Summary of main results

This review included evidence from 35 randomised controlled trials involving 3429 participants, primarily covering vasopressin receptor antagonists. Generally, the evidence for patient‐important outcomes was limited. Vasopressin receptor antagonists had very uncertain effects on mortality and health‐related quality of life. There was low certainty evidence for a reduction in hospital stay (1.6 days) and for the absence of an important improvement in cognitive function.

Most studies have focused on the intermediate biochemical outcome and there was moderate certainty evidence that vasopressin receptor antagonists increased the serum sodium concentration (4 mmol/L). But they were also associated with an increased risk of rapid serum sodium correction and commonly caused adverse effects such as thirst and polyuria. On average, treating 1000 people would cause 290 additional people to have an increase in serum sodium concentration of at least 5 mmol/L, but it would come at a cost of an additional 29 people having an increase exceeding 8 to 12 mmol/L/d, considered as the threshold from which on there is an increased risk for osmotic demyelination, be it that no such cases were documented in any of included studies. Effects were generally consistent across the different drugs in this class. RCT data for other interventions such as fluid restriction, urea, mannitol, loop diuretics, corticosteroids, demeclocycline, lithium and phenytoin were largely absent.

Overall completeness and applicability of evidence

Although the FDA and the EMA have approved certain vasopressin receptor antagonists for treating people with hyponatraemia, regulatory approval was largely based on intermediate outcomes in short‐term studies. To date, clinically important outcomes (e.g. reduction in all‐cause mortality or improvement in health‐related quality of life, cognitive and general functional status) remain insufficiently investigated. The data we had, indicated with moderate confidence that vasopressin receptor antagonists modestly increased the serum sodium concentration, but there is insufficient evidence to conclude this truly translates in the improvement of patient‐important outcomes.

In the context of intervention studies, a surrogate or intermediate is a measurable outcome such as a laboratory test, which responds to an intervention (e.g. lowering of cholesterol with statins) and is causally associated a clinically important outcome (e.g. reduction in mortality with statins) (Ballinger 2014). Investigators often use surrogates instead of important health outcomes because surrogates can substantially reduce the cost, sample size and duration of a randomised trial. However, not all are valid proxies of clinically important outcomes. It is true that in acute and profound hyponatraemia, the evidence from observational studies is so overwhelming that we readily accept that increasing the serum sodium concentration is life‐saving. But as we move further from the extreme of acute, profound hyponatraemia into chronic, mild hyponatraemia, the evidence for such a consistent causal link weakens. Although it is true that a chronically low serum sodium concentration is strongly and consistently associated with increased mortality and risk of bone fractures (Wald 2010), there is currently insufficient evidence that aside from affecting the surrogate (e.g. increase in serum sodium concentration with a vasopressin receptor antagonist) treatment in case of more chronic hyponatraemia also changes the patient‐important outcomes downstream of the surrogate in the same causal pathway (e.g. reduction in mortality as a consequence of raising serum sodium concentration with a vasopressin receptor antagonist). There are intuitive reasons to assume that such causality can reasonably be extrapolated to the entire spectrum of hypotonic hyponatraemia. But likewise, there are reasons for caution. Lixivaptan primarily failed to obtain regulatory approval by the FDA for hyponatraemia in chronic heart failure due to a numeric ‐ be it statistically non‐significant ‐ imbalance in early deaths. The clinical review team argued that 'while the early death in participants with chronic heart failure and hyponatraemia could reflect the underlying disease, they could not exclude the possibility that some subjects with hyponatraemia associated with acute worsening congestive failure were exquisitely sensitive to intravascular free water shifts and did not tolerate even a small change in intravascular volume status or osmolality, induced by lixivaptan and/or effects resulting from a compensatory neurohumoral activation' (BALANCE 2010).

Studies contributing to this review included mostly participants with mild to moderate hyponatraemia (mean serum sodium concentration at study level ≥ 123 mmol/L). Meta‐regression revealed a modifying effect of the serum sodium concentration at baseline, with lower values associated with larger increases in natraemia. Extrapolation of meta‐regression data would suggest higher increases, but possibly higher risks of rapid correction as the baseline serum sodium decreases. Although no study reported osmotic demyelination, it is unclear what would happen if vasopressin receptor antagonists were used on a larger scale and in people with sodium concentrations below those included in the RCTs that contributed to this review. Likewise, it is possible that while improvements in cognitive function were not readily detected, they would emerge for people with lower serum sodium concentrations at baseline.

Finally, studies evaluating the effectiveness of alternative interventions for increasing serum sodium concentration in people with chronic, non‐hypovolaemic hyponatraemia are largely absent.

Quality of the evidence

Overall, we considered the data evaluating the effects of vasopressin receptor antagonists for people with chronic, non‐hypovolaemic, hypotonic hyponatraemia on patient‐important outcomes such as mortality, health‐related quality of life, and hospital stay limited and of low certainty. Low certainty evidence suggests that additional studies are likely to change our confidence in the effects (GRADE 2008). According to GRADE, RCT data is considered high quality, but may be downgraded for several reasons (Guyatt 2010). These reasons varied for each outcome. For death it was largely driven by the width of the confidence interval, which captured both appreciable benefit and harm, and left us with considerably uncertainty around the treatment effect. For health‐related quality of life, there was some evidence that vasopressin receptor antagonists increased the mental component summary score of the SF‐12, and unclear effects on the physical component summary score. However, we serious questioned the validity of this tool for measuring quality of life in the context of hyponatraemia. For one thing, many items included in the SF‐12 do not reflect readily acknowledged signs and symptoms of hyponatraemia (e.g. bodily pain, anxiety) and similarly, the neurocognitive signs of hyponatraemia are not assessed by the tool. In addition to the general content validity issues, we also judged effects for this self‐reported outcome may have been inflated because participants were likely unblinded due to polyuria as a side‐effect; results were selectively reported for one in two measurement time‐points; and more than one‐third of the data were missing.

We had more extensive data for outcomes related to serum sodium concentration, resulting in moderate confidence in the effect estimates all around. Downgrading occurred mainly for missing outcome data, which we judged could have somewhat overestimated the treatment effect; as well as indirectness, in that tight follow‐up during dose‐titration, is likely to have avoided rapid correction somewhat in the trial setting. It may result in higher risk of rapid correction when used in clinical practice.

Of note, all studies assessing benefits and harms of vasopressin receptor antagonists were likely instigated and sponsored by the pharmaceutical company developing or seeking to commercialise the compound. For the studies that provided a declaration of interest for the authors of study reports, all save one had author lists that featured people who had received money for presentations or consultancy, or were employed by the sponsor. Industry sponsorship does not necessarily introduce bias into the design and conduct of clinical trials, but empiric evidence does show that pharmaceutical industry‐sponsored studies are more likely to have favourable efficacy results (RR 1.32, 95% CI 1.21 to 1.44) and harm results (RR 1.87, 95% CI 1.54 to 2.27) than studies not sponsored by industry (Lundh 2012).

Data for other interventions such as fluid restriction, change in medical regimens, captopril or albumin were sparse and inconclusive.

Potential biases in the review process

Although this review was conducted by two or more independent authors, used a comprehensive search of the published and unpublished research designed by a specialist librarian, and examined all potentially relevant clinical outcomes, potential biases exist in the review process. The major weakness of this review is the paucity of data for treatments other than vasopressin receptor antagonists. First, summary of existing evidence therefore focusses the discourse on new, and expensive, treatments, rather than focusing on existing, and cheaper alternatives. RCTs are extraordinarily expensive and consequently often conducted by the pharmaceutical industry. This results in a catch‐22 situation of evidence being mostly created, and thus only available, for newer interventions in general. Many other interventions are or have been used in clinical practice, but were driven to the background because RCT data were largely absent. Notably, also for fluid restriction, despite it being the currently accepted first‐line treatment for both hypervolaemic and euvolaemic hyponatraemia, there are no randomised trial data available. Currently, there are limited opportunities for use of vasopressin receptor antagonists in practice. Only two vasopressin receptor antagonists have obtained large scale regulatory approval. Conivaptan is FDA approved for euvolaemic and hypervolaemic hyponatraemia in hospitalised patients. It is available only as an intravenous preparation and treatment duration is limited to a maximum duration of four days because of drug‐interaction effects with other agents metabolized by the cytochrome P450 3A4 hepatic isoenzyme (FDA 2012). Tolvaptan, an oral agent, is also FDA‐approved for the treatment of euvolaemic and hypervolaemic hyponatraemia. Although theoretically available for long‐term treatment, recent concerns around the potential for severe liver injury in patients with autosomal dominant polycystic kidney disease ‐ be it at doses far exceeding the ones used in hyponatraemia ‐ has caused the FDA to restrict the use of tolvaptan to 30 days and issue a contra‐indication for patients with underlying liver disease (Mirski 2013). Canadian regulatory authorities mandated monitoring liver injury tests at regular intervals, but did not limit duration of use. In the European Union, tolvaptan is approved only for the treatment of euvolaemic hyponatraemia, due to safety concerns stemming from a numeric ‐ be it statistically non‐significant ‐ difference in treatment‐emergent fatalities in patients with hypervolaemia.

Agreements and disagreements with other studies or reviews

This review largely agreed with the findings of two earlier systematic reviews. However, neither of these previous reviews sought to examine any other treatments beyond vasopressin receptor antagonists for treating chronic non‐hypovolaemic hypotonic hyponatraemia. The first, including 15 RCTs and 1619 participants found vasopressin antagonists on average increased the serum sodium concentration by approximately 5 mmol/L at one week (13 studies, 1119 participants: MD 5.27 mmol/L, 95% CI 4.27 to 6.26), and approximately 3.5 mmol/L beyond the first week (8 studies, 793 participants: MD 3.49 mmol/L, 95% CI 2.56 to 4.41), but it came at a cost of increased risk of overly rapid correction of the serum sodium concentration (8 studies, 860 participants: RR, 2.52, 95% CI 1.26 to 5.08) (Rozen‐Zvi 2010).The second review, including 11 RCTs and 1094 participants, found vasopressin antagonists on average increased the serum sodium concentration by approximately 5 mmol/L at day 4 (11 studies, 1094 participants: MD 4.90 mmol/L, 95% CI 4.10 to 5.80), but was associated with a 6% increased risk of overly rapid correction of the serum sodium concentration (9 studies, 995 participants: RD 0.06, 95% CI 0.03 to 0.10) (Jaber 2011).

We also largely agreed with the interpretation of these data by the respective study authors as presented in their discussion: vasopressin receptor antagonists effectively raise the serum sodium concentration. But whether this translates to meaningful changes in outcomes that matter to patients is not clear. At this stage the available RCTs have insufficiently studied clinical outcomes to conclude that people whose serum sodium concentration increases under treatment will experience changes in end‐points such as general well‐being, cognitive function, gait stability, bone fractures and survival.

In 2013, a guideline group consisting of seven authors (six Americans and one Irishman), published expert recommendations identifying several alternative treatments, including fluid restriction, demeclocycline, urea and vasopressin receptor antagonists. Driven by the emergence of RCT evidence for vasopressin receptor antagonists as effective means for increasing the serum sodium concentration, and the absence of such evidence for other treatments, they projected vasopressin receptor antagonists were likely to become a mainstay of treatment for euvolaemic hyponatraemia and probably represented the best approach to treating hyponatraemia in most oedema‐forming states (Verbalis 2013).

A European guideline published in 2014 ‐ to which four authors of this review contributed ‐ suggested that in moderate or profound hyponatraemia (serum sodium concentration < 130 mmol/L), restricting fluid intake was first‐line treatment (2D); increasing solute intake with urea or a combination of low‐dose loop diuretics and oral sodium chloride were equal second‐line treatments (2D); and vasopressin receptor antagonists were not recommended (1C). (Spasovski 2014). There is almost no RCT evidence for treatments other than vasopressin receptor antagonists. The group's rationale to advocate some of these treatments anyway stemmed from the premise that in the absence of unequivocal evidence that increasing the serum sodium concentration leads to an improvement in patient‐important outcomes, the focus should be on avoiding harm. They gave much weight to avoiding rapid increases in the serum sodium concentration, for it's ‐ albeit very rare ‐ association with osmotic demyelination syndrome. And based on the absence of observational evidence for rapid increases or other important harms for fluid restriction, urea and low‐dose loop diuretics with oral sodium chloride, they formulated a weak recommendation to consider these for treating people with sodium concentrations below 130 mmol/L. Largely because of the documented risk of rapid increase and some concern around its potential for liver disease, a negative recommendation was formulated for vasopressin receptor antagonists.