Optimisation of biopsy practice to facilitate timely diagnosis and management of kidney disease

Abstract

Background: A kidney biopsy is an invasive investigation to diagnose intrinsic forms of kidney disease such as glomerulonephritis (GN). There is a rare risk of significant complications including bleeding requiring transfusion, embolisation or death. Previous literature is limited in terms of the patient experience and studies of practice patterns have demonstrated significant national and international variation. Immunoglobulin A Nephropathy (IgAN) is the most common form of GN diagnosed by kidney biopsy. Outcomes in IgAN are variable however affected patients have a high lifetime risk of end stage kidney disease (ESKD). Risk tools with and without data obtained on biopsy can be utilised to risk stratify affected patients so that treatments are reserved for individuals at greatest risk of progressive disease. Objectives: 1. Explore the experience and opinions on kidney biopsy practice from the perspective of patients, nephrologists and healthcare systems 2. Describe the longitudinal course of IgAN and examine the utility of tools to predict the risk of progressive disease Methods: A mixed methods approach was taken. Individuals who had a kidney biopsy in the past were invited for semi-structured interview to describe their experience. This led to a co-designed questionnaire for nephrologists to explore attitudes on kidney biopsy practice. Nephrologists were contacted by email and social media to complete the questionnaire. A biopsy-propensity score was generated from responses to assess willingness to biopsy and tolerance of potential risks. A subgroup analysis of US clinicians was also undertaken. Large language model outputs were compared with the human consensus from this questionnaire. An anonymous retrospective cohort of incident cases of IgAN from 2011-2022 in Northern Ireland were analysed to describe the longitudinal disease course and identify risk factors for progressive kidney disease. A systematic review was undertaken to compare the Kidney Failure Risk Equation (KFRE) and Oxford classification (OC) found on kidney biopsy for prediction of progressive disease. These risk tools were then assessed using the Northern Ireland IgAN cohort to assess this differentiation using receiver operating curves and calibration curves at three annual intervals within the first two years from diagnosis. Results: Patients varied in their experience of kidney biopsy. The main themes were: the health condition was subtle, resilience was tested and a patient-centred approach was valued. 1181 nephrologists from 83 countries participated in the questionnaire. There was international variation in practice within and between countries. Urine abnormalities were the main driver of biopsy propensity and reduced kidney volume was the main inhibitor. An adjusted linear regression model was applied to the propensity score, which identified that younger male doctors who frequently performed the procedure were more likely to recommend biopsy. In the United States, increased biopsy propensity was reported in more affluent states with higher nephrologist coverage. LLM performance of this task varied widely and Open AI’s Chat GPT 3.5 and GPT 4.0 were the optimal models in recreating typical human clinician practice. Longitudinal analysis in IgAN demonstrated a significant reduction in proteinuria in most participants within the first year of biopsy. Higher OC score and time-averaged proteinuria were associated with more ESKD events. Individuals with over 1g/day of proteinuria at biopsy had significantly less ESKD events if this level was successfully reduced to less than 1g/day at one year from biopsy (p<0.001). A systematic review found only one eligible study which compared the KFRE and OC, which had significant limitations. Comparison of the KFRE and IgA Risk Prediction Tool, incorporating the OC in the IgAN cohort demonstrated excellent discrimination in these tools particularly at one year from biopsy, but modest calibration, with persistent underprediction of true risk. Discussion: Kidney biopsy has been in common nephrology practice for many decades, however this may change in the future due to the developments in precision medicine. Patients and clinicians value the importance of a kidney biopsy to guide management, however there was variation in experience and opinions on when this should be used. LLMs varied in their ability to replicate typical clinical practice. IgAN is a common cause of CKD and ESKD and proteinuria levels correlated closely with kidney outcomes. Risk tools could accurately differentiate the individuals at the greatest risk of ESKD however they were prone to underprediction and observed outcomes exceeded those that were predicted. IgAN is a condition undergoing rapid changes in terms of diagnostics and treatment options. As these changes continue, identifying which patients require new treatments will likely become of increased clinical importance. Thesis is embargoed until 31 December 2027.