Healthcare providers are cognizant of performing tests that have limited utility. The balance between diagnostic value and necessary risk is considered to varying degrees based on the invasiveness of a test.
Kidney biopsies may be one of the more invasive tests that nephrologists perform. Although many nephrologists would consider a transplanted allograft biopsy to be of lower risk than that for a native kidney, the risk for infection and hemorrhage remain present and require careful consideration between value and risk.
Moreover, many nephrologists (both globally and increasingly in the United States) practice in work environments where there is difficulty in obtaining routine transplant kidney biopsies. Given these considerations, it is often helpful to utilize a prediction model to quantify the need for an invasive diagnostic test. TheRenal BiopsyIndex (1-RBI) is a new model that helps providers determine the need to perform a 1-year surveillance biopsy in deceased-donor kidney transplant recipients.
The index was developed using an internal and external validation cohort. Recipients of a deceased-donor kidney transplant were evaluated from French and Belgian cohorts (the French DIVAT cohort for the internal validation and the Belgian Leuven cohort). Over 1000 patients were included in both validation groups. Patients had a 1-year surveillance biopsy performed in which results were grouped as normal/subnormal histology versus abnormal histology based on grade of interstitial fibrosis/tubular atrophy, T/B cell-mediated rejection, and recurrent/de novo glomerulonephritis.
These patients also had various demographic and blood testing performed, which were incorporated into a logistic regression model to predict the likelihood of abnormal histology. The area under the curve (AUC), positive predictive value (PPV), and negative predictive value (NPV) were ascertained to establish a threshold level above which the likelihood of abnormal histology was sufficiently great to warrant a 1-year surveillance biopsy. The internal and external validation cohorts were used to create and test the model, respectively.
Of the various parameters included in the model, only five were useful in predicting abnormal histology: recipient sex; anti-class II immunization status; andcreatininelevels at 3, 6, and 12 months posttransplant. A logistic regression model using these five variables had an AUC of 0.71 and a PPV of 71% in the internal validation cohort when the 1-RBI threshold was set to 2.81. Externally, the results were similar (AUC, 0.62; PPV, 70%) at the same threshold.
The authors concluded that the 1-RBI had enough statistical power to persuade an undecided physician to perform a 1-year surveillance biopsy if the index returned a value of 2.81 or more. Unfortunately, with NPVs under 60% in both cohorts, the 1-RBI did not have enough power to dissuade a convinced physician to avoid a 1-year surveillance biopsy.
Thus, the 1-RBI model itself has limited predictive power, but can be useful for "constrained" providers who don't have an easy method of performing/obtaining transplant biopsies. In theory, these providers can use the 1-RBI to judiciously use their limited resources to identify the most at-risk transplant recipients using relatively available data.