Abstract
Waise and Price 1 have discussed the importance of setting the cut-off concentration of thyroid stimulating hormone (TSH) to assist in identifying individuals with ‘subclinical hypothyroidism’. This is important because subclinical hypothyroidism is often diagnosed in the absence of specific symptoms or signs of thyroid dysfunction solely on the basis of a TSH concentration slightly above the upper limit of the reference range in the face of normal circulating thyroid hormone levels. Furthermore, at a TSH concentration above a predetermined cut-off level, thyroid hormone replacement treatment has been suggested. 2,3 The prevalence of subclinical hypothyroidism in different population studies varies from 1% to 17%, with the highest prevalences being reported in the elderly. 2,4,5
Analytical bias and its impact on the upper limit of TSH reference ranges have been eloquently discussed. 1 However, I wish to further highlight another compounding and confusing factor in the diagnosis of subclinical hypothyroidism, namely analytical interference from endogenous antibodies. This source of error alone can lead to erroneously elevated TSH results with a prevalence of ∼0.4%. This indicates that in every 250 TSH measurements, one could be analytically wrong and clinically misleading. 6,7 However, this is purely an analytical and not a clinical error rate: laboratorians and clinicians alike must distinguish between analytical error rate and clinical error rate. In fact, to calculate the probability of a correct diagnostically elevated TSH, the prevalence of subclinical hypothyroidism must be taken into account. 8 The prevalence of this condition in both sexes and at all ages may be reasonably assumed to be ∼2%. To compute the probability of an accurately raised and truly diagnostic TSH result, the number of expected true cases of subclinical hypothyroidism in 1000 patients would be 20 (assuming no false negatives for simplicity); this figure of 20 would then be divided by the sum of true cases plus falsely elevated TSH cases caused only by analytical interference, i.e. 20 + 4 patients. In this case, the probability of a raised TSH result being accurate, and therefore diagnostically correct, may be ∼83% (the range that encompasses extremes of quoted age-related prevalence is 71.4–97.7%). 8 This well-established statistical concept 9 illustrates that, when the analytical false-positive rate of ∼0.4% in TSH measurement, together with the population prevalence of subclinical hypothyroidism are taken into account, the wrong diagnosis could potentially occur in ∼17% of cases. 8–10
It should therefore be emphasized that the clinical error rate in using a TSH cut-off level in the diagnosis of subclinical hypothyroidism is not insignificant when both the disease prevalence and the error rate of analyses caused by interference (both of which increase with age) are taken into account. The error rate from interference can be further compounded by a misguided cut-off TSH value. As pointed out by Waise and Price, 1 the notion of treating each individual patient on his or her own merit regardless of the mildly raised TSH concentration has been suggested as an option. An alternative is to monitor an initial finding of modest TSH elevation at ∼6 monthly intervals to establish a progressive rising trend (e.g. >40%) before initiation of treatment is considered. 1,2,8 This may be prudent because apart from the prompt need to treat children and pregnant women, the potential benefits and risks of therapy in adults with mild TSH elevation identified as subclinical hypothyroidism have been debated over the last two decades with a consensus still lacking. 4,11,12
Declarations