There are no potential conflicts or disclosures to report.
Correspondence: Dr Soham Rej MD, Irving Ludmer Research & Training Building, McGill University, 1033 Pine Avenue West, Montreal, QC, Canada H3A 1A1. Email: firstname.lastname@example.org
Diabetes insipidus (DI) is a recognized adverse effect of lithium use, and studies have shown an association between decreased renal function and DI in patients using lithium. We hypothesize that hypernatraemic events that occur in DI predict decreased renal function in elderly patients on lithium.
We conducted a retrospective cohort study involving 55 geriatric psychiatry patients using lithium between 1985 and 2010. Patients who always had sodium levels ≤146 mmol/L were compared to patients with one or more episodes of hypernatraemia (serum sodium level ≥147 mmol/L) for estimated glomerular filtration rate (eGFR) levels and prevalence of severe chronic renal failure (eGFR ≤30 mL/min/1.73 m2).
eGFR was found to be less in the hypernatraemic group than in the non-hypernatraemic controls (41 vs 56 mL/min/1.73 m2; P = 0.0074). Severe chronic renal failure appeared more prevalent in hypernatraemic patients (4/14 (28.6%) vs. 3/41 (7.3%)), but this did not achieve statistical significance (P = 0.061). The two groups did not differ for age, sex, medical comorbidities or other clinical variables, except antidepressant use. Hypernatraemic patients appeared less likely to use antidepressants than non-hypernatraemic patients, odds ratio = 0.69 (P = 0.020). However, in multivariate analysis, hypernatraemia correlated with decreased eGFR (β = −0.39, P = 0.004), while antidepressant use did not (P = 0.81).
These results suggest that hypernatraemic events may predict reduced renal function in geriatric patients using lithium. The role of hypernatraemia and DI in renal failure in this population requires further study. Health professionals should be aware of the risks of renal failure in older patients treated with lithium, especially in the context of sodium level abnormalities.
Chronic renal failure (CRF) is a well-established side effect of lithium use, occurring in roughly 1% of adults using lithium for 15 years. The elderly are already at elevated risk for renal dysfunction: roughly 30%–35% of community-dwelling Americans over 70 years of age have moderate-to-severe CRF (estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m2). It appears that geriatric lithium users may be at a higher risk of renal failure than similarly aged controls, but this needs to be replicated in larger studies. Given the world's ageing population, the number of geriatric lithium users will likely continue to rise and concerns about their renal safety will become of increasing importance.
Diabetes insipidus (DI) is another adverse effect of lithium use. It involves excessive urination (>3 L/24 h) and is associated with dilute urine (osmolality <300 mOsm/Kg) and/or polydipsia. DI is also known to have caused cases of severe hypernatraemia requiring hospitalization in the elderly,[4, 5] although systematic studies of DI and hypernatraemia have not yet been performed.
It has long been believed that DI is a precursor condition for lithium-induced CRF.[6, 7] In some studies, reduced glomerular filtration rate correlated with decreased maximal urine osmolality and urine volume.8,9 However, However, the link between CRF and either DI or hypernatraemia has not been well-studied in the elderly.
Because DI may be associated with CRF in adult populations and because hypernatraemia is a known manifestation of DI in the elderly, a correlation between serum sodium (Na+) levels and CRF may exist in geriatric lithium patients. We hypothesize that hypernatraemic events, which can occur in DI, can predict decreased renal function in elderly patients using lithium.
Design and setting
We performed a retrospective study in geriatric psychiatry clinics of three tertiary care hospitals in Montreal, Quebec, Canada. Ethics approval for chart review was obtained from all centres.
Study population and exposures
We included all geriatric psychiatry outpatients (age >65 years) with current or previous lithium use >1 month with at least two non-zero (>0.2 mEq/L) lithium levels. Laboratory test results (serum Na+ and Li+ levels) were available from January 1998 to June 2011. Our primary exposure of interest was peak serum Na+ levels. We compared renal function (eGFR) and the prevalence of severe chronic renal failure (CRF) between hypernatraemic patients (serum Na+ level ≥147 mmol/L) and controls (serum Na+ ≤146 mmol/L). Serum Na+ level ≥147was used as our threshold, as such moderate levels of hypernatraemia have been known to cause symptoms and necessitate hospitalization in certain geriatric patients. Historical records of psychiatric and medical diagnoses as well as medication use (e.g. diuretics, NSAIDs psychotropic medications) were also recorded for each patient.
The main continuous outcome was ‘minimum eGFR’, defined as the lowest eGFR reading occurring between January 1998 and June 2011. The main binary outcome was severe CRF (eGFR < 30 mL/min/1.73 m2 on two occasions separated by >3 months).
Patients with hypernatraemia (serum Na+ level ≥147 mmol/L) were compared with controls (serum Na+ ≤146 mmol/L). Dichotomous and continuous variables were analyzed for significance with Fisher's exact test (two-tailed) and Student's t-test, respectively. Stepwise multiple linear regression was then used to determine the effects of potential confounding variables on eGFR in hypernatraemic patients and controls. All analyses were performed using SPSS 15.0 (SPSS, Chicago, USA).
We identified 55 lithium-using geriatric outpatients: 14 with hypernatraemia (serum Na+ ≥147 mmol/L) and 41 controls (serum Na+ ≤146 mmol/L). The groups did not differ for age, sex, medical comorbidities or other clinical variables, except for peak serum Na+ level and antidepressant use (Table 1). Hypernatraemic patients were less likely to use antidepressants than non-hypernatraemic patients, odds ratio = 0.69 (P = 0.020).
N.B. Groups did not differ significantly in their use of other medications (diuretics, antipsychotics, anxiolytics, cholinesterase inhibitors, methylphenidate).
Age, mean age ± SD (years)
79.4 ± 7.71
76.9 ± 7.29
Peak serum Na+, mean ± SD (mmol/L)
148.6 ± 1.94
142.9 ± 2.78
Lithium (Li) use parameters, mean ± SD
Mean Li duration (years)
14.5 ± 5.85
10.7 ± 10.4
Mean Li dose (mg/day)
521.4 ± 339.2
519.5 ± 253.0
Peak Li level (mmol/L)
1.05 ± 0.54
0.87 ± 0.44
Psychiatric diagnoses (n)
Medical diagnoses (n)
Medications used (n)
Our results showed that in geriatric psychiatry outpatients, hypernatraemia was associated with reduced renal function (Table 2). Hypernatraemic patients had decreased eGFR when compared with controls (41 vs 56 mL/min/1.73 m2; P = 0.0074). Also, a trend towards increased prevalence of severe renal failure was observed in hypernatraemic patients (4/14 (28.6%) vs 3/41 (7.3%); P = 0.061). In multivariate analysis, peak serum Na+ continued to correlate significantly with decreased eGFR (β = −0.39, P = 0.004), while antidepressant use did not (P = 0.812) (Table 3). After accounting for the effect of antidepressant use, our statistical model predicted that for each increase in peak serum Na+ by 1 mmol/L, eGFR would decrease by 2.01 mL/min/1.73 m2 (Fig. 1).
Table 2. Renal function in hypernatraemic patients and controls
Hypernatraemic patients (n = 14)
Controls (n = 41)
eGFR, estimated glomerular filtration rate.
Minimum eGFR (mL/min/1.73 m2)
40.6 ± 12.3
55.6 ± 18.8
Severe renal failure (eGFR <30 mL/min/1.73 m2) (n)
R2 = 0.162. B, slope of regression line; eGFR, estimated glomerular filtration rate; SE B, standard error of the regression line's slope.
Peak sodium level
DI had been thought of as a precursor for CRF,[6, 7] and some studies in adults have found a correlation between DI and CRF.[6-9] It had previously been thought that DI caused tubular damage in the kidney, which then predisposed patients to glomerular dysfunction, ultimately leading to CRF.[6, 7] Only one previous study has investigated DI systematically in the elderly. This paper found that in geriatric lithium users (mean lithium duration = 9.2 years, n = 48), the prevalence of decreased urine osmolality (<300 mOsm/Kg), increased urine volume (>2.5 L/24 h), and decreased renal function (GFR <40 mL/min/1.73 m2), was 19%, 33%, and 19%, respectively. However, there was no indication of whether an association existed between DI parameters and renal function. Furthermore, although past studies have reported that DI can cause severe hypernatraemia in the elderly and require hospitalization,[4, 5] whether hypernatraemia correlates with other parameters of DI, such as urine osmolality, remains unknown.
Our study demonstrated that hypernatraemic episodes correlated with decreased renal function within a 14-year period. One potential explanation for our results is that hypernatraemic events are reflective of incidents of DI in our population. If DI indeed increases the risk of CRF in the elderly, we would expect hypernatraemia to correlate with decreased renal function, as was the case in our study. However, we must be cautious about claiming that there is an association between DI and CRF in the elderly. Urine osmolalities and urine volumes results were not available retrospectively to confirm whether patients had DI. Also, the relationship between peak Na+ levels and better understood parameters of DI have not yet been systematically investigated. Nonetheless, we can still be relatively confident that hypernatraemic events may be a useful marker for reduced renal function in elderly lithium patients.
There were several limitations to this study. Its retrospective methodology did not allow for serum Na+ and eGFR testing at regular intervals in controlled testing conditions. Despite this, attempts were made to include all patients with previous lithium use and systematically record patients' minimum eGFR and peak serum Na+ levels historically. Another limitation of our study was its descriptive design. Even though our study revealed that patients with a hypernatraemic event had an increased risk of decreased renal function during the 14-year study period, it remains difficult to attribute causality; whether patients actually develop CRF following hypernatraemia is still unclear. An additional limitation was our small sample size. With a larger sample, our results and the effect of confounding variables on renal failure could be verified with multivariate analysis incorporating other known risk factors (e.g. hypertension, diabetes).
Despite these limitations, we encourage clinicians to follow a systematic approach to screen for renal side effects of lithium use in the elderly. Current guidelines suggest routine screening for lithium levels and renal function in lithium patients every 3 and 6 months, respectively. The results of our study suggest that elderly patients with hypernatraemic events are at higher risk of having renal dysfunction, although this needs to be confirmed in future studies. Though very inexpensive to perform, screening for serum Na+ levels is not part of the current guidelines. We therefore suggest regular screening of serum Na+ levels in addition to systematic follow-up of renal function in this population, which may help alert physicians to patients at increased risk for renal failure.
We would like to acknowledge Dr Silvia Monti De Flores and Ms Nicole Bissonnette for their help with this project.