A reply


We are grateful for Dr Fawcett's comments on our article [1]. We can confirm that when blood gas and laboratory samples were being sampled from the arterial line flushed with sodium chloride 0.9% with glucose 5%, no dramatic changes of sodium, potassium or chloride occurred and these stayed within the laboratory reference ranges. As described in our article, the patient was deteriorating before the drug error occurred and had developed marked sepsis, a new chest infection and acute kidney injury. Consistent with this, levels of inflammatory markers rose significantly and pH fell modestly. Even in retrospect, the changes in his blood chemistry and arterial blood gases were entirely consistent with worsening sepsis and kidney injury and do not raise concerns about sample contamination.

When two fluids containing concentrations A and B of a substance are mixed in proportions X and Y (where Y = 1 – X), the final concentration of the substance (F) can be calculated as follows:

display math

In our case, blood glucose samples taken from the arterial line had variable glucose concentrations, ranging from normal to 27 mmol.l−1, with several recordings in the mid-teens. Assuming a true blood glucose of 5 mmol.l−1 and a measured blood glucose of 16 mmol.l−1, we can calculate this implies a 4% contamination (0.04 ml in 1 ml). If we assume the same degree of contamination, it is easy to see how far this contamination (with sodium chloride 0.9% with glucose 5%) will alter normal blood chemistry (Table 1).

Table 1. Effect of contamination on blood sample biochemistry. The volume of blood (X, above) is 0.96 ml and that of flush (Y, above) is 0.4 ml.
 Concentration or value in bloodConcentration or value in flush contaminantFinal concentration ((A × X) + (B × Y)) or value
Glucose; mmol.l−1528016
Sodium; mmol.l−1145154145
Chloride; mmol.l−1100154102
Potassium; mmol.l−1504.8
Hydrogen ion; nmol.l−1403160165

It will be noted that only glucose and hydrogen ion concentrations (and hence pH) are markedly different from those in the blood (or indeed outside normal laboratory reference ranges). Based on physiological considerations, and our observations in this case, buffering of hydrogen ions in the metabolically active blood markedly reduces any observed alteration in pH.

As we emphasised in our paper, sodium chloride 0.9% with glucose 5% appears to be a particularly dangerous fluid in these circumstances. It has the capacity to be readily mistaken for sodium chloride 0.9% and a high enough glucose concentration to lead to markedly elevated glucose levels with very low levels of sample contamination. Other blood biochemistry is highly unlikely to be dramatically altered.