Turning Routine Data to Predictive Care: Dr. Navdeep Tangri on CKD Risk Stratification using Algorithms

What drove the development of the Kidney Failure Risk Equation (KFRE), and how does it address the current unmet needs in CKD, specifically within the broader Cardiovascular-kidney-metabolic (CKM) context?

Kidney disease is a global public health problem. 90% of patients with chronic kidney disease are not managed by nephrologists [1]. They are managed by cardiologists, endocrinologists and family doctors, largely. And the problem is that we have an asymptomatic disease, and the patients are not aware and the physicians are not recognising it. And that’s because we live in a creatinine or eGFR-based paradigm of kidney disease. And people only start to react when the creatinine values become abnormal. And by that time, more than half of kidney function is lost [2]. And highly effective therapeutics are no longer indicated. In APAC, it’s actually worse than the rest of the world [3]. In places like Thailand and Malaysia, the prevalence of chronic kidney disease is 15 to 17% [4], compared to 10 to 12% in North America and European countries [5].

We built the KFRE right around 2011. And at that time I was struck as a nephrologist that all my patients always wanted to know, why do I have kidney disease and am I going to be on dialysis? Most of the time, the nephrologist could tell them why they had kidney disease. But when it came time to the second most important question on their list, which is, am I going to be on dialysis? Nephrologists were just guessing. We realised that this could be predicted. And nephrology, which is in a lab based disease can be predicted almost entirely based on lab data [6]. So we created the KFRE to fulfil this urgent unmet need.

The uptake of the KFRE incorporated as a 1A recommendation in the clinical practice guidelines [7] in 2024, speaks to how slow science moves to practice. The KFRE has been successfully adopted is because it’s easy to implement. It can be integrated in labs, it can be integrated in electronic medical records. And because it does not require a lot of data that’s outside the lab [8]. So you don’t need blood pressures, you don’t need complex variables like race, ethnicity. You don’t even need to define things like diabetes, hypertension and heart disease, where each doctor may have a different definition.

Moving from reactive to predictive care is a core theme of your work. What are the unmet needs in early CKD detection, and how can routine laboratory data help address them?

Early detection in chronic kidney disease was not important up until about the year 2015. The landscape of treatment for chronic kidney disease has evolved. There’s an asymptomatic disease which is picked up late. And now we have highly effective treatments [9]. So I felt there was a new unmet need. The unmet need now is early detection and connecting the dots from detection to triage to treatment. When you combine state of the art machine learning methods with lab data, you can really predict really important outcomes to patients and providers.

After the development of the KFRE score, you moved on to explore AI-powered tools. How is the integration of AI shaping your approach to risk prediction?

So I think of artificial intelligence applications in medicine. We’ve been actually using these methods for more than 20 years. And what’s really changed is the speed at which we can do work now and the speed of implementation as well as the acceptance by the medical community. Now we always adhere to the principles of transparent AI. So when we’re building predictive algorithms. We’re being really transparent in our methods. We’re showing them the weights of the variables. But what we’ve seen in the last 5 years is that physicians are now waking up to this idea more and more. For KFRE, it took 13 years to go from publication to 1A recommendation in guidelines. I think with the societal acceptance of AI and with the coding tools and the implementation tools we now have, we hope to shrink that gap remarkably.

The KFRE is a great tool for nephrologists when they need to make nephrologist decisions. Nephrologists decisions are: who needs education about dialysis? Who needs to be seen more frequently? Who needs to have planning for transplant? Those are very late stage kidney disease decisions. Where the KFRE falls short, is in detection of earlier stages of kidney disease and prognostication and guidance for treatment in earlier stages. There’s a really nice gap, I would say somewhere in stage 3 to 4 to 5, we have the KFRE. And then stage 1 to 3 and earlier, even more upstream with patients with diabetes and hypertension, we have Klinrisk. With these highly effective treatments, the best time is to deliver early. If you act when kidney function is normal, you can delay dialysis by 20 to 30 years or prevent it entirely [10]. Klinrisk enables earlier intervention [10] and that’s why we needed to move from KFRE to Klinrisk.

What is the impact of using digital tools for earlier detection amongst those at risk of CKD?

In our early uptake, we’ve seen most of the testing being driven by primary care. That means that primary care is finding kidney disease earlier and finding high risk kidney disease earlier. Also, there’s more prescriptions of RAAS inhibitors, SGLT2 inhibitors, MRAs and GLP-1s, all of which are highly effective therapies that prevent the progression of kidney disease and prevent heart failure [11]. Now if you connect the dots to the randomised trials, better treatment will result in better outcomes [12].

So one of the earliest benefits that’s been seen with the use of these algorithms is actually resource management. So in places like Canada and the United Kingdom and other single-payer universal healthcare systems, there’s a real shortage of nephrology resources [13]. What algorithms do is they get you beyond eGFR or creatinine as the cutoff for referral, and they help you focus on the patients at highest risk. So high risk patients get to nephrologists earlier and lower risk patients tend to stay with primary care or cardiology or endocrinology for longer. And what that does is optimises the resource. Now what that means is not that the lower and intermediate risk patients don’t need care. It means that that care can be safely delivered by a non-nephrologist.

I started out this conversation by saying that 90% of patients with kidney disease are followed outside nephrology. What we need nephrology to do is focus on the highest risk patients and we need our other colleagues to treat the patients who are at lower risk. And that’s because the treatments are readily available and almost everyone in the audience is comfortable using them. They’re just RAAS inhibitors, SGLT2 inhibitors, MRAs and GLP-1s. You don’t need a nephrologist to prescribe them.

What is your key advice to other clinicians when managing CKD within the CKM spectrum using such predictive tools?

I think for all my colleagues in nephrology and certainly for my colleagues who are listening, who are outside of nephrology, I want them to realise that the landscape for kidney disease has changed completely. It is no longer a chronic, progressive disease. We can now stop the disease entirely and put CKD into remission [14]. The best way to make remission happen is when we act early. When we act when eGFR is normal and there’s only microalbuminuria or less than 30mg/mmol of albumininuria [14]. And we need equations like Klinrisk and others to help us find these patients early so that remission is actually possible.

For all my colleagues, these patients are not in the nephrology clinic. Patients with a normal GFR and microalbuminuria are in your practice. So you actually have the best opportunity to make remission happen. So let’s stop being nihilistic about kidney disease and let’s instead be optimistic. It really is a disease which can be treated and put into remission. And let’s pass that language on to our patients because that really matters. 20 years ago, we used to call chronic kidney disease chronic renal failure. And we changed the words because patients don’t want to be in failure. Failure is depressing. Let’s change the narrative around chronic kidney disease to one of slowing progression to achieving remission.