Malondialdehyde levels in diabetic retinopathy patients: a systematic review and meta-analysis : Chinese Medical Journal

Secondary Logo

Journal Logo

Meta Analysis

Malondialdehyde levels in diabetic retinopathy patients: a systematic review and meta-analysis

Jiang, Fanwen; Zhou, Lin; Zhang, Chun; Jiang, Hui; Xu, Zhuping

Editor(s): Li, Jinjiao; Ji, Yuanyuan

Author Information
Chinese Medical Journal 136(11):p 1311-1321, June 5, 2023. | DOI: 10.1097/CM9.0000000000002620
  • Open



Diabetic retinopathy (DR) is a progressive asymptomatic microvascular complication of diabetes that triggers irreversible retinal damage. It remains a major cause of sight-loss among the working-age populations of industrialized countries.[1] The prevalence of DR has continued to increase in many areas, such as East Asia, high-income America, Oceania, and southern Africa, while other diseases that cause blindness have decreased.[2] A global systematic review showed that there were approximately 103.12 million adults with DR worldwide in 2020, and the number would be 160.50 million by 2045.[3] As in China, we also are facing a challenge by the obesity pandemic.[4] Considering the close relationship between obesity and diabetes, it makes the research on diabetes and its complication such as DR in urgent need.

The pathophysiological mechanisms underlying the development of DR are still controversial. Currently, we believe that the expose of retina to hyperglycemia and other causal risk factors initiate a cascade of biochemical and physiological changes,[5] which finally bring about microvascular damage and retinal dysfunction.[6] There are several pathways of biochemical mechanisms modulating the pathogenesis of retinopathy: the accumulation of sorbitol and advanced glycation end-products, oxidative stress, protein kinase C activation, inflammation, and upregulation of the renin-angiotensin system and vascular endothelial growth factor, which are interrelated with each other.[7] Lipid peroxidation is one of the most important components of the response to oxidative stress.

Malondialdehyde (MDA) is one of the most widely used biomarkers for evaluating oxidative stress, as a secondary product of lipid peroxidation. Many animal trials have also supported the idea that higher circulating MDA levels could act as a predictor of DR.[8–10] Many studies have suggested that increased circulating MDA levels may be a risk factor in people with DR.[11–39] However, it is still unclear whether circulating MDA levels play a role in the evolution of DR. To clarify this relationship, we designed a meta-analysis to critically examine MDA levels in people with DR compared to people with diabetic mellitus (DM) but not DR.


Literature research strategies

We conducted this systematic review on MDA levels in people with DR according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses systematic review protocols. Several electronic databases were used, including PubMed, Medline (Ovid), Embase (Ovid), and Web of Science. The following MeSH search terms were used: (“malondialdehyde” or “thiobarbituric acid reactive substances [TBARS]” or “lipid peroxidation” or “oxidative stress”) and “diabetic retinopathy.” All English studies were conducted before May 2022, with no specified time span.

Criteria for inclusion and exclusion

The inclusion criteria for this study were as follows: (a) the study should be published in a journal in English; (b) the assessment of MDA levels (method description is provided for MDA estimation with mean and standard deviation [SD] values or any other data to calculate the mean and SD) should be feasible from the sample matrix, which should be available for both people with DR and DM subjects; (c) the study should be a case-control study design; and (d) we could obtain the full texts of the reported studies (not reviews, abstracts, posters, protocols, letters, comments, or editorial papers). The exclusion criteria were as follows: (a) they did not contain clear original data (efforts have been made to obtain data); (b) the subjects had a history of other diseases; (c) before MDA measurement, people received other treatments; (d) studies that do not provide information concerning clear clinical DR diagnoses.

Data extraction and management

We imported all obtained studies into an EndNote 20 library (EndNote™ 20, Camelot UK Bidco Limited, Clarivate, UK) and removed duplicate studies. Two authors (JFW and ZL) independently screened eligible articles based on their abstracts and titles. If relevant, two investigators independently reviewed all the full articles. Furthermore, we manually reviewed all bibliographies of the published articles to define additional studies and identify one additional record. Both authors independently utilized standard extraction spreadsheets to extract data from the selected articles and listed them in a table. After the extraction, we cross-checked the data tables and discussed options to resolve conflicts and inconsistencies.

Quality assessment

We used the Newcastle–Ottawa Quality Assessment Scale (NOS) for case-control studies to evaluate the quality of the included studies. The NOS provides three main parts in nine points for each study: four for selection, two for comparability, and three for exposure. Two authors independently assessed all the included studies and discussed them to resolve discrepancies. NOS scores of 1 to 3 indicate low quality, 4 to 6 indicate moderate quality, and 7 to 9 indicate high quality.[40]

Statistical analysis

Statistical software named STATA 16 (Stata Corp, College Station, TX, USA) and Review Manager V5.4 (Cochrane Collaboration, Copenhagen, Denmark) were used for our meta-analysis. We calculated the standardized mean difference (SMD) with the corresponding 95% confidence interval (CI) for each parameter using the random-effects model (DerSimonian–Laird method) when I2 was >50%. Statistical significance was set at P < 0.05. The overall effect size for SMD was presented using a Z-test (calculating each mean value and SD). We used I2 to estimate the existence of heterogeneity and used chi-squared tests to examine the resultant P values. I2 values of 25%, 50%, and 75% were defined as low, moderate, and high heterogeneity, respectively. Furthermore, forest plots adopted as the pooling method were used to evaluate the differences in MDA levels between people with DM but not DR and people with DR. If I2 showed high heterogeneity, a subgroup analysis was performed to explore the source of heterogeneity. We also used a sensitivity analysis to assess robustness by omitting each study. To investigate publication bias, we visually evaluated the asymmetry of the funnel plot and used Egger's and Begg's biases. We used the trim-and-fill method to import potentially missing studies if publication bias was suspected.

Two authors (JFW and ZL) independently evaluated the quality of pooled evidence at the outcome level using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework. Any conflict was resolved by mutual consensus. Our study was conducted following the Meta-analysis of Observational Studies in Epidemiology guidelines. This study was registered in PROSPERO (; No. CRD42022352640).


Literature research and study characteristics

As shown in Figure 1, we searched four databases and found 7357 records, and also identified 1 record from searching citation. After reviewing the titles and abstracts, 7276 were excluded because they were duplicates, unrelated, reviews, animal studies, or meeting abstracts. Subsequently, 49 articles were excluded after reading their full text. In total, 29 case-control studies were enrolled in our meta-analysis.[11–39] A total of 1680 people with DR and 1799 people with diabetes but not DR were included in our meta-analysis. The basic characteristics and NOS scores of the studies are presented in Table 1.

Figure 1:
Flowchart for study selection. DR: Diabetic retinopathy; MDA: Malondialdehyde.
Table 1 - Characteristics of the included studies.
DM without DR DR
Study Country Samp le Matrix Assay Type NOS score/9 Continent MDA unit n Mean Age Sex (Male/Female) MDA levels
Mean ± SD
DM type n Mean Age Sex (Male /Female) MDA levels
Mean ± SD
DR type Significant results of MDA levels
Dave et al [11] India Serum Colorimetric 7 Asia nmol/mL 44 52.05 46/34 (N DR + DR) 3.655 ± 2.278 T2DM 4PDR 32NPDR PDR: 57.00 NPDR: 56.13 NR PDR: 3.820 ± 1.965; NPDR: 2.919 ± 1.142 PDR and NPDR It was not different in patients with and without DR, and thus it cannot be used as a marker for prediction of development of DR.
Mondal et al [12] India Serum Colorimetric 9 Asia nmol/mL 100 NR 55/45 2.61 ± 1.10 T2DM 42 NR NR 3.95 ± 1.40 PDR In this study, the estimation of MDA demonstrated significantly higher levels in PDR participants as compared to diabetics with no retinopathy.
Sanz-González et al [13] Spain Plasma Spectrophoto metry 8 Europe μmol/L 52 65.0 46%/54% 3.0 ± 0.2 T2DM 69 NR NR 3.7 ± 0.2 NPDR Furthermore, significantly higher plasma levels of MDA/TBARS were observed in the T2DM + DR with respect to the NDR patients.
Verma et al [14] India Serum Colorimetric 5 Asia μmol/L 54 NR NR 2.12 ± 1.55 T2DM 54 NR NR 4.25 ± 1.03 NR MDA was highly significantly (P = 0.0001) higher among cases (4.25 ± 1.03) than controls (2.12 ± 1.55).
Fahmy et al [15] Saudi Arabia Serum Spectrophoto metry 9 Asia μmol/mL 12 52.41 NR 0.08 ± 0.0320 T2DM 12 46.83 NR 0.08 ± 0.0157 NR GST and lipid peroxides did not show any significant difference between the three studied groups.
Dai et al [16] China Serum Colorimetric 9 Asia μmol/L 52 53.21 29/23 11.48 ± 3.19 T2DM 52 52.64 27/25 17.80 ± 4.34 PDR The MDA levels in the NDR group were significantly lower than the DR group, with statistical significance (P < 0.05).
Kuppan et al [18] India Plasma Spectrophoto metry 7 Asia μmol/L 22 51 14/8 72.37 ± 44.15 T2DM 22PDR
PDR: 52 NPDR: 58 PDR: 19/3; NPDR: 15/6 PDR: 115.34 ± 114.18; NPDR: 98.79 ± 53.99 PDR and NPDR There was a significant increase in the levels of plasma MDA in the DR cases compared to that of control.
Roig-Revert et al [19] Spain Plasma Spectrophoto metry 8 Europe μm/L 68 62.3 51.4%/48.6% 2.37 ± 1.39 T2DM 62 65.1 47.3%/52.7% 3.63 ± 1.30 NR The MDA/TBARS displayed significantly higher in the T2DMG + DR with respect to those diabetics without DR and the CG.
Shawki et al [20] Egypt Serum Colorimetric 7 Africa μmol/L 40 45.4 14/26 12.00 ± 4.99 NR 70 43.0 30/40 18.60 ± 3.67 NR The levels of serum MDA were significantly higher in type 2 diabetics than healthy controls.
Choudhuri et al [21] India Serum ELISA 8 Asia Pmol/mL 100 49.4 54/46 118.3 ± 35.6 T2DM 100 51.8 56/44 134.7 ± 36.3 NPDR Further, subjects with NPDR showed higher levels of serum MDA protein adduct than the NDR group, and the difference was statistically significant (P = 0.0009).
Vivian Samuel et al [22] India Serum Spectrophoto metry 6 Asia μmol/L 30 50.6 NR 4.63 ± 0.58 T2DM 30 49.5 NR 6.72 ± 0.20 DR and DME The levels of serum MDA levels were significantly higher compared with those of the control group (P < 0.001). The levels of serum MDA of the group with retinopathy were significantly higher (P < 0.001) than the group without retinopathy.
Aldebasi et al [17] Saudi Arabia. Plasma spectrophoto metry 8 Asia μmol/L 33 51.82 14/19 2.79 ± 0.5745 T2DM 21 60.38 8/13 3.43 ± 0.4583 PDR Moreover, the lipid peroxidation marker, MDA, was significantly higher (P = 0) in diabetics with proliferative retinopathy than in subjects in control groups. Data showed that, when compared to diabetics without retinopathy, patients with PDR were characterized by significantly higher MDA (P = 0).
Kurtul et al [23] Turkey Leuk ocytes Spectrophoto metry 5 Asia nmol/mg protein 16 59.00 NR 3.95 ± 0.98 T2DM PDR: 10; BDR (NPDR): 7; prePDR
8; T2DR: 25
PDR 57.00 BDR (NPDR): 55.85; prePDR: 54.25; T2DR: 55.80 NR PDR: 5.54 ± 1.72; BDR (NPDR): 4.19 ± 1.11; prePDR: 4.30 ± 0.76; T2DR: 4.76 ± 1.42 BDR (NPDR); prePDR; PDR; T2DR MDA concentrations rose with increasing severity of DR. However, MDA levels and SOD and CAT activities were not different in type 2 diabetic patients with retinopathy compared to those without retinopathy.
Gupta et al [24] India Blood samp les Condense with 1 methyl
2 phenyl indole
8 Asia μmol/L 40 50 27/13 1.72 ± 0.27 NR PDR: 22; NPDR: 20 PDR: 46; NPDR: 47 PDR: 10/ 12; NPDR: 9/ 11 PDR: 1.93 ± 0.25; NPDR: 1.98 ± 0.20 PDR and NPDR The levels of MDA (P < 0.001 and P < 0.05) were significantly increased in diabetics with proliferative retinopathy when compared with the controls and the diabetic group without complications. The levels of MDA (P < 0.001 and P < 0.01) were significantly increased in diabetics with non-proliferative retinopathy when compared with the control and the diabetic group without complications.
Losada and Alio [25] Spain Plasma Spectrophoto metry 6 Europe mmol/L 28 26.92 12/16 1.80 ± 0.81 T1DM 32 38.81 17/25 2.65 ± 1.00 NR We have found that the diabetics with retinopathy show MDA levels that are significantly higher than in the control group and in the diabetics without retinopathy (P < 0.001). We have not found any significant statistical differences among the other groups (P > 0.05).
Sharma et al [26] India Plasma Spectrophoto metry 7 Asia μmol/L 20 52.4 12/8 2.56 ± 4.1452 T2DM PDR: 20 NPDR: 20 PDR: 53.5 NPDR: 56.2 PDR: 13/ 7 NPDR: 13/7 PDR: 3.01 ± 2.8632 NPDR: 3.1100 ± 3.8674 PDR and NPDR Analysis revealed that increased levels of plasma LPO were associated with increased severity of DR.
Kundu et al [27] India Plasma Spectrophoto metry 9 Asia μmol/L 50 56.20 39/11 6.44 ± 1.53 T2DM 50 58.56 36/14 6.65 ± 0.30 NR We found increased lipid peroxidation in terms of MDA in diabetics compared with controls and this increase was larger in DR.
Vidya et al [28] India Plasma Spectrophoto metry 6 Asia nmol/dL 25 48.6 NR 473.01 ± 111.65 T2DM 50 53.6 NR 1041.90 ± 664.20 NR The levels of MDA were significantly higher in the diabetics without retinopathy as compared to those of the control group (P < 0.001). The levels of MDA were significantly higher in the diabetics with retinopathy, as compared to those in the diabetes without retinopathy group (P < 0.001).
Khalili et al [29] Iran Plasma Spectrophoto metry 8 Asia μmol/L 100 55.9 28/72 6.6 ± 1.9 T2DM 100 58.5 35/65 6.7 ± 2.0 NR A significantly higher level of MDA (6.7 ± 2.0 nmol/mL) was observed in patients with retinopathy compared to the control group.
Mandal et al [30] India Serum and vitre ous Spectrophoto metry 8 Aisa nmol/L 32 49.5 22/10 Serum: 3.10 ± 1.83
Vitreous: 1.21 ± 0.32
T2DM 45 51.7 26/19 Serum: 4.65 ± 1.79
Vitreous: 1.65 ± 0.67
PDR A noticeable increase in serum and vitreous MDA level was observed among PDR subjects compared with NDR (P = 0.0004 and P = 0.0058, respectively) and HC individuals (P < 0.0001 and P = 0.0003, respectively). Yet again, the serum MDA level was significantly low in healthy individuals even compared with NDR subjects (P = 0.027).
Gaonkar et al [31] India Plasma Colorimetry 8 Asia μmol/L 148 59.3 NR 0.55 ± 0.08 T2DM PDR: 74
NPDR: 148
PDR: 62.0 NPDR: 58.6 NR PDR: 0.71 ± 0.12 NPDR: 0.60 ± 0.12 PDR and NPDR A significantly higher plasma MDA concentration was observed in NPDR and PDR groups.
Kumari et al [32] India Plasma Spectrophoto metry 8 Asia μmol/L 36 55.30 NR 1.67 ± 0.14 T2DM 50 63.04 NR 2.02 ± 0.14 NR Lipid peroxidation marker MDA was significantly elevated (P < 0.001) in both the diabetic groups as compared to controls.
Turk et al [33] Turkey Serum Spectrophoto metry 7 Asia μmol/L 35 55.74 13/22 0.325 ± 1.4268 T2DM 35 60.00 15/20 0.244 ± 1.3032 NR The MDA levels of patients with DR were lower than those without DR. In ROC analysis conducted for MDA levels, we observed that MDA did not reflect DR at a sufficient level of specificity and sensitivity.
Longo-Mbenza et al [34] South Africa Serum HPLC 8 Africa mmol/L 84 56.6 39/45 9.1 ± 5.2 T2DM 66 53.4 26/40 8.7 ± 2.5 NR The highest levels of TBARS among the present T2DM group with DR confirmed the significant impact of the imbalance of oxidant/antioxidant status in the pathophysiology of DR from several studies.
Gürler et al [35] Turkey Serum Spectrophoto metry 9 Asia μmol/L 34 51.79 21/13 3.38 ± 0.95 NR 25 54.60 17/8 4.38 ± 1.31 NR LPO assessment of the groups suggested that all subjects with DM showed an increased LPO, and there were statistically significant differences between the DR and NDR groups.
El-Mesallamy et al [36] Egypt Serum Spectrophoto metry 6 Africa μmol/L 23 54.65/57.92 11/12 5.28 ± 1.6306 T2DM PDR: 22 NPDR: 22 PDR: 59.65/5 9.13 NPDR: 54.11/58.82 PDR: 14/ 8 NPDR: 9/ 11 PDR: 6.81 ± 2.2983 NPDR: 5.57 ± 2.0169 PDR and NPDR MDA levels in PDR are significantly different from the type II DM group at P ≤0.03, and PDR is significantly different from the NPDR group at P ≤0.03.
Hartnett et al [37] USA Serum HPLC 9 America mmol/L 18 NR NR 2.76 ± 0.89 T1DM and T2DM Advance d risk: 23; Moderat e risk: 31 NR NR Advanced risk: 2.86 ± 1.45; Moderate risk: 2.33 ± 1.19 (1) low risk (no retinopathy), (2) moderate risk (mild and moderate NPDR), or (3) advanced risk (severe NPDR, very severe NPDR, and all PDR). We found that levels of TBARS were significantly elevated in patients with diabetes compared with those in control subjects, which supports the findings by others. We were unable to support the hypothesis of an association between severity of DR and TBARS as a measure of oxidative stress.
Icel et al [38] Turkey Serum Spectrophoto metry 7 Asia μmol/L 39 57.56 12/27 2.53 ± 1.13 T2DM PDR: 55 NPDR: 42 PDR: 60.00 NPDR: 59.50 PDR: 22/ 33 NPDR: 20/22 PDR: 5.79 ± 1.16; NPDR: 2.74 ± 1.63 PDR and NPDR There is a significant increase in MDA levels with an advance in DR.
Kumari et al [39] India Plasma Spectrophoto metry 8 Asia μmol/L 30 56.30 NR 1.69 ± 0.15 T2DM 42 64.07 NR 2.02 ± 0.14 NR MDA registered a higher value in DR than the control group.
TBARS: Thiobarbituric acid reactive substances; T1DM: Type 1 diabetic mellitus; T2DM: Type 2 diabetic mellitus; MDA: Malondialdehyde; NDR: Diabetic mellitus without diabetic retinopathy; PDR: Proliferative diabetic retinopathy; NPDR: Non-proliferative diabetic retinopathy; CG: Control group; HC: Healthy control; ELISA: Enzyme-linked immunosorbent assay; HPLC: High performance liquid chromatography; GST: Glutathione S-transferase; DME: Diabetic macular edema; BDR: Background diabetic retinopathy; prePDR: Preproliferative diabetic retinopathy; T2DR: Type 2 diabetic retinopathy; SOD: Superoxide dismutase; CAT: Catalase; LPO: Lipid peroxidation; ROC: Receiver operating characteristic; NR: Not reported; DR: Diabetic retinopathy; DM: Diabetic mellitus; NOS: Newcastle–Ottawa Quality Assessment Scale; SD: Standard deviation.

Overall and subgroup analysis

The forest plot for MDA concentrations in people with DR and diabetic mellitus without diabetic retinopathy (NDR) from the random-effects meta-analysis that combined SMD is presented in Figure 2. MDA levels were significantly higher in the DR group than in the NDR group (SMD, 0.897; 95% CI, 0.631−1.162), but we observed an obvious heterogeneity among the 29 studies (I2 = 92.03%). Subgroup and meta-regression analyses were performed to analyze the source of heterogeneity. Subgroup analysis was conducted by year of publication, sex, age, duration of DM, study size, NOS score, continent, MDA assay type, MDA sample matrix, MDA absorption, and DR type. The results of almost all subgroups showed that MDA levels in the DR group were significantly higher than those in the NDR group [Table 2]. Meta-regression results indicated that the type of MDA assay and MDA absorption spectrum might slightly contribute to the heterogeneity (R2 = 7.50% and R2 = 14.82%, respectively).

Figure 2:
Forest plot of studies examining MDA level and DR. CI: Confidence interval; DR: Diabetic retinopathy; MDA: Malondialdehyde; SMD: Standardized mean difference.
Table 2 - Subgroup meta-analysis and meta-regression on the relationship between MDA levels and risk of DR.
MDA levels in DR and NDR
Subgroups No. of study SMD (95% CI) I 2 (%) P for heterogeneity P for Meta- regression P for Group differences
Overall 29 0.897 (0.631, 1.162) 92.03 <0.001
Year of publication 0.414 0.511
 ≤2010 7 1.066 (0.450, 1.682) 90.82 <0.001
 >2010 22 0.836 (0.537, 1.136) 92.59 <0.001
Gender 0.668 0.696
 Male/Female≥1 10 0.814 (0.267,1.360) 94.24 <0.001
 Male/Female<1 13 0.693 (0.431, 0.955) 81.23 <0.001
Age (years) 0.847 0.816
 ≥55 12 0.857 (0.467, 1.247) 92.77 <0.001
 <55 15 0.925 (0.503, 1.348) 92.10 <0.001
Duration of diabetes mellitus 0.239 0.905
 >10 13 0.819 (0.393, 1.245) 87.66 <0.001
 ≤10 9 0.853 (0.493, 1.213) 92.97 <0.001
Study size (total number) 0.999 <0.001
  (51,100] 16 0.810 (0.415, 1.204) 90.86 <0.001
  (101,150] 8 1.385 (0.936, 1.834) 89.98 <0.001
  (151,200] 3 0.144 (−0.179, 0.466) 72.44 0.027
NOS score 0.602 0.267
  5 2 0.872 (0.287, 1.457) 71.06 0.008
  6 4 1.474 (0.378, 2.570) 93.80 <0.001
  7 6 0.555 (−0.056, 1.165) 91.92 <0.001
  8 11 1.154 (0.703, 1.606) 94.73 <0.001
  9 6 0.530 (−0.016,1.162) 87.71 <0.001
Continent 0.698 0.424§
 Asia 22 0.948 (0.643, 1.253) 92.25 <0.001
 Africa 3 1.439 (0.496, 2.382) 91.25 <0.001
 America 1 −0.167 (−0.632, 0.298) NA NA
 Europe 3 0.599 (−0.234, 1.432) 92.08 <0.001
MDA sample matrix 0.440 0.439
 Serum 15 0.820 (0.396, 1.245) 93.13 <0.001
 Plasma 12 1.065 (0.611, 1.520) 93.93 <0.001
Types of DR 0.590 0.593
 PDR 12 0.989 (0.494, 1.485) 91.26 <0.001
 NPDR 12 0.803 (0.336, 1.270) 92.22 <0.001
MDA assay type 0.054 0.998||
 Spectrophotometry 19 1.004 (0.629, 1.378) 92.03 <0.001
 Colorimetric 6 1.009 (0.484, 1.533) 92.80 <0.001
 HPLC 2 −0.122 (−0.379,0.135) 0 0.530
 1-methyl-2-phenyl-indole 1 0.914 (0.523, 1.305) NA NA
 ELISA 1 0.456 (0.175, 0.737) NA NA
MDA absorption spectrum 0.057 <0.001
 400nm 2 −0.122 (−0.379, 0.135) 0 0.530
 530nm 2 1.710 (0.082, 3.338) 96.83 <0.001
 532nm 5 0.516 (0.183, 0.850) 57.64 0.015
 535nm 6 1.186 (0.360, 2.012) 93.35 <0.001
As some of the articles did not clearly indicate the relevant information, only articles with clear relevant information were included in the subgroup analysis.
Male/Female≥1: The ratio of the number of males to the number of females is greater than or equal to 1; Male/Female<1: The ratio of the number of males to the number of females is less than 1.
Significant differences.
§Only compared group African, group Asian, and group European for a sufficient number of studies.
||Only compared group spectrophotometry and group colorimetric for a sufficient number of studies.
Only compared group 532 nm and group 535 nm for a sufficient number of studies. MDA: Malondialdehyde; CI: Confidence interval; DR: Diabetic retinopathy; NDR: Diabetic mellitus without diabetic retinopathy; ELISA: Enzyme-linked immunosorbent assay; HPLC: High performance liquid chromatography; ELISA: Enzyme linked immunosorbent assay; SMD: Standard mean difference; PDR: Proliferative diabetic retinopathy; NPDR: Non-proliferative diabetic retinopathy; NA: Not available; NOS: Newcastle–Ottawa Quality Assessment Scale.

Sensitivity analysis

To assess the stability and reliability of our results, we performed a sensitivity analysis. SMD was not affected by the removal of each study from the pooled analysis [Table 3].

Table 3 - Sensitivity analysis.
95% CI
Study SMD Lower CI limit Higher CI limit
Turk et al [33] 0.921 0.653 1.189
Dave et al [11] 0.929 0.664 1.194
Dave et al [11] 0.912 0.645 1.180
Gürler et al [35] 0.896 0.626 1.167
Longo-Mbenza et al [34] 0.923 0.656 1.190
Gaonkar et al [31] 0.874 0.608 1.140
Gaonkar et al [31] 0.908 0.629 1.187
Vidya et al [28] 0.892 0.622 1.163
Kundu et al [27] 0.915 0.644 1.185
Icel et al [38] 0.916 0.646 1.185
Icel et al [38] 0.846 0.590 1.103
Khalili et al [29] 0.919 0.649 1.188
Shawki et al [20] 0.878 0.610 1.146
El-Mesallamy et al [36] 0.914 0.645 1.183
El-Mesallamy et al [36] 0.899 0.629 1.169
Kuppan et al [18] 0.906 0.636 1.176
Kuppan et al [18] 0.905 0.635 1.175
Mandal et al [30] 0.899 0.627 1.170
Mandal et al [30] 0.897 0.626 1.168
Mondal et al [12] 0.891 0.619 1.163
Dai et al [16] 0.876 0.608 1.144
Hartnett et al [37] 0.916 0.647 1.184
Hartnett et al [37] 0.928 0.662 1.194
Losada and Alio [25] 0.895 0.625 1.166
Madhur [24] 0.899 0.628 1.169
Madhur [24] 0.892 0.622 1.162
Verma et al [14] 0.877 0.609 1.145
Roig-Revert et al [19] 0.895 0.622 1.169
Kurtul et al [23] 0.889 0.621 1.157
Kurtul et al [23] 0.902 0.633 1.172
Kurtul et al [23] 0.907 0.639 1.176
Kurtul et al [23] 0.910 0.642 1.178
Fahmy et al [15] 0.916 0.648 1.184
Sharma et al [26] 0.914 0.646 1.183
Sharma et al [26] 0.915 0.646 1.184
Sanz-González et al [13] 0.829 0.583 1.074
Kumari et al [32] 0.855 0.594 1.116
Choudhuri et al [21] 0.909 0.633 1.184
Kumari et al [39] 0.861 0.598 1.125
Vivian Samuel et al [22] 0.818 0.566 1.069
Aldebasi et al [17] 0.888 0.619 1.158
Overall 0.896 0.631 1.160
CI: Confidence interval; SMD: Standardized mean difference.

Publication bias

From the Begg and Egger tests, we found no publication bias in our study (P = 0.568 and P = 1.000, respectively). Visual examination of the funnel plot revealed publication bias on the right side. Therefore, trim-and-fill analysis was performed, and it was confirmed that the contour-enhanced funnel plot after trim-and-fill showed that all inputted studies were located in the statistically significant area, which indicated that visual asymmetry would have some extra reasons in addition to publication bias.

GRADE quality of evidence

Using the GRADE framework, we judged the overall quality of evidence for our outcome to be moderate.


This meta-analysis assessed the association of MDA levels with DR. The results showed that circulating MDA levels were significantly higher in people with DR than in people with DM but not DR. Observing the extreme heterogeneity in our study, we should explain the results with prudence.

The complexity and diversity of mechanisms make it important to identify more biomarkers to stratify the risk of DR. MDA, as a popular biomarker of oxidative stress with many available measurement methods, was chosen to be a potential predictor of DR.

When oxidative stress is not controlled, the hydroxyl radical oxidizes lipids that contain carbon–carbon double bonds, particularly polyunsaturated fatty acids (PUFAs)[41] and produces various bioactive aldehydes with toxicity. The secondary products of lipid peroxidation, MDA, 4-Hydroxynonenal (4-HNE), and acrolein, can adduct with cellular proteins, and protein carbonylation participates in the development of diabetes by activating the other pathways or factors and causes harm to the structure and function of the retina.[42] The uncontrolled oxidative stress would account for the increased MDA levels in people with DR compared to people with DM but not DR.

We have found evidence that circulating MDA might have some combination with sex, age, and diabetes duration.[43–45] However, in our subgroup analysis of sex, age, and duration of DM, we found no significant differences among subgroups. We supposed that it is the limit on the number of studies that resulted in us being able to collect only the data of the mean of ages, the proportion of sex, and the mean of diabetes duration of each study. More specific data and more meticulous divided subgroups might lead to more reliable results.

In the field of statistics, we considered that the publication year, the study location, and the study size might affect our results, but we only found significant differences between subgroups in analyzing the number of people included in the studies (study size). Furthermore, heterogeneity slightly decreased when divided into different groups by study size (changed to 90.86%, 89.98%, and 72.44%). We also compared different studies on NOS scores to evaluate whether the quality of the studies is relevant. In addition, there were no significant differences between the groups. These results suggest that statistical differences between our involved studies have slight effects on the extreme heterogeneity mainly because of study size.

Based on the absence or presence of neovascularization, people with DR are classified as having non-proliferative diabetic retinopathy (NPDR) or proliferative diabetic retinopathy (PDR).[46] Therefore, in the study design, we suspected that MDA levels in NPDR should be lower than those in PDR. Surprisingly, we did not observe a significant difference between the NPDR and PDR groups. At the same time, meta-regression did not find a relationship between DR type and high heterogeneity. This finding might show that lipid peroxidation has limited effect on neovascularization compared with other mechanisms, and thus the differences of MDA levels between NPDR and PDR could not be found in our study.

In meta-analysis, SMD is used as a summary statistic when all the studies evaluate the same outcome, but they do that in different ways. However, the different approaches of MDA measurement may lead to different results. We found that the subgroup of the TBARS assay with different absorption spectra showed statistically significant differences, and in the meta-regression, both the MDA assay type and absorption spectrum contributed to high heterogeneity. Some studies have also reported that TBARS assays lack specificity because thiobarbituric acid can react with other compounds,[47,48] and the difference in total MDA in human plasma was up to three-fold when measured by TBARS compared to that by high performance liquid chromatography (HPLC) in human plasma.[49] However, we found only two studies using HPLC in our included studies.[34,37] The most widely used method in our meta-analysis is TBARS. In general, the high heterogeneity of our study might have been caused by different MDA assays and non-specific TBARS methods.

Our study had several strengths and limitations. The current meta-analysis of case-control studies evaluated the circulating MDA levels in people with DR vs. people with DM but not DR. The sensitivity analysis revealed that none of the studies individually influenced the overall SMD, indicating that after removing any single study, an increase in the MDA levels of DR people was still evident. We also predesigned a couple of subgroup analyses to assess the role of MDA in DR in various aspects. Then, the high heterogeneity might be attributable to various environmental factors, diet and lifestyle habits, TBARS methods, and methodological limitations of case-control study design. We suspect that the TBARS method has a low specificity, which could be a major source of heterogeneity. But our included studies are all cross-sectional studies with a case control designed. This type of design usually cannot establish the temporal relationship between exposure and disease; thus, the causation and its underlying mechanisms between MDA and DR can be unclear. To ascertain the cause–effect correlation, more prospective studies are needed, especially prospective cohort studies.

In summary, this study supports that the view that circulating MDA levels increased in people with DR compared to those with DM but not DR. However, the presence of marked heterogeneity makes this finding debatable. Further high-quality epidemiologic studies in cohort design with more specific methods are necessary to determine the concrete mechanisms and effects of MDA. At this stage, no recommendation can be made regarding routine investigation and treatment of elevated MDA in DR.


We thank Dr. Sheyu Li from the Department of Guideline and Rapid Recommendation, Chinese Evidence-Based Medicine Center, Cochrane China Center and MAGIC China Center, West China Hospital for his statistical advice.


This work was supported by grants from Chengdu Science and Technology Program (No. 2021-YF09-00024-SN), 1·3·5 project for disciplines of excellence–Clinical Research Incubation Project, West China Hospital, Sichuan University (No. 2021HXFH026), and Natural Science Foundation of Sichuan (No. 2022NSFSC1370).

Conflicts of interest



1. Kobrin Klein BE. Overview of epidemiologic studies of diabetic retinopathy. Ophthalmic Epidemiol 2007; 14:179–183. doi: 10.1080/09286580701396720.
2. Steinmetz JD, Bourne RRA, Briant PS, Flaxman SR, Taylor HRB, Jonas JB, et al. Causes of blindness and vision impairment in 2020 and trends over 30 years, and prevalence of avoidable blindness in relation to VISION 2020: the right to sight: an analysis for the Global Burden of Disease Study. Lancet Glob Health 2021; 9:e144–e160. doi: 10.1016/S2214-109X(20)30489-7.
3. Teo ZL, Tham Y-C, Yu M, Chee ML, Rim TH, Cheung N, et al. Global prevalence of diabetic retinopathy and projection of burden through 2045: systematic review and meta-analysis. Ophthalmology 2021; 128:1580–1591. doi: 10.1016/j.ophtha.2021.04.027.
4. Li J, Shi Q, Gao Q, Pan XF, Zhao L, He Y, et al. Obesity pandemic in China: epidemiology, burden, challenges, and opportunities. Chin Med J 2022; 135:1328–1330. doi: 10.1097/cm9.0000000000002189.
5. Cai XL, Wang F, Ji LN. Risk factors of diabetic retinopathy in type 2 diabetic patients. Chin Med J 2006; 119:822–826. doi: 10.1097/00029330-200605020-00005.
6. Cheung N, Mitchell P, Wong TY. Diabetic retinopathy. Lancet 2010; 376:124–136. doi: 10.1016/S0140-6736(09)62124-3.
7. Das A, McGuire PG, Rangasamy S. Diabetic macular edema: pathophysiology and novel therapeutic targets. Ophthalmology 2015; 122:1375–1394. doi: 10.1016/j.ophtha.2015.03.024.
8. Wang Y, Zhai WL, Yang YW. Association between NDRG2/IL-6/STAT3 signaling pathway and diabetic retinopathy in rats. Eur Rev Med Pharmacol Sci 2020; 24:3476–3484. doi: 10.26355/eurrev_202004_20806.
9. Li X, Deng A, Liu J, Hou W. The role of Keap1-Nrf2-ARE signal pathway in diabetic retinopathy oxidative stress and related mechanisms. Int J Clin Exp Pathol 2018; 11:3084–3090.
10. Zhang L, Yu J, Ye M, Zhao H. Upregulation of CKIP-1 inhibits high-glucose induced inflammation and oxidative stress in HRECs and attenuates diabetic retinopathy by modulating Nrf2/ARE signaling pathway: an in vitro study. Cell Biosci 2019; 9:67doi: 10.1186/s13578-019-0331-x.
11. Dave A, Kalra P, Gowda BH, Krishnaswamy M. Association of bilirubin and malondialdehyde levels with retinopathy in type 2 diabetes mellitus. Indian J Endocrinol Metab 2015; 19:373–377. doi: 10.4103/2230-8210.152777.
12. Mondal LK, Bhaduri G, Bhattacharya B. Biochemical scenario behind initiation of diabetic retinopathy in type 2 diabetes mellitus. Indian J Ophthalmol 2018; 66:535–540. doi: 10.4103/ijo.IJO_1121_17.
13. Sanz-González SM, García-Medina JJ, Zanón-Moreno V, López-Gálvez MI, Galarreta-Mira D, Duarte L, et al. Clinical and molecular-genetic insights into the role of oxidative stress in diabetic retinopathy: antioxidant strategies and future avenues. Antioxidants 2020; 9:1101doi: 10.3390/antiox9111101.
14. Verma MK, Singh SP, Alam R, Verma P. Comparative study on MDA, SOD and HBA1C Levels in patients of type 2 diabetes mellitus with retinopathy and without retinopathy. Int J Pharm Sci Res 2016; 7:4184–4190. doi: 10.13040/IJPSR.0975-8232.7(10).4184-90.
15. Fahmy R, Almutairi NM, Al-Muammar MN, Bhat RS, Moubayed N, El-Ansary A. Controlled diabetes amends oxidative stress as mechanism related to severity of diabetic retinopathy. Sci Rep 2021; 11:17670doi: 10.1038/s41598-021-96891-7.
16. Dai L, Wang Y, Liu Y, Wu Y. Correlation of serum LXA4 and VEGF levels with inflammatory factors and oxidative stress indexes in patients with diabetic retinopathy. Acta Med Mediterr 2021; 37:2843–2847. doi: 10.19193/0393-6384_2021_5_438.
17. Aldebasi YH, Mohieldein AH, Almansour YS, Almutairi BL. Dyslipidemia and lipid peroxidation of Saudi type 2 diabetics with proliferative retinopathy. Saudi Med J 2013; 34:616–622. doi: 10.1016/j.pop.2013.04.001.
18. Kuppan K, Mohanlal J, Mohammad AM, Babu KA, Sen P, Das Undurti N, et al. Elevated serum OxLDL is associated with progression of type 2 Diabetes Mellitus to diabetic retinopathy. Exp Eye Res 2019; 186:107668doi: 10.1016/j.exer.2019.05.008.
19. Roig-Revert MJ, Lleó-Pérez A, Zanón-Moreno V, Vivar-Llopis B, Marín-Montiel J, Dolz-Marco R, et al. Enhanced oxidative stress and other potential biomarkers for retinopathy in type 2 diabetics: beneficial effects of the nutraceutic supplements. Biomed Res Int 2015; 2015:408180doi: 10.1155/2015/408180.
20. Shawki HA, Elzehery R, Shahin M, Abo-Hashem EM, Youssef MM. Evaluation of some oxidative markers in diabetes and diabetic retinopathy. Diabetol Int 2021; 12:108–117. doi: 10.1007/s13340-020-00450-w.
21. Choudhuri S, Roy PK, Mitra B, Sen S, Sarkar R, Das M, et al. Hyperlipidemia-mediated increased advanced lipoxidation end products formation, an important factor associated with decreased erythrocyte glucose-6-phosphate dehydrogenase activity in mild nonproliferative diabetic retinopathy. Can J Diabetes 2017; 41:82–89. doi: 10.1016/j.jcjd.2016.07.007.
22. Vivian Samuel T, Jayaprakash Murthy DSK, Dattatreya, Suresh Babu P, Smilee Johncy S. Impaired antioxidant defence mechanism in diabetic retinopathy. J Clin Diagnostic Res 2010; 4:3430–33436.
23. Kurtul N, Bakan E, Aksoy H, Baykal O. Leukocyte lipid peroxidation, superoxide dismutase and catalase activities of type 2 diabetic patients with retinopathy. Acta Medica 2005; 48:35–38. doi: 10.14712/18059694.2018.26.
24. Gupta MM, Chari S. Lipid peroxidation and antioxidant status in patients with diabetic retinopathy. Indian J Physiol Pharmacol 2005; 49:187–192.
25. Losada M, Alio JL. Malondialdehyde serum concentration in type 1 diabetic with and without retinopathy. Doc Ophthalmol 1996; 93:223–229. doi: 10.1007/bf02569062.
26. Sharma S, Saxena S, Srivastav K, Shukla RK, Mishra N, Meyer CH, et al. Nitric oxide and oxidative stress is associated with severity of diabetic retinopathy and retinal structural alterations. Clin Exp Ophthalmol 2015; 43:429–436. doi: 10.1111/ceo.12506.
27. Kundu D, Mandal T, Nandi M, Osta M, Bandyopadhyay U, Ray D. Oxidative stress in diabetic patients with retinopathy. Ann Afr Med 2014; 13:41–46. doi: 10.4103/1596-3519.126951.
28. Vidya D, Shekhar R, Prabodh S, Chowdary NVS, Das MC, Joji Reddy M. Oxidative stress in diabetic retinopathy. J Clin Diagnostic Res 2011; 5:994–997. doi: 10.1016/S0168-8227(00)81893-8.
29. Khalili F, Vaisi-Raygani A, Shakiba E, Kohsari M, Dehbani M, Naseri R, et al. Oxidative stress parameters and keap 1 variants in T2DM: association with T2DM, diabetic neuropathy, diabetic retinopathy, and obesity. J Clin Lab Anal 2022; 36:e24163doi: 10.1002/jcla.24163.
30. Mandal LK, Choudhuri S, Dutta D, Mitra B, Kundu S, Chowdhury IH, et al. Oxidative stress-associated neuroretinal dysfunction and nitrosative stress in diabetic retinopathy. Can J Diabetes 2013; 37:401–407. doi: 10.1016/j.jcjd.2013.05.004.
31. Gaonkar B, Prabhu K, Rao P, Kamat A, Rao Addoor K, Varma M. Plasma angiogenesis and oxidative stress markers in patients with diabetic retinopathy. Biomarkers 2020; 25:397–401. doi: 10.1080/1354750x.2020.1774654.
32. Kumari S, Panda S, Mangaraj M, Mandal MK, Mahapatra PC. Plasma MDA and antioxidant vitamins in diabetic retinopathy. Indian J Clin Biochem 2008; 23:158–162. doi: 10.1007/s12291-008-0035-1.
33. Turk A, Nuhoglu I, Mentese A, Karahan SC, Erdol H, Erem C. The relationship between diabetic retinopathy and serum levels of ischemia-modified albumin and malondialdehyde. Retina 2011; 31:602–608. doi: 10.1097/IAE.0b013e3181ed8cd1.
34. Longo-Mbenza B, Mvitu Muaka M, Masamba W, Muizila Kini L, Longo Phemba I, Kibokela Ndembe D, et al. Retinopathy in non diabetics, diabetic retinopathy and oxidative stress: a new phenotype in Central Africa? Int J Ophthalmol 2014; 7:293–301. doi: 10.3980/j.issn.2222-3959.2014.02.18.
35. Gürler B, Vural H, Yilmaz N, Oguz H, Satici A, Aksoy N. The role of oxidative stress in diabetic retinopathy. Eye 2000; 14 (Pt 5):730–735. doi: 10.1038/eye.2000.193.
36. El-Mesallamy HO, Rizk KA, Hashad IM. Role of oxidative stress, inflammation and endothelial dysfunction in the pathogenesis of diabetic retinopathy. IIOAB J 2011; 2:91–97.
37. Hartnett ME, Stratton RD, Browne RW, Rosner BA, Lanham RJ, Armstrong D. Serum markers of oxidative stress and severity of diabetic retinopathy. Diabetes Care 2000; 23:234–240. doi: 10.2337/diacare.23.2.234.
38. Icel E, Icel A, Mertoglu C, Tasli NG, Karakurt Y, Ucak T, et al. Serum SCUBE-1 levels in patients with diabetic retinopathy. Int Ophthalmol 2020; 40:859–865. doi: 10.1007/s10792-019-01249-8.
39. Kumari S, Pradhan T, Panda TK. Trace minerals and oxidative stress in diabetic retinopathy. Bangladesh J Med Sci 2014; 13:175–179. doi: 10.3329/bjms.v13i2.14963.
40. Wells GA, Shea B, O’Connell D, Peterson J, Welch V, Losos M, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses In: Oxford. 2000;
41. Ito F, Sono Y, Ito T. Measurement and clinical significance of lipid peroxidation as a biomarker of oxidative stress: Oxidative stress in diabetes, atherosclerosis, and chronic inflammation. Antioxidants 2019; 8:72doi: 10.3390/antiox8030072.
42. Fritz KS, Petersen DR. An overview of the chemistry and biology of reactive aldehydes. Free Radic Biol Med 2013; 59:85–91. doi: 10.1016/j.freeradbiomed.2012.06.025.
43. Li G, Chen Y, Hu H, Liu L, Hu X, Wang J, et al. Association between age-related decline of kidney function and plasma malondialdehyde. Rejuvenation Res 2012; 15:257–264. doi: 10.1089/rej.2011.1259.
44. Choromańska B, Myśliwiec P, Dadan J, Maleckas A, Zalewska A, Maciejczyk M. Effects of age and gender on the redox homeostasis of morbidly obese people. Free Radic Biol Med 2021; 175:108–120. doi: 10.1016/j.freeradbiomed.2021.08.009.
45. Nakhjavani M, Esteghamati A, Nowroozi S, Asgarani F, Rashidi A, Khalilzadeh O. Type 2 diabetes mellitus duration: an independent predictor of serum malondialdehyde levels. Singapore Med J 2010; 51:582–585. doi: 10.1590/S1516-31802010000400013.
46. Solomon SD, Goldberg MF. ETDRS grading of diabetic retinopathy: still the gold standard? Ophthalmic Res 2019; 62:190–195. doi: 10.1159/000501372.
47. Tsikas D. Assessment of lipid peroxidation by measuring malondialdehyde (MDA) and relatives in biological samples: analytical and biological challenges. Anal Biochem 2017; 524:13–30. doi: 10.1016/j.ab.2016.10.021.
48. Mas-Bargues C, Escrivá C, Dromant M, Borrás C, Viña J. Lipid peroxidation as measured by chromatographic determination of malondialdehyde. human plasma reference values in health and disease. Arch Biochem Biophys 2021; 709:108941doi: 10.1016/
49. Moselhy HF, Reid RG, Yousef S, Boyle SP. A specific, accurate, and sensitive measure of total plasma malondialdehyde by HPLC. J Lipid Res 2013; 54:852–858. doi: 10.1194/jlr.D032698.

Malondialdehyde; Diabetic retinopathy; Lipid peroxidation; Oxidative stress

Copyright © 2023 The Chinese Medical Association, produced by Wolters Kluwer, Inc. under the CC-BY-NC-ND license.