Patterns of the low‐dose dexamethasone suppression test in canine hyperadrenocorticism revisited

Abstract Background The low‐dose dexamethasone suppression test (LDDT) is considered an accurate screening and valuable differentiation test in dogs with suspected hyperadrenocorticism (HAC). A recent study showed that the different response patterns not only provide complementary information about etiology, but also the probability of HAC in these patients. Objectives We aimed to determine the diagnostic test performance of LDDT response patterns in a population of dogs from an animal hospital. Methods The electronic database was retrospectively searched for dogs suspected of HAC that were given an LDDT. Dogs with acute non‐adrenal illnesses during the test were excluded. Response patterns were classified as complete suppression, lack of suppression, partial suppression, escape, inverse, and increasing patterns. Cortisol concentrations ≥ 27.59 nmol/L (≥1 µg/dL) 8 hours after dexamethasone administration were considered positive results irrespective of the patterns observed. Calculations included likelihood ratios (LRs) and predictive values (PVs). Results HAC and non‐adrenal illness were diagnosed in 115 (54%) and 62 (46%) dogs, respectively. The positive (+) LRs (95% CI) for the lack of suppression, partial suppression, escape, and an inverse pattern to diagnose HAC were infinite, 8.09 (2‐32.72), 3.23 (0.75‐14), and 0.2 (0.06‐0.73), respectively. Conclusions The study confirms that the “lack of suppression” pattern strongly supports a diagnosis of HAC. It shows that the “partial suppression” pattern moderately increases, and the “inverse” pattern decreases the likelihood of HAC. The fact that the study found no association between the “escape” pattern and a diagnosis of HAC, does not support its integration into decision making.


| INTRODUC TI ON
In 2013, the American College of Veterinary Internal Medicine (ACVIM) published a consensus statement on the diagnosis of spontaneous hyperadrenocorticism (HAC) in dogs. 1 The low-dose dexamethasone suppression test (LDDT) was propagated as the screening test of choice with a recommendation to re-evaluate the cutoff values. Further endocrine testing was strongly recommended for dogs with an "inverse" LDDT pattern, that is, a cortisol concentration > 27.59 nmol/L (>1 µg/dL) 4 hours (t 4 ), and < 27.59 nmol/L (<1 µg/dL) at 8 hours (t 8 ) after dexamethason administration, which was traditionally interpreted as a negative test result.
Mueller et al were the first to describe this pattern in five dogs with pituitary-dependent hyperadrenocorticism (PDH). 2 The authors hypothesized that it could reflect a new type of HAC, but also detected this pattern in two of 29 dogs with initially suspected, but later excluded, HAC. 2 Five different LDDT patterns and the respective positive predictive values (PPV) to diagnose HAC were investigated in a subsequent study. 3 The patterns were defined as complete suppression (t 4  The "inverse" and the "escape" patterns had very low PPVs, and the authors raised a concern that these patterns might not be supportive of HAC. The "escape" pattern is currently considered a classic and common pattern of HAC. 4 Additionally an "increasing" pattern, defined by a >50% increase in cortisol concentrations between any time point, was found to be potentially useful for differentiating pituitary-dependent (PDH) and adrenal tumor hyperadrenocorticism (ATH). 3 The differentiation between HAC subtypes is important, as the choice of treatment and prognosis could differ significantly. 1 Using currently established criteria, approximately 60% of the dogs with HAC can be identified as having PDH with the LDDT alone. 4 For dogs without a suppression pattern, additional tests, such as high-dose dexamethasone tests, endogenous adrenocorticotropic hormone (ACTH) measurements, or diagnostic (adrenal or pituary) imaging, are necessary. Bennaim et al described an "increasing" pattern in 6 of 31 dogs without suppression, all of which were diagnosed with PDH. 3 The authors speculated that transient increases could be a consequence of ACTH stimulation, not expected in dogs with ATH, and recommended additional studies be performed, including the use of more dogs.
The unexpected poor performance of the "escape" pattern in the study by Bennaim et al and the presentation of the new "increasing" pattern for differentiation, 3 prompted us to perform this retrospective study. The primary aim was to investigate the diagnostic performance, primarily looking at the likelihood ratios (LRs) of various LDDT patterns to diagnose HAC in a population of dogs with suspected HAC. The hypothesis was that individual LDDT patterns, especially the lack of suppression and partial suppression patterns, have a high likelihood of diagnosing HAC and, thus, providing additional diagnostic support.

| Endocrine tests and assays
For UCCR determinations, urine was collected at home, preferably, but not necessarily, in the morning. UCCRs < 26.5 × 10 −6 , between 26.5 and 161.2 × 10 −6 , and > 161.2 × 10 −6 were considered negative, suspect, or supportive for HAC, respectively. 11 Cortisol was measured using a competitive chemiluminescent immunoassay (Immulite 1000 and Immulite 2000xpi Cortisol, respectively, before and after 2017; Siemens Healthcare Diagnostics) validated for use in dogs. 12,13 Endogenous ACTH was measured in EDTA-plasma using the Immulite 1000 and Immulite 2000 xpi ACTH assay (Siemens Healtcare Diagnostics) before and after 2017, respectively. An ACTH concentration in dogs with confirmed HAC of ≥2.2 pmol/L and later ≥1.1 pmol/L was considered diagnostic for PDH. 6,7 Total ALP and HS-ALP activities were analyzed in lithium-heparin plasma. Total ALP was measured using the Cobas ALP2-assay on a Roche/Hitachi Cobas c502 analyzer (Roche Diagnostics), and HS-ALP was assessed using the heat stability method at 65°C.

| Statistical analysis
Distribution of the data was assessed with the Kolmogorov-Smirnov test and data were given as the median and range. The

| Study population
Three-hundred and thirty-four LDDT test results were retrieved.

| D ISCUSS I ON
The results of the present study agree with earlier studies assigning a high sensitivity to the LDDT for the diagnosis of HAC, 3,14-18 As suggested by Bennaim et al, our results confirm that the "inverse pattern" provides no support for HAC. 3 The sensitivity or true-positive rate of the LDDT (not including the inverse pattern as a positive test) to diagnose HAC in our patients was 89%, comparable to the 85%-100% reported in the literature. 3,[14][15][16][17][18] Possible explanations for sensitivities >95% in earlier studies could, in part, be attributed to the different criteria used to confirm the diagnosis. In at least two studies, the LDDT, as a sole test, was used to confirm the diagnosis, which is questionable. 16,17 In another study, only dogs with a complete necropsy report were included, which likely selected for animals with more severe or advanced disease. 18 Although the -LR of 0.12 supports an LDDT result that can eliminate an HAC diagnosis in unaffected dogs, the results also show that false-negative results are possible and that diagnosis should not be based on LDDT-testing alone. Traditionally, a t 4 cortisol concentration was used to discriminate between PDH and ATH exclusively, whereas the t 8 concentration was used for both screening and discrimination. Mueller et al described an "inverse" dexamethasone response pattern with a high cortisol concentration at t 4 , but physiologic suppression of cortisol concentrations was seen at t 8 in 5 (6.25%) of 80 dogs with confirmed HAC. 2 The authors speculated that the "inverse" pattern might represent a new HAC type and that the t 8 interpretation might not be accurate. A later retrospective study investigated the diagnostic performance of the LDDT, looking at various patterns in 123 canine patients with suspected HAC and found the "inverse" pattern in 5/123 (4%) dogs. 3 In that study, the authors recommended using alternative diagnostic criteria for dogs with this specific pattern since there were only 2 (3.4%) and 3 (4.7%) dogs in the HAC-and non-adrenal illness group, respectively, with that pattern. The number of dogs was deemed too small to draw final conclusions. This recommendation is clearly supported by the results of the current study, as the integration of the inverse pattern reduced the +LR and the PPV of the LDDT from 13.75 to 4.72 and from 96.2% to 88.8%, respectively, without a considerable change of the −LR and NPV. LRs, in contrast to PVs, are independent of the disease prevalence and are thought to constitute one of the best ways to measure and express diagnostic test accuracy. The +LR is used as a pre-test probability multiplier of the respective disease to estimate its post-test probability. +LRs can range from 0 to infinity, and the higher the number, the more likely the findings suggest the presence of the disease. A +LR >10 depicts substantial changes of post-test probability estimates and increases the probability of disease by approximately 45%. 19 The specificity of the LDDT overall, or in other words, the probability of a negative test result in a dog without HAC, was 93% and clearly higher than in all earlier studies. Comparable specificities were found in only one study, where healthy dogs were used as controls. 17 All other studies reported specificities between 67% and 73%. 3 nine dogs with this pattern in the present study, which was considered a low number. Therefore, the significance of this finding was reduced. However, the +LR near 1 and the 95%CIs that included 1 suggest that an association between the increasing pattern and a pituitary origin for HAC was unlikely.
An increased ALP activity, frequently exceeding 1000 U/L, is the most common biochemical finding in dogs with HAC. 21 Corticosteroid-induced ALP is synthesized by the liver after exposure to glucocorticoids and is unique to dogs. This analyte can be easily measured using a routine laboratory procedure due to its heat stability at 65°C. 25 Although the HS-ALP analyte is commonly measured in dogs with suspected HAC; its diagnostic value is ambiguous. The fact that the AUC-ROC of the HS-ALP and total ALP tests were comparable in this study suggested that the HS-ALP determination is redundant. Further studies on this topic are in progress.
Besides the retrospective study design and the low number of patients with specific patterns widening the 95% CIs, other limitations of the study must be recognized. First, the final diagnosis of pituitary HAC was based primarily on the response to therapy and was rarely verified by histopathology. Thus, placebo effects 26  In conclusion, this retrospective study is in line with earlier studies that assigned LDDT with a good sensitivity for diagnosing HAC in dogs with appropriate clinical signs. The "inverse" pattern decreased the likelihood of HAC, although it remains unclear to what degree, considering the wide 95%CIs in the present study. This study confirmed that the "lack of suppression" and "partial suppression" patterns strongly and moderately supported a diagnosis of HAC, respectively. As no association could be shown between the "escape" pattern and a diagnosis of HAC or non-adrenal illness, additional diagnostic tests are strongly encouraged in dogs with this pattern.
Finally, the low +LR of the recently proposed "increasing" pattern does not support its use as a discriminatory test between PDH and ATH in dogs with HAC.