New statistical model predicts AAV prognosis with good accuracy

Lung involvement, hemoglobin levels, subtypes predictive factors in survival

Marisa Wexler, MS avatar

by Marisa Wexler, MS |

Share this article:

Share article via email
A risk dial is turned to

A new statistical model that considers age, disease type, patterns of organ involvement, and lab findings may help predict long-term survival in ANCA-associated vasculitis (AAV) better than available prognostic systems, a study suggests.

The study “Development and internal validation of a model to predict long-term survival of ANCA associated vasculitis,” was published in Rheumatology and Immunology Research.

AAV can be a fatal disease, though outcomes continue to improve as advances in care and new treatments are developed. While “risk stratification and prognosis prediction are critical for appropriate management of” AAV, a model for prognosis prediction is still in need,” wrote researchers in China who analyzed clinical data from 653 people with AAV treated at a single center in the country between 1999-2019 to create one.

Regarding the types of AAV, 303 patients (46.4%) had microscopic polyangiitis (MPA), 245 (37.5%) had granulomatosis with polyangiitis (GPA), and 105 (16.1%) had eosinophilic granulomatosis with polyangiitis (EGPA). Over a median follow-up time of nearly three years, 120 patients (18.4%) died.

After evaluating 20 parameters as potential risk factors, and therefore candidate predictors, the researchers selected the six most influential factors for their final model. These included age at admission, AAV type, organ involvement patterns (lung and cardiovascular involvement), and laboratory information, including blood levels of creatinine and hemoglobin. Creatinine is a marker of kidney damage and hemoglobin is the protein in red blood cells that carries oxygen through the body.

Recommended Reading
S100 proteins | ANCA Vasculitis News | illustration of researcher using microscope in lab

Immune Cells Work as ‘Friend and Foe’ in ANCA-associated Vasculitis

Factors in predicting AAV prognosis

In the final model, the prognosis was generally poorer for those who were older, had lung and/or heart involvement, had high blood levels of creatinine, or low hemoglobin levels. It was also poorest for MPA patients, followed by those with GPA and then EGPA.

To evaluate the model’s predictive potential, the researchers calculated the C-index, which reflects how the number of expected deaths (based on the model) lines up with the actual number. C-index values can range from 0.5 to 1, with higher numbers reflecting better accuracy at discriminating between low- and high-risk patients.

“As a rule of thumb, C-index over 0.7 indicates modest or acceptable discriminative ability,” the researchers wrote.

Results showed the 5-year C-index for the model was about 0.75, suggesting a fairly good predictive ability. Its C-index scores were generally higher than those of two currently used AAV prognostic systems — the revised five factor score (rFFS) and the Birmingham vasculitis activity score (BVAS) system.

“Compared with the rFFS and the BVAS system, the current prediction model had higher predictive power for overall survival,” the researchers wrote, adding that “in a quite wide range of threshold probabilities, our prediction model had higher net benefits, implying probable better clinical values.”

“In addition to the factors included in the widely used rFFS system, our prediction model showed that lung involvement, hemoglobin levels, and AAV subtypes” can help predict survival, they wrote.

The rFSS system considers an age over 65 and the presence of heart, kidney, and/or gastrointestinal involvement as risk factors of poorer survival, and the presence of ear, nose, throat (ENT) involvement as a protective factor.

As most patients included as part of the analysis had MPA in whom ENT involvement was less common, “ENT signs were not included in our model,” the researchers said.

While the new model may be useful for predicting AAV prognosis, the researchers acknowledged their analysis was limited by the use of retrospective data from a single specialty center in China and that more research is needed to test whether their model performs well for a broader population.