Monday, December 2, 2019
There are some changes to the article: Gordon G. Hyperthyroidism in cats: should we be routinely tesing for early diagnosis? The Veterinary Nurse. 2019; 10(8):444–452.
Having had an expert review and analyse the original raw data of this study (post publication), a few questions have come to light. Due to human error on the researchers behalf, the expert found a very small error in the original data in that one cat with a T4 value of 48 nmol/litre had been counted twice in the analyse thus making the overall prevalence of cats with a T4 value >48nmol/litre being 36.6% and not 37.6%.
The main finding of the study (that a large proportion of T4 levels in cats not showing any signs of hyperthyroidism were above the normal range) is unexpected and the expert had three questions:
‘You say in your Methods section that each test on the inhouse analyser that came back as hyperthyroid (i.e., was above 48 nmol/litre) was sent to an external laboratory for confirmation. I assume that confirmation was another measurement of T4 on the same blood sample. Which external laboratory did you send the blood to, and do you know what analyser they used? I assume (I don't think you actually say in your paper) that the external lab result actually did confirm each high result on the in-house analyser?’
Idexx was the external laboratory that the blood sample was sent off to for a further confirmation using the same sample. The researcher did not follow up on treatment due to limited time.
‘What is the difference between the T4 measurements in Figure 4 (plotted against age) and those in Figure 5 (plotted against weight) — it is clear that the T4 values in Figure 4 are, on average, much higher than those in Figure 5. (It looks like much more than 37.6% of the measures in Figure 4 are above the top of the normal range of 48 nmol/litre.)’
The graphs were generated by statistical package for social sciences (SPSS). The same T4 result was measured against weight and age and the results showed the overall prevalence of elevated T4 levels to be 37.6% (now 36.6%).
Having analysed the original raw data, the expert found an error in Figure 4. The T4 data in Figure 4 is different to those in Figure 5. In Figure 4, far more of the T4 values are plotted >48 nmol/litre than in Figure 5, and in Figure 4, 16 T4 values are >100 nmol/litre. In Figure 4, all the T4 points are plotted too high. The expert suspects that either somehow all the T4 values got multiplied by a number around 1.4 during the plotting of Figure 4 – that would cause this type of error. The expert has recreated a correct graph using the original raw data:
Figure 5 was correct.
For reasons unknown, Figure 3 has an error. The median T4 level was 45 nmol/litre, but in the ‘box & whiskers’ plot, the median has plotted it >50 nmol/litre. As well as the median being plotted too high, you can see that cat 71, with T4 = 153 nmol/litre, was also plotted too high. The expert has recreated a histogram using the original data to represent the range of T4 results to be correct:
‘In hyperthyroid cats, usually — but not always — the liver enzymes (especially ALT) are raised. Presumably, in your pre-anaeasthetic profiles, the liver enzymes were measured? If so, did the cats with high T4 in your data set have higher ALT values than those with normal T4?’
Although access to full pre-anaesthetic blood results were available, due to the researchers clinical experience as a veterinary nurse and not a veterinary surgeon, the liver enzyme results were not analysed and only the T4 measurements were recorded.
One possible explanation for the apparent discrepancy was that the in house analyser used was sometimes giving artificially high T4 readings. Due to time constraints Idexx T4 values that were sent off as a confirmation of hyperthyroidism were not confirmed by the researcher thus further studies need to be done.