LN remains an essential manifestation of jSLE and plays a role in the long-term prognosis and complications. In Korean and recent Japanese studies, subsequent LN was associated with poorer prognosis and treatment outcomes than early LN [15, 16]. This study found that elevated average anti-dsDNA antibody, low average serum complements, and high average ESR levels were associated with subsequent LN in patients with jSLE. After adjustment, high average anti-dsDNA antibody and ESR levels remained key factors for subsequent LN during serial disease follow-ups.

Studies of patients with SLE reported that 50% to 80% of patients were affected by LN during the disease course [19, 21,22,23,24,25, 32]. Among 103 patients with jSLE in our study, 75 cases (72.8%) had ever developed LN. The variation in LN prevalence in different jSLE cohorts may be due to patient ethnicity, different methods used for LN diagnosis, and length of follow-up period. Studies concerning subsequent renal involvement in patients with SLE were mainly conducted in patients with aSLE. According to different definitions, subsequent LN accounts for 20% to 40% of all SLE patients with LN [13,14,15,16]. In our data, 55 patients (53.4%) had evident renal involvement at TD or within one month after TD, and the other 20 patients, 19.4% (20/103) of all jSLE, further developed LN.

Several studies focused on determining the risk factors for delay-developed LN or incident proteinuria, which was used as a surrogate for LN. Most of the research targeted patients with aSLE, and the results varied in different studies due to different definitions. Our study focused on patients with jSLE and defined those with definite LN during follow-up as having subsequent LN instead of incident proteinuria only, which may represent a more profound renal involvement in SLE. Unlike most previous studies, we compared not only the initial manifestations and laboratory data at TD but also the levels of serum serologic markers during a follow-up period.

The status of anti-dsDNA antibody at TD or case enrolment was the most frequently mentioned. Anti-dsDNA IgG, classified as one of the nephritogenic autoantibodies, was widely discussed in earlier studies [33] and is believed to play a role in the development of LN. Previous studies focused on mechanism elucidation suggested that anti-dsDNA antibody may contribute to LN development through direct or non-direct binding to chromatin materials or cross-reactive antigens; however, these results are mainly based on murine models or in vitro studies [34,35,36,37,38]. Reports have indicated that an elevation of anti-dsDNA antibody levels is often observed prior to SLE flares [39, 40] but decreases during lupus flares, including renal flares [39, 41], which may be explained by the deposition of anti-dsDNA immune complexes in tissues. In previous studies, initial positive anti-dsDNA antibody were related to subsequent LN or incident proteinuria [7, 42,43,44,45]. The Korean data revealed that aSLE patients developing subsequent LN had a higher anti-dsDNA antibody titre at the TD with an adjusted hazard ratio of 1.004 (95% CI: 1.000–1.007, p = 0.026) [45]. In our study, there was no difference in the positivity of anti-dsDNA antibodies at the TD between the two groups. The higher positivity rate of anti-dsDNA antibodies in jSLE than in aSLE, regardless of proteinuria, may explain this discrepancy [46, 47]. However, when we compared the average levels of anti-dsDNA antibody during the following periods, higher mean anti-dsDNA antibody titres correlated strongly with subsequent LN development. For jSLE patients with higher average anti-dsDNA antibody levels, the risk of subsequent LN increased by a 1.29-fold scale fashion every time with an increment of 100 IU/ml of anti-dsDNA antibody. Although the examination kits may vary in different hospitals, and our unit for anti-dsDNA antibody presentation is not universal, the data can be shifted to other universal units using the manufacturer's equation. The hazard ratio increased significantly, primarily in patients with strongly positive anti-dsDNA antibodies. This caused a 5-year LN-free survival of 37.5% after SLE diagnosis, compared with the 100% survival rate in those with nearly negative anti-dsDNA antibody during follow-ups. The findings indicated that persistently or fluctuating high anti-dsDNA antibody levels during clinical follow-ups were important signs of subsequent LN.

In our study, patients with jSLE with a lower average complement level, both C3 and C4, tended to have a higher risk of subsequent LN in the long-term follow-up. A Korean study for patients with aSLE [45] and the UK JSLE Cohort Study for jSLE [32] both reported that a lower C3 level at TD was a risk factor for developing subsequent LN. Lower complement levels have also been associated with increased renal disease activity [48] and renal flares [41] in patients with LN. The core of LN development lies in the intra-renal immune complex deposition that causes complement activation, inflammation, and further kidney damage [49].The alternative complement pathway was also implicated by previous studies that it plays a crucial role in complement-mediated damage in LN [50, 51]. Persistent low C3 levels in our patients with subsequent LN may reflect activation of the alternative complement pathway in the kidneys. Differences in C3 or C4 levels at TD between the two groups were not noted in our study as other studies did, which may be due to our patients’ active disease status at TD. Several studies have also pointed out that patients with jSLE have more episodes of complement level decline and a higher percentage of low C3 compared with patients with aSLE, regardless of the presence of LN [20, 23, 32, 52, 53].

ESR, an inflammation marker, is one of the indices evaluated in the Systemic Lupus Activity Measure (SLAM). Elevated ESR is associated with renal [54] and overall disease activity in SLE [54, 55]. We found that an elevated ESR was related to subsequent LN, and the effect remained after adjusting for other factors. None of the other studies on subsequent LN reported the role of ESR, but active disease status, including higher ACR scores at TD, or higher SLEDAI scores at TD, has been reported to be a risk factor for subsequent LN in patients with jSLE in previous studies [21, 32]. Although we lacked serial data on disease activity in our cases, we found that patients with subsequent LN received a higher accumulated steroid dosage before LN onset than the other group. The effect diminished after adjustment for other factors related to disease activity, such as complement levels and anti-dsDNA antibody titres, indicating higher disease activity and the requirement for higher steroid dosage in Group 1 patients. For other medications, previous studies revealed that antimalarial use was associated with a lower rate of LN [5, 56, 57]. Our data showed no correlation between the other treatments and subsequent LN. However, nearly all our patients using antimalarial for disease control unless contraindicated that we were unable to evaluate the protective effect of using antimalarial as previous studies mentioned.

Regarding the skin manifestations at TD, we found that those with discoid rashes were less likely to have subsequent LN overall. Several studies have reported that patients with SLE with mucocutaneous lesions tend to have less vital organ involvement [58, 59]. A British study that included 241 patients with jSLE found that those without skin manifestations tended to have more haematological and renal involvement [60]. Another cohort of 47 jSLE patients in the U.S. also pointed out the inverse trend of the mucocutaneous lesions and renal involvement in patients with jSLE (OR: 0.5, 95% CI: 0.2–0.9) [22]. Our findings are consistent with those of the above studies. However, the mechanism and relationship of this inverse correlation between discoid rash and LN requires further research.

Our study had several limitations. First, we had a relatively small sample size due to the restriction of the juvenile-onset population, which is generally one-tenth of aSLE [47]. Second, based on the retrospective data collection design, we lacked the serial disease activity score and objective assessment for drug adherence in our study, and there were some incomplete data, such as the status of autoantibodies other than anti-dsDNA and anti-Sm antibodies. The relationship between subsequent LN or proteinuria and other autoantibodies, such as anti-Sm [44, 61], anti-histone [62], anti-RNP [44], and anti-cardiolipin antibodies [10, 43], have also been reported in a few studies. Establishing a relationship between these autoantibodies and subsequent LN development in jSLE may require more comprehensive data. Third, because of a lack of available nephrologists or refusal by the family, only a small portion of our LN cases received renal biopsy, which is currently the gold standard for LN diagnosis. To make a firm LN diagnosis, we used a relatively rigid definition of LN for patients without renal biopsy, which included persistent proteinuria plus the presence of RBC or WBC cellular casts. Last, the patients with jSLE in our study were observed for various durations. Those who remained free from LN may have had subsequent LN if we extended the observing periods. To minimise this possibility, all patients in Group 2 had regular follow-up records for more than one year, and the shortest follow-up duration was 1.65 years. In addition, the mean follow-up durations of the two groups were similar.

Previous studies that searched for risk factors of subsequent LN checked the clinical features or serologic biomarkers at a single period; nevertheless, they did not further consider the change in these markers in the follow-ups. Our study is novel in that we focused on patients with jSLE and collected data not only from TD but also from a full-length follow-up period. To understand the course and prevent the subsequent LN more comprehensively, we still require further investigations involving the dynamic change in disease activity of SLE and other more sensitive biomarkers for early LN detection.