In our Perspectives article (The case for uric acid-lowering treatment in patients with hyperuricaemia and CKD. Nat. Rev. Nephrol. 15, 767–775; 2019)1 we proposed that individuals with hyperuricaemia and chronic kidney disease (CKD) with worsening kidney function should be considered for urate-lowering therapy (ULT) to slow progression of CKD1. When reviewing the literature, we argued that a trial was only interpretable if the control group showed clinically meaningful worsening of kidney function, which we defined as a reduction in estimated glomerular filtration rate (eGFR) of at least 4–5 ml/min/1.73 m2. In their Correspondence article (The case for evidence-based medicine for the association between hyperuricaemia and CKD. Nat Rev. Nephrol. https://doi.org/10.1038/s41581-020-0288-3 (2020))2, Steiger and colleagues state that we omitted a post-hoc analysis of two negative randomized clinical trials (RCTs)3 but these trials would be considered to be ‘non-interpretable’ by our criteria (ΔeGFR 2.7 ml/min/1.73 m2 in controls)3. Steiger et al. also expressed concerns about our inclusion of five RCTs published in peer-reviewed Chinese journals, which were summarized in a systematic review and meta-analysis4. Although one of these articles contained an error in the English abstract5, we contend that their data should not be ignored simply because they are not indexed in PubMed.

Steiger et al. also argue that CKD progression should be evaluated according to the rate of eGFR decline rather than the absolute change in eGFR during the trial. However, we believe that absolute changes in ΔeGFR are most relevant to clinicians. Our study was not ‘self-fulfilling’ as it was based on whether the controls progressed. Indeed, it is self-fulfilling to report that a study is negative when it is inconclusive.

Preliminary data from two randomized placebo-controlled studies — CKD-FIX and PERL — indicated that treatment with allopurinol did not prevent worsening of CKD. Full reports from these trials are pending but these studies were not designed to test whether allopurinol is beneficial in patients with CKD and hyperuricaemia as patients with normal serum urate were included in each study. Both studies were intention-to-treat analyses in which large numbers of patients (19% in CKD-FIX and 30% in PERL) stopped treatment yet were included in the final analysis. Although this approach might be statistically correct for the analysis of a specific treatment effect, the inclusion of individuals in the treatment group who were non-compliant with therapy and therefore had persistent hyperuricaemia, or of treated individuals with normouricaemia, is scientifically incongruous with testing the hypothesis that ULT can slow eGFR decline in patients with hyperuricaemia and progressive CKD.

Evidence-based medicine should evaluate the totality of evidence available and all approaches carry assumptions that deserve scrutiny6,7. A clinical trial can be well designed but uninterpretable. For example, an anti-hypertensive cannot be deemed ineffective if tested in patients who are normotensive. The effects of a drug used to treat hyperuricaemia cannot be tested in individuals with normal uric acid levels, and a clinical trial cannot be interpreted as having failed if no patient in either group gets the disease. Alas, the problem with evidence-based medicine is not the evidence, but the interpretation, especially if the underlying assumptions are not recognized.