Introduction

Sodium disturbances are frequent and serious complications in neurointensive care [1–6]. Both hyponatremia and hypernatremia cause brain injury, primary in patients without brain damage and secondary in patients with various primary brain injuries. Hyponatremia occurred more frequently than hypernatremia, which is, however, prognostically worse [7–9].

Sodium as the main extracellular cation affects effective extracellular fluid (ECF) osmolality. The osmotic gradient between ECF and intracellular fluid (ICF) spaces is balanced by a shift in water. The shift in water between ECF and ICF is caused by dysnatremias, which are connected with disturbances of serum osmolality. It is in hypernatremia which is always related to serum hyperosmolality, in contrast to hyponatremia. In some cases, hyponatremia has no relation to low serum osmolality, for example during osmotherapy by mannitol or hyperglycemia. Hypernatremia causes hypertonicity of ECF, and the following shift leads to dehydration of cells [10]. On the other hand, hypoosmolal hyponatremia gives rise to cell edema [11–13]. For these reasons it is necessary to pay attention to dysnatremias in neurointensive care and actively search for them. Daily monitoring of serum sodium should be part of the daily care of every patient with acute brain disease.

The aim of this study was to analyse outcome and frequency of sodium disturbances in relation to measured serum osmolality in neurologic-neurosurgical critically ill patients.

Methods

Collection of patients

The 5-year retrospective study was carried out in an eight-bed neurologic-neurosurgical intensive care unit (NNICU) in the Neurocenter of Regional Hospital Liberec. During this period there were 1,440 patients with various acute brain diseases, both neurological and neurosurgical, admitted. Collections of patients and laboratory data were made from the database of Laboratory Information System Stapro (LISSTA) Pardubice in the department of clinical biochemistry in Regional Hospital Liberec with the aid of the programme LisBed.

The criteria for patients’ inclusion was acute brain disease and serum sodium (SNa+) <135 mmol/l (hyponatremia) or SNa+ >150 mmol/l (hypernatremia). We defined hypernatremia according to Aiyagari’s study [2], which was conducted in neurologic/neurosurgical intensive care units in the same population as in our NNICU, and Miulli’s definition [14]. Hypoosmolality was defined as measured serum osmolality SOsm <275 mmol/kg, and hyperosmolality as SOsm >295 mmol/kg.

Observed clinical parameters

We observed the parameters related to the patient’s prognosis: mortality in NNICU, Glasgow Outcome Scale (GOS) upon discharge from NNICU, bad outcome marked as GOS 1–3, incidence of cerebral complications; parameters to the onset of dysnatremia: Glasgow Coma Scale (GCS), operation, relation to hospitalisation, diuretic and antiedematic therapy, fluid balance: fluid intake (ml/day), fluid output (ml/day), fluid balance (ml/day). Polyuria was defined as diuresis above 4,000 ml/day. We calculated all fluids given by mouth, tube and parentally into fluid intake. Fluid output was from diuresis and drainage.

Observed biochemical parameters

The following parameters were measured: serum sodium (SNa+), serum potassium (SK+), serum calcium (SCa2+), serum magnesium (SMg2+), serum chloride (SCl−), serum phosphorus (SP), serum osmolality (SOsm), urine osmolality (UOsm), serum protein (SProt), serum albumin (SAlb), serum glucose (SGlu), serum urea (SUrea), serum creatinine (SCr), blood pH (BpH), daily output of creatinine (dUCr), sodium (dUNa+) and potassium (dUK+). Sodium, potassium and chloride measurements in serum and urine were carried out on the COBAS Integra 800 system (Roche, Diagnostics, Switzerland) using selective ion electrodes. Creatinine in serum and urine, protein, albumin in serum were also measured on this equipment, but photometrically. Osmolality was gauged on the cryoscopic osmometer Fiske 210 (Advanced Instruments, Inc, Norwood, MA, USA). BpH was measured on the blood gas analyzer ABL 625 (Radiometer, Denmark). Blood samples for biochemical examination were collected using standard protocol in the NNICU, from extremities without application of infusion.

We used these calculated biochemical parameters from the department of clinical biochemistry. Clearance of creatinine was calculated according to general formulae for clearance with correction for the body’s surface.

Values of urine volume (V) are in litres and time in seconds. Serum osmolality calculated: SOsmC = 2 × Na+ + glucose + urea. Creatinine clearance: CCr = (UCr × V)/(time × SCr × surface). Osmotically active substances clearance: COsm = (UOsm × V)/(time × SOsm). Electrolyte clearance: CEl = V × [2× (UNa+ + UK+) + UGlu]/[2 × (SNa+ + SK+) + SGlu] × time. Sodium clearance: CNa+ = (UNa+ × V)/(time × SNa+). Solute free water clearance: CH2O = (V/time) − COsm. Electrolyte free water clearance: EWC = (V/time) − CEl or (V/time) − V × [2× (UNa+ + UK+) + UGlu]/[2 × (SNa+ + SK+) + SGlu] × time. Fractional excretion of osmotically active substances: FEOsm = (UOsm × SCr)/(SOsm × UCr × 1,000). Fractional excretion of sodium: FENa+ = (UNa+ × SCr)/(SNa+ × UCr × 1,000). Fractional excretion of free water: FEH2O = SCr/(UCr × 1,000). Biochemical parameters from urine were processed only from urine collected within 24 h.

Statistical processing

STATISTICA 10.0 software (StatSoft CR s.r.o) was used for statistical analysis.

The parametric t tests or non-parametric M–W U tests were used for comparison of continuous variables. Comparison of categorical parameters was carried out using Fisher tests. Univariate logistic regression was used for identifying prognostic factors of mortality and poor outcome during NNICU stay. Factors from univariate analysis with level of significance defined as p < 0.1 were used for multivariate regression analysis with forward stepwise method of final model building (p value for model enter was <0.05, p value for remove from model was >0.05). p values of less than 0.05 were considered significant.

Univariate and subsequent multivariate logistic regression analysis was used for identifying significant predictors of mortality during NNICU stay. Studied factors were age as a continuous predictor, hyponatremia or hypernatremia on admission, administration of steroids, antiedematic therapy, initial GCS less than 9, surgical intervention, cerebral complications, and diffuse or focal brain lesion.

The study was carried out with the approval of the Hospital Ethical Committee.

Results

During the 5-year period there were, out of 1,440 pts with acute brain diseases, 251 (17 %) pts with hyponatremia (mean SNa+ 131.78 ± 2.89 mmol/l) and 75 (5 %) pts with hypernatremia (mean SNa+ 154.38 ± 3.76 mmol/l). Hypoosmolal hyponatremia occurred in 50 (20 %) pts (mean SNa+ 129.62 ± 4.15 mmol/l; mean SOsm 267.35 ± 6.28 mmol/kg).

There was no difference between hyponatremic and hypernatremic patients in length of stay in NNICU (p = 0.825), length of dysnatremia (p = 0.699), operation (p = 0.163) and onset of dysnatremia after operation (p = 0.061), but upon admission to the NNICU there was a significantly higher presence of hyponatremia (p = 0.035) than hypernatremia. Hypernatremic pts more frequently received antiedematic therapy (p < 0.001), had lower GCS on onset of hypernatremia (p < 0.001), more cerebral complications (p = 0.001), higher mortality in NNICU (p < 0.001) and worse GOS upon discharge from NNICU (p = 0.024) (Table 1). In the hypernatremia group we found higher serum urea (p < 0.001). There were no differences either in fluid intake (p = 0.240) or infusion (p = 0.097). Significantly higher diuresis (p = 0.008) and fluid output (p = 0.007) with negative fluid balance (p = 0.001) were seen in hyponatremic patients. Further biochemical parameters can be seen in Table 2. Multivariate logistic regression analysis showed that hypernatremia compared to hyponatremia is a significant predictor of mortality during NNICU stay (OR 5.3, p = 0.002) but not a predictor of bad outcome upon discharge from NNICU, defined as GOS 1–3. Results of multivariate regression analysis are summarised in Table 3.

Table 1 Characteristics of the hyponatremic and hypernatremic patients
Full size table
Table 2 Parameters in hyponatremia and hypernatremia groups of patients
Full size table
Table 3 Results of multivariate logistic regression analysis in hyponatremia and hypernatremia
Full size table

Subanalysis between hypernatremic patients and hypoosmolal hyponatremic patients showed more antiedematic therapy (p < 0.001) and lower GCS on onset of dysnatremia (p = 0.026) in hypernatremic patients, but no significant differences in cerebral complications (p = 0.846), mortality in NNICU, (p = 0.061) and worse GOS upon discharge from NNICU (p = 0.100), (Table 4).

Table 4 Parameters in hypoosmolal hyponatremia and hypernatremia groups of patients
Full size table

Discussion

Dysnatremia is a frequent dysbalance of effective osmolality in acute brain diseases [1–6] which was also shown in our study. Dysnatremia occurred in 23 % of patients with acute brain diseases hospitalised in our neurologic-neurosurgical care unit. Hyponatremia was found more often (17 %) than hypernatremia (5 %), but hyponatremia connected with low serum osmolality was less frequent (3 %), only in 20 % of all hyponatremias. This means that this serious type of hyponatremia which influences the shift of water between ICF and ECF and causes cell edema was not a frequent type of sodium disturbance in our neurologic-neurosurgical critically ill patients, and even less common than hypernatremia (5 %). For this reason it is important while managing hyponatremia to find out the value of measured serum osmolality [15] as the first step. Hyponatremia had a significantly higher presence upon admission to the intensive care (p = 0.035) than hypernatremia, which was also seen in Funk’s study [16]. Hypernatremia arises more often during stay in intensive care [17]. In literature, it is known that antiedematic therapy is a risk factor that causes hypernatremia [2]. In our study hypernatremic patients received more antiedematic therapy. Another cause of onset of hyponatremia or hypernatremia is fluid therapy, but we did not find differences either in fluid intake (p = 0.240 in hyponatremia; p = 0.142 in hypoosmolal hyponatremia) or infusion (p = 0.097 and p = 0.156). Significantly higher diuresis (p = 0.008 and p = 0.014) and then fluid output (p = 0.007 and p = 0.008) were seen in hyponatremic or hypoosmolal hyponatremic patients. Another reason for occurrences of dysnatremia is diuretic therapy. In our study, we found significant differences in diuretic therapy only in comparison of hypernatremic with hypoosmolal hyponatremic patients (p = 0.006). The occurrence of dysnatremia has an influence on the type of diuresis. Water diuresis leads to hypernatremia and salt diuresis to hyponatremia, which was seen in these results. Renal function parameters [18, 19] of water diuresis EWC and FEH2O were significantly higher in hypernatremia, on the other hand in hyponatremia and hypoosmolal hyponatremia there was a higher daily output of sodium (dUNa+, p < 0.001, p < 0.001, respectively), COsm (p = 0.002 and p < 0.001), CEl (p < 0.001, and p < 0.001), CNa+ (p < 0.001 and p < 0.001) and FENa+ (p < 0.001 and p < 0.001).

From observed parameters influencing sodium homeostasis, in hypernatremic patients we found significantly higher serum urea concentration, which is another factor that can cause hypernatremia [20]. On the other hand we found significant differences in serum glucose concentration (p < 0.001) only between the hypoosmolal hyponatremia and hypernatremia groups.

Hypernatremia is a prognostically serious complication in critically ill patients [21, 22] which was also evident in our study. Multivariate logistic regression analysis showed that hypernatremia in comparison with hyponatremia is a significant predictor of mortality during NNICU stay (OR 5.3, p = 0.002), Further predictors are presence of cerebral complication (OR 6.4, p = 0.019) and GCS less than 9 (OR 11.0, p < 0.001). Different results were found when bad outcome upon discharge from NNICU (defined as GOS 1–3) was studied. In this analysis risk factors such as cerebral complication (OR 2.9,  p < 0.001) and GCS less than 9 (OR 21.0, p < 0.001) remained. New predictors of bad outcome were surgical intervention (OR 1.8, p = 0.042), steroid administration (OR 0.4, p = 0.005) and antidiuretic therapy (OR 0.2, p = 0.028). These results have the limitations of a retrospective study and further prospective cohort studies are needed to confirm these findings.

In conclusion, this study showed that hypernatremia occurred less frequent than all hyponatremias, but more often than hypoosmolal hyponatremia. Hypernatremia in comparison with hyponatremia is a significant predictor of NNICU mortality.