Tara Haelle

December 09, 2024

132

Added to Email Alert

WASHINGTON — An investigational 12-protein panel of urinary biomarkers predicted histologically active lupus nephritis (LN) with 86% accuracy, according to research presented at the American College of Rheumatology (ACR) 2024 Annual Meeting.

The noninvasive biomarker panel “robustly predicts meaningful and actionable histological findings” in patients with active proliferative LN, Andrea Fava, MD, assistant professor of medicine in the Division of Rheumatology at Johns Hopkins Medicine in Baltimore, told attendees.

“In contrast to proteinuria, which can’t differentiate inflammation from damage, this panel for histological activity includes a set of 12 proteins linked to intrarenal inflammation,” said Fava, who is also director of Lupus Translational Research at Johns Hopkins. A decline in the biomarker score at 3 months predicted a clinical response at 1 year, and persistent elevation of the score at 1 year predicted permanent loss of kidney function, “which makes it tempting as a treatment endpoint,” Fava said. “Upon further validation, this biomarker panel could aid in the diagnosis of lupus nephritis and guide treatment decisions.” 

Alfred Kim, MD, PhD, an associate professor of medicine at Washington University in St. Louis, was not involved in the research but noted the potential value of a reliable biomarker panel.

“If we have urinary biomarkers that strongly associate with histologic activity, this would be a game changer in the management of LN,” Kim told Medscape Medical News. “Right now, the gold standard is to perform another kidney biopsy to determine if therapy is working. But this is invasive, and many patients do not want to do another kidney biopsy. Conversely, the easiest way to assess lupus nephritis activity is through a urinalysis, focusing on urinary protein levels,” but relying on proteinuria has limitations as well.

“The most important [limitation] is that proteinuria cannot distinguish treatable inflammation from chronic damage,” Fava said. Persistent histologic activity in patients without proteinuria predicts flares, but tracking histologic activity, as Kim noted, requires repeat biopsies.

“So we need better biomarkers because biomarkers that can reflect tissue biology in real time can guide personalized treatment, and that’s one of the main goals of the Accelerating Medicines Partnership [AMP],” he said. The AMP is a public-private partnership between the National Institutes of Health (NIH), the US Food and Drug Administration (FDA), multiple biopharmaceutical and life science companies, and nonprofit and other organizations. Lupus is one of the AMP’s funded projects.

Kim agreed that “effective biomarkers are a huge unmet need in LN.” Further, he said, “imagine a world where the diagnosis of LN can be made just through urinary biomarkers and obviate the need for biopsy. Both patients and providers will be ecstatic at this possibility.” 

Fava described the background for how his research team determined what biomarkers to test. They had previously enrolled 225 patients with LN undergoing a clinically indicated kidney biopsy and collected urine samples from them at baseline and at 12, 24, and 52 weeks after their biopsy.

Of the 225 patients included, 9% with only mesangial LN (class I-II), 25% with pure membranous LN (class V), 24% with mixed LN (class III or IV with or without V), 38% with proliferative LN (class III or IV), and 4% with advanced sclerosis LN (class VI). From these samples, they quantified 1200 proteins and looked at how they correlated with histologic activity.

“What was interesting was that in patients who were classified as responders after 1 year, there were many of these proteins that declined as early as 3 months, suggesting that effective immunosuppression is reducing intrarenal inflammation, and we can capture it in real time,” Fava said.

Biomarker Panel Predicts Histologically Active LN

So they set to determining whether they could develop a urinary biomarker for histologically active LN that could be useful in clinical decision-making. They focused on one that could detect active proliferative LN with an NIH activity index score > 2. Their 179 participants included 47.5% Black, 27.9% White, and 14.5% Asian participants, with 10.1% of other races. The predominantly female (86.6%) cohort had an average age of 37 years. Among the LN classes, about one third (34.6%) had pure proliferative disease, 17.9% had mixed proliferative, 27.9% had pure membranous, 11.7% had class I or II, and 5% had class VI. Just over half the participants (55.7%) had not responded to treatment at 12 months, whereas 25% had a complete response, and 19.3% had a partial response.

However, both the 78 participants with an NIH activity index score > 2 and the 101 with a score ≤ 2 had a median score of 3 on the NIH chronicity index. And the urine protein-to-creatinine ratio — 2.8 in the group with an NIH activity index score > 2 and 2.4 in the other group — was nearly indistinguishable between the two groups, Fava said.

They then trained multiple algorithms on 80% of the data to find the best performing set of proteins (with an area under the curve [AUC] of 90%) for predicting an NIH activity index score > 2. They reduced the number of proteins to maximize practicality and performance of the panel, Fava said, and ultimately identified a 12-protein panel that was highly predictive of an NIH activity index score > 2. Then, they validated that panel using the other 20% of the data. The training set had an AUC of 90%, and the test set was validated with an AUC of 93%.

The 12-protein panel score outperformed anti-dsDNA, C3 complement, and proteinuria, with a sensitivity of 81%, a specificity of 90%, a positive predictive value of 87%, a negative predictive value of 86%, and an accuracy of 86%. The proteins with the greatest relative importance were CD163, cathepsin S, FOLR2, and CEACAM-1.

“In contrast to proteinuria, these proteins were related to inflammatory processes found in the kidneys in patients with lupus nephritis, such as activation of macrophages, neutrophils and monocytes, lymphocytes, and complement,” Fava said.

When they looked at the trajectories of the probabilities from the biomarker panel at 3, 6, and 12 months, the probability of the NIH activity index score remaining > 2 stayed high in the nonresponders over 1 year, but the trajectory declined at 3 months in the responders, indicating a decrease in kidney inflammation (P < .001).

Can the Biomarker Panel Serve as a Treatment Endpoint?

Then, to determine whether the panel could act as a reliable treatment endpoint, the researchers followed the patients for up to 7 years. One third of the patients lost more than 40% of their kidney function during the follow-up. They found that a high urinary biomarker score at 12 months predicted future glomerular filtration rate loss, independent of proteinuria.

This panel was tested specifically for proliferative LN, so “we may need distinct panels for each [LN type] to capture most of these patients,” Kim said. “I think that’s where the gold mine is: A personalized medicine approach where a large biomarker panel identifies which smaller panel that patient best fits, then use that for monitoring.”

Kim did note an important potential limitation in the study regarding how samples are used in biomarker discovery and validation vs in clinical practice. “Most samples in research studies are frozen, then thawed, while urine is assayed within a couple hours after collection in the clinical setting,” he said. “Do sample processing differences create a situation where a biomarker works in a research project but not in the clinical setting?” But more likely, he said, the opposite may be the case, where frozen samples allow for more degradation of proteins and potentially useful LN biomarker candidates are never detected.

Another challenge, Kim added, albeit unrelated to the study findings, is that diagnostic companies are finding it difficult to get payers to cover new tests, so that could become a challenge if the panel undergoes further validation and then FDA qualification.

The research was funded by Exagen. Fava reported disclosures with Arctiva, AstraZeneca, Exagen, Novartis, UCB, Bristol Myers Squibb, Annexon Bio, and Bain Capital. His coauthors reported financial relationships with numerous pharmaceutical and life science companies, including Exagen, and some are employees of Exagen.

Kim reported research agreements with AstraZeneca, Bristol Myers Squibb, Novartis, and CRISPR Therapeutics; receiving royalties from Kypha; and receiving consulting/speaking fees from AbbVie, Amgen, Atara Bio, Aurinia, Cargo Tx, Exagen, GlaxoSmithKline, Hinge Bio, Kypha, and UpToDate.

Tara Haelle is a Dallas-based science/health journalist.

1