| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Letters to the Editor |
Department of Diabetes and Endocrinology City General Hospital Stroke on Trent, ST4 6QG United Kingdom
Dr. Oelkers raises some important issues. We can summarize these with three main questions: 1) What is the best cut-off to be used for LDSST? 2) Is there any difference between LDSST and SST? 3) What are the potential technical pitfalls with LDSST?
1. What is the best cut-off?
We do not agree with Dr. Oelker’s method of setting cut-off points. Deriving a fixed cut-off from responses in normal subjects is not really the correct way of doing it. Cut-off points for screening tests are normally determined by comparing performance to a "gold standard" in the population of interest. A most important factor in choosing the desired sensitivity and specificity of any screening test is the nature of the disease or condition under investigation. The cut-off should be set to maximize specificity if a false positive result is extremely undesirable, either in terms of anxiety or discomfort to the subjects (e.g. a false diagnosis of cancer), or because it would lead to further expensive investigations or a dangerous procedure (e.g. major surgery). On the other hand if a false negative result must be avoided, (i.e. when the consequences of missing the condition or disease would be deleterious), then the cut-off must be set to maximize sensitivity (1). HPA-axis impairment is a condition that has potentially serious consequences if missed, yet is easily treatable when detected. Hence a high sensitivity, not specificity, is the priority. A false positive result, in this case, will only lead to further clinical evaluation and, at most, an insulin tolerance test. By using the cut-off derived from responses in normal individuals, Dr. Oelkers found that both the LDSST and the SST have a similar sensitivity and specificity (65% and 96% respectively) for detection of HPA-axis impairment and therefore concluded that there is no need to replace the SST with LDSST. We would argue that a test with a sensitivity of 65% is not good enough for screening for HPA axis failure, a 35% false negative (false reassurance) rate would not be acceptable. The sensitivity of either test can be improved (at the expense of specificity) by increasing the cut-off value.
We further analyzed our data and determined the sensitivity and specificity of the LDSST and SST at eight different plasma cortisol cut-off points (300, 400, 450, 500, 550, 600, 650 and 700 nmol/L) compared with the gold standard of ITT and clinical evaluation in the 64 patients. Sensitivity is plotted as a function of 1-specificity (false positive rate) in the Receiver Operator Characteristic (ROC) curve (see figure). The curve portrays the trade-off involved between improving either tests’ sensitivity or their specificity (2).
An ideal test with 100% sensitivity and 100% specificity is one that reaches the upper left corner of the graph. In the steep part of the graph the sensitivity increases a great deal while the specificity (and false positive rate) hardly change. Moving up the curve in this section increases sensitivity without substantially reducing specificity. As we move along the graph, in the flat part of the curve, the sensitivity does not change but the specificity decreases progressively. The best cut-off point is therefore where the ROC curve is just on the plateau of sensitivity. By inspecting both curves, a cut-off of 600 nmol/L is optimal, ensuring maximal sensitivity (100% for LDSST and 91% for SST) and acceptable specificity (83% for both tests).
Although we chose the cut-off in our paper from other published work, the ROC analysis does confirm that a cut-of of 600 nmol/L is optimal.
2. Is there any difference between LDSST and SST?
The answer to this question from our data is yes, there is enough evidence to suggest a small but definite superiority of LDSST over SST.
From our data the diagnostic efficiency of the two tests can be
compared by comparing the area under the ROC curves (see fig. 1
), the
test with greater area under the curve having better performance
(2).
|
When we analyzed our results in the 42 patients who had ITT as well as LDSST and SST, we found similar sensitivity and specificity. However in the remaining 22 patients who only underwent LDSST and SST the SST failed to identify 2 patients with clinically evident cortisol deficiency, yet both were correctly identified by the LDSST. Overall LDSST performed slightly better than the SST. This is in keeping with most of the data published in this field (6, 7, 8, 9).
Although in absolute terms the superiority of LDSST over SST is small, the fact that low dose test can save a number of patients from potentially life threatening adrenal crises is a strong argument in favor of this test.
3. What are the potential pitfalls?
We agree with Dr. Oelkers that dilution of the contents of a 250-µg ACTH 1-24 ampoule could be troublesome. We did not advocate that inexperienced personnel carry out the dilution. In our center the dilution is performed by the hospital pharmacist under the very strict protocol according to published evidence. The availability of a 1-µg ampoule would certainly be an advantage.
We believe there is enough evidence to recommend that LDSST replace SST for assessment of HPA-axis. Like any new test our experience with the LDSST test procedure performance will improve with more usage.
Footnotes
1 Address correspondence to: Tarig A.M. Abdu, Department of Diabetes and
Endocrinology, City General Hospital, Stroke on Trent, ST4 6QG United
Kingdom. 
Received April 18, 1999.
References
This article has been cited by other articles:
 |
|
 |
|
  R. I. Dorin, C. R. Qualls, and L. M. Crapo Diagnosis of Adrenal Insufficiency Ann Intern Med, August 5, 2003; 139(3): 194 - 204. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
