes), records including pregnant or breastfeeding women, non-human studies (animal, in vitro, and in vivo studies), non-English language publications, and reviews, editorials, and opinion pieces.

Two independent reviewers (P.R. & M.A.) conducted a multi-step screening process for the identified studies to ensure the inclusion of relevant and high-quality data. Initially, they performed a title and abstract review to eliminate studies that did not meet the predefined criteria quickly. Studies deemed potentially relevant underwent a detailed full-text assessment, where the reviewers evaluated them against specific inclusion and exclusion criteria, such as study design, population characteristics, interventions, and outcomes. Any discrepancies between reviewers were resolved through discussion, and if consensus was not reached, a third reviewer (M.K.) provided an independent evaluation. Throughout the selection process, the reasons for excluding studies were documented to ensure transparency and reproducibility.

The screening process involved three steps: removing duplicates with reference management software, reviewing titles, abstracts, and keywords, followed by full-text assessment based on PICO criteria, and finally, selecting relevant studies for analysis.

Two individual reviewers (P.R. and M.A.) independently performed data extraction, ensuring accuracy and consistency through a cross-over verification process. In cases of disagreement or conflict, resolution was achieved through discussion and consultation with a third reviewer (M.K.), followed by double-checking the extracted data to confirm accuracy. Data from each included article were systematically compiled across five key categories: general information (such as the first author, publication year, country of origin, and study design), population characteristics (age and gender), intervention features (form, dose, and duration of boswellia administration), comparative analysis (control groups including placebo, no treatment, or alternative interventions), and primary outcomes (glycemic markers and lipid profile). This rigorous and comprehensive approach assured accurate capture and analysis of all relevant data, providing a robust basis for evaluating the effects of boswellia on specified health outcomes.

The results were interpreted in the context of existing literature, highlighting the clinical implications of boswellia on lipid profile and glycemic markers. Recommendations for future research were proposed based on the gaps identified in the current evidence.

Two researchers (M.K. & K.V.) evaluated the methodological quality of each included study through the Revised Cochrane risk-of-bias tool for randomized trials (RoB 2). This tool was built to assess the risk of bias in randomized clinical trial studies (23). The tool consists of 5 items that concern five domains, including selection bias, reporting bias, performance bias, attrition bias, and other sources of bias.

We used Stata version 18 (Stata Corp, College Station, TX, USA) software to calculate the pooled standard mean difference (SMD) [95% confidence interval (CI)] of serum lipid profile levels in diabetic patients receiving boswellia treatments versus those who received placebo or no treatment in both pretreatment and posttreatment. We also performed SMD analysis to compare the pretreatment blood levels of lipid profiles to the posttreatment levels in both groups. In addition, we used the inconsistency index (I ² ) and Cochrane Q to examine the heterogeneity of the included studies. An I² greater than 50% or a p-value less than 0.05 indicated significant heterogeneity among the studies. Pooled prevalences were estimated using a random-effect model. In addition, we assess the effect of potential factors. The confidence intervals for I² were reported to determine the degree of heterogeneity among the studies (24).

A comprehensive systematic search across online databases, including PubMed (n=60), Web-of-Science (n=117), and Scopus (n=213), initially identified 390 studies. Also, to find any gray literature, Google Scholar was manually searched. After removing 134 duplicates, 256 studies were screened based on the [Titles and Abstract]. During this screening process, 242 additional records were excluded due to non-human studies (in vivo, in vitro, etc.), review studies, irrelevant studies, and population. This left 14 studies for full-text review to assess their eligibility. Of these, 9 studies were excluded primarily due to insufficient data, other outcomes, or without appropriate treatment or control groups. Finally, 5 papers published between 2014 and 2024 were included in the meta-analysis, as shown in Figure 1 .

The five studies included in the meta-analysis involved 287 patients with T2DM. A total of 144 individuals received the boswellia supplementation, and 143 patients in the control group received either a placebo or no treatment (in 3 of 5 studies, subjects received a placebo; in the other 2, they received nothing). The dosage regimen for the boswellia supplementation ranged from 500 to 1200 mg per day, with treatment durations spanning 6, 8, or 12 weeks. The form of Boswellia, in one study, was powder (16), and in the remaining 4 studies were resin (12–15) ( Table 2 ).

The study population was pretty balanced in terms of gender, with 52.32% (95%CI: 41.26-63.27; I ² = 26.98%) of cases and 55.04% (95%CI: 45.03-64.86; I ² = 10.48%) of controls being men. The age range of all studies varied from 18 to 65 years, with most patients falling in their 30-50th decades of life. In the intervention group, the mean age and BMI were 54.57 (95%CI: 51.34-57.80; I ² = 52.9%) and 27.01 (95%CI: 24.93 – 29.09; I ² = 78.6%), respectively. Controls’ pooled mean age and BMI were 53.66 (95%CI: 50.2 -57.30; I ² = 56.6%) and 26.58 (95%CI: 23.77 -29.40; I ² = 81.1%), respectively. The mean BMI between the two groups showed no significant difference (SMD=0.17; 95%CI: -0.12 to 0.46; I ² = 0.0%, P=0.251) ( Table 3 ).

Risk of bias assessment.

Our results showed that four studies had “Some concerns,” suggesting the potential for bias in certain domains, though not significant enough to fully compromise the findings. One study exhibited a “High risk of bias,” indicating serious issues that could affect the reliability of its results. More details regarding quality assessment are mentioned in Table 4 .

In 3 out of 5 studies, we retrieved the mean and SD of the fasting blood glucose (FBG) level in both intervention and control groups before and after the supplementation. The SMD of before and after supplementation levels of FBG in control and intervention groups was -0.12 (95%CI: -0.41 - 0.16; I² = 0.0%; P=0.396) ( Figure 2A ) and -1.34 (95%CI: -2.68 - 0.00; I² = 94.0%; P=0.049) respectively ( Figure 2B ). We also conducted an SMD analysis between the treatment and control groups in both time points and retrieved 0.05 (95%CI: -0.23 - 0.34; I² = 0.0%; P=0.711) ( Figure 2C ) and -1.34 (95%CI: -2.76 - 0.08; I² = 94.6%; P=0.065) ( Figure 2D ) for before and after supplementation levels, respectively.

In 4 out of 5 studies, we collected data on the mean and SD of HbA1C levels in both the intervention and control groups before and after supplementation. The SMD for the before versus after supplementation HbA1C levels in the control group was -0.19 (95%CI: -0.45 - 0.07; I² = 0.0%; P=0.159) ( Figure 3A ), while in the intervention group, it was -1.01 (95%CI: -1.55-(-0.46); I² = 73.5%; P=0.000) ( Figure 3B ). We also compared the SMD between the treatment and control groups at both time points and found values of -0.14 (95%CI: -0.40 - 0.12; I² = 0.0%; P=0.882) ( Figure 3C ) and -1.18 (95%CI: -1.83-(-0.54); I² = 79.8%; P=0.000) ( Figure 3D ) for the before and after supplementation levels, respectively.

We obtained data on the mean and SD of all lipid profile values (including TC, TG, LDL, and HDL) in the intervention and control groups before and after supplementation in all 5 included studies, except for the HDL reported in 4 out of 5 studies.

The SMD for the control group’s pre- vs. post-supplementation TC levels was 0.01 (95%CI: -0.22 - 0.24; I² = 0.0%; P=0.930) ( Figure 4A ). In the intervention group, the SMD was -0.44 (95%CI: -0.68 – (-0.21); I² = 0.0%; P=0.00) ( Figure 4B ). We also compared the SMD between the treatment and control groups at two different time points. The SMD for the pre-treatment level was -0.05 (95%CI: -0.37 to 0.27; I² = 46.7%; P = 0.746) ( Figure 4C ), while for the post-treatment level, it was -0.59 (95%CI: -1.20 to 0.03; I² = 84.1%; P = 0.060) ( Figure 4D ).

For TG levels, the SMD between the pre- and post-treatment in the control group was -0.07 (95%CI: -0.30-0.16; I² = 0.0%; P=0.541) ( Figure 5A ). The SMD in the treatment group was -0.42 (95%CI: -0.66 – (-0.19); I² = 0.0%; P=0.000) ( Figure 5B ). We analyzed the SMD between the treatment and control groups at two periods. The SMD for the pre-treatment level was -0.02 (95%CI: -0.27-0.22; I² = 10.0%; P=0.854) ( Figure 5C ). For the post-treatment level, the SMD was -0.39 (95%CI: -0.62-(-0.15); I² = 0.0%; P=0.001) ( Figure 5D ).

The SMD in HDL levels between the pre- and post-treatment in the control group was 0.18 (95%CI: -0.21 -0.56; I² = 51.3%; P=0.367) ( Figure 6A ), while in the treatment group was 0.56 (95%CI: -0.14 -1.26; I² = 83.8%; P=0.118) ( Figure 6B ). We also compared the SMD between the treatment and control groups at two-time points. The SMD for the pre-treatment level was 0.02 (95%CI: -0.25 -0.28; I² = 0.0%; P=0.903) ( Figure 6C ), while for the post-treatment level was 0.50 (95%CI: -0.47 -1.47; I² = 91.3%; P=0.311) ( Figure 6D ).

In the treatment group, the SMD in LDL levels between pre- and post-treatment was -0.43 (95%CI: -0.73 – (-0.12); I² = 39.4%; P=0.006) ( Figure 7A ), while in the control group, it was 0.03 (95%CI: -0.20 -0.27; I² = 0.0%; P=0.775) ( Figure 7B ). Additionally, we compared the SMD at two different periods between the treatment and control groups. SMD was -0.06 (95%CI: -0.56 -0.44; I² = 70.2%; P=0.805) ( Figure 7C ) for the pre-treatment level and -0.61 (95%CI: -1.32 -0.11; I² = 88.2%, P=0. 0.097) ( Figure 7D ) for the post-treatment level.

Due to the insufficient number of papers, we did not perform a subgroup analysis. However, we used meta-regression as an alternative to address heterogeneity. The meta-regression results for the effect size (SMD) with the dose and duration of treatment did not show significant results, except for the dose of therapy in post-treatment SMD between intervention and control groups in TC (P=0.041) and LDL (P=0.039) levels ( Table 5 ; Figures 8A, B ).

Figure 9 displays Egger’s publication bias plot for TC. The analysis of this plot indicated the absence of publication bias (p=0.054), indicating that both positive and negative findings have been reported.