On the optimal sodium correction rate in hyponatraemia and clinical outcome: a meta-analysis

Abstract

Background

Hyponatraemia is associated with greater mortality in emergency patients. Therefore, the correction of low serum sodium levels remains an important field in clinical practice.

Methods

Eligible studies were searched through a literature query in PubMed and Web of Science in August 2024. Sodium correction rates and mortality data were extracted. The articles were grouped based on the threshold used to distinguish between slow and rapid correction. For each group, a random effects model was used for meta-analysis. The sodium correction rate thresholds were 8, 10 and 12 mmol/L/24 hours. Additionally, a random effects model was used to estimate the difference in osmotic demyelination syndrome (ODS) incidence.

Results

11 retrospective studies with a total of 27 672 cases were included in the meta-analysis, each scoring at least 8* on the Newcastle-Ottawa quality assessment score. Mortality was significantly lower in the rapid correction group compared with the slow correction group for the 8 mmol/L/day threshold (OR: 0.398, p<0.001), for 10 mmol/L/day (OR: 0.489, p<0.001) and for 12 mmol/L/day (OR: 0.57, p<0.001). Overall incidence of ODS was low (0.085%), but it was higher in the rapid correction group (OR: 3.959, p=0.002).

Conclusion

Within all three thresholds of sodium correction, rapid correction was associated with improved survival. Although no causal relation can be concluded, a more liberal approach to sodium correction rate in hyponatraemia may be beneficial. However, osmotic demyelination remains a serious, although very rare, issue.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• This meta-analysis shows that slow serum sodium correction is associated with increased mortality in hyponatraemia.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These results have the potential to change hyponatraemia guidelines as they hint towards a more favourable outcome when sodium levels are corrected rapidly in hyponatraemia.

Introduction

Hyponatraemia, defined as a serum sodium concentration below 135 mmol/L, is the most common electrolyte disorder, affecting around 35% of hospitalised patients.¹ Its significance extends beyond mere laboratory abnormality, since it is associated with increased morbidity, mortality and healthcare cost across all patient groups and should therefore not be underestimated.² The underlying mechanisms are manifold and not fully understood. Hyponatraemia is often associated with migration of water into neurons due to an increased osmotic gradient, leading to brain oedema.³ Clinically, this can present as neurological symptoms ranging from mild problems such as nausea and dizziness to life-threatening conditions such as seizures and coma.⁴ Due to interactions between the brain and other organs, secondary conditions such as pulmonary oedema can arise.⁵ The alteration of the transmembrane potential due to decreased extracellular cation concentration has been shown to predispose for cardiac arrhythmias.⁶ Furthermore, hyponatraemia is associated with increased systemic inflammation.⁷ The epidemiology of hyponatraemia is complex, with prevalence rates varying widely depending on the clinical setting and patient characteristics.

Determining the underlying cause of hyponatraemia is often difficult and requires an understanding of complex pathophysiology, evaluation of the patient’s volume status and interpretation of a combination of laboratory values. The treatment of hyponatraemia is not without risk. Current guidelines recommend correcting the serum sodium levels with a correction rate of 4–8 mmol/L and by no more than 10–12 mmol/L within the first 24 hours due to the increased risk of osmotic demyelination syndrome (ODS).^{8 9} ODS is a neurological disorder characterised by symmetric demyelination in the pons, potentially occurring as a complication of rapid correction of chronic hyponatraemia.^{10 11} While ODS is a serious adverse outcome, death can be caused by the complications of not treating hyponatraemia aggressively enough. Currently, it is unclear whether a slower or more rapid correction is associated with lower mortality. The purpose of this meta-analysis was to answer that question by synthesising the available evidence.

Methods

Search strategy and inclusion criteria

We conducted a structured literature search in PubMed and Web of Science on 4 August 2024. We searched for articles reporting the outcome (death, ODS, length of stay) based on the correction rate. The exact search terms are found in the online supplement. The articles were screened for suitability for the research question based on their abstract using the artificial intelligence-based tool ‘ASReview’.¹² Articles were considered appropriate for the purpose of this analysis if they reported a dichotomous distinction between a rapid correction group with a correction rate in mmol/L/day as a cut-off value. Furthermore, they needed to include in-hospital mortality data differentiating between the two groups. The quality of the studies was graded using the Newcastle-Ottawa quality assessment scale.¹³ Only studies with at least 8* were included to ensure high-quality primary data. Data regarding the serum sodium correction rate and mortality as well as group sizes and patient demographics were extracted from the articles.

Additionally, we extracted data concerning the incidence of ODS as well as the length of the hospital stay. In the original articles, it is not explained consistently how ODS was diagnosed.

Statistical analysis

Statistical analysis was performed using the R-based tool jamovi. The threshold distinguishing between slow and rapid correction was not the same in the original studies. Therefore, for the mortality analysis, the underlying studies were grouped based on the cut-off value used to differentiate between slow and rapid sodium correction rates. This resulted in three groups (8, 10 and 12 mmol/L/day, respectively). Wherever studies reported trichotomisation of the underlying data, two groups were merged into one, resulting in a dichotomised data set (eg, if a study reported <10, 10–12 and >12 mmol/L/day as three groups, we would merge them into two data sets, resulting in one data set with 10 mmol/L/day and another one with 12 mmol/L/day as cut-off values). The study by Macmillan et al was included both in the 8 and the 12 mmol/L/day analysis, since the way the authors reported their data, it was suitable for both cut-off values.¹⁴

The analysis was carried out using the log OR as the outcome measure. A random-effects model was fitted to the data. The Q-test for heterogeneity and the I² statistic are reported. Studentised residuals and Cook’s distances are used to examine whether studies may be outliers and/or influential in the context of the model. Studies with a studentised residual larger than the 100x (1–0.05/(2 X k))th percentile of a standard normal distribution are considered potential outliers (ie, using a Bonferroni correction with two-sided alpha=0.05 for k studies included in the meta-analysis). Studies with a Cook’s distance larger than the median plus six times the IQR of the Cook’s distances are considered to be influential. The regression test, using the SE of the observed outcomes as predictor, is used to check for funnel plot asymmetry. For the ODS analysis, the studies were not analysed separately based on the cut-off value due to a lower number of studies, since not all authors reported the ODS incidence. Another random effects model was fitted to the data. The amount of heterogeneity was estimated using the DerSimonian-Laird estimator. For the length of stay analysis, the raw mean difference in days was the main outcome. Tests and CI were computed using the Knapp and Hartung method. Model estimates are found in online supplement.

Patient and public involvement

Patients or the public were not involved in the design, or conduct, or reporting or dissemination plans of our research.

Results

Search results and study characteristics

The initial search yielded 4932 results (excluding editorials and reviews). After screening (full text and abstract), eleven articles remained for inclusion in the final analysis. This screening process is illustrated in figure 1. The articles identified in the research are presented in table 1.

Figure 1. STROBE (Strengthening The Reporting of Observational Studies in Epidemiology) diagram of the search process. Abstracts were excluded if they suggested the article did not report data appropriate to our research question. In the full-text screening, we excluded articles not reporting mortality data differentiating between a clearly defined cut-off value for slow/rapid correction. Additionally, studies with a Newcastle Ottawa quality assessment scale of <8* were excluded to ensure high quality primary data.

Table 1. Search results and study characteristics.

Authors	Country	Study design	Cut-off value	Number of patients slow/fast	Age slow/fast (± SD)	Baseline sodium slow/fast (mmol/l)	Deaths slow/rapid	ODS slow/rapid
Giordano et al15	Italy	Retrospective, patients with sodium<120 mmol/L and neurologic symptoms in ED	0.3 mmol/l/h	35/32	78±3/74±6	116/117	26 (74.3%)/13 (40.6%)	N/A
Kinoshita et al16	USA	Retrospective, patients with sodium<120 mmol/L in the ICU	8 mmol/L/d	451/573	46±13.3/50±13	117/116.4	77 (17.1%)/38 (6.6%)	0/0
MacMillan et al14	Canada	Retrospective, patients with sodium<130 mmol/L in ED	8/12 mmol/L/d (trichotomous)	16940/3632 19416/1156	68.6±16.5/64.3±17.1 68.2 ±16.6/62.5±17.3	125.2/120.3 124.7/118.9	1566 (9.2%)/186 (5.1%) 1691 (8.7%)/61 (5.3%)	7/5
Geoghegan et al18	USA	Retrospective, patients with sodium<120 mmol/L in ED	10 mmol/L/d	298/114	70.4/63	117/115	20 (6.7%)/6 (5.3%)	0/1
Gottlieb and Celi19	USA	Retrospective, patients with sodium<120 mmol/L in the ICU	10 mmol/L/d	332/87	69/63	117/117	39 (11.7%)/6 (6.9%)	N/A
Seetapathy et al20	USA	Retrospective, patients with sodium<120 mmol/L on admission	10 mmol/L/d	2207/1067	67/62	117/114	231 (10.5%)/56 (5.2%)	N/A
Pham et al23	USA	Retrospective, hospitalised patients with sodium<130 mmol/L	12 mmol/L/d	86/14	73/63	N/A	13 (15.1%)/0 (0%)	0/0
George et al17	USA	Retrospective, patients with sodium<120 mmol/L on admission	8 mmol/L/d	884/606	68±15/63±15	117/115	167 (18.9%)/46 (7.6%)	1/7
Turkmen et al21	Turkey	Retrospective, patients with sodium<115 mmol/L	10 mmol/L/d	87/58	N/A	N/A	14 (16.1%)/4 (6.9%)	N/A
Arshad et al22	UK	Retrospective, patients treated with hypertonic saline	10 mmol/L/d	59/48	71.67±15.6/60.17±13.47	114.35/112.31	5 (8.5%)/1 (2.1%)	0/0
Sterns et al24	USA	Retrospective, patients with sodium<116 mmol/L	12 mmol/L/d	19/43	N/A	N/A	2 (10.5%)/3 (7%)	N/A

ED, emergency department; ICU, intensive care unit; ODS, osmotic demyelination syndrome.

Since the statistical analysis concerning mortality was performed for each cut-off value independently, we present the results of the analysis below.

8 mmol/L/day as cut-off value

A total of four studies reported mortality based on a cut-off value of 8 mmol/L/day and were therefore included in this analysis (see figure 2).^14–17 Mortality was significantly lower in the rapid correction group. 1836 out of 18 310 (10%; 95% CI 9.6% to 10.5%) died in the slow sodium correction group and 283 out of 4843 (5.8%; 95% CI 5.2% to 6.5%) in the rapid correction group. This equals an OR of 0.398 (95% CI 0.29 to 0.547). The average outcome differed significantly from zero (z=−5.67, p<0.0001). According to the Q-test, the true outcomes appear to be heterogeneous (Q(3)=8.79, p=0.03, I²=65.872%). An examination of the studentised residuals revealed that one study (MacMillan et al)¹⁴ had a value larger than ±2.4977 may be a potential outlier in the context of this model. According to the Cook’s distances, none of the studies could be considered to be overly influential. The regression test did indicate funnel plot asymmetry (p=0.0062).

Figure 2. Forest plot of the random effects model using 8 mmol/l/day as cut-off value. The plot displays the mortality in the slow and rapid sodium correction group using the log OR as outcome measure. Less mortality is seen in the rapid correction group. RE, random effects.

10 mmol/L/day as cut-off value

A total of five studies reported mortality based on a cut-off value of 10 mmol/L/day and were therefore included in the analysis (see figure 3).^18–22 Mortality again was significantly lower in the rapid correction group. 309 out of 2983 (10.4%; 95% CI 9.3%–11.5%) died in the slow sodium correction group and 73 out of 1374 (5.3%; 95% CI 4.2%–6.6%) in the rapid correction group. This equals an OR of 0.489 (95% CI 0.376 to 0.637). Therefore, the average outcome differed significantly from zero (z=−5.3061, p<0.0001). According to the Q-test, there was no significant amount of heterogeneity in the true outcomes (Q(4)=1.6493, p=0.7999, I²=0.0000%). One study (Seetapathy 2023) had a relatively large weight compared with the rest of the studies, (ie, a weight at least three times as large as having equal weights across studies). An examination of the studentised residuals revealed that none of the studies had a value larger than ±2.5758 and hence there was no indication of outliers in the context of this model. According to the Cook’s distances, one study (Seetapathy 2023) could be considered to be overly influential. The regression test did not indicate any funnel plot asymmetry (p=0.9737).

Figure 3. Forest plot of the random effects model using 10 mmol/l/day as cut-off value The plot displays the mortality in the slow and rapid sodium correction group using the log OR as outcome measure. Less mortality is seen in the rapid correction group. RE, random effects.

12 mmol/L/day as cut-off value

A total of three studies reported mortality based on a cut-off value of 12 mmol/L/day and were, therefore, included in the analysis (see figure 4).^{14 23 24} Once more, mortality was significantly lower in the rapid correction group. 1706 out of 19 521 (8.7%; 95% CI 8.4% to 9.2%) died in the slow sodium correction group and 64 out of 1213 (5.3%; 95% CI 4.1% to 6.7%) in the rapid correction group. This equals an OR of 0.570 (95% CI 0.440 to 0.738). Therefore, the average outcome differed significantly from zero (z=−4.2561, p<0.0001). According to the Q-test, there was no significant amount of heterogeneity in the true outcomes (Q(2)=1.2729, p=0.5292, I²=0.0000%). An examination of the studentised residuals revealed that none of the studies had a value larger than ±2.3940 and hence there was no indication of outliers in the context of this model. According to the Cook’s distances, none of the studies could be considered to be overly influential. The regression test did not indicate any funnel plot asymmetry (p=0.2605).

Figure 4. Forest plot of the random effects model using 12 mmol/l/day as cut-off value. The plot displays the mortality in the slow and rapid sodium correction group using the log OR as outcome measure. Less mortality is seen in the rapid correction group. RE, random effects.

Osmotic demyelination syndrome

A total of six studies reported the incidence of ODS and were therefore included in this analysis (see figure 5).^{14 16–18 22 23} The risk of ODS was significantly increased in the rapid correction group. ODS occurred in 8 out of 18 718 (0.0427%; 95% CI 0.0216% to 0.0841%) in the slow sodium correction group and in 13 out of 4987 (0.2607%; 95% CI 0.1525% to 0.4454%) in the rapid correction group. The observed log ORs ranged from −0.2392 to 2.334, with the majority of estimates being positive (83%). According to the Q-test, there was no significant amount of heterogeneity in the true outcomes (Q(5)=2.0971, p=0.8356, τ²=0.0000, I²=0.0000%). One study (MacMillan 2023) had a relatively large weight compared with the rest of the studies (ie, a weight at least three times as large as having equal weights across studies). An examination of the studentised residuals revealed that none of the studies had a value larger than ±2.6383 and hence there was no indication of outliers in the context of this model. According to the Cook’s distances, none of the studies could be considered to be overly influential. The regression test did not indicate any funnel plot asymmetry (p=0.8898).

Figure 5. Forest plot of the random effects model displaying osmotic demyelination syndrome risk. The plot displays the incidence of ODS in the slow and rapid sodium correction group using the log OR as outcome measure. A greater incidence of ODS can be observed in the rapid correction group. ODS, osmotic demyelination syndrome; RE, random effects.

Discussion

To our knowledge, this is the largest meta-analysis comparing slow and rapid correction of hyponatraemia on different outcomes. We have shown that mortality is lower in rapid correction compared with slow correction with 8, 10 and 12 mmol/L/day as cut-off values.

Comparison to previous research

An earlier meta-analysis by Ayus et al had shown similar results with higher survival rates in the rapid correction group.²⁵ In this paper, we were able to include other studies (Gottlieb and Celi¹⁹, Pham et al²³, Sterns et al²⁴) as well as a more diverse cohort of hyponatraemia patients and therefore analyse more than two times the number of patients, further validating the results in a larger patient cohort. Moreover, we included more studies in the >12 mmol/L/day analysis, where the previous meta-analysis incorporated one paper. In contrast to their study, we included only studies reporting mortality as an outcome variable, while they also included studies reporting only ODS incidence or length of hospital stay.

Limitations

This meta-analysis has a number of limitations.

Limitations of the review methodology

The study of Gottlieb et al has not undergone peer review before publication. We decided to include their data, because we did not build on statistical evaluation or conclusions, but rather on raw data on sodium levels, which they extracted from the reputable MIMIC-IV database.²⁶

Limitations of the included studies

All the studies included in this meta-analysis were conducted in a retrospective manner. While the association of rapid sodium correction with decreased mortality is clear, causation cannot be inferred. On the contrary, an association of rapid sodium correction with decreased mortality might be observed due to accelerated normalisation of serum sodium in patients who are less severely ill. Hyponatraemia might be corrected more easily in healthy patients than in more morbid patients. Adding to that, there might be heterogeneity in the exact degree of sodium correction in both groups, which would become apparent with tighter sodium controls. Also, healthcare providers might assume that patients in good health can tolerate more rapid sodium correction than more morbid patients, leading them to choose to correct hyponatraemia more slowly in sicker patients, risking treatment bias. Indeed, in studies that also reported comorbidity, there seems to be a trend towards increased comorbidity in the slow correction group, although this cannot be statistically evaluated due to inconsistent reporting.^{15 18 20 27} Chronic kidney disease, liver disease and congestive cardiac failure were reported more often in the slow correction group.^{16–20 23} In general, the rapid correction group was also younger.

These studies included in this meta-analysis do not distinguish between the different aetiologies and duration of hyponatraemia. However, the cause of hyponatraemia is generally considered to be relevant for its optimal therapy. In acute hyponatraemia, rapid sodium correction is generally considered safe and favourable, so it is likely that there were more patients with acute hyponatraemia in the rapid correction group.^{8 28} Also, there was no differentiation regarding the methods of sodium correction, which might be more important than the correction rate itself.

There is considerable heterogeneity in terms of sodium correction measurement. The distinction between slow and rapid correction is therefore not always straightforward across studies, as correction rates may be reported in different units. For example, within the 8 mmol/L/day correction rate group, we included the study of Giordano et al, which reported the cut-off as 0.3 mmol/L/hour—technically equivalent to only 7.2 mmol/L/day. This heterogeneity in how correction rates are expressed complicates the comparison between studies.¹⁵ This also proves true in reference to hyponatraemia definition and thereby patient inclusion.

Within the 8 mmol/L/day group, the regression test suggests heterogeneity. This prompts caution concerning the generalisability of the results and could potentially be attributed to publication bias, systemic differences in the study population or design as well as the subsumption of the cut-off values for the analysis.

The follow-up regime was not consistent among all studies. Therefore, there might have been some under-reporting, especially in regards to ODS occurrence. There may be systemic bias in the inclusion of patients in the primary studies. Arshad et al included only patients treated with hypertonic saline.²² Kinoshita et al and Gottlieb et al only included patients in the intensive care unit (ICU).^{16 19} These patients are more likely to be severely symptomatic, for example, with seizures or low Glasgow Coma Score, since hypertonic saline is recommended to use in these patient groups and the ICU treats patients with high morbidity. Therefore, the generalisability of the results might be further lowered.

Implications for clinical practice, policy and research

The results presented in this paper challenge the current paradigm for the treatment of hyponatraemia. Current guidelines published by leading European academic societies recommend limiting the sodium correction rate to 10 mmol/L/day within the first 24 hours and to 8 mmol/L/day within the next day.⁸ These recommendations, however, are based on low-quality evidence (grade D recommendation). Similarly, American experts recommend correction rates of not more than 10–12 mmol/L within the first 24 hours, citing case series of increased rates of ODS with rapid sodium correction.²⁸ The cautious approach advised in current guidelines is understandable, given that they were published before the studies cited in this article.^{8 28} Recently, the Austrian Society of Nephrology published their recommendations on hyponatraemia treatment.²⁹ Despite acknowledging studies suggesting reduced mortality risk with rapid sodium correction, they maintained their recommendations for slower correction rates to reduce the risk of ODS, citing low validity of the supporting data due to the retrospective nature of the studies. This argument has been proposed by Sterns and Rondon-Berrios in another consensus paper.^{30 31}

While current guidelines favour a cautious approach using slow serum sodium correction, the studies analysed within this meta-analysis have shown that rapid sodium correction is associated with decreased mortality.^{8 28} In the three correction thresholds assessed, there was a clear decrease in mortality for the fast correction group. Between the groups, the greatest benefit for the rapid correction could be observed when using 8 mmol/L/day as cut-off value. The smallest benefit was observed in the 12 mmol/L/day analysis. This suggests that there may be a critical threshold above 12 mmol/L/day, above which rapid correction rates may be associated with higher rather than lower mortality.

Given the limitations mentioned above, our findings should be interpreted with caution. The observed association between rapid sodium correction and reduced mortality cannot be interpreted as evidence of causation due to the retrospective nature of all included studies and the potential for confounding by baseline patient characteristics and treatment decisions. The heterogeneity in study populations, definitions and measurement approaches limits the precision of our estimates and the applicability of our findings to specific clinical scenarios.

While the criticism of the quality of studies supporting a more rapid correction rate may be legitimate, it is notable that no prospective randomised controlled trial dividing patients into a rapid or slow sodium correction target group has ever shown ODS to occur more often following rapid sodium correction. Studies demonstrating an association between rapid correction and ODS are case reports or retrospective observational trials with the same low validity. It is unclear why the same quality of data should be rated differently for ODS.

Although ODS is a severe complication potentially leading to fatal outcomes, this complication is only seen in 0.26% of rapidly corrected patients. The overall mortality risk in hyponatraemia is 5%–10%. The magnitude and consistency of the observed mortality reduction across all three correction thresholds—despite the retrospective nature of the included studies—suggests that the potential harms of very slow correction may have been underestimated and that existing guideline recommendations may be overly conservative for some patient populations.

Our results suggest potential benefit from more rapid correction rates up to certain thresholds. For clinical practice, these data support careful individualisation rather than a uniformly restrictive approach. While there seems to be consensus among physicians that the risk–benefit analysis favours rapid sodium correction in acute symptomatic hyponatraemia, for example, leading to seizures, generally patients with assumed chronic hyponatraemia tend to be corrected slowly in order to avoid ODS. In clinical practice, it is often unclear whether hyponatraemia is acute or chronic due to a lack of outpatient laboratory data of the previous days. This could lead to a tendency to misclassify acute hyponatraemia patients as chronic and therefore might lead to unnecessarily slow hyponatraemia correction. A rapid increase in sodium levels despite minimal or no active correction therapy could indicate such a misclassification. In these cases, tolerating a rapid increase of sodium levels might be beneficial. On the other hand, risk factors for the development of ODS such as hypokalaemia, low baseline sodium, malnutrition and alcohol use disorder should be screened for and taken into account, potentially warranting a more cautious approach to sodium correction.

Policymakers and guideline makers should acknowledge that the same type and quality of evidence (retrospective observational data) is currently used both to support slow-correction limits for ODS prevention and to suggest potential mortality benefit with faster correction. This asymmetry in how evidence is weighted deserves explicit discussion.

Future research should prioritise well-designed prospective studies with clearer and standardised data collection.

Both prospective cohort studies with rigorous outcome standardisation and randomised controlled trials comparing different sodium correction targets should be pursued. Such studies must stratify by acute versus chronic hyponatraemia, standardise correction rate definitions and treatment strategies, uniformly prespecify correction rate categories and rigorously and uniformly capture both mortality and ODS outcomes with consistent definitions. Only with such prospective data can the true risk–benefit balance of different correction rates be defined and robust, evidence-based updates to hyponatraemia guidelines be made.

Conclusions

Currently, there is limited evidence to determine the optimal sodium correction rate in hyponatraemia. This article cannot provide a definitive answer to this question. However, we have shown that there is a clear and consistent reduction in the risk of death associated with rapid serum sodium correction across 8, 10 and 12 mmol/L/day as cut-off values. Although retrospective data cannot establish causation, the consistent and substantial mortality reduction observed across all analysed correction rates suggests this finding merits prospective validation. These findings suggest that sodium correction rates recommended in current guidelines may be too low. A clear answer to this important question can only be given through a properly conducted prospective randomised controlled study.

Data availability statement

Data are available upon reasonable request.