Association between Lipid Profile and Apolipoproteins with Risk of Diabetic Foot Ulcer: A Systematic Review and Meta-Analysis

Background and Aims Biomarkers are necessary to stratify the risk of diabetic foot ulcers (DFUs). This systematic review and meta-analysis aimed to evaluate the association between the lipid profile and apolipoproteins with the risk of DFU. Methods A systematic search was conducted in PubMed, Scopus, Cochrane Library, and Web of Science among adult patients. Cohort and case-control studies were included. Random-effects models were used for meta-analyses, and the effects were expressed as odds ratio (OR) and their 95% confidence intervals (CIs). We evaluated publication bias through Egger's test and funnel plot. Results A total of 12 cohort studies and 26 case-control studies were included, with 17076 patients. We found that the higher values of total cholesterol (TC), low-density lipoprotein (LDL), triglycerides, and lipoprotein(a) (Lp(a)) were associated with a higher risk of developing DFU (OR: 1.47, OR: 1.47, OR: 1.5, OR: 1.85, respectively). Otherwise, the lower values of HDL were associated with a higher risk of developing DFU (OR: 0.49). Publication bias was not found for associations between TC, HDL, LDL, or TG and the risk of DFU. Conclusions The high values of LDL, TC, TG, and Lp(a) and low values of HDL are associated with a higher risk of developing DFU. Furthermore, we did not find a significant association for VLDL, ApoA1, ApoB, and ApoB/ApoA1 ratio.


Introduction
In 2019, the global prevalence of diabetes was estimated at 9.3%, which translates to 463 million people affected by this disease. It is expected that by 2030, the prevalence will increase to 10.2% and, by 2045, to 10.9% [1]. Among the most important complications of diabetes, the diabetic foot ulcer (DFU) emerges as a growing problem for public health systems since it is a leading cause of hospitalization and amputation in patients with diabetes [2]. Globally, the prevalence of DFU is estimated at 6.3%, and the highest prevalence has been found in North America (13.0%) [3]. Hence, adequate measures are necessary to tackle these data of concern.
ree factors are usually involved in DFU formation: diabetic neuropathy, trauma with secondary infection, and arterial occlusive disease [4]. Patients with diabetes often have an altered lipid profile and apolipoproteins, which contributes to the appearance of these factors [5]. ese alterations are multifactorial; nevertheless, one of the most important alterations involves protein glycosylation, such as low-density lipoprotein (LDL).
is event impedes the recognition of lipoproteins by cell receptors, and as a result, they remain in circulation for a more extended period [6]. Later, lipoproteins would be phagocytized by macrophages, and the synthesis of cholesterol esters will thus increase. When lipoproteins accumulate, they cause the transformation of macrophages into foam cells, which will form the fatty streak and, consequently, initiate the atherosclerotic process [7]. e most feared consequence of patients with DFU is amputation, which occurs in 6 to 43% of cases [8]. Currently, health care aimed at the care of DFU represents an average cost of $ 8,659 per person [9]. However, this pathology is preventable through certain measures such as lifestyle modification and lipid management [10]. Additionally, it is necessary to have tools to identify diabetic patients at higher risk of developing this pathology early, to prioritize surveillance. Biomarkers are valuable tools for these purposes. To date, multiple biomarkers focused on stratifying the risk of developing DFU have been studied, such as creatinine, lipid profile, apolipoproteins, hemoglobin A1c, platelet-to-lymphocyte ratio, and others [11,12]. e lipid profile and apolipoproteins have been associated with complications in cardiovascular and neuropathic diseases, the same ones associated with the development of DFU; given this, several studies have evaluated their association [13]. Although a systematic review was published in 2014, it only included four studies [14], so an update is needed. erefore, we aim to systematize the evidence regarding the association between the lipid profile and apolipoproteins with the risk of DFU.

Methods
is systematic review was registered in the International Prospective Register of Systematic Reviews (PROSPERO) (CD42022308926). In addition, we followed the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) statement [15] (see PRISMA checklist in Supplementary Table S1) and the Cochrane Handbook of Systematic Reviews [16].

Data Source and Search Strategy.
A systematic search on articles evaluating the association between lipid biomarkers and apolipoproteins for the risk of developing DFU was carried out on December 20, 2021, in the following databases: PubMed, Scopus, Cochrane Library, and Web of Science. e search strategy was originally built in PubMed and adapted to the other databases (see Search Strategy in Supplementary Appendix 1). Additionally, we performed a manual search on preprint databases (Research Square and medRxiv) and other databases (China National Knowledge Infrastructure, Wangfang Data, VIP Chinese Science Journals Database, and CINAHL). e Peer Review of Electronic Search Strategies (PRESS) checklist was used to develop the search strategy. No language restrictions were applied. e biomarkers included in this systematic review are the following: total cholesterol (TC), LDL, high-density lipoprotein (HDL), triglycerides (TGs), very-low-density lipoprotein (VLDL), apolipoprotein B (ApoB), apolipoprotein A1 (ApoA1), ApoB/ApoA1 ratio, and lipoprotein(a) (Lp(a)). ese markers were selected due to the lipid profile, as it is mainly comprised of them. Also, these were found in the systematic search of the available evidence for our research question.

Study Selection and Data Extraction.
We included studies that met the following criteria: (1) evaluation of the association between lipid profile and the risk of developing a DFU, (2) case-control and cohort studies, and (3) adult patients (≥18 years). We excluded studies that were (1) case reports, (2) studies carried out in animals, (3) cross-sectional studies, (4) scoping reviews, (5) narrative reviews, (6) systematic reviews, (7) conference abstracts, and (8) letters to editors.
Regarding the selection process, after applying the search strategy to each database, the results were exported to the Rayyan QCRI program [17]. After eliminating duplicate studies, four authors independently analyzed the titles and abstracts of each article (RRU-B, MDM-R, EAA-B, and EAH-B). After identifying the potential literature to be included in the review, two reviewers analyzed the full text of each study independently (VAB-Z and PH-A) and verified that these met the selection criteria in their entirety. After this process, the articles were pooled, and duplicate studies were eliminated. In case of missing information, the authors of the articles were contacted. Finally, a secondary bibliographic search was carried out from the articles read in full text.
ere was consensus among the authors in case of discrepancies about the inclusion/exclusion of an article at any stage of the selection process. We used a data extraction sheet built in Microsoft Excel for the data extraction. e following information was extracted from the selected articles independently by four authors (JRU-B, MDM-R, EAA-B, and EAH-B): number of participants, year of publication, study design, population characteristics, and association or exposure measures.

Evaluation of Study Quality and Publication Bias.
Quality assessment was evaluated with the Newcastle-Ottawa Scale (NOS) [18] independently by two authors (AA-C and AVH). e maximum score was nine stars, and scores greater or equal to 6 were considered studies with low risk of bias (high quality), while studies with less than six stars were considered high risk of bias (low quality). e publication bias assessment was evaluated through funnel plots and the Egger test [19]. A p value >0.1 was considered as indicative of no publication bias.

Data Synthesis and Analysis.
e statistical analysis was performed using Review Manager 5.4 (RevMan 5.4) ( e Cochrane Collaboration, Copenhagen, Denmark) and STATA Release 17.0 (College Station, TX: StataCorp LLC). Odds ratio (OR) and corresponding 95% confidence intervals (CIs) were the only effective measures used. Values expressed as medians and their interquartile ranges (IQRs) were transformed into means and their corresponding standard deviations (SDs) using Hozo's method [20]. Chinn's method was employed to transform standard mean differences (SMDs) to their corresponding natural logarithm of the OR (ln(OR)) and its standard error [21]. e primary outcome analyzed was the risk of developing DFU.
We performed a random-effects meta-analysis of the reported OR in all cases. e heterogeneity of the selected studies was analyzed using the I 2 statistic and Cochran's Q statistic. For the I 2 statistic, p values ≥60% were considered a sign of severe heterogeneity, and for the Cochran's Q test, p values <0.1 were considered a sign of heterogeneity. Subgroup analyses were carried out by study location. Sensitivity analyses were performed using only studies with a low risk of bias. A p value <0.05 was considered statistically significant.

Study Characteristics.
e systematic search yielded 1325 records, and 820 duplicates were removed. According to the eligibility criteria, after excluding articles by title and abstract and assessing their full-text documents, 38 were identified as eligible for this systematic review . A PRISMA flow diagram summarizes the study selection process (Figure 1).

Evaluation of Study Quality.
Regarding the evaluation of the quality of the studies with the NOS, 13 studies were at a low risk of bias (high quality), and the remaining 25 studies were at a high risk of bias (low quality) (Supplementary Table S2).

Association between TC and Risk of DFU.
is association was evaluated in 30 studies (n = 9951). We found that higher values of TC were associated with higher risk of developing DFU (OR: 1.47; 95% CI: 1.09 to 1.97; p < 0.05; I 2 = 88%) ( Figure 2). Due to the severe heterogeneity, a subgroup analysis by countries was performed (Supplementary Figure S1). e Turkish studies subgroup  Figure S2).

Association between HDL and Risk of DFU.
is association was evaluated in 33 studies (n � 10982). We found that lower values of HDL were associated with a higher risk International Journal of Clinical Practice International Journal of Clinical Practice 5   (Figure 3). Due to the severe heterogeneity, a subgroup analysis by countries was performed (Supplementary Figure S3). All subgroups kept the statistically significant association, and just the Turkish studies subgroup did not show a decrease in heterogeneity. In sensitivity analysis, heterogeneity decreased, and the association remained (OR: 0.51; 95% CI: 0.39 to 0.67; p < 0.05; I 2 � 37%) (Supplementary Figure S4).

Association between LDL and Risk of DFU.
is association was evaluated in 31 studies (n � 15570). We found that higher values of LDL were associated with a higher risk of developing DFU (OR: 1.47; 95% CI: 1.08 to 2.01; p < 0.05; I 2 � 92%) (Figure 4). Due to the severe heterogeneity, a subgroup analysis by countries was performed (Supplementary Figure S5). e statistically significant association was lost in Chinese subgroup  Figure S6).

Association between Lp(a) and Risk of DFU.
is association was evaluated in two studies (n � 277). We found that higher values of Lp(a) were associated with a higher risk of     Figure S9).

Association between ApoB and Risk of DFU.
is association was evaluated in three studies (n � 870). However, no statistically significant association was found between ApoB and risk of DFU (OR: 2.48; 95% CI: 0.66 to 9.29; p � 0.18; Figure S10).

Association between VLDL and Risk of DFU.
is association was evaluated in three studies (n � 160). However, no statistically significant association was found between high VLDL values and risk of DFU (OR: 1.38; 95% CI: 0.74 to 2.57; p � 0.31; I 2 � 0%) (Supplementary Figure S13).

Discussion
We found that a low level of HDL and a high level of TG and Lp(a) were associated with the development of DFU. However, there were significant regional variations regarding these associations. Most of the studies showed a high risk of bias. Moreover, in the sensitivity analysis, only the association between HDL and TG with DFU remained statistically significant, with consistently less heterogeneity.
Diabetic  International Journal of Clinical Practice prevalence is higher than 75% and is mainly mixed [60]. DD is linked to insulin resistance and can be an early manifestation prior to developing the diseased [61]. e main lipoprotein quantitative abnormality is the rise of TG. Meanwhile, the main qualitative abnormality of the DD is the increase in the sub-fraction of large VLDL and small and dense LDL particles [60]. DD is associated with a higher risk of cardiovascular events and peripheral vascular disease compared with subjects without diabetes, since it plays a central role in the genesis and progression of atherosclerosis [60][61][62]. Likewise, DD is associated with microvascular complications such as peripheral neuropathy related to the effect of fatty acids on mitochondrial traffic [63]. e deposition of these substances associated with lipid metabolism causes oxidative stress followed by increased expression of pro-inflammatory cytokines and neuronal apoptosis [64].
is neuropathic compromise is even associated with the initial lipid profile of the patient with a recent diagnosis of diabetes. Interestingly, an Anglo-Danish-Dutch study found that baseline waist circumference, body mass index, HDL, and LDL were associated with peripheral neuropathy 13 years after diabetes diagnosis [63]. ese data are relevant to our research because both neuropathy and peripheral vascular disease are involved in the development of DFU [65]. In this sense, various studies found that a low level of HDL and a high level of TG were associated with an increase in diabetic peripheral neuropathy, while LDL did not show any association, which could explain our results [66,67]. In addition, a study in diabetic patients in Iran found that serum Lp(a) level was positively correlated with the development of CVD, neuropathy, and diabetic nephropathy [68]. In particular, in patients with DFU, some studies showed that a low TG level was an independent risk factor for amputation of the affected limbo [69]. Similarly, another study showed that a low HDL level was associated with a lower incidence of lower limb amputation and death related to DFU [70].
is aspect is relevant because the participants evaluated in the studies included in our review include patients with some degree of ulceration, that is, with a higher risk of amputation.
A previously published systematic review only included four studies [14]. However, unlike our study, it did not include the evaluation of VLDL, Lp(a), or apolipoproteins. On the other hand, the search for articles was based on only two databases. Probably due to the small number of articles   found, since the search lasted until September 2013, and the small sample in each of them, they did not perform a sensitivity analysis as in our case [14]. Although our results show that low HDL and high TG and Lp(a) levels can be used to predict the development of DFU, regional variations must be considered. ere is a possibility that part of the explanation is due to variation in DD prevalence; for instance, DD prevalence is higher in the South Asian population [60]. Additionally, it is possible that regional variations in lipid profile could explain these differences. Although in general, the prevalence of total cholesterol is lower in Southwest Asian countries than in European and North or South American countries, their variations depend on the income level of the countries [71,72]. Although in high-income countries the mean plasma cholesterol value decreased, in low-and middle-income countries it increased alongside with TG values [73]. Likewise, these values can vary between regions of large countries such as China [71].
Consequently, there are also regional variations in the use of statins for cholesterol management whose effects, in addition to lipid-lowering, could explain our findings. Indeed, statins are associated with cholesterol-independent effects by modulating the immune response, decreasing DD-associated oxidative stress, and stimulating fracture and wound healing [74]. In this sense, in addition to preventing macrovascular diseases, statins would also slow the progression of microvascular complications of diabetes by improving the ability of endothelial nitric oxide synthase to generate nitric oxide in endothelial cells regardless of lipid-lowering effects. Also, in addition to lowering lipid levels, statins can improve endothelial function and reduce oxidative stress, in turn, improving microvascular function [74,75]. us, their use is increased in high-income countries [76], but in contrast, only a small percentage of patients in low-income countries have access to statins [77]. Although we have not evaluated the effect of nutritional status on our results, its influence on the lipid profile is wellknown [78]. In this sense, as well as the regional variations of the lipid profile and the use of statins, it is possible that the regional variations of obesity and malnutrition explain the variations between countries in our findings [79].
However, aspects such as the possible effect of gender of the patients included in the studies have not been considered. For example, a study in Chinese diabetic patients   International Journal of Clinical Practice showed that the ApoB/ApoA1 ratio predicted cardiovascular disease risk in men. In contrast, TC, LDL, ApoB, LDL/HDL, and TC/HDL were better predictors in women [80]. In addition, a systematic review found that there is a correlation between poor nutritional status and the presence of ulceration or delayed healing in patients with DFU [81]. Despite these biases, our results show the association of commonly used laboratory values with public health implications in this disease.

Limitations.
Our study has several limitations. First, the studies did not adjust the values of the lipid markers with the confounding variables affecting the outcome under study. To avoid confounding bias, lifestyles, sociodemographic, and comorbidity variables should be considered and adjusted in future studies. Second, most of the studies have been carried out in the Asian continent, and very few have been found in other continents. us, it would be essential to analyze the value of these biomarkers in the risk of DFU in other latitudes. ird, high statistical heterogeneity was found due to the methodological and clinical differences between the studies. However, heterogeneity decreased when the sensitivity analysis was performed, which only included a low risk of bias studies. Finally, due to the lack of information in the studies, the values of sensitivity, specificity, and an optimal cutoff point of lipids for the risk of DFU were not estimated in a meta-analysis, so it should be addressed in future research.

Conclusions
High values of LDL, TC, TG, and Lp(a) and low values of HDL are associated with a higher risk of developing DFU. Furthermore, we did not find a statistically significant association for VLDL, ApoA1, ApoB, and ApoB/ApoA1 ratio. However, primary studies are needed to define the optimal cutoff point for these biomarkers according to the profile of different diabetic populations worldwide.
Data Availability e data that support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e authors declare that there are no conflicts of interest regarding the publication of this study.

Authors' Contributions
All authors have accepted responsibility for the entire content of this manuscript and approved its submission.

Supplementary Materials
Appendix 1: Search strategy. Table S1. PRISMA checklist. Table S2. Newcastle-Ottawa Quality Assessment scale for included studies. Figure S1. Subgroup analysis according to the countries of the association between TC and risk of DFU. Figure S2. Sensitivity analysis according to the risk of bias of the association between TC and risk of DFU. Figure S3. Subgroup analysis according to countries of the association between HDL and risk of DFU. Figure S4. Sensitivity analysis according to the risk of bias of the association between HDL and risk of DFU. Figure S5. Subgroup analysis according to countries of the association between LDL and risk of DFU. Figure S6. Sensitivity analysis according to the risk of bias of the association between LDL and risk of DFU. Figure S7. Subgroup analysis according to countries of the association between TG and risk of DFU. Figure S8. Sensitivity analysis according to the risk of bias of the association between TG and risk of DFU. Figure S9. Association between Lp(a) and risk of DFU. Figure S10. Association between ApoB and risk of DFU. Figure S11. Association between ApoA1 and risk of DFU. Figure S12. Association between ApoB/ApoA1 ratio and risk of DFU. Figure S13. Association between VLDL and risk of DFU. Figure S14. A funnel plot of the studies that evaluated the association between TC and risk of DFU. Figure S14. B Egger test of all the studies that evaluated the association between TC and risk of DFU. Figure S15. A Funnel plot of the studies that evaluated the association between HDL and risk of DFU. Figure S15. B Egger test of all the studies that evaluated the association between HDL and risk of DFU. Figure S16. A Funnel plot of the studies that evaluated the association between LDL and risk of DFU. Figure S16. B Egger test of all the studies that evaluated the association between LDL and risk of DFU. Figure S17. A Funnel plot of the studies that evaluated the association between TG and risk of DFU. Figure S17. B Egger test of all the studies that evaluated the association between TG and risk of DFU. . (Supplementary Materials)