Distinguishing between salt poisoning and hypernatraemic dehydration in childrenBMJ 2003; 326 doi: https://doi.org/10.1136/bmj.326.7381.157 (Published 18 January 2003) Cite this as: BMJ 2003;326:157
- Malcolm G Coulthard (), consultant paediatric nephrologista,
- George B Haycock, professor of paediatricsb
- a Department of Paediatric Nephrology, Royal Victoria Infirmary, Newcastle upon Tyne NE1 4LP
- b Guy's, King's, and St Thomas's School of Medicine, London SE1 9RT
- Correspondence to: M G Coulthard
- Accepted 4 October 2002
The criteria most widely used to diagnose salt poisoning in children were formulated by Meadow.1 These criteria focus on hypernatraemia with high urinary concentrations of sodium and chloride, but this combination may also be found in children with dehydration caused by diarrhoea. The medical and legal management of the two conditions fundamentally are different, so reliable ways to distinguish them are needed.
We consider the physiology of salt overload and hypernatraemic dehydration.2 We explain how to differentiate the two situations on the basis of history; examination; and biochemical analysis of serial, paired, blood, and “spot” urine samples. We illustrate the method with two cases that have been tested in court.
Medical causes of hypernatraemia other than salt poisoning and dehydration are persistent and easy to exclude
Urinary sodium concentrations may be high in cases of salt poisoning and dehydration, so they cannot distinguish between the two conditions
Fractional excretions of sodium and water can differentiate salt poisoning from dehydration
Serial pairs of spot plasma and urine samples should be taken during recovery
“Spot balances” for salt and water estimated from fractional excretion values give a clear picture of the physiology in individual cases
In case 1, a 7 year old boy had severe colitis, for which he underwent a colostomy and received intravenous and nasogastric nutrition, for five years. He presented twice with confusion, hypernatraemia, and weight gain but without fever, diarrhoea, or vomiting. Paired plasma and urine samples collected during his second presentation led to a diagnosis of salt poisoning (table). His mother then confessed to inducing the colitis with purgatives and twice giving him salt solutions nasogastrically.
In case 2, a boy born at 28 weeks' gestation with aortic arch and bronchial anomalies needed prolonged ventilation and remained oxygen dependent. Fundoplication was considered because he had severe gastro-oesophageal reflux and was failing to thrive. He was treated with domperidone, a compound alginate preparation (Gaviscon Infant), and feed supplements that contained glucose polymers. The boy's plasma biochemistry was normal, except when he was 5 and 8 months old, when he was acutely ill with high fever, profuse diarrhoea, exacerbation of vomiting, hypotonia, weight loss, and poor perfusion (table). Intravenous fluids dramatically improved his condition and restored his body weight.
The boy was taken into foster care after he was admitted at age 8 months, because doctors believed his biochemistry results confirmed salt poisoning and because his siblings' medical histories were considered suspicious. Three siblings had become hypernatraemic under similar circumstances: one died after remaining at home (on the general practitioner's advice) with pyrexia, explosive diarrhoea, and weight loss but without haemorrhagic encephalopathy.3 In addition, one sister, who was eunatraemic, died suddenly and unexpectedly. The odds of a second innocent death were suggested (incorrectly) to be 73 million to one.4 A court found that the mother was not guilty of manslaughter.
Diagnostic pathway for childhood hypernatraemia
Exclusion of other medical causes
When a child presents with hypernatraemia, a defect in the system that controls urinary concentration must be excluded as a cause.1 Children with hypernatraemia caused by such a defect usually have obvious polyuria and polydipsia that result from central or nephrogenic diabetes insipidus, chronic renal failure, or rare renal problems. Such children can easily develop negative water balance (when the volume of water ingested is less than the volume excreted in urine), which reduces solute excretion and induces hypernatraemic dehydration. Osmolality typically is less then 300 mmol/kg—sometimes as low as 50 mmol/kg. In all other children, urine osmolality can exceed 700 mmol/kg—as in both of our illustrative cases. A water deprivation test may be needed in children for whom the diagnosis is uncertain.
Essential hypernatraemia is a rare disorder of children and adults in which hypertonicity causes neither thirst nor the release of antidiuretic hormone but the release of antidiuretic hormone mediated by baroreceptors is normal. 5 6 The mechanism behind the condition is almost certainly selective destruction of the anterior hypothalamic osmoreceptor. Plasma concentrations of sodium >170 mmol/l are not unusual in this condition.
In practice, children who can maintain completely normal plasma concentrations of creatinine and electrolytes when they are well, eating a normal diet, and drinking a normal amount of fluids are extremely unlikely to have significant, persistent, underlying medical causes for hypernatraemia.
Salt overload is usually deliberate, so a story to explain the hypernatraemia, such as intravenous sodium bicarbonate treatment or accidental contamination of feeds, rarely is convincing. Speculation that only infants will take salty drinks1 was disproved by an accident in 1963, in which 14 babies up to 10 weeks old became severely hypernatraemic after drinking feeds made with salt instead of sugar.3 Six of these babies died of haemorrhagic encephalopathy.
The notion that severe hypernatraemia cannot be caused by water depletion alone is also incorrect, because it is seen in babies who are inadequately breast fed 7 8 or who had undiagnosed oesophageal atresia.8 In healthy people, hypernatraemia causes vomiting and intense thirst, which may be recognised even in babies,3 and neurological signs, such as irritability, drowsiness or coma, and fits.
In patients with hypernatraemic dehydration, dehydration results from a child having negative water balance (losing more water than is replaced). If the child is also losing more salt that he or she is receiving (negative sodium balance), the plasma concentrations of sodium may remain stable. For hypernatraemia to develop, the water loss must exceed any loss of salt. The patient's history usually reveals causative factors. Typically, diarrhoea causes water loss, and vomiting prevents intake of water. This condition was common in the 1960s and 1970s, when infant milks were more calorie dense than they are today9 (not because they had higher sodium concentrations, as previously suggested10).
In patients with infective diarrhoea, undigested proteins and carbohydrates reach the colon, where bacteria metabolise them to small, osmotically active molecules that draw water but not sodium from the plasma into the colon; this makes the plasma hypernatraemic.11 The continued administration of glucose polymers in the presence of diarrhoea greatly increases this risk12—as happened in the boy described in case 2 and in three of his siblings. Parents find it difficult to recognise exacerbations caused by concomitant illnesses because of persistent vomiting caused by severe gastro-oesophageal reflux. The ratios of surface area to weight in infants that weigh 5 kg are twice those of adults and the skin is more permeable in infants; this means that water loss is much greater in infants than adults—especially in children with pyrexia.13
Dehydration associated with hypernatraemia undoubtedly is under-recognised. Even babies who lose up to 27% of their weight may not be diagnosed correctly outside hospital.8 Only half of such babies' paediatricians seem to notice this hypernatraemia.8
Clinical observations: acute changes in body weight
Body weight changes acutely in children with salt poisoning and in those with dehydration, but in opposite directions. Accurate weighings during treatment therefore provide important evidence. Weight change was not included in previous diagnostic schemes.1
Salt poisoning —Salt ingestion causes weight gain by inducing thirst and stimulating release of antidiuretic hormone—the baby in case 1 gained more than 4% of his original body weight. This volume expansion triggers excretion of the excess salt by suppressing release of aldosterone and stimulating release of atrial natriuretic peptide; this leads to restoration of normal body weight. Weight gain may be slower in babies who cannot regulate their fluid intake.
Dehydration —Dehydration is synonymous with water loss and therefore weight loss; clinical grading is defined by the percentage loss. In case 2, the boy lost 7.6% of his weight and his siblings lost 4.8-14%. These changes are much larger than the normal fluctuations associated with recent feeds, a full bladder, or faecal loading. Occasionally, in infective enteritis, fluid is lost into the lumen of the bowel so rapidly that intravascular hypovolaemia occurs before the diarrhoea is clinically apparent; this means that weights measured at admission underestimate the extent of dehydration.
Measurements of plasma concentrations
Dehydration causes underperfusion, which reduces the glomerular filtration rate. Conversely, the hyperdynamic circulation found in children with salt overload maintains the filtration rate. The rate is best estimated from the plasma creatinine concentration, although changes often are not appreciated for two reasons:
The absolute creatinine concentration is low in infants14
The concentration increases slowly after the glomerular filtration rate falls and may not have reached a plateau when it is first measured.
The paediatrician in case 2 considered the boy's values to be unremarkable, although they indicated that his glomerular filtration rate was only two thirds of the normal value.15 The rate recovered after the boy was rehydrated. The plasma concentration of urea is influenced by too many factors other than hydration to be a reliable indicator of renal function.1 Nevertheless, the differences in plasma concentrations of urea in cases 1 and 2 at presentation and after treatment are striking, and they certainly support the diagnosis of volume repletion and expansion in case 1 and volume contraction in case 2.
The plasma concentration of bicarbonate usually is normal in salt poisoning (as in case 1), but tissue underperfusion from dehydration induces metabolic acidosis and reduces the plasma concentration of bicarbonate (as in case 2). Concentrations of potassium in plasma are unhelpful, as they remain normal in salt poisoning and vary in dehydration, when release of aldosterone makes potassium concentrations fall and reduced glomerular filtration rate makes them rise.
Calculations of fractional excretions of salt and water
Dehydration drives avid reabsorption of sodium in the renal tubular regardless of the concentration of sodium in plasma, because preservation of volume predominates over tonicity. Dehydrated infants are assumed to have low urinary concentrations of sodium, and high urinary concentrations of sodium are assumed to indicate that excess salt is being excreted—but this is only half the story. Hypovolaemia also stimulates maximal water conservation, which increases the concentration of urine, so the final urinary concentration of sodium is unpredictable. In one study of infants with hypernatraemic dehydration caused by infective diarrhoea, the urinary sodium concentration ranged from 35 mmol/l to 232 mmol/l,16 and in dogs with dialysis-induced hypernatraemic dehydration, the concentration ranged from 36 mmol/l to 280 mmol/l.2 The values from our illustrative cases and from previous reports of apparent salt poisoning all fell within these ranges.1 Clearly, high urinary concentrations of sodium alone cannot distinguish salt poisoning from dehydration.
Fractional excretions of sodium and water, which are calculated from the sodium and creatinine concentrations of paired plasma and “spot” urine samples, can distinguish the two situations (see box A on bmj.com). The values should be ≥2% or more in a child who has been salt poisoned and is volume replete and ≤1% in a dehydrated child with viable renal tubules (table). Frequent serial data provide a dynamic picture—for example, the fractional excretion values for the boy in case 2 decreased after he was admitted.
Estimates of salt and water balance
Serial “instantaneous” rates for sodium and water excretion can be estimated from the values of fractional excretion and glomerular filtration rate calculated at each collection time (see box B on bmj.com). Approximate “point” balances can be deduced if the salt and water administration rates are known. These estimates will be imprecise because their calculation requires several assumptions; however, salt poisoning and hypernatraemic dehydration produce such different physiological patterns that the estimates can confirm the diagnosis (table).
On admission, the child in case 1 had an estimated urine output in the normal range, but his sodium excretion was about 15 mmol/kg/day—enough to reduce the plasma concentration of sodium by over 1 mmol/hour and to allow it to correct fully by the next day. In contrast, sodium balance in case 2 was approximately neutral, but he conserved water avidly and retained most of his administered fluids, which fully corrected his plasma sodium concentrations within 17 hours.
Reliance on urinary concentrations of sodium to diagnose salt poisoning in hypernatraemic children is unsafe because dehydrated infants may be diagnosed falsely as poisoned. The child's clinical history and acute changes in their weight provide important evidence that, combined with fractional excretions of sodium and water calculated from serial paired “spot” blood and urine samples and estimates of net sodium and water balances, allow the correct diagnosis to be made.
Contributors: MGC and GBH were involved equally in conceiving the idea for the paper and in developing and writing it. MGC is the guarantor of the paper.
Competing interests None declared.
Two boxes with worked examples are available on bmj.com