Associations between lipid metabolism and Henoch–Schönlein purpura nephritis in children: a systematic review and meta-analysis

Abstract

Background

Henoch-Schönlein purpura (HSP)is the most common childhood vasculitis, and renal involvement (HSPN) significantly worsens its prognosis. Noninvasive correlates of renal involvement are therefore of interest. Dyslipidemia often coexists with renal inflammation. Clarifying this association is essential to understand the metabolic perturbations in HSPN. However, although dyslipidemia has been associated with HSPN, its exact mechanism remains unclear from an evidence-based.

Methods

A comprehensive search was performed using PubMed, Embase, Web of Science, China National Knowledge Infrastructure (CNKI), Wanfang, China Science and Technology Journal (VIP), and the Chinese Medical Journal full-text databases from their inception to December 25, 2024. Primary outcomes compared levels of total cholesterol (TC), total triglycerides (TG), low-density lipoprotein (LDL), and high-density lipoprotein (HDL) between the HSPN group and the HSP group; secondary outcomes included study methodology, publication date, sample size, and quality assessment scores. Study quality was assessed using modified NOS scales, with inter-rater reliability measured by Cohen’s kappa. Statistical analysis was performed using Stata 16.0 to calculate standardized mean differences (SMDs) with 95% confidence intervals (CIs). Subgroup analyses were limited by the paucity of data regarding the timing of blood sampling relative to corticosteroid initiation and renal onset.

Results

Eighteen studies were included. Meta-analysis revealed that patients with HSPN had significantly higher levels of TC, TG, and LDL compared with patients with HSP without nephritis. The pooled estimates were TC (SMD = 0.55, 95% CI = 0.36 to 0.75; I² = 89.9%, P < 0.001), TG (SMD = 0.47, 95% CI = 0.29 to 0.65; I² = 69.8%, P = 0.001), and LDL (SMD = 0.37, 95% CI = 0.12 to 0.61; I² = 76.1%, P < 0.001). No significant difference was observed for HDL (SMD = − 0.87, 95% CI = − 1.89 to 0.14; I² = 97.9%, P < 0.001).

Conclusion

Elevated levels of TC, TG, and LDL are observed in patients with HSPN, underscoring a robust association between lipid dysregulation and renal involvement. However, significant heterogeneity, coupled with the lack of pretreatment baseline data, precludes causal inference and precludes utility in early prediction. Prospective studies measuring pretreatment baseline lipids prior to renal onset are warranted to validate these findings.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13052-026-02330-6.

Introduction

Henoch–Schönlein purpura (HSP), also known as IgA vasculitis, is characterized by skin purpura; however, its pathogenesis remains unclear and is considered to be mainly related to infections, medications, and autoimmune dysfunction [1–3]. Although the prognosis of this disease is generally good [4–6], 1–2% of patients can develop end-stage renal disease [7]. Consequently, careful clinical and laboratory monitoring of renal involvement remains a cornerstone of pediatric management.

Some studies have identified several high-risk factors for renal involvement, including older age, hypertension, abdominal pain, persistent purpura, recurrent rash, increased hemoglobin and blood urea nitrogen levels, and decreased albumin levels [8]. Studies have also reported that dyslipidemia may be closely associated with HSP [9].

Dyslipidemia is a metabolic disorder caused by disturbances in lipid metabolism and abnormal lipid transport in vivo [10, 11]. It is characterized by elevated serum levels of total triglycerides (TG), total cholesterol (TC), and low-density lipoprotein (LDL), and reduced levels of high-density lipoprotein (HDL) [12]. In 1982, Moorhead described lipids as nephrotoxic [13], and since then, researchers have proposed that mechanisms of lipid-related renal injury include defective autophagy, oxidative stress, endoplasmic reticulum stress, inflammatory responses, and the accumulation of toxic lipids. These mechanisms interact with each other and may lead to organelle dysfunction, cellular damage, or cell death [14–16]. While dyslipidemia commonly coexists with HSPN, its temporal relationship to renal involvement remains unclear. This meta-analysis therefore aims to characterize the association between lipid metabolism and HSPN.

Materials and methods

This meta-analysis strictly adhered to the pre-specified study protocol (registered on the PROSPERO platform with registration number CRD42025641677). There were no deviations from the protocol throughout the study process, and all procedures were consistent with the pre-planned design. Ethical approval was not required because this meta-analysis was a statistical analysis based on previously published studies.

Search methodology

The PubMed, Embase, Web of Science, CNKI, Wanfang, VIP, and the Chinese Medical Journal Full Text databases were searched from their inception to December 25, 2024. The detailed search strategies are listed in Supplementary Table 2.

Criteria for literature inclusion and exclusion

The eligibility criteria were as follows:

1. The patients were children;

2. All patients included in the present study were diagnosed with Henoch–Schönlein purpura. The diagnosis was based on the presence of palpable purpura as a mandatory criterion, together with at least one of the following features: diffuse abdominal pain, biopsy showing predominant IgA deposition, arthritis or arthralgia, or renal involvement [17, 18].

3. The diagnostic criteria for Henoch–Schönlein purpura nephritis included: Within 6 months after the onset of HSP, the presence of hematuria (≥ 3 red blood cells per high‑power field, either macroscopic or microscopic) and/or meeting any of the following proteinuria criteria: positive urine protein in 3 out of 7 weekly tests, 24‑hour quantitative protein excretion > 150 mg, or elevated microalbuminuria in 3 out of 3 weekly assessments; [19]

4. The literature provides the HSP group and the HSPN group, and the measurements of TC, TG, and other indices, which are continuous-type variables, are expressed as (‾X ± S).

5. The minimum sample size for inclusion should not be less than 20 cases.

The exclusion criteria were as follows:

1. studies published in the form of reviews, case reports, or conference abstracts;

2. studies with no data on research indicators;

3. Studies involving other interventions or lacking a control group will also be excluded.

Study selection and data extraction

Initially, all databases were searched using the predefined strategy, and duplicate studies were removed. In the first phase, two researchers independently screened the titles and abstracts of the remaining articles according to the inclusion and exclusion criteria. In the second phase, the full texts of potentially relevant articles were reviewed to determine eligibility, exclude irrelevant studies, and record reasons for exclusion. In cases of disagreement between the two researchers, the article was assessed by a third reviewer.

Finally, information and characteristics from the included studies were extracted, including first author, publication year, study type, diagnostic criteria, number of cases (HSPN group/HSP group), sex (HSPN group/HSP group), age (HSPN group/HSP group), and post-diagnostic measurements (TC, TG, HDL, and LDL) for each group. These data are summarized in Supplementary Table 1.

Crucially, we systematically extracted study-level methodological data to address potential biases. This encompassed the timing of lipid measurements, corticosteroid exposure, proteinuria quantification, nephritis severity grading, and fasting status, as detailed in Supplementary Table 1. Notably, the systematic absence of these data precluded stratified analyses based on treatment status or disease severity.

Risk of bias assessment

The Newcastle–Ottawa Scale (NOS) was used to assess the risk of bias in the included studies [20]. The NOS evaluates study quality across three domains: selection of the study population (4 stars), comparability between groups (2 stars), and assessment of outcomes (3 stars), with a maximum score of 9 stars. Studies scoring ≥ 8 stars were considered low risk, 6–7 stars moderate risk, and < 6 stars high risk. Studies with a high risk of bias are more likely to overestimate or underestimate the true effect, which may affect the accuracy of pooled results and the strength of the evidence.

The NOS assessment was conducted independently by two researchers. Discrepancies were evaluated using Cohen’s kappa coefficient, and disagreements were resolved through discussion within the research team to reach consensus. This process helped ensure objectivity in the quality assessment and supported the reliability of the meta-analysis results [21].

Statistical methods

Given the systematic absence of study-level data on measurement timing and corticosteroid exposure, the pooled estimates reflect association rather than causality. All analyses were performed using Stata version 16.0. The standardized mean difference (SMD) with 95% confidence interval (CI) was used for continuous variables. Heterogeneity among the included studies was assessed using the I² statistic. A fixed-effects model was applied when heterogeneity was low (I² < 50%); otherwise, a random-effects model was used. Due to sparse reporting, pre-defined subgroup analyses were limited to readily available characteristics (Supplementary Table 4).Sensitivity analyses were conducted by sequentially excluding individual studies to assess the robustness of the pooled results. Publication bias was evaluated using Begg’s and Egger’s tests (p < 0.05). When bias was detected, the trim-and-fill method was applied, and Rosenthal’s fail-safe N (Nfs) was calculated; results were deemed robust if Nfs exceeded 5k + 10. Statistical significance was set at p < 0.05.

Results

Literature search

The study selection and exclusion processes are shown in Fig. 1. A total of 1052 records were retrieved, including 861 Chinese publications and 191 English publications. Duplicate studies, case reports, literature reviews, and other irrelevant articles were excluded. After successive rounds of screening, 18 studies were finally included [22–39], of which 16 reported TC levels, 9 reported TG levels, 9 reported LDL levels, and 8 reported HDL levels.

Fig. 1.

Flow chart showing inclusion and exclusion of literature

Basic characteristics and quality assessment of the included studies

A total of 5440 children were included, comprising 3388 in the HSP group and 2052 in the HSPN group. Detailed characteristics are presented in Table 1. Study quality was assessed using the NOS scale, with detailed results shown in Table 2. Cohen’s kappa coefficient was 0.97, indicating high consistency and reliability between the reviewers.

Table 1.

Basic characteristics of included studies

Study	Publication Year	Method	Sample (cases)		Age (years)		Sex (m/f)		TC(mmol/L)		TG(mmol/L)		HDL(mmol/L)		LDL(mmol/L)		Diagnostic Criteria
			HSPN	HSP	HSPN	HSP	HSPN	HSP	HSPN	HSP	HSPN	HSP	HSPN	HSP	HSPN	HSP
Cao Tingting	2019	Retrospective	170	230	NR	NR	NR	NR	4.15±1.06	3.30±0.78	1.58±0.91	1.04±0.75	NR	NR	NR	NR	NR
Fang Hongxing	2023	Prospective	57	81	NR	NR	31/26	39/42	1༎22±0༎26	1.14±0.23	NR	NR	1༎21±O༎23	1.25±0.29	2༎96±0༎68	2༎79±0༎69	Practical Pediatrics by Zhu Futang
Fu Zhongying	2024	Prospective	22	54	9.47±1.96	9.81±1.65	9/13	30/24	3.94±1.21	3.16±1.44	2.44±0.97	2.12±1.11	1.54±0.87	1.25±1.01	2.65±0.96	2.93±1.28	Practical Pediatrics by Zhu Futang
Gao Jizhao	2007	Retrospective	78	128	9.47±2.82	8.68±2.88	NR	NR	4.39±1.32	3.95±1.03	1.55±1.05	1.38±0.92	NR	NR	NR	NR	Practical Pediatrics by Zhu Futang
Gu Nana	2012	Retrospective	116	180	9.20±2.70	7.46±2.90	NR	NR	4.05±1.21	3.27±0.57	1.42±0.92	0.98±0.56	NR	NR	NR	NR	EULAR/PReS
Li Hui	2024	Prospective	41	95	NR	NR	17/24	53/42	3.64±0.73	3.55±0.66	0.95±0.36	0.96±0.36	1.26±0.31	1.20±0.25	1.95±0.56	1.91±0.47	EULAR/PReS
Li Xinye	2019	Prospective	39	67	9.5	5	24/15	42/25	5.34±2.45	4.15±1.08	NR	NR	NR	NR	NR	NR	Chinese Medical Association
Li Yan	2020	Retrospective	64	52	8.23±1.28	8.22±1.18	38/26	29/23	NR	NR	NR	NR	1.08±0.12	1.62±0.21	2.92±0.82	2.73±0.91	Chinese Medical Association
Liu Cuiyun	2018	Retrospective	38	35	7.61±2.0	5.77±1.9	NR	NR	3.83±0.48	3.70±0.60	NR	NR	1.11±0.08	1.16±0.105	2.47±0.57	2.33±0.47	Practical Pediatrics by Zhu Futang
Liu Hongxia	2015	Retrospective	32	70	9.9±3.4	6.7±3.8	20/12	41/29	4.89±1.25	3.98±0.87	1.08±0.34	0.96±0.31	NR	NR	NR	NR	Practical Pediatrics by Zhu Futang
Liu Lijun	2018	Retrospective	35	57	NR	NR	18/17	35/22	NR	NR	NR	NR	1.19±0.25	1.56±0.23	2.94±1.14	2.42±1.29	Practical Pediatrics by Zhu Futang
Ma	2017	Prospective	103	147	NR	NR	NR	NR	4.82±1.26	3.65±1.19	NR	NR	NR	NR	3.25±1.09	2.29±0.83	NR
Wu Lijuan	2013	Retrospective	92	148	NR	NR	56/36	103/45	4.98±1.84	3.81±0.85	1.16±0.30	0.85±0.37	1.50±0.32	1.20±0.26	2.95±1.75	2.23±0.5	Chinese Medical Association
Xi Xiaojuan	2011	Prospective	80	159	NR	NR	55/25	110/49	4.6±1.1	4.5±0.9	NR	NR	NR	NR	NR	NR	EULAR/PReS
Xu	2017	Retrospective	256	413	8.9±2.7	6.8±2.6	130/126	238/175	4.66±1.60	3.91±0.80	NR	NR	NR	NR	NR	NR	EULAR/PReS
Zhang Jingfu	2024	Retrospective	38	63	NR	NR	22/16	21/42	1.06±0.15	1.05±0.19	1.15±0.14	1.10±0.13	1.07±0.15	1.61±0.18	2.91±0.31	2.78±0.29	Chinese Medical Association
Zhang Xia	2013	Retrospective	654	1228	9.97±3.46	8.41±3.27	NR	NR	4.57±1.59	4.43±1.04	NR	NR	NR	NR	NR	NR	Practical Pediatrics by Zhu Futang
Zhang Xiaohua	2017	Retrospective	137	181	9.43±1.74	8.05±2.28	81/56	105/76	4.13±1.03	3.29±0.77	1.51±0.84	1.03±0.72	NR	NR	NR	NR	EULAR/PReS

Table 2.

Basic characteristics of included studies

Study	Selection (up to 4stars)				Comparability (up to 2stars)	Outcome (up to 3stars)
	Representativeness of the exposed cohort	Selection of the nonexposed cohort	Ascertainment of exposure	Demonstration that outcome	Comparability of cohorts	Assessment of outcome	Was follow up long enough	Adequacy of follow up of cohorts
Cao Tingting 2019		*	*	*	**	*	*	*
Fang Hongxing 2023	*	*	*	*	**	*
Fu Zhongying 2024	*		*	*	**	*
Gao Jizhao 2007		*	*	*	**	*
Gu Nana 2012	*	*	*	*	**	*
Li Hui 2024	*	*	*	*	*	*	*	*
Li Xinye 2019	*	*	*	*	**	*
Li Yan 2020	*	*	*	*	**	*
Liu Cuiyun 2018	*	*	*	*	**	*
Liu Hongxia 2015	*	*	*	*	**	*
Liu Lijun 2018	*		*	*	**	*	*	*
Ma 2017		*	*	*	**	*	*	*
Wu Lijuan 2013		*	*	*	**	*
Xi Xiaojuan 2011	*	*	*	*	**	*	*	*
Xu 2017	*	*	*	*	**	*
Zhang Jingfu 2024	*	*	*	*	**	*
Zhang Xia 2013	*	*	*	*	**	*
Zhang Xiaohua 2017	*	*	*	*	**	*

Meta-analysis results

TC

This analysis included 5232 participants, comprising 3279 individuals with HSP and 1953 with HSPN. Significant heterogeneity was observed between the HSPN and HSP groups (I² = 89.9%, P < 0.001). Using a random-effects model, the results showed that TC levels were higher in the HSPN group than in the HSP group, with a statistically significant difference (SMD = 0.55, 95% CI = 0.36 to 0.75). However, the prediction interval ranged from − 0.27 to 1.38, crossing zero, suggesting that the true effect size may vary substantially across different populations and that the direction of the effect may differ in future studies. Further details are shown in Fig. 2A.

Fig. 2.

Meta-analysis of lipid levels between children in the HSPN group and the HSP group (A) Comparative forest plot of TC levels, (B) Comparative forest plot of TG levels, (C) Comparative forest plot of LDL levels, and (D) Comparative forest plot of HDL levels. Note: This forest plot compares serum levels of TC, TG, LDL, and HDL between the HSPN and HSP groups. The diamond represents the pooled effect size. The HSPN group showed significantly higher levels of TC, TG, and LDL, while no statistically significant difference was observed in HDL levels

Sensitivity analyses were conducted for studies reporting TC levels (Fig. 3A). After sequential exclusion of individual studies, the pooled effect size shifted from 0.55, but the 95% CI for the SMD (0.32–0.79) remained positive and unchanged in direction.

Fig. 3.

Sensitivity analyses (A) TC, (B) TG, (C)LDL, and (D) HDL. Note: Each row in the figure represents the recalculated combined effect size (SMD) and 95% confidence interval after excluding one corresponding study. Symbols (circles or squares) indicate the point estimate, and the horizontal lines denote the 95% confidence intervals

TG

A total of 1875 patients were included, with 1149 in the HSP group and 726 in the HSPN group. Substantial heterogeneity was observed between the two groups (I² = 69.8%, P = 0.001). Using a random-effects model, TG levels were significantly higher in the HSPN group than in the HSP group (SMD = 0.47, 95% CI = 0.29 to 0.65). The prediction interval ranged from − 0.11 to 1.05. Further details are presented in Fig. 2B.

Sensitivity analyses for TG (Fig. 3B) showed that the pooled effect size ranged from 0.47, but the 95% CI for the SMD (0.23–0.70). All corresponding 95% CIs remained significant, and the direction of effect was consistently positive. Additional sensitivity analyses identified studies by Li Hui and Wu Lijuan as potential sources of heterogeneity.

After excluding these two studies, the heterogeneity in the TG group decreased from 69.8% to 42.8%, which may be related to the type of study design and year of publication.

LDL

The study involved 1222 patients, 732 in the HSP group and 490 in the HSPN group, suggesting heterogeneity between the HSPN and HSP groups (I² = 76.1%, P < 0.001). Therefore, the results were analyzed using a random effects model, revealing that patients in the HSPN group had significantly higher LDL levels than those in the HSP group (SMD = 0.37, 95% CI = 0.12, 0.61). However, the forecast range was (–0.46, 1.19) (Fig. 2(C)).

Given the small number of included studies and very high heterogeneity, sensitivity analyses were performed. The pooled effect size was 0.37 (95% CI: 0.04–0.67). However, all 95% CIs after exclusion remained significant, and the effect had a positive direction. The results also identified studies by Ma and Wu Lijuan as sources of heterogeneity. After the two studies were excluded, the heterogeneity decreased from 76.1% to 0%, which may be related to differences in study methodology and year of publication (Figs. 3B).

HDL

This analysis included 972 patients, with 585 in the HSP group and 387 in the HSPN group. Very high heterogeneity was observed between groups (I² = 97.9%, P < 0.001); therefore, a random-effects model was applied. The results showed no statistically significant difference in HDL levels between the two groups (SMD = − 0.87, 95% CI = − 1.89 to 0.14), with a wide prediction interval ranging from − 4.63 to 2.89. These findings indicate that HDL has no consistent association with renal involvement in HSP and therefore cannot be considered a reliable indicator for assessing renal injury. (Fig. 2D).

Sensitivity analysis indicated that the pooled estimate for HDL was not robust, largely due to the substantial heterogeneity among the included studies (Fig. 3D).

Publication bias

The Egger test results for TC levels (t [bias] = 1.61, P = 0.130) were consistent with the Begg test results, indicating no significant publication bias (Fig. 4A).

Fig. 4.

Funnel plot for the evaluation of publication bias (A) TC, (B) TG, (C)LDL, and (D) HDL. Note: Each point represents a study; the vertical line indicates the pooled effect size, and the diagonal lines show 95% confidence limits. Asymmetry suggests potential publication bias

The Egger test results for TG levels (t [bias] = − 1.84, P = 0.108) were consistent with the Begg test results, also indicating no significant publication bias (Fig. 4B).

For LDL levels, the Begg test (P = 0.144) and Egger test (t = − 3.23, P = 0.014) suggested possible publication bias. The fail-safe N (Nfs) was 88.37, exceeding Rosenthal’s criterion of 55, indicating that 89 unpublished negative studies would be required to overturn the current findings. However, the trim-and-fill analysis did not identify any studies requiring adjustment, suggesting that publication bias likely had a limited impact on the pooled results.

For HDL levels, the Begg test (P = 0.035) and Egger test (t [bias] = − 4.22, P = 0.006) indicated possible publication bias. The trim-and-fill analysis showed no need to add missing studies, suggesting a largely symmetrical funnel plot. The Nfs value was 112.28, exceeding the Rosenthal threshold (5k + 10 = 50), indicating that 112 unpublished negative studies would be required to change the statistical significance of the findings (Fig. 4C and D).

Regression analysis

The results indicated that, among TC indicators, mean platelet difference showed a significant negative moderating effect on effect size (β = − 0.005, P = 0.007). Similarly, mean platelet difference exhibited a significant negative association for TG indicators (β = − 0.008, P = 0.003). For LDL indicators, publication year showed a significant negative correlation with effect size (β = − 0.4662, P = 0.040), suggesting a decreasing trend in effect size in more recent studies. For HDL indicators, all examined covariates (publication year, sample size, quality score, methodology, gender ratio, and mean platelet difference) showed P values > 0.05, indicating no significant moderating effects, as shown in Table 3.

Table 3.

Regression analysis

TC
Variable	β	se	p	95%CI	I²_res	Adj R ²
Publication (y)	-0.381	0.203	0.081	(-0.8156,0.0542)	0.903	0.161
Sample size	0.175	0.305	0.575	(-0.4799, 0.8302)	0.906	-0.042
Quality score	-0.007	0.212	0.973	(-0.4613, 0.4465)	0.900	-0.087
Method	-0.151	0.214	0.492	(-0.6095,0.3078)	0.906	-0.038
Gender ratio (male ≥ 50%/<50%)	0.465	0.393	0.267	(-0.4249,1.3549)	0.901	0.034
PLT mean difference	-0.005	0.002	0.007	(-0.008, -0.002)	0.821	0.444
TG
Variable	β	se	p	95%CI	I²_res	Adj R²
Publication (y)	-0.374	0.188	0.087	(-0.818, 0.070)	59.02%	40.47%
Sample size	0.184	0.372	0.636	(-0.696, 1.063)	72.69%	-9.29%
Quality score	-0.139	0.24	0.581	(-0.705, 0.428)	73.38%	-16.95%
Method	-0.447	0.218	0.079	(-0.962, 0.068)	58.53%	42.69%
Gender ratio (male ≥ 50%/<50%)	0.084	0.412	0.848	(-1.060,1.228)	77.60%	-27.47%
PLT mean difference	-0.008	0.002	0.003	(-0.012, -0.004)	0.91%	100.00%
LDL
Variable	β	se	p	95%CI	I²_res	Adj R²
Publication (y)	-0.4662	0.1847	0.040	(-0.903, -0.030)	54.43%	57.03%
Sample size	0.2907	0.2607	0.302	(-0.326, 0.907)	75.48%	4.55%
Quality score	0.2485	0.253	0.359	(-0.350, 0.847)	75.09%	0.33%
Method	-0.028	0.272	0.922	(-0.672, 0.616)	78.34%	-19.04%
Gender ratio (male ≥ 50%/<50%)	-0.176	0.32	0.606	(-0.999, 0.647)	58.74%	-28.74%
PLT mean difference	-0.0027	0.0013	0.084	(-0.006, 0.001)	59.85%	44.80%
HDL
Variable	β	se	p	95%CI	I²_res	Adj R²
Publication (y)	-0.8603	1.2254	0.509	(-3.859, 2.138)	97.97%	-8.13%
Sample size	0.4444	1.2633	0.737	(-3.536, 2.647)	98.15%	-14.90%
Quality score	0.2862	1.4204	0.847	(-3.189, 3.762)	98.18%	-16.43%
Method	1.515	1.284	0.283	(-1.626, 4.656)	98.10%	5.10%
Gender ratio (male ≥ 50%/<50%)	2.635	1.704	0.183	(-1.747,7.016)	97.84%	19.42%
PLT mean difference	-0.007	0.008	0.413	(-0.028, 0.013)	97.73%	-3.41%

Subgroup analyses

To explore potential sources of heterogeneity, predefined subgroup analyses were performed (Table 4). In the TC group, heterogeneity was not adequately explained by study methodology or quality scores, whereas publication year and sample size appeared to be important contributors to heterogeneity. In the TG group, study design type and publication year were likely the main sources of heterogeneity. In the LDL group, study methodology and publication year were key contributors. In the HDL group, differences in study methodology, particularly retrospective study design, were the primary sources of heterogeneity.

Table 4.

Subgroup analysis

TC
Project		Number of studies	I2 (%)	P	SMD (95%CI)
Method
	Prospective cohort	4	87.30%	< 0.001	0.44(-0.03,0.91)
	Retrospective cohort	12	91.20%	< 0.001	0.59(0.36,0.82)
Quality score
	Medium	12	89.40%	< 0.001	0.56(0.33,0.78)
	Low	4	91.70%	< 0.001	0.54(0.08,1.01)
Publication (y)
	< 2020	12	92.20%	< 0.001	0.64(0.40,0.88)
	≥ 2020	4	0.00%	0.392	0.24(0.05,0.44)
Sample size
	< 100	2	0.00%	0.35	0.39(0.05,0.73)
	≥ 100	14	91.20%	< 0.001	0.57(0.36,0.79)
TG
Project		Number of studies	I2 (%)	P	SMD (95%CI)
Method
	Prospective cohort	2	6.50%	0.301	0.09(-0.22,0.40)
	Retrospective cohort	7	62.00%	0.015	0.55(0.38,0.72)
Quality score
	Medium	7	62.60%	0.014	0.51(0.31,0.70)
	Low	2	90.30%	0.001	0.33(-0.34,1.00)
Publication (y)
	< 2020	6	66.10%	0.011	0.57(0.39,0.76)
	≥ 2020	3	14.00%	0.313	0.19(-0.07,0.45)
Sample size
	< 100	1	NA	NA	NA
	≥ 100	8	72.70%	0.001	0.48(0.29,0.68)
LDL
Project		Number of studies	I2 (%)	P	SMD (95%CI)
Method
	Prospective cohort	3	92.70%	< 0.001	0.31(-0.48,1.09)
	Retrospective cohort	6	0.00%	0.441	0.41(0.26,0.55)
Quality score
	Medium	6	52.50%	0.062	0.30(0.08,0.52)
	Low	3	88.60%	< 0.001	0.52(-0.08,1.11)
Publication (y)
	< 2020	4	71.90%	0.014	0.62(0.30,0.94)
	≥ 2020	5	14.40%	0.323	0.18(-0.01,0.37)
Sample size
	< 100	3	50.30%	0.133	0.17(-0.21,0.55)
	≥ 100	6	79.60%	< 0.001	0.45(0.16,0.75)
HDL
Project		Number of studies	I2 (%)	P	SMD (95%CI)
Method
	Prospective cohort	2	0.00%	0.811	0.25(-0.05,0.54)
	Retrospective cohort	6	98.40%	< 0.001	-1.26(-2.26,0.10)
Quality score
	Medium	6	98.30%	< 0.001	-0.95(-2.27,0.38)
	Low	2	97.00%	< 0.001	-0.66(-2.40,1.08)
Publication (y)
	< 2020	3	98.00%	< 0.001	-0.34(-1.96,1.28)
	≥ 2020	5	98.00%	< 0.001	-1.20(-2.61,0.21)
Sample size
	< 100	3	92.80%	< 0.001	-0.60(-1.64,0.44)
	≥ 100	5	98.60%	< 0.001	-1.04(-2.55,0.46)

Discussion

The incidence of HSP in children is rising, and renal involvement remains a major determinant of long-term prognosis [40]. Although renal biopsy is the diagnostic gold standard [41], its invasiveness, technical demands, and sampling variability limit routine use. Therefore, understanding noninvasive correlates of renal involvement is of interest, albeit with clear recognition of their current limitations. Notably, lipid disturbances are well described in chronic kidney disease [42], providing a relevant context for interpreting the metabolic alterations observed in the present meta-analysis.

In this study, we examined the association between lipid metabolism disorders and renal involvement in HSP. The results showed that TC, TG, and LDL levels were significantly higher in the HSPN group than in the HSP group, indicating a robust association between dyslipidemia and established renal disease. However, due to the observational nature of the included studies and the lack of pretreatment baseline data, causal inference is precluded.

Despite substantial heterogeneity, sensitivity analyses confirmed the associations were robust to individual outliers. Nevertheless, wide prediction intervals and sparse reporting of covariates preclude definitive interpretation. The pooled estimates should therefore be considered exploratory benchmarks, highlighting the need for standardized, prospective data collection.

The role of HDL remained uncertain, with extreme heterogeneity (I² > 98%) limiting interpretability. Variability in HDL measurement methods, unmeasured metabolic and inflammatory confounders, and population differences likely contributed to these inconsistencies.

Experimental and clinical studies in chronic kidney disease have identified several biological pathways that could, in principle, link dyslipidemia to renal injury. First, impaired HDL maturation may result from lecithin-cholesterol acyltransferase deficiency, leading to dysfunctional HDL particles [43]. Second, reduced LDL clearance may occur due to decreased LCAT activity, hepatic LDL receptor dysfunction, or reduced hepatic esterase activity [42]. Experimental evidence suggests that increased uptake of LDL cholesterol and inflammatory responses induced by oxidized LDL contribute to proteinuria and accelerate chronic kidney disease progression [44]. Third, decreased triglyceride catabolism [42], often caused by lipoprotein lipase inactivation, may lead to elevated TG levels in renal disorders [45]. Recent studies have also highlighted lipid-induced inflammation and oxidative stress, showing that dyslipidemia can activate the NLRP3 inflammasome in renal cells, promoting the release of pro-inflammatory cytokines such as IL-1β and IL-18 and exacerbating tissue injury [46, 47]. Concurrent activation of the Nrf2/ARE signaling pathway may further amplify inflammatory responses and aggravate glomerular and tubular damage [48].

A critical limitation is the systematic lack of standardized timing relative to nephritis onset and corticosteroid initiation, raising the possibility of reverse causality. Coupled with inconsistent reporting of covariates and geographic concentration within Chinese populations, generalizability is limited. Future prospective studies should prioritize pretreatment baseline measurements to clarify whether dyslipidemia is a driver or merely a concomitant feature of HSPN.

Conclusion

In children with established HSPN, elevated TC, TG, and LDL levels are prevalent, establishing a robust association between lipid dysregulation and renal involvement. However, due to the systemic lack of standardized reporting on lipid measurement timing at the study level, causal inference is precluded, and utility for early prediction cannot be supported. Given the substantial heterogeneity, the specific role of HDL remains uncertain. Future prospective studies are warranted to measure fasting lipid profiles prior to both renal involvement and corticosteroid initiation, aiming to disentangle whether dyslipidemia acts as a primary driver or merely a secondary concomitant phenomenon in HSPN pathogenesis. Such studies must employ standardized protocols with rigorous control for confounding factors to generate high-quality evidence to inform clinical practice.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

HSP

Henoch-Schönlein purpura

HSPN

Henoch-Schönlein purpura nephritis

NOS

Newcastle-Ottawa scale

SMD

Standardized mean difference

CI

Confidence interval

TC

Total cholesterol

TG

Total triglycerides

LDL

Low-density lipoprotein

HDL

High-density lipoprotein

Author contributions

Data curation: Xu Zhang, Resources: Xu Zhang, Zhenlin Yang. Software: Xu Zhang, Jingjing Guo. Supervision: Jinzi Li. Writing – original draft: Xu Zhang, Zhenlin Yang. Writing – review & editing: Xu Zhang, Jinzi Li.

Funding

The present study was funded by the Jilin Health Science and Technology Capacity Enhancement Program (No. 2023JC022), and the Scientific Research Program of Jilin Provincial Department of Education (No. JJKH20240693KJ).

Data availability

The datasets used and/or analyzed in this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Conflict of interest

The authors declare that they have no conflict of interest.