From Mice To Men

Is That Fancy New Biomarker Clinically Relevant?

Januvi Jegatheswaran; Manish M. Sood


Nephrol Dial Transplant. 2019;34(6):899-900. 

A quick PubMed search reveals more than 200 000 articles published involving the term 'biomarker' in the last 5 years alone. In this climate, how does a researcher navigate whether the latest biomarker is going to contribute to existing risk prediction models? At the bedside, how does a physician know which one is the best for clinical application? An ideal biomarker should predict risks more effectively compared with known and existing risk models, demonstrate internal and external validity, be easily applicable and be cost-effective. It is an unrealistic expectation to demonstrate all of these criteria in the early stages of biomarker discovery; however, it should be mandatory prior to widespread acceptance and use. A logical first step is the demonstration of a novel biomarker's independent and strong association with the outcome of interest. As chronic kidney disease is often a progressive condition with a highly heterogeneous natural history, it follows that a successful biomarker to delineate individuals at risk of progression from those who do not would be a top research priority. In this issue of NDT, Han et al.[1] examined an emerging potential biomarker to assess the risk of chronic kidney disease (CKD) progression, the Aquaporin 11 (AQP11) rs2276415 variant.

Aquaporins (AQ) are a diverse set of channels that are responsible for transport of water within cells. There are seven isoforms of aquaporins in the kidney, with the new isoform AQP11 also being present in the testis, thymus, liver and intestine.[2]AQP11 was first characterized in mice when AQP11-null mice were noted to develop polycystic kidneys. The responsible mutation of AQP11 leading to the rs2276415 variant, a single-nucleotide polymorphism with substitution of glycine for serine, has been implicated in kidney disease in a small number of studies in humans.[3] Park et al.[4] demonstrated that the presence of the AQP11 rs2276415 variant in a transplanted kidney was independently associated with a reduction in graft survival. Furthermore, Choma et al.[5] reported a higher risk for acute kidney injury, chronic kidney disease (both defined by administrative data codes) and CKD progression [defined by estimated glomerular filtration rate (eGFR) slopes] with the AQP11rs2276415 variant among individuals with diabetes mellitus compared with those without. Interestingly, as there was no association with the AQP11 rs2276415 variant in individuals without diabetes, it is suggested that the gene variant may be the link between hyperglycemia and kidney injury. Many pertinent questions remained, including whether the association persisted after accounting for other important factors (such as proteinuria), whether it was truly limited to diabetes and hyperglycemia, the magnitude and consistency of the association and whether the variant added information above and beyond existing known risk factors for CKD.

It is in this context that Han et al.[1] examined the association of the AQP11rs2276415 variant with CKD progression in the Chinese population. Using a prospective cohort study design, the investigators assessed the association of AQP11rs2276415 polymorphism on CKD progression in 620 adult community-based Chinese patients with CKD Stages II–V. Dialysis, previous kidney transplant, malignancy or serum creatinine >177 μmol/L for men and >159 μmol/L for women were exclusion criteria. Genotyping of peripheral blood leukocytes was conducted to determine the patient's AQP11 gene variant at the start of the study. Study follow-up was for 3 years, with the primary study outcome being a composite of an increase in serum creatinine ≥35 μmol/L from baseline, initiation of hemodialysis or need for transplantation. Of the 620 patients enrolled in the study, 121 (19.5%) patients had the rs2276415 polymorphism. Among individuals with the AQ polymorphism, the majority were female, hypertensive, on a greater number of antihypertensive medications and were more likely to have a history of congestive heart failure and more advanced CKD (both a lower eGFR and higher proteinuria) at baseline. A total of 170 patients experienced the study outcome of CKD progression (114 by creatinine criteria, 54 initiated dialysis and 2 required kidney transplants). The rs2276415 polymorphism was more prevalent in the patients who had CKD progression (40.6%) than those who did not (16%). Crude renal event-free survival was longer in patients without the rs2276415 polymorphism, with very early separation between the two groups. In models adjusted for demographics, comorbid illnesses, baseline eGFR, lipids, inflammation high-sensitivity C-reactive protein (hs-CRP)], serum phosphate and proteinuria, the AQP11 rs2276415 variant was associated with a composite outcome {hazard ratio [HR] 1.92 [95% confidence interval (CI) 1.31–2.84]}. The magnitude of the association was stronger among those requiring dialysis or transplantation [HR 3.19 (95% CI 1.47–6.94)], albeit with a limited number of events. In multivariate models, higher baseline hs-CRP and proteinuria (g/24 h) and a lower baseline eGFR were significantly associated with the composite outcome, whereas age, diabetes mellitus, cardiac disease and hypertension were not. The investigators conclude that the AQP11 polymorphism is associated with CKD progression or end-stage kidney disease.

We commend the study investigators for advancing the search for a potential prognostic biomarker in CKD. The study has a number of notable strengths. It was a prospectively conducted study with an adequate length of follow-up (3 years) and patients were consecutively recruited, minimizing selection bias. There was minimal loss to follow-up or missing data. The AQP11 polymorphism assay was conducted consistently, as the validation between duplicate samples was high (99%). A reasonable number of known variables associated with CKD progression were accounted for in their multivariate models. The composite outcome included actual creatinine measures and the need for renal replacement therapies as opposed to diagnostic codes, which are fairly inaccurate in CKD.[6] Most importantly, it adds to the evolving literature regarding the role of the AQP11 rs2276415 variant in kidney disease.

Nevertheless, some potential important questions remain. It remains unclear if the AQP11 variant/composite outcome association would persist if other notable known risk factors for CKD were included, such as peripheral vascular disease, hemoglobin A1c and perhaps most importantly, the cause of the CKD. As the AQP11 variant directly affects renal tubular function, it could differentially impact glomerular, vascular and tubular-based kidney disease. Indeed, this may explain the heterogeneity of outcomes observed among patients with the AQP11 variant. For example, this may explain the association of the AQP11 variant and long-term poor graft survival in transplanted kidneys (a primarily tubulointerstitial process). Furthermore, in the current study, Han et al.[1] detected no association between the AQP11 variant and diabetes mellitus, in contrast to Choma et al..[5] The key to this discrepancy may lie in the fact that the former accounted for proteinuria whereas the latter did not. Perhaps proteinuria is required for the AQP11 variant to induce injury (as opposed to diabetes), again with the common pathway being proteinuria-induced tubulointerstitial fibrosis.[7]

Some peculiarities were noted in the study design. First, the exclusion of individuals with arbitrary levels of serum creatinine (men >177 μmol/L, women >159 μmol/L) is not justified and unclear. Second, the serum creatinine criterion for CKD progression is also arbitrary and not evidence-based. Great strides have been made to evaluate acceptable declines in eGFR as a surrogate kidney outcome.[8] Third, it is uncommon for CKD Stage II patients to be referred to a nephrologist and it is unclear what proportion of the study cohort that comprises. Lastly, in the multivariable model there were a number of variables that are highly similar and potentially collinear (e.g. four lipid measures).[9]

Returning to the potential role of the AQP11 variant as a prognostic biomarker, there is much work remaining. The question of the strength of the association, its causal role and whether it is specific to existing subpopulations remain unknown. With three studies to date, demonstrating a potential association, investigators should go further to determine if it adds incremental predictive value above and beyond well-established CKD prognostic factors. This could be examined using traditional measures of model performance, such as the C-statistic, or more recently developed net reclassification indices.[10] From there, the clinical utility of the biomarker needs to be assessed for ease of applicability and cost–benefit analysis before incorporation into current risk prediction tools.[11] Taken together, the current study adds an important further step in the development of the AQP11 variant as a predictor of kidney disease, but much work remains.