Classically, we are taught to look out for an elevated anion gap in patients with a metabolic acidosis. Although much less common, a low anion gap can also be a useful sign and there are a variety of causes.
The commonest cause is lab error, particularly in the measurement of the serum sodium. As previously discussed, high serum lipids or high serum proteins can lead to spuriously low serum sodium measurements thus altering the AG. Severe hypernatremia can also lead to errors in measuring the sodium concentration (usually underestimating the real result) and will lower the AG. Similarly, errors in measuring the chloride or the HCO3 will alter the perceived AG. The bicarbonate is usually measured indirectly and allowing the sample to sit without separating the cells can lead to increased production of CO2 and thus lower the AG.
Apart from lab errors, the commonest cause of a low AG is due to alterations in serum protein levels. Most of the AG is due to negative charges on circulating proteins, primarily albumin so that if the albumin concentration falls, the AG will fall also. It is generally accepted that the AG should be corrected upwards by 2.5 for every 1g/dl fall in the serum albumin. This applies also for patients with an elevated serum albumin; the AG should be corrected downwards in that scenario. Although they do not normally contribute significantly to the AG, immunoglobulins can be important in patients with paraproteinemias. IgG tends to be cationic while IgA is an anion. Thus, patients with an IgG paraproteinemia and a high tumor burden can have a low or even negative AG. In contrast, patients with IgA paraproteinemia will have an elevated AG.
Calcium and magnesium could theoretically decrease the AG if they are significantly increased. In practice, however, hypercalcemia does not usually cause a lowered AG unless it is due to hyperparathyroidism. Other causes of hypercalcemia are not associated with changes in the AG. It is uncertain why this is the case. Hypermagnesemia usually does not affect the AG because it is normally accompanied by sulphates and as these are unmeasured anions, they balance each other out.
Several drugs are associated with reductions in the AG. As mentioned by Nate before, bromide intoxication is a rare cause of a negative anion gap. At first, this does not appear to make sense; bromide is an anion, similar to chloride. As a result, elevated bromide levels should cause an increase in the AG. However, bromide interferes with the chloride analyzer – every 1 mEq increase in bromide leads to a reported 3 mEq increase in chloride. Thus, patients with bromide intoxication can have extremely negative AG. Iodide can also interfere with the chloride assay and lead to a negative AG.
Lithium is a cation, in the same family as sodium and thus lithium poisoning will lead to a reduction in the AG although usually only when the level is above 4. This can be a clue to a lithium overdose in a patient with suspected poisoning where there is no ready access to lithium levels.
I would recommend an excellent review of the uses of the anion gap published in CJASN in 2007 that addresses all of these in great detail.