Lupus nephritis (LN) is often characterised by flares of activity alternated with periods of quiescence, generally induced by therapy. While complete remission is a strong predictor of a good renal outcome,1 the persistence of nephritic signs and/or the development of renal flares are associated with a poor renal prognosis.2 The diagnosis of LN activity relies on active urine sediment, worsening of proteinuria and deterioration of renal function but, in some cases, it may be difficult to recognise whether these signs are actually caused by the activity of systemic lupus erythematosus (SLE) or by other non-immunological causes. Renal biopsy may be necessary in these difficult cases but this invasive procedure is not devoid of complications. Thus, an immunological biomarker that reliably measures disease activity, could be of great value. However, whether and which serological test can reliably measure the clinical activity of SLE and in particular of LN is still a matter of debate.

A number of studies have demonstrated that anti-dsDNA antibodies (anti-dsDNA), as well as complement fractions may be useful in assessing the disease and the renal activity in some3 4 but not in all studies.5 More recently a significantly higher frequency of anti-C1q antibodies (anti-C1q) has been reported in patients with LN than in those without.6^–10 Moreover, the presence of anti-C1q in patients with SLE has been found to herald the development of LN.6 9 10 In patients with established LN the presence and the titres of anti-C1q significantly correlated with renal activity in many7 8 10^–12 but not all studies.13

In this prospective study the levels of anti-dsDNA, anti-C1q, C3 and C4 complement fractions were serially measured throughout 6 years in a large cohort of patients with LN to evaluate their role in differentiating between active and inactive LN.

PATIENTS AND METHODS

Participants

From January 2000 to December 2005, all patients with SLE, defined according to the American College of Rheumatology criteria14 and LN, followed by Renal Units of Fondazione Ospedale Maggiore, and Azienda Ospedaliera Ospedale San Carlo Borromeo, Milano were enrolled.

Objectives

The aim of this prospective study was to assess the possible role of C3, C4, anti-dsDNA and anti-C1q in differentiating patients with active and inactive LN.

Definition of activity

At each clinical examination the activity of LN was classified as follows:2

0 = Complete renal remission: normal renal function for at least 6 months, proteinuria <0.5 g/24 h, urinary red blood cells <5/hpf.

1 = Partial renal remission: for nephritic flare: improvement of at least 30% of serum creatinine but persistence of active urinary sediment. For proteinuric flares improvement of 50% of proteinuria.

2 = Nephritic flare: increase of 30% of serum creatinine over the basal value and active urinary sediment (>10 red blood cells/hpf, cellular casts) with or without an increase in proteinuria.

3 = Proteinuric flare: Increase of proteinuria of at least 2 g/day in patients with non-nephrotic syndrome or the doubling of nephrotic proteinuria with stable renal function.

4 = Persistent renal activity: the lack of achievement of remission after induction therapy.

Renal biopsies were classified according to the ISN/RPS classification.15

Laboratory investigations

Anti-dsDNA antibodies were measured by a commercial quantitative enzyme-linked immunosorbent assay (Varelisa dsDNA Antibodies, Phadia GmbH, Freiburg, Germany); C3 and C4 plasma levels by nephelometry (Nephelometer Analyser II, Behring, Marburg GmbH, Germany).

Anti-C1q antibodies were detected using a home-made enzyme-linked immunosorbent assay as described by Sinico et al.8

Statistical analysis

Mean and standard deviation, together with median and interquantile (IQ) range (25–75th percentile) were used as descriptive statistics. For continuous variables, the non-parametric Wilcoxon test was used for assessing any difference between the two groups of patients, while the χ² test was used for dichotomised variables.

Multivariate logistic regression analysis has been used to find predictors of the probability of flare occurrence. Odds ratios (OR) and their 95% confidence interval (CI) for the covariates were derived as the antilogarithm of the regression coefficients.

The statistical package S-Plus (MathSoft, Cambridge, Massachusetts, USA), was used for all the analyses and plots.

RESULTS

We enrolled 228 patients with a diagnosis of LN. LN was diagnosed 101 (102) months before the beginning of the study. The main clinical and histological features at diagnosis of LN are summarised in table 1.

During the study these patients were submitted to 1523 evaluations: 704 on patients in complete remission, 471 on patients in partial remission, 45 at diagnosis of a nephritic flare, 97 at diagnosis of a proteinuric flare and the remaining 206 in patients with persistent renal activity after the beginning of induction therapy for flares.

Immunological tests in active and inactive lupus nephritis

One thousand and 81 tests were performed on the 144 patients with proliferative LN (fig 1). During complete or partial remission the median values of each test were in the normal range or approaching the normal value. During nephritic and proteinuric flares, C3 and C4 levels significantly decreased (p = 0.00001), while anti-dsDNA and anti-C1q titres significantly increased (p = 0.00001). No significant differences were documented in median values of C3 and C4 between tests performed during flares and those evaluated after the beginning of the induction therapy in patients with persistent renal activity, while the median values of anti-C1q and anti-dsDNA, although still in a pathological range, significantly decreased after the beginning of therapy (p = 0.0001).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Median values and interquartile ranges of the four immunological parameters in the different clinical settings in patients with proliferative glomerulonephritis. (A) C3, (B) C4, (C) anti-dsDNA, (D) antiC1q. Boxplots have been used as a graphical tool. In a boxplot a line shows the median, a box shows the IQ range, and whiskers are drawn to the nearest value not beyond a standard span (ie, 1.5×IQ range) from the quartiles; points beyond (outliers) are drawn individually. 0 = complete renal remission/inactive disease; 1 = partial renal remission; 2 = nephritic flare; 3 = proteinuric flare; 4 = persistent renal activity. Normal values: C3 >80 mg/dl, C4 >15 mg/dl, anti-dsDNA <100 U/ml, antiC1q <100 UA.

For the 364 tests performed in 49 patients with the non-proliferative forms of LN, the median values of C3 and C4 and that of anti-dsDNA did not change significantly in the different settings of renal activity (fig 2). The only significant differences were between the median value of anti-C1q at proteinuric flares and those at complete remission (p = 0.02), even if 46% of flares in this group occurred with normal value of anti-C1q in comparison with only 20% that developed in proliferative forms (p = 0.02).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Median values and interquartile ranges of the four immunological tests in the different clinical settings in patients with non-proliferative glomerulonephritis. (A) C3, (B) C4, (C) anti-dsDNA, (D) antiC1q. Boxplots have been used as a graphical tool. In a boxplot a line shows the median, a box shows the IQ range, and whiskers are drawn to the nearest value not beyond a standard span (ie, 1.5×IQ range) from the quartiles; points beyond (outliers) are drawn individually. 0 = complete renal remission/inactive disease; 1 = partial renal remission; 2 = nephritic flare; 3 = proteinuric flare; 4 = persistent renal activity. Normal values: C3 >80 mg/dl, C4 >15 mg/dl, anti-ds DNA <100 U/ml, anti-C1q <100 UA.

In patients with antiphospholipid antibodies the median value of anti-C1q at renal flares was significantly lower (median 155; 25th and 75th percentile 66–298) than those of patients without (median 267; 25th and 75th percentile 116–320; p = 0.03). As a matter of fact, in patients with antiphospholipid antibodies, 33% of flares occurred with a normal value of anti-C1q, in comparison with only 14.5% flares that occurred in negative patients (p = 0.02).

We computed the sensitivity, specificity and their positive and negative predictive values of the four tests for detecting a flare in patients with LN. Although anti-C1q assured slightly better results in terms of sensitivity (anti-C1q: 80.5%, anti-dsDNA: 70%, C3: 79%, C4: 74%), and specificity (anti-C1q: 71%, anti-dsDNA 67%, C3: 51%, C4: 64%), 20% of patients developed a flare in the presence of normal value anti-C1q, while 30% of patients in remission continued to have pathological values of anti-C1q. None of the four immunological tests had a satisfactory positive predictive value.(anti-C1q: 38%; anti-dsDNA: 31%; C3: 28%; C4: 31%). All four tests had a good negative predictive value (anti-C1q: 94%; anti-dsDNA: 91%; C3: 93%; C4: 92%), underlying that, in the presence of a normal value of each, active LN is quite improbable. At univariate analysis the best renal flare predictor was anti-C1q (p<0.0005; OR: 12.7; 95% CI: 6.3 to 25). At multivariate analysis the best model for renal flares prediction was obtained by combining anti-C1q (p<0.0005; OR: 11.8; 95% CI: 4.9 to 8.1) with C3 (p<0005; OR: 2.99; 95% CI: 1.5 to 5.8) and C4 (p<0.005; OR: 3.3; 95% CI: 1.7 to 6.5).

DISCUSSION

We have shown that, in patients with proliferative LN, all the four tests were able to differentiate between renal flares and quiescent renal disease. All of them had good negative, but unsatisfactory positive predictive values. A practical message from these data is that, in the presence of normal values of all these tests, the activity of LN is quite improbable. Anti-C1q antibodies achieved a slightly better performance in terms of sensitivity and specificity and, at univariate analysis anti-C1q is the best renal flare predictor. Nevertheless, a small number of patients developed renal flares in the presence of normal values of anti-C1q. At multivariate analysis, the association of anti-C1q with C3 and C4 provided the best prediction of renal flares.

Noteworthy, in the non-proliferative forms only anti-C1q was able to differentiate between renal flares and quiescent renal disease. However, only 54% of flares that developed in the non-proliferative forms were associated with high titres of anti-C1q in comparison with 80% of those that occurred in patients with the proliferative forms. These data outline that anti-C1q is less reliable for assessing the activity of the non-proliferative forms. In patients with antiphospholipid antibodies, anti-C1q seem to be less reliable in predicting renal flares as a high number of flares in these patients occurred with normal values of anti-C1q.

In conclusion, our data underline that renal exacerbations seem to be quite improbable in the presence of normal values of C3, C4, anti-dsDNA and anti-C1q. None of these tests are completely reliable in confirming the clinical diagnosis of renal relapses; however, our results (which were obtained in a large number of patients with SLE prospectively evaluated for 6 years) suggest that anti-C1q is slightly better than the other tests to confirm the clinical activity of LN, particularly in patients with focal and diffuse proliferative LN and in the absence of antiphospholipid antibodies.