Literature search
The detailed flowchart of the screening and selection process in the PRISMA flow diagram is shown in Fig. 1. Afterward 1055 potentially relevant studies exclusion, 1055 of papers eligible for inclusions were confirmed with the initial search strategy. After initial screening, 517 studies were removed due to duplication studies, Of the 538 candidate studies, 46 studies were excluded due to unrelated study design, and 488 articles were left for abstract assessment. After carefully reviewing titles and abstracts, 385 studies were precluded for obvious irrelevance disease, cell or animal studies data, lack of comparative group. Then, 103 studies were chosen for full-text valuation. Of the remaining 103 full-text candidate articles, 65 potential studies were excluded, due to insufficient and useable data. Finally, 38 studies were selected to find a relationship between LDL and SBP levels and risk of diabetic peripheral neuropathy. Of the 38 finalized studies, 2 studies [22, 33] and 7 studies [6, 10, 14-16, 34-37] were excluded involving insufficient data to find a relationship between LDL and SBP levels and risk of DPN, respectively. Hence, in this current meta-analysis, only 29 articles were attempts to find a relationship between SBP levels and DPN in type 2 diabetes mellitus patients.
Fig 1.
Quality assessment
To evaluate the methodological quality of selected studies, we applied the NOS and AHRQ criterion. The detailed quality assessment of eligible studies according to design, enrollment scheme for participants, the credibility of results, assessment of confounding factors and made their individual reports, were summarized individually in Table S1 (Additional file 1: Table S1). Overall, all 38 selected studies in the current meta-analysis were judged to be at moderate to high risk of bias, with scores ≥7 points. The average NOS score was 7.68 out of 10, that was pretty categorized in the high-quality evaluation standards of the Cochrane Reviewers’ Handbook. Furthermore, QUADAS-2 results confirmed that significant bias was not presented in the current meta-analyses (Fig. 2). The reviewers' decisions about each risk of bias and applicability concerns graph presented as percentages across selected studies. More than half of the included studies were rated as low risk for most parameters of the bias risk (49.83%) and applicability concerns (60.54%). In this study, no significant bias (Fig. 2a) and applicability concerns (Fig. 2b) were found in all selected studies.
Fig 2.
Study and patient's characteristics
Characteristics of all relevant studies included in this systematic review and meta-analysis are summarized in Table 1. A total of 355,438 patients were included in these studies, and the median trial sample size was 161,734 patients, between 2004 and 2018. Gender subgroups among 354,088 patients, 168,095 (47.4%) and 185,993 (52.6%) patients were male and female, respectively. Most studies used a cross-sectional study (63.10%) deign for measuring LDL and SBP, respectively. The mean age of the participants was 60.11 ± 10.00 years. The mean duration of diabetes was 6.50 ± 2.80 years; and the mean level of LDL and SBP were 2.82 ± 0.80 mmol/L and 134.81 ± 15.10 mmHg, respectively (Table 1).
Table 1 Demographics information of included studies
The participants were divided into 2 groups: that T2DM without neuropathy (n = 309,197) and patients with DPN without pain (n = 44,891), representing an overall DPN prevalence of 12.67% (Table 2). According to Table 2, a total of 36 studies were included in the analysis of LDL comprising of 354,088 patients (309,197 cases without and 44,891 cases with DPN). Subgroup analyses for LDL were based on the continent were done by diving studies done from Asia (n=29, 80.6%), Europe (n=4, 11.2%), and America (n=3, 8.4%). Based on study type, there were three subgroups: cross-sectional studies (n=22, 61.2%), case-control studies (n=11, 30.6%) and cohort studies (n=3, 8.4%). Analysis for SBP comprised of 22 studies. For SBP, a total number of subjects, in the final analysis, was 25,022 cases including 16,969 cases without and cases 8053 with DPN. For the sake of subgroup analyses of SBP, most of the studies were conducted in people of the Asian race, tracked by 26 studies (72.2), 2 studies (5.6%) in European countries, one study in the USA (2.8%). In our study, there were no data from African populations. Of all the studies, 21 cross-sectional studies (58.4%), 6 case-control studies (16.67%) and 2 cohort studies (5.56%) were attempts to find a relationship between SBP and risk of type 2 diabetic peripheral neuropathy. The analysis of these 36 studies was performed consistently without any studies from African populations.
Table 2 Main clinicopathological characteristics of all relevant studies
Meta-analysis results
Association of serum LDL level with DPN
We collected the mean and SD of LDL in both DPN and non-DPN patients of T2DM (Table 3). The Meta-analysis report on LDL (DPN = 2.78 ± 0.98 and non-DPN = 2.86 ± 0.91 mmol/L) showed that there was no significant difference between two groups’ age-matched participants at a high effect level (SMD = 0.08, 95 % CI: 0.03-0.130.05; p < 0.001). In order to assess the influence of LDL level in the development DPN in T2DM patients, we collected SMD with 95% CIs of LDL level of 36 included studies, after analyzed all the studies; there was obvious heterogeneity in those 36 studies (I2 = 88.10%, Cochran Q-test P < 0.001) (Fig. 3a).
The subgroup's analysis conducted regarding the type of race and study design (Table 2). Fig. 3b reveals none of the above covariates contributed to the heterogeneity (all P > 0.05). As shown in Table 2, the European DPN patients (SMD = 0.16, 95 % CI: -0.06 - 0.38) have higher serum level of LDL in compare with the American and Asian DPN patients (SMD = 0.07, 95 % CI: -0.24 - 0.37 and SMD = 0.07, 95 % CI: 0.01 - 0.12; respectively). Furthermore, subgroup analysis of different study design showed more accuracy of cohort study design for the evaluation of serum LDL in type 2 diabetic peripheral neuropathy (Fig. 3c).
Table 3 Meta-analysis results for assassinate of association between serum LDL level and type 2 diabetic peripheral neuropathy risks
Fig. 3.
Association of SBP level with DPN
We tried to evaluate the difference SBP level between DPN and non-DPN in patients with T2DM in Table 4. Also, the associations between SBP and DPN level in patients with T2DM are shown in Fig. 4. This combined analysis of 29 studies indicates SBP level range with a higher heterogeneity I2 = 84.88% (Fig. 4a). In comparison with the T2DM patients with DPN, the participants without DPN had significantly lower SBP levels (138.45 ± 18.50 mmHg vs 141.08 ± 19.10 mmHg). SBP level was associated with a statistically significant 2.6 fold reduce in non-DPN patients of T2DM when compared to the DPN group (SMD = -2.63, 95% CI: -4.00 - -1.27, P < 0.001; Table 4). Heterogeneity was noticeably decreased after the analysis of study design and race subgroup. There is low heterogeneity between studies which are cross-sectional study design (SMD = -3.06, 95% CI: -4.55 - -1.57, I2 = 83.99%; Fig, 4b) and performed in the Europe population (SMD = -5.61, 95% CI: -12.66 - 1.45, I2 = 78.38; Fig. 4c). Meanwhile, the highest significant SMD of SBP is more precise in the cross-sectional study design model. Thus, SBP may be a high-risk factor for the occurrence of DPN in European diabetic patients.
Table 4 Associations between SBP levels with the type 2 diabetic peripheral neuropathy risk
Fig. 4.
Meta-regression results
Meta-analysis regression was applied to investigate which factors determine heterogeneity among included individual studies in the meta-analysis. Meta-regression finding tried to clear the effects of the age of patients and the year of publication of articles on the average difference between the levels of LDL and SBP in both groups of study: T2DM patients with and without peripheral neuropathy (Fig. 5).
The difference in collision risk of DPN drivers over time
By performing a meta-regression using the publications year, we tried to monitored change in collision risk of DPN drivers over time (Fig 5a and 5b). The collision risk tends to decrease in the last 15 years (2004-2019) compared with the initial studies dealing specifically with this issue. In details, one-year increase in the average year of study, the difference of LDL level between the two groups were reduced 0.02 (B = - 0.02, 95% CI: -0.03 - -0.01, P < 0.001; Fig. 5a) and difference in SBP levels between the two groups was reduced 0.32 (B = -0.32, 95% CI: - 0.50 - - 0.14, P < 0.001; Fig. 5b). By running a meta-regression analysis, we found that there is a significantly decreased collision risk of DPN by drivers over time.
Effect of age on collision risk in DPN
Effect of age on collision risk of DPN between both groups of T2DM with and without peripheral neuropathy is shown in Fig. 5c and 5d, respectively. Meta-regression results show that age had a significant influence on the collision risk in DPN drivers. By dividing the studies depending on their mean age into two groups, a one-unit increase in the average age of the patients, the difference in LDL level between the two groups increased 0.17 unit. (Β = 0.001, 95% CI: -0.004 - 0.006, P = 0.74; Fig. 5c); as well as a one-unit increase in the average age of the patient’s difference in SBP level between the two groups increased 33.0 unit. (Β = 0.33, 95%CI: 0.21 - 0.45, P < 0.001; Fig. 5d). In total, the elderly (47-75 years old) T2DM patients have a higher collision risk of DPN.
Fig. 5.
Publication bias and sensitive analysis
Investigations of publication bias and sensitivity were analyzed in the included literature with Begg's and Egger's regression test. The analysis was carried out by precluding a single study at a time and significant differences between events and hypnotizes (difference of mean LDL and SBP levels) (Fig. 6) [38]. Then, the Trim and Fill test was performed to find the studies missing impact on our results. This test indicates that the addition of three studies does not have any significant effects on our main findings (n = 3, Adjusted Mean Difference = 0.09, 95%CI: 0.04 - 0.14). The shape of funnel plot and Egger’s test provided no statistical evidence for publication bias of the LDL (t = -0.92, P = 0.148, 38 study; Fig. 6a) and SBP (t = 1.11, P = 0.27, 29 study; Fig. 6b). Hence, there is no obvious proof for significant publication bias in our meta-analysis and meta-regression, which implies our stable and credible finding.
Fig. 6.