Cum grano salis

In this issue of Nephrology, Dialysis and Transplantation, Ayus et al. [1] report on the association between chronic hyponatraemia and the risk of hip fracture in an elderly population. The authors use a retrospective analysis of data derived from a health maintenance organization administrative database. The study includes 31 527 adults, whereby ‘elderly’ is defined as age >60 years and chronic hyponatraemia as two consecutive measurements of serum sodium <135 mmol/L. Elderly patients with chronic hyponatraemia had a 4.5 times higher relative risk for hip fracture.

Whereas this result seems impressive, and is in line with previous reports describing an association between chronic hyponatraemia and mortality [2], the clinical relevance of this observation is less obvious. First, only 228 of the 31 527 patients actually fulfilled the criterion of ‘chronic hyponatraemia’, implying that the incidence of this condition in the population treated at a tertiary-level hospital is quite low, even though the cut-off for hyponatraemia is set at 135 mmol/L and the mean serum sodium was 132 mmol/L in this study. According to the recent European Renal Best Practice (ERBP) guideline on hyponatraemia, most of these patients would be considered to have mild hyponatraemia [3]. The current ERBP recommendation for these patients is not to take any interventional action [4]. Would the results of this observational study persuade the ERBP to change its view on this issue? This is not very likely.

First, the current data are both retrospective and observational. There are many common risk factors for both mild hyponatraemia and hip fracture; age, gender, congestive heart failure, chronic kidney disease, chronic use of antidepressant medications (selective serotonin reuptake inhibitors, tricyclic antidepressants), anticonvulsants and thiazide diuretics are among the most relevant ones. It is not infrequent that some, or even all, of these risk factors are present in a single patient [5]. In the current cohort, patients with hyponatraemia were older (mean 78 versus 70 years) and had more diabetes (12.3 versus 10.4%), chronic kidney disease (12 versus 10%), heart failure (8 versus 3%) and alcohol abuse (2.6 versus 0.8%). Authors use statistical techniques, such as a propensity score, to adjust for the confounding created by this pre-existing unequal distribution of risk factors for hip fracture. However, it is uncertain whether, in this setting, such a propensity score approach can adequately adjust for the inequalities between the two groups. This is a major concern, especially in view of the low event rate, as is witnessed by the C-statistic of 0.73. In other words, most of the patients in the chronic hyponatraemia group had other risk factors that increased the chances of suffering a hip fracture. Also, the absolute risk for hip fracture in this group is very low: only 7 of 282 patients had an event of hip fracture over an 8-year follow-up period, with a median of only 510 days. The fractures occurred quite early after detection of hyponatraemia, which puts into question, from a pathophysiological perspective, the potential role of hyponatraemia in the event. It is thus highly likely that the observed increased relative risk is a finding of statistical rather than clinical relevance.

Let us for a moment accept that the 4.5% increased relative risk for hip fracture is a true association with chronic hyponatraemia and not merely a statistical artefact. Under these conditions, would it then be warranted to change the recommendation of the ERBP? This is very unlikely.

It has been well established that observational data do not allow causal inferences. Moreover, even if a causal relation were present, it is not certain that eliminating the cause would result in improved outcomes. In diabetes, for example, lower HbA1C values have been associated with better outcomes in the general population. In patients with chronic kidney disease stage 3b or higher, this association is only true after adjusting for comorbidities, mainly malnutrition and inflammation [6]. However, active intervention aimed at lowering HbA1C does not reduce overall or cardiovascular mortality [7]. The most likely underlying reason for this disparity is that the potential benefit (prevention of microvascular damage) occurs slowly and only becomes apparent in the (far) future, whereas the adverse effects of such a strict glycaemia-lowering strategy (mainly hypoglycaemia) are immediate [8]. So, even if we accept that there is a causal relationship between chronic hyponatraemia and the risk of hip fracture, we would need to test whether reversing chronic hyponatraemia results in a reduction of that risk.

So, do we now need to set up a randomized clinical trial to determine whether correcting chronic hyponatraemia results in a reduction in hip fractures?

Let us assume there is a drug that would effectively restore hyponatraemia to normonatraemia in 100% of patients and that the only additional risk of hip fracture in the chronic hyponatraemia group was completely, and causally, attributable to the presence of the hyponatraemia. In this highly unlikely situation, treating 282 patients with this drug would result in avoidance of seven hip fractures, and the number needed to treat would be 40. If we take a somewhat more realistic, but still optimistic, scenario where half of the fractures are causally related to hyponatraemia and a drug is effective in avoiding chronic hyponatraemia in half of the patients, the number needed to treat increases to 160. A randomized study to demonstrate a statistically significant result under these assumptions would need to include 8000 patients in each arm, which would result in a total of 16 000 patients (50/50 coffee rule) [9]. This calculation does not yet take into account the negative impact of the potential side effects of such a drug.

So, what can we then learn from the new data in order to improve our daily practice?

When confronted with a patient with mild (chronic) hyponatraemia (serum sodium 130–135 mmol/L), we should check for underlying causes. Also, if there are drugs present that can induce mild hyponatraemia and that can be safely switched or stopped, this should be done [10]. If there is underlying malnutrition or alcohol abuse, this should be treated. However, such measures would be taken irrespective of whether or not chronic hyponatraemia is considered to be an additional risk factor for hip fracture. As explained previously, the likelihood that a randomized trial will ever provide conclusive evidence that, in this setting, a drug could reduce the risk of hip fracture is close to zero. Therefore, we should not embark on a journey of disease mongering inspired by observational data.

CONFLICT OF INTEREST STATEMENT

The authors declared to have no conflict of interest with regard to the content of this manuscript. (See related article by Ayus et al. Mild prolonged chronic hyponatremia and risk of hip fracture in the elderly. Nephrol Dial Transplant 2016; 31: 1662–1669)

REFERENCES