Somatostatin receptor ligands (SRLs) are the first-line medical treatment of patients with acromegaly (1–5). However, their efficacy is approximately 50% and treatment response assessment requires about 6 months (6–11). Furthermore, there have been more medical therapy options for acromegaly, and personalized medicine will be the focus of all treatment decisions in the near future. Then, it is of utmost importance to investigate predictive factors of the individual response for each patient (12–14).

As originally described, the acute octreotide test (AOT) is a functional test that consists of the administration of 100mcg of subcutaneous octreotide and the determination of the GH nadir or the decrease of GH during the following 6 hours. Its results have been related to long-term SRL response, and its utility has been extensively evaluated with some controversial results probably related to methodological differences: the use of 100 or 50mcg of octreotide for the test, the definition of the long-term response according to different parameters (GH levels, GH decrease, and IGF1 levels) and the use of different SRL presentations for long-term treatment (15–28). Moreover, the long duration of the originally described procedure limits its usefulness in clinical practice (4). In 2016, Wang et al. (15) reassessed the AOT predictive capacity with a very stringent methodology, obtaining remarkably good sensitivity and specificity values, but the methodology was still complex, time-consuming, and expensive. In 2008, we described a short version (sAOT) of the classic AOT where 100mcg of subcutaneous octreotide was administered, and the GH was determined at 2 hours post-administration (GH_2h), as we found that this time point concentrated most if not all the nadir values. With this testing modality, we were able to predict the long-term response (more than 6 months) to SRLs with the cutoff for GH_2h of 3.6 ng/mL, which presented a negative predictive value (NPV) of 89% to identify non-response. This cutoff would help clinicians rule out the use of SRLs in monotherapy as initial medical therapy, and introduce pasireotide or pegvisomant earlier, alone or in combination with SRLs, thus shortening the time to control (18).

AOT results have also been linked to some particular molecular specificities such as the expression of SSTR2 and E-Cadherin (19, 29), both highly related to good SRL response mostly in terms of normalization of IGF1 (30).

Employing current GH ultra-sensitive assays, this study aimed to reevaluate prospectively the capacity of the sAOT to predict long-term response to SRLs in a cohort of patients with acromegaly included in the ACROFAST trial. We also aimed to evaluate the correlation of the functional predictive results of the sAOT with the expression of those molecular tumor biomarkers with high predictive capacity, such as E-cadherin and SSTR2.

We analyzed a cohort of 47 patients with acromegaly (24 men and 23 women); the mean age at diagnosis was 53.6 ± 13.8 years. Of the total, 37 subjects corresponded to recently diagnosed patients and 10 to non-cured cases after surgery. A total of 30 out of 47 patients were identified as responders (26 CR and 4 PR) and 17 as NR. The group of 10 patients non-cured after surgery presented a slightly better response than the newly diagnosed patients (6 CR, 3 PR, and 1 NR, vs 19 CR, 1 RP, and 16 NR, respectively; p=0.01). Information and differences in clinical, hormonal, and radiological features between responders and NR patients are presented in Table 1 . Responder patients were older and presented lower basal GH and IGF1 concentrations and lower tumor diameter. Apart from a better biochemical response, they presented a higher tumor volume shrinkage [Responders: 49.7 (12.2-89.7) % vs NR 12.5 (0-23.2) %; p=0.01].

Prediction of therapeutic response to a given compound with high accuracy for an individual patient is the cornerstone of precision medicine. Functional tests have been extensively used for diagnosis, evaluation of disease activity, and assessment of cure in most endocrine diseases. To be useful in clinical practice, a predictive test must be highly accurate, but it must also be easily implementable in the day-to-day clinical practice, and, if possible, inexpensive. The prediction of response to SRLs started historically in the early 90s when regular octreotide was given two to four times per day for controlling GH hypersecretion in patients with acromegaly (28). The GH_nad was used to schedule the number of doses to be delivered per day in each case by evaluating the GH drop after delivery of 100 mcg of regular octreotide in the subsequent hours. During the 1990s and 2000s, different authors evaluated the ability of the original 6 hours AOT - with sample collection every 60 minutes - to predict GH control in patients treated with long-acting SRLs with variable results and conclusions. Wang et al. reported in 2016 that this classic version of the AOT has a consistent capacity for prediction of SRL response in acromegaly (15), but it has never been included in the recommendations of clinical guidelines. We have previously reported that a short version of the classic AOT, in which GH_nad was established at 2 hours after octreotide injection, was reasonably predictive of the long-term response to SRLs (18). In that study, we found that a GH_2h cutoff of 3.6ng/mL was a measure of non-response with an NPV of 89% and that GH_2h presented a correlation of 0.76 with 6 months IGF1-SDS.

In the present study, we aimed to reevaluate our previously described cutoff points and their capacity to predict the response to SRLs by using a prospective cohort and current GH immunoassays. Our results confirmed the robustness of the AOT in this shorter version, with an accuracy of 81% (PPV for response 96%) and 74% (PPV for non-response 86%) for a cutoff of 1.4ng/mL and 4.3ng/mL, respectively. We also examined if tumor shrinkage was also predictable in parallel to biochemical response and we found that those cases showing a poor hormonal response were also those in which less marked volume reduction were found. Our results thus confirm that non-responder patients present an extremely low shrinkage compared to hormonal-responsive patients. The accuracy values obtained ranged from 81% for responsiveness to 71% for non-responsiveness, depicting the high heterogeneity of somatotroph tumors, especially those NR cases. However, the predictive ability shown in the present study may be more than sufficient to apply the sAOT in clinical practice, or at least indicates that it is much better to use it than not use it. In addition, this short version is easy to perform and inexpensive, and the results can be obtained quickly enough to promptly decide on medical treatment.

As expected, in the present study the cutoff values obtained are lower than those described in our previous study, which was published more than 12 years ago. The current values were obtained according to new immunoassays available, which have higher sensitivities. They go in parallel with the new, much lower cutoff values established in the latest clinical guidelines for the definition of acromegaly control and acromegaly remission after surgical treatment [GH<0.14ng/mL (4, 37) and <0.4ng/mL (5, 38)].

Most previous studies regarding AOT demonstrated its usefulness (15–18, 21–26, 28). However, some had methodological limitations that could have biased their interpretation, including retrospective studies, a small number of subjects, patients treated with radiotherapy, cumbersome octreotide presentations that could interfere with the adherence, and different response criteria (GH < 5mU/L, IGF1 normalization, IGF1 50% decrease or IGF1-SDS, among others). One of the relevant findings of the present study is the robust correlations observed, as in other studies, between the GH decrease observed at AOT and the long-term GH decrease (16, 21); these decreases were observed between GH_2h and basal GH (19, 22), and between post-treatment GH (17, 22, 28) and post-treatment IGF1 (17, 18), with mixed results regarding GH_2h and basal IGF1 presenting a positive correlation in some studies (17, 22) but not in others (18, 21). A different GH_nad between responders and non-responders has also been previously described (15, 17, 23), which is in fact, the key to its application in clinical practice. Our correlation analysis and the differences found in GH_2h and the %ΔGH_2h among responder and non-responder patients are concordant with the results described in these studies.

We also investigated the ability of basal GH and IGF1 to predict SRL response, as previously described by other researchers (17–19, 21, 39–41). The predictive ability of both parameters can be useful, but only IGF1 is able to define the cutoff point for responders and non-responders. Moreover, our data show that GH_2h is the best predictor tool, being able to identify both responders and non-responders, showing the best accuracy and predictive values, and being consistent enough for use in clinical practice.

Patients with smaller tumors presented a lower GH_2h and a better response as reported by other studies (16). Wang et al. described an AUC of 0.794 for the AOT ability to predict tumor volume reduction (15). In the present study and concordant with these previous findings, GH_2h also correlated with volume decrease.

Non-cured after-surgery patients presented better responses to SRL treatment than the newly diagnosed ones, probably due to the debulking surgery effect as previously described (42, 43). However, we did not observe in our cohort a lower tumor volume in this group prior to sAOT and treatment initiation. GH_2h in post-surgical active disease cases did not show statistical differences compared with newly diagnosed patients, although the results could be influenced by the small number of non-cured after-surgery patients. When we performed the statistical analysis focused only on newly diagnosed patients, we obtained the same results as when pooling all the cases. This data reinforces that sAOT can be useful either in post-surgery patients or in newly diagnosed cases.

Interestingly, GH_2h was lower among positive E-cadherin tumors. Patients with positive E-cadherin did present a higher SSTR2 expression confirming that both biomarkers are usually expressed concordantly (30, 44). Casar-Borota described an association between SSTR2 IRS-score and AOT results (r_s -0.29; p=0.02) (45), which was not replicated in our study. Moreover, the differences in GH_2h obtained according to E-cadherin but not to SSTR2 immunostaining showed a higher predictivity of the former biomarker, as previously reported (30). When E-cadherin is present in GH-secreting tumors, it seems to reflect a greater phenotypic somatotropic differentiation (46–48), confirming that it is a comprehensive molecular marker not just reflecting intercellular adhesion, but also including the information regarding the responsiveness to SRL treatment (30, 49, 50). The IHC determination of E-cadherin is a cheaper, easily evaluable, robust, and daily-used technique in all pathology clinical laboratories, which can facilitate its implementation for somatotropinomas theragnostic assessment; and, interestingly, with no inferior (or even superior) properties to predict SRL response compared with SSTR2 (30). For these reasons, we consider that E-cadherin should be implemented in clinical practice even before SSTR2 immunostaining.

The present work has some strengths: it is prospective, and the number of patients is sufficient for a proof-of-concept study. One limitation to defining the GH_2h cutoff is the use of different GH assays (51). To overcome this aspect, we have harmonized GH values as Müller et al. described (36). They made comparisons between GH determinations obtained from children and adolescents with suspected GH deficiency through eight different assays, three of which were used in the present study (Immulite i2000, Liaison XL, and UniCel DxI 800 Access). Assay standardization and different statistical harmonization strategies were performed. Using the linear regression equations, inter-laboratory assay variability was reduced from a CV of 68.21 ± 45.6% to 32.3% ± 29.0% for GH <1 ng/mL and from 28.2% ± 11% to 15.4% ± 11.7% for GH 1-4.99 ng/mL, thus, enhancing the possibility to use our GH cutoffs with all the GH assays included in the present work. On the other hand, some inconsistencies or relatively unusual clinical situations must be clarified, such as those subjects with a basal GH value lower than the theoretical cutoff point for responsiveness. We identified three cases with a basal GH <1.4 ng/mL. Of these three, two were included after surgery and presented a %ΔGH_2h of 89% and 76%. Both presented a complete response to SRLs. The third case was a newly diagnosed acromegalic patient with a %ΔGH_2h of 19% who did not respond to SRLs. This information exemplifies that even if basal GH is relatively low, the sAOT retains a good capacity to predict long-term response to SRLs and to assess a predictive response to medical treatment. In addition, the use of different assays for measuring IGF1 may limit the interpretation of our results. Even if the use of SDS or %ULN may allow comparability among centers using different IGF1 assays, the upper limit of normality of the different assays we used shows relatively different values (52), thus, the final upper cutoff from assays to define responsiveness may also be affected.

Currently, SRLs are the established first-line medical treatment for all patients. Tools such as sAOT can help clinicians overcome therapeutic ineffectiveness when a GH_2h is high and a non-response to SRLs is expected, which can be quickly confirmed with this functional test. Thus, in those cases with a prediction of non-response, the primary use of SRLs could be avoided and the use of pasireotide or pegvisomant as first-line medical treatment would be recommended according to the specific characteristics of the patient and the tumor. Time and cost would be potentially saved with such an approach. In conclusion, we report that the short version of the AOT performed within 2 hours from the time of diagnosis or soon after surgical failure, and using the GH immunoassays included in the present paper, is robust enough for its implementation as a predictive SRL response tool at the clinical practice level and is capable of guiding the medical treatment of acromegaly.

The raw data supporting the conclusions of this article are available from the corresponding author on reasonable request.

The study was approved by the Comite d’ètica i investiació de l’Hospital Germans Trias i Pujol. The study was conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

MM-P: Formal Analysis, Methodology, Writing – original draft. JG: Formal Analysis, Supervision, Writing – review & editing. EV: Writing – review & editing. MH: Writing – review & editing. BB: Writing – review & editing. OG-P: Writing – review & editing. SM: Writing – review & editing. CC: Methodology, Writing – review & editing. LP: Methodology, Writing – review & editing. RV-T: Writing – review & editing. MA-C: Writing – review & editing. CB: Writing – review & editing. ISi: Writing – review & editing. AS: Writing – review & editing. GX: Writing – review & editing. FV: Writing – review & editing. IP: Writing – review & editing. RG-C: Writing – review & editing. RZ: Writing – review & editing. FH: Writing – review & editing. MMo: Writing – review & editing. AA: Writing – review & editing. NV: Writing – review & editing. SL: Writing – review & editing. MC: Writing – review & editing. PM: Writing – review & editing. CA-E: Writing – review & editing. AP: Writing – review & editing. MS: Writing – review & editing. ISa: Writing – review & editing. CF-M: Writing – review & editing. RC: Writing – review & editing. IB: Writing – review & editing. MJ: Writing – review & editing. SW: Writing – review & editing. MMa: Writing – review & editing. MP-D: Conceptualization, Funding acquisition, Supervision, Visualization, Writing – review & editing.