Chronic Kidney Disease-Mineral and Bone Disorder

Changing Insights Form Changing Parameters?

Marc G. Vervloet


Nephrol Dial Transplant. 2020;35(3):385-389. 

The exceedingly high mortality risk of patients on haemodialysis has been the incentive for a wide range of therapeutic interventions, besides dialysis itself. Lifestyle and dietary changes, anaemia treatment, modifications of dialysis modality and duration, and optimizing traditional risk factors including hypertension and dyslipidaemia have all failed to substantially improve outcome, except for kidney transplantation and possibly high-volume haemodiafiltration in selected patients,[1] although this latter option is still debated. Therefore, hope for improvement currently is to a large extent based on a presumed benefit of optimizing disturbances in mineral metabolism, traditionally defined by abnormal serum concentrations of calcium, parathyroid hormone (PTH), phosphate and vitamin D. Indeed, a large proportion of the daily pill burden for patients on dialysis that is being prescribed is aiming to modify these values.[2,3] The recently updated Kidney Disease Improving Global Outcome (KDIGO) guideline, while acknowledging the absence of evidence from prospective clinical trials, generally supports this approach, in particular for hyperphosphataemia, extreme values for PTH and the avoidance of hypercalcaemia.[4] The rationale is based on both experimental data, which support the assumption that PTH and phosphate may act as uraemic toxins, and observational mainly cross-sectional data from large cohorts, as was recently summarized.[5] In this issue of Nephrology Dialysis Transplantation, Lamina and co-workers[6] report on the associations not only of single time point concentrations of chronic kidney disease-mineral and bone disorder (CKD-MBD) biomarkers with mortality, but importantly also the implications of changes over time, which revealed some important insights as will be outlined below.[6] These data are desperately needed, because it is not obvious at all that interventions which change these parameters towards or even into the 'optimal' range are paralleled by risk reduction. This lack of proof of benefit of improved biochemical profile may have at least three principle explanations.

  1. Residual confounding of unknown or unmeasured factors may mitigate the association between MBD biomarkers and risk, and therefore MBD interventions do not target the true cause of complications. This has been demonstrated for hypophosphataemia and all-cause mortality, with malnutrition and inflammation being confounders.[7] No-one would supplement phosphate in hypophosphataemic patients on dialysis with the goal to improve his or her mortality risk, since hypophosphataemia merely identifies risk, but is not causing morbidity, unless it is extreme. In turn, it would be naïve to assume that currently all other potential confounders that may influence the association between MBD biomarkers and risk have been established. As a more recent example, magnesium concentrations importantly modify the risk of hyperphosphataemia, a confounder frequently neglected in previous cohort analyses: for those patients with higher concentrations of magnesium, the association between hyperphosphataemia and mortality risk was nullified.[8]

  2. MBD-attributable risk may be better reflected by the constellation of all markers together, whereas interventions usually target a single component.[9,10] Inclusion criteria of the EVOLVE trial (a randomized controlled trial among 3883 haemodialysis patients, examining the reduction in occurrence of a composite primary endpoint by cinacalcet, a PTH-lowering calcimimetic) was based on PTH value.[11] Elevated PTH concentration itself is not a disease (hyperparathyroidism is) and judgement of PTH in isolation as proxy for high bone turnover generally performs poorly, unless extreme.[12,13]

  3. The assumption that interventions only modify the parameter for which they have been prescribed probably is an oversimplification of reality. Most interventions have off-target effects, like an increase of Fibroblast Growths Factor 23 (FGF23) following active vitamin D prescribed for secondary hyperparathyroidism.[14] In turn, active vitamin D may increase α-klotho[15] and downregulate cardiac expression of Fibrobast Growth Factor Receptor 4 (FGFR4),[16] the receptor alleged to mediate cardiotoxic effects of FGF23.[17] For an individual patient, it will be difficult to predict which aspect of all treatment effects will dominate. Different phosphate binders, despite achieving comparable control of hyperphosphataemia, may be associated with differences in outcome, generally in favour of non-calcium containing binders.[18] Whatever may explain these differences, additional benefits of a specific class, unforeseen harm of the intervention or applicability of findings only in specific circumstances or for selected patients, most likely is all are of importance, arguing against a simple parallel between modifiability of a biomarker and modifiability of risk.

The ultimate way to provide clarity in the opaque field of CKD-MBD should come from well-designed trials that study the effect of reduction of MBD-associated risk in high-risk patients with relatively low background (non-MBD) risk. Unfortunately, such trials are unlikely to be available in the near future. Therefore, for the coming years, novel knowledge might come from novel analyses of improved phenotyped cohorts with prospective and longitudinal data collection.

The study by Lamina et al.[6] reports on the results of a meticulous analysis of the second recruitment wave of the 'Analysing data, Recognizing excellence, Optimising outcomes in end-stage renal disease' (AROii) cohort.[6] The study population consisted of 8817 incident patients on haemodialysis, recruited 10–12 years ago among 334 sites in 15 countries from Fresenius Medical Care dialysis facilities. The cohort had prospective data collection, which included demographics, comorbidity, clinical and laboratory values, and medication use, most to a large extent depend on different time points per individual patients. Outcome variables were all-cause and cardiovascular mortalities. The analysis consisted of several ways to study the association between single time point measurement of calcium, phosphate and PTH with mortality. In addition, the association of change of these MBD markers over time with mortality was studied. Key analyses are summarized in Figure 1. Several notable findings emerged.

Figure 1.

Key analyses from the AROii cohort. Baseline was set at 3 months after dialysis initiation. Time-updated analysis was performed, studying the association between events (baseline events, blue arrow), with time-updated concentration of calcium, phosphate and PTH. To mimic a prevalent cohort, the same analysis was performed, but with a 2-year delay (T2 events, green arrow). In addition, varying follow-up time was studied. Next, association of change of either calcium, phosphate or PTH and mortality follow-up time were studied. For this, as indicated on top, the difference between the average of the biomarker in the early period of dialysis, indicated by the yellow-shaded area and future time point (T2 and T3) was associated with subsequent events, T2 events and T3 events, respectively. Lines A, B and C represent individual patients with spontaneous fluctuations around the upper range of the target or within target, as examples.

With regard to single time point measurements, the study in general confirmed U-shaped associations between the three biomarkers and risk for both all-cause and cardiovascular mortalities. Of relevance, these relationships were similar when ignoring values and outcomes in the first 2 years, thereby mimicking a prevalent instead of an incident cohort. In addition, also varying follow-up time did not essentially change the findings. A strength of this study is that it defined its own range of optimal values for calcium, phosphate and PTH based on the lowest hazard ratio (HR) for each analyte. For most of these ranges from AROii it can be said that they more or less confirm the optimal ranges from other observational studies and KDIGO guidelines, with two exceptions. First, the high margin of optimal serum calcium concentration was 2.36 mmol/L, which is well below the upper limit of normal (ULN), a value proposed by the KDIGO guideline. Secondly, the lower margin of optimal PTH was 239 ng/L, which appears to be substantially higher than twice the ULN as suggested by the KDIGO guideline from 2009, which was recently reinforced.[19] Unfortunately, in AROii, PTH measurement was not performed centrally and different assays may have been used. However, for almost all contemporary assays, this lower margin of optimal would still be higher than twice its ULN. One of the most clinically relevant findings from this study is the very high percentage of patients having a too low PTH concentration. This applied to an astonishing 60%, all facing an increased mortality with an HR of 1.18. Despite the fact that the HR for hyperphosphataemia was higher (1.34) than that for low PTH, the numbers at risk associated with high phosphate levels were much lower (19%). This suggests that on a population level the more important epidemic is 'hypo-PTH-ism', not hyperphosphataemia. Clearly, for most patients in AROii, one can assume that measured phosphate is under treatment for that, and the notion that the low-PTH epidemic may have more impact on mortality should not be an incentive to relax on phosphate management. One additional remarkable finding of the single time point analysis is that risks associated with high phosphate and high PTH differed for men and women, denoting much less impact for woman. The most prominent other established gender-associated difference of MBD-related morbidity is the incidence rate of fractures,[20] and it is possible that this confounded this gender difference observation from AROii because findings were not adjusted for history of fracture or fracture risk.

Most novelty from this study comes from the analysis that addressed the impact on mortality risk of changing calcium, phosphate and PTH. Change was defined by the difference from a measurement during follow-up compared with the averaged value of measurements in follow-up Months 3–6, so relatively early after dialysis initiation (Figure 1). Analysis consisted of an absolute change and categorical change, with categories defined as within, below or above the optimal range. The first message is straightforward: those within the optimal range for any analyte should stay there. For those that start with calcium >2.36 mmol, the analysis suggests it should come down. The incidence of a too low calcium in the early phase was too low to draw any conclusion. This is good news for the AROii population since hypocalcaemia in CKD generally is a feature of the untreated natural history of kidney failure. Probably, vitamin D or other measures that prevent hypocalcaemia from developing had been instituted, although the study does not report on medication use.

Slightly more than 20% of patients had a phosphate concentration below the minimal risk range in the early phase of dialysis. In line with expectations, moving over time into the optimal risk range was associated with risk reduction. When speculating about what may have caused this improved risk, the most likely explanation is improved nutritional status compared with the early period of dialysis. Although down-titrating phosphate binder treatment or lowering dialysis dose may also have caused this restoration of hypophosphataemia, these are unlikely explanations for this observation. Very few patients will have had hypophosphataemia due to binders, and decreasing dialysis dose will probably not lead to improved mortality risk. More puzzling is the finding that those patients who had hyperphosphataemia in the early phase did not benefit from a decline. This finding requires a cautious interpretation. While in prospective clinical trial it is usually possible to ascribe changes in outcomes observed to a single change in exposure (the intervention under study, such as a pharmaceutical), in observational studies, it is a black box, and usually changes are the consequence of the net effect of numerous known, but also unknown variables. Two major factors that may promote phosphate concentrations to decline are intensified phosphate-lowering treatments and the development of malnutrition, but these factors conceivably have divergent effects on mortality risk (Figure 2). In AROii, it was not possible to make a distinction between these two: medication use was not modelled and body mass index (BMI) was not adjusted for in the main adjustment model used for the changing parameters. For this reason, there is no justification for nihilism with regard to treatment for hyperphosphataemia, which the authors rightfully do not claim. In clinical practice, it is reasonable to search for causes in an individual patient if phosphate declines, and to look back to what may explain this trend.

Figure 2.

Conceptual implications of changes of concentrations of phosphate and PTH. Upper panel: restored hyperphospataemia (red line, A) may be caused by a variety of circumstances. Development of inflammation or malnutrition may induce a seemingly beneficial change of phosphate concentration, but it increases mortality risk (dashed red line). In turn, improved dietary habits or improved pharmacological management may also decline phosphate, and in addition improve risk. Restored hypophosphataemia (blue line, B) is usually caused by improved nutritional status, a beneficial dominating effect. Theoretically, the risk of phosphate toxicity increases (dashed blue line). Lower panel: stable or declining PTH when it is above target range at baseline has no clear net effect on mortality (green line, C). This may be the result of counterbalancing effect of the absence of malnutrition inflammation complex (MICS), but suboptimal treatment when it is stable (solid green line), and improved management but occurrence of MICS when its declining (dashed green line). An increasing PTH, when it is below target at baseline has beneficial effects, possibly because of improved MICS, less PTH-suppressing therapy or less calcium loading (yellow line). Persistently suppressed PTH is associated with higher mortality risk (dashed yellow line).

The most striking finding with regard to changing parameters comes from the data on PTH. In addition to the previously described observation that so many patients are at risk because of having a low PTH, this study shows that increasing PTH is associated with improvement of mortality risk. This presumably beneficial effect is most pronounced for those that were below the optimal range in the early phase of dialysis, but appeared also to be present for those that actually were in the optimal range and had a slight increase. Here, the big question is: what made PTH go up? A recent report found that the use of higher dialysate calcium is associated with both PTH suppression and cardiovascular risk.[21] Other iatrogenic causes of low PTH may include overzealous use of calcimimetics and active vitamin D. With regard to active vitamin D, a recent randomized trial among Japanese haemodialysis patients without hyperparathyroidism (based on the definition of the Japanese Society of Dialysis Therapy) indeed refuted any benefit of alfacalcidol.[22] Adjusting for the malnutrition inflammation complex or protein energy wasting was shown to push the optimal PTH range downwards.[23] As outlined, in AROii such an adjustment was not performed on the analysis of changing parameters, due to too many missing values of BMI and C-reactive protein. Other variables generally associated with lower PTH are the presence of diabetes and older age, but for both factors results had been adjusted. The translation to the clinic of this possible benefit of increasing PTH could be to search for modifiable, especially iatrogenic, factors that suppress PTH, such as the use without clear indication of active vitamin D, calcimimetics, too high dialysate calcium, malnutrition and hidden causes of inflammation. Obviously, such an approach should be studied in trials as well because hitherto unknown factors may importantly confound the residual association between PTH and mortality. Like hyperphosphataemia, restoration of increased PTH into the optimal range also did not go hand in hand with improved risk [Hazard ratio = 0.91 (95% confidence interval 0.47–1.77)]. As outlined earlier, a wide range of factors may underlie declining PTH concentration, and here too, these factors may have a diverging effect on mortality risk. As an example, it is probable that intentionally lowering PTH by an intervention (like active vitamin D, calcimimetics or surgery) has a different effect on risk than a 'spontaneous' decline by inflammation or calcium loading (Figure 2).

Some more aspects complicate the interpretation of changing parameters in the AROii analysis. Changes were all relative to the values at Months 3–6, a time point that may lie several years in the past for some, and likely has a different meaning than more short-term changes. Concentrations of 25-hydroxyvitamin D and FGF23 were not measured. Although the causal role of FGF23 as inducer of complications might be argued, its importance as risk predictor appears straightforward.

Despite these limitations, the current analysis of the AROii provides another important piece of the puzzle in understanding CKD-MBD. Our focus can be sharpened in identifying those at highest risk, and some research attention should shift towards the epidemic of suppressed PTH. Moreover, this important study provides hope that raising PTH might improve outcome, once we know how to do so.