Normal Insulin-like Growth Factor 1 During Somatostatin Receptor Ligand Treatment Predicts Surgical Cure in Acromegaly

Abstract

Context

Treatment with somatostatin receptor ligands (SRLs) is often given before pituitary surgery to patients with acromegaly.

Objective

To study whether the response to treatment with SRLs is predictive of surgical outcome.

Design

Retrospective, observational study.

Setting

Tertiary care center.

Patients

We investigated 272 patients with acromegaly who had been treated with an SRL between 1990 and 2018.

Intervention

All patients underwent pituitary surgery performed by a skilled neurosurgeon.

Main outcome measure

Outcome of pituitary surgery in patients who had normalization of insulin-like growth factor 1 (IGF-1) levels during SRL therapy in comparison with patients who did not normalize IGF-1 levels.

Results

Normalization of IGF-1 levels during SRL treatment occurred in 62 patients (22.8%) and was similar for the 3 different types of SRL (P = .88). Surgical remission occurred in 59.6% of the patients. Patients who normalized IGF-1 levels during SRL treatment had a higher probability of surgical cure than patients without IGF-1 normalization (83.9% vs 52.4%, respectively; P < .001). Multivariate analysis confirmed that lack of cavernous sinus invasion, small maximum tumor diameter, and IGF-1 normalization during SRL therapy were the only factors independently associated with a favorable surgical outcome.

Conclusions

Our study demonstrates that the normalization of IGF-1 levels during treatment with SRLs is an independent predictive factor of a favorable surgical outcome. The underlying mechanisms remain unclear, but an optimal response to medical therapy may be a characteristic of less aggressive tumors that are more likely to be entirely removed at surgery.

The first-choice treatment in most patients with active acromegaly is transsphenoidal surgery (TSS) performed by a pituitary-dedicated neurosurgeon (1-3). However, even in the more experienced hands, the overall surgical remission rate is 60% to 70% (4-6). The probability of hormonal remission is particularly low in patients with a very large tumor and invasion of 1 or both cavernous sinuses. It has long been hypothesized that obtaining a reduction of tumor size before surgery could translate into a better surgical outcome. Somatostatin receptor ligands (SRLs), the cornerstone of the medical treatment of acromegaly, act directly at the pituitary level by inhibiting the secretion of growth hormone (GH) by the tumor cells. Moreover, they also exert an antitumoral effect, as suggested by a significant shrinkage of the tumor that occurs in more than 50% of treated patients and is more evident in treatment-naïve patients (7, 8). Considering these premises, it is not surprising that the role of presurgical treatment with SRLs has been investigated for several years. Opposite conclusions have been reached in different studies: some authors found an improvement of surgical results in SRL-pretreated patients (9-12), while others found no significant difference (13-15). This uncertainty led to the suggestion against the routine use of presurgical SRL therapy to improve biochemical control after pituitary surgery (2). The authors underlined that the improved results reported in patients with macroadenomas in small prospective trials may have been exaggerated due to a carryover effect of SRL treatment on the postoperative hormonal evaluation at 3 months (9-11). Indeed, an extension study of the Norwegian trial found that at 1 year the results favoring presurgical SRL treatment were no longer significant (16).

Interestingly, all the published literature in this field has centered on the comparison between SRL-pretreated patients and untreated controls. No single paper reported or analyzed the data within pretreated patients. In particular, the question of whether responsiveness to SRL may serve to predict surgical outcome has never been investigated.

We decided to perform an observational and retrospective analysis of all patients who had undergone pituitary surgery at our center and that had received presurgical treatment with SRLs to verify whether the response to medical treatment has prognostic relevance for surgical outcome.

Materials and Methods

Patients

Between 1990 and 2019, 889 consecutive patients with active acromegaly received surgery for removal of a GH-secreting pituitary adenoma at the Department of Neurosurgery of the Istituto Scientifico San Raffaele in Milan, which serves as a referral tertiary care center for the surgical treatment of patients with pituitary adenoma. Diagnosis of active acromegaly was based on the clinical picture, demonstration of elevated levels of GH and insulin-like growth factor 1 (IGF-1), and a pituitary tumor visualized on magnetic resonance imaging (MRI). All surgical procedures were performed by 2 dedicated neurosurgeons using a standard microsurgical technique.

Four hundred and forty patients (49.5%) never received SRL treatment before surgery and are excluded from the present study. From the remaining 449 cases, we excluded 40 patients (8.9%) who had already received pituitary surgery at another center. The decision to treat or not to treat the patients with SRLs, the choice of the drug and its dose, as well as the referral of the patient to us for the surgical treatment, was usually taken by the caring endocrinologist.

In keeping with our previous work on this topic (14, 15), from the 409 patients, we excluded 33 cases (8.1%) who received SRL therapy for less than 3 months and 60 patients (14.7%) in whom SRL treatment was stopped for more than 10 days before surgery in the case of octreotide subcutaneously (n = 27), more than 2 months in the case of lanreotide (n = 4), or more than 3 months in the case of octreotide long acting release (LAR) (n = 20), or lanreotide autogel (n = 9). Moreover, to avoid the possibility of a carryover effect of SRL pretreatment on postsurgical results, we excluded 45 additional patients (11.0%) for whom we did not have information on the postsurgical GH status at 6 months or later. After the exclusion of the patients for the aforementioned reasons, 272 eligible patients (66.5%) were included in the analysis.

Standard informed consent for the collection of data and surgical procedures was obtained from all the patients.

Data collection, subgroup classification, and hormonal criteria

Serum GH and IGF-1 levels at diagnosis and before any therapeutic intervention were the baseline hormonal values. Response to SRL therapy was considered as the nadir hormone levels after at least 3 months from initiation of the highest dose of SRL used by the patient. Patients were considered as fully responsive to SRL if they had normalization of IGF-1 levels while on medical treatment. Patients who did not normalize IGF-1 levels but showed a reduction of GH levels of 50% or more compared with the baseline value were categorized as partial responders. All patients had pituitary MRI performed at diagnosis, but not all patients had MRI repeated while receiving SRL therapy. All MRIs have been reviewed by 1 of the 2 senior authors (M.L. and P.M.) and tumor diameters, as well as signs of tumor invasiveness into the cavernous sinus, have been recorded. When available, the MRI performed during SRL treatment was compared with the diagnostic MRI by one of us (M.L.) to verify the change, if any, induced by medical therapy. Tumor volume was approximately calculated using the following formula: 0.5 × anteroposterior dimension × vertical dimension × horizontal dimension. The reduction of tumor volume of at least 20% during SRL treatment compared with the baseline imaging was considered indicative of tumor shrinkage.

Surgical remission required demonstration of a normal age- and sex-adjusted IGF-1 concentration together with a GH level lower than 1 µg/L in a random sample or after a standard 75-g oral glucose load (2). Until 1994, GH levels were measured by an assay not specific for the 22K moiety of GH. Accordingly, the cut-off for a normal GH level from 1990 through 1994 was set at 2 µg/L. To avoid a possible carryover effect of SRL pretreatment on surgical results, the classification of surgical results was based on the biochemical results obtained at least 6 months after surgery.

Hormone assays

For patients in whom acromegaly was diagnosed and followed up in our center, GH levels before 1994 were measured by a commercially available radioimmunoassay not specific for the 22K GH moiety. From 1994 through 2005, GH levels were measured by a 2-site immunofluorimetric assay (AIA-PACK HGH, Tosoh, Tokyo, Japan) specific for the 22K GH moiety and from 2005 onward by an automated immunometric assay (Immulite 2000, Siemens Healthcare, Diagnostics Products Ltd, Llanberis, Gwynedd, UK).

Serum IGF-1 concentrations were measured until 2005 by a specific commercially available radioimmunoassay (Bioclone, Marrickville, Australia) after acid extraction. Afterward, IGF-1 concentrations were measured by an automated immunometric assay (Immulite 2000, Siemens Healthcare, Diagnostics Products Ltd).

Most patients either at diagnosis or during follow-up had their hormonal measurement performed in several laboratories that use different assay kits. To obviate the variability of measurements, we converted IGF-1 levels into the multiple of the upper normal limit (mULN) specific for that assay.

Statistical analysis

The primary endpoint of our study was to investigate whether the response to SRL therapy was independently associated with surgical outcome. One secondary analysis was performed to include also the patients who had stopped SRL therapy before surgery outside the abovementioned exclusion criteria.

Continuous data were examined for homogeneity of variance with the Kolmogorov–Smirnov test. Because the distribution of baseline GH levels at diagnosis were markedly skewed, we performed logarithmic transformation of the data before statistical analysis. Continuous data are expressed as mean ± standard error of the mean or median values and interquartile range (IQR). Statistics of means were carried out using parametric or nonparametric tests, as appropriate. Categorical variables are reported as number (percentage) and compared with the Pearson’s chi-squared test. Multiple logistic regression analysis was used to determine which variables independently predicted surgical remission of acromegaly. The final model included all the variables with a P ≤ .1 value in univariate analysis. P < .05 was considered to be statistically significant and all reported analyses are 2 tailed. All calculations were performed using IBM SPSS Statistics, version 23.0 (IBM Corp., Armonk, NY, USA).

Results

Type and duration of SRL treatment

The type of SRL used to treat the 272 patients with acromegaly before surgery varied according to the different periods of the study. Forty-two patients (15.4%) received the subcutaneous octreotide formulation, 146 (53.7%) received octreotide LAR, and the remaining 84 patients (30.9%) received lanreotide. The median duration of SRL treatment before surgery was 11 months (IQR 6-20 months, range 3-156 months) and was similar in the 3 different types of SRL (P = .58; Table 1). The estimated median total dose of SRL received by the patients before surgery was 112 mg (IQR 54-218 mg, range 27-1710 mg) for subcutaneous octreotide, 315 mg (IQR 160-605 mg, range 60-4680 mg) for octreotide LAR, and 915 mg (IQR 360-1440 mg, range 120-7200 mg) for lanreotide. The median interval between the last administration of SRL and surgery was 2 days (IQR 1-6 days, range 1-10 days) for subcutaneous octreotide, 20 days (IQR 15-30 days, range 1-90 days) for octreotide LAR, and 20 days (IQR 13-30 days, range 5-90 days) for lanreotide. The other clinical characteristics of the patients are summarized in Table 1. Despite the lack of randomization, the clinical characteristics of the patients were not significantly different among the three different SRL formulations.

Clinical response to SRL treatment

Treatment with SRL reduced median basal GH level from 16 µg/L (IQR 8-32 µg/L, range 2-346 µg/L) to 4 µg/L (IQR 2-12 µg/L, range 0-101 µg/L; P < .001). During SRL treatment, GH levels decreased to a median nadir of 25% of the baseline (IQR 11-57%, range 2-100%). The median GH nadir was 51% (IQR 24-85%, range 2-100%), 19% (IQR 9-44%, range 2-100%), and 33% (IQR 15-64%, range 2-100%) of baseline in patients treated with subcutaneous octreotide, octreotide LAR, and lanreotide, respectively (P < .001). Pairwise comparison showed that the inhibition of GH secretion was better in patients treated with octreotide LAR than in patients treated with lanreotide (P = .009) or subcutaneous octreotide (P < .001), while it was not significant (P = .27) between subcutaneous octreotide and lanreotide. The overall rate of normalization of IGF-1 levels during medical treatment occurred in 22.8% of the patients and was achieved in 23.8%, 24.0%, and 20.2% of patients treated with subcutaneous octreotide, octreotide LAR, or lanreotide, respectively (P = .80). The reduction of GH secretion to more than 50% of baseline levels occurred in 48.7%, 80.1%, and 61.0% of patients receiving subcutaneous octreotide, octreotide LAR, and lanreotide, respectively (P < .001). Pairwise comparison among the 3 groups showed that patients receiving octreotide LAR had a significantly higher probability to reach GH inhibition more than 50% of baseline compared with patients receiving subcutaneous octreotide (P < .001) or lanreotide (P < .006), while the difference between subcutaneous octreotide and lanreotide was not significant (P = .24).

Paired MRI examinations at baseline and during treatment with SRL were available for review in 148 patients (54.4%). A reduction of tumor volume of at least 20% or greater occurred in 78 patients (52.7%) and was observed in 32.0%, 59.0%, and 52.5% of patients treated with subcutaneous octreotide, octreotide LAR, and lanreotide, respectively. This difference was of borderline significance (P = .06). The reduction of tumor size was independent of IGF-1 normalization (57.1% in patients with IGF-1 normalization vs 51.3% in those without normalization; P = .55).

Table 2 summarizes the baseline clinical characteristics of the patients irrespectively of the type of SRL received and according to IGF-1 status during SRL treatment. Patients who had normalization of IGF-1 levels during SRL treatment had a significantly lower baseline GH level (P < .001) and a smaller maximum tumor diameter (P < .001) than patients who did not normalize IGF1 levels. The difference in tumor characteristics at baseline is also confirmed by the higher frequency of microadenomas (P ≤ .001) and the lower frequency of invasion of the cavernous sinus (P = .002) in SRL-responsive patients than in resistant patients. In contrast, age at diagnosis, estimated duration of disease, gender, and invasion of the sphenoid sinus were not significantly different between the 2 groups of patients. Interestingly, basal IGF-1 levels, calculated as mULN, were not different between responsive and not responsive patients, even though this analysis was weakened by the high number of missing data (88 patients) due to the impossibility to retrieve the normal range for the IGF-1 level at diagnosis.

Surgical outcome and response to SRL treatment

Surgical remission of acromegaly, defined according to the criteria outlined in the Methods section, was achieved in 162 of the 272 patients (59.6%). To verify whether the rate of surgical remission was changing over time, we compared the results in patients operated in the period 1990-2005 (n = 132) with those in patients operated from 2006 onward (n = 140). The remission rates were 56.1% and 62.9%, respectively (P = .25). The main characteristics of the study groups according to surgical outcome are reported in Table 3. As expected, several variables were significantly associated with a poor surgical outcome in univariate analysis. Thus, patients with an unsuccessful outcome were younger (P = .03), had a higher basal GH level at diagnosis (P < .001), a lower frequency of microadenomas (P = .004), a larger maximum tumor diameter (P < .001), and a tumor that invaded more frequently the cavernous sinus (P < .001) than patients who achieved surgical remission. Moreover, patients who normalized IGF-1 levels during SRL treatment had a better surgical outcome (83.9%) than patients who did not normalize IGF-1 levels (52.4%; P < .001). On the contrary, inhibition of GH levels of 50% or more during SRL treatment and reduction of tumor size induced by SRL were not significantly associated with a favorable surgical outcome.

Table 4 reports the odds ratio for the association between the clinical characteristics of the patients and an unsuccessful surgical outcome in the univariate and multivariate model. The final multivariate model included 258 of the 272 patients (94.9%; 151 surgically cured and 107 not cured) because of missing values. The variables that were independently predictive of an unsuccessful surgical outcome were invasion of the cavernous sinus (odds ratio, 10.71; 95% CI 4.39-26.12; P < .001), maximum tumor diameter (odds ratio, 1.08 per mm increase; 95% CI 1.02-1.15; P = .008), and lack of remission during SRL therapy (odds ratio, 3.29; 95% CI 1.28-8.44; P = .013). Baseline GH level at diagnosis was of borderline significance (P = .075), whereas classification of adenoma size into micro- or macroadenoma and age at surgery had no independent effect on surgical outcome.

We performed a secondary analysis that did not exclude the patients who stopped SRL treatment outside the time constraints described in “Materials and Methods.” This analysis included 304 patients and replicated the same results of the primary analysis. In particular, the odds ratio for lack of remission during SRL therapy and unfavorable surgical remission remained significant (odds ratio, 3.3; 95% CI 1.4-7.8; P = .008).

Discussion

The main goal of our study was to investigate whether the response to preoperative treatment with SRL is associated with the surgical outcome. All published studies on presurgical treatment with SRL in acromegaly compared pretreated vs naïve patients. Therefore, an analysis including only SRL-pretreated patients has never been presented.

An unequivocal definition to assess the response to SRL therapy is of paramount importance. GH and IGF-1 levels are known to be the best indicator of disease activity, with IGF-1 being the best reflection of disease control (17, 18). The response to medical treatment is usually classified according to the age- and sex-adjusted IGF-1 levels during an adequate period of SRL treatment. In our study, normalization of IGF-1 levels during medical treatment was achieved in 62 patients (22.8%) independently of the type of SRL used. The similar efficacy of the different SRLs preparation has long been recognized (19-21). A previous meta-analysis reported a higher overall rate of IGF-1 normalization (approximately 55%) but with a great heterogeneity among the different studies (21). The relatively low percentage of patients who normalized IGF-1 levels in our study might be an underestimation because of selection bias, that is, patients who are in remission during SRL therapy may not be referred to surgery. This possibility cannot be excluded because we do not know whether and how many patients who had normalization of IGF-1 levels during SRL therapy were not referred to our department during the same period of the study. However, a prospective, randomized trial comparing the second-generation SRL pasireotide with octreotide LAR reported that the normalization of IGF-1 levels occurred in 21.2% of 182 consecutive and unselected acromegalic patients in the Octreotide LAR group (22). Our results confirm that several clinical variables may predict a good response to first-line SRL treatment (Table 2). Older age and female sex were of borderline importance, at variance with other studies (17), while lower GH levels at diagnosis, smaller tumor size, and no tumor extension into the cavernous sinus were strongly associated with a better hormonal response to SRL treatment. Taken together, these results suggest that the mechanism underlying a favorable response to medical treatment is in some way dependent on the biological behavior of the tumor. Interestingly, a similar relationship exists between prolactinomas and the therapeutic response to dopaminergic drugs (23).

Another very important component of the efficacy of SRL therapy is its effect on tumor shrinkage. A reduction of tumor diameter greater than 20% occurs in approximately 50% of surgically naïve patients within the first 4 to 6 months of treatment with SRLs (7, 8, 18, 19). We detected a similar percentage of patients (52%) with significant tumor shrinkage, again suggesting that our study population does not seem to have a particular selection bias. The analysis of tumor shrinkage in our study was partially weakened by the high number of missing paired MRI examinations we could study.

The overall surgical remission rate in this group of SRL-pretreated patients was 59.6%, which is in keeping with the results of large surgical series (4, 5, 24, 25). We used the criteria proposed by the Endocrine Society Guidelines (2), but a strict requirement in our study was to include only patients with hormone evaluation at least 6 months after surgery, thus minimizing any possible carryover effect of presurgical SRL treatment on GH secretion, especially in patients who had a very good response to medical treatment.

It is known that different variables can affect the surgical outcome in acromegaly. In keeping with previous data (4, 26), our multivariate analysis confirms that greater tumor size and tumor invasiveness of the cavernous sinus were the principal variables associated with an unfavorable surgical outcome. The main and novel finding of our study is that the response to SRL treatment is also significantly associated with the surgical outcome. At first sight, this result can be viewed as obvious because the phenotype of the patients that are not cured by surgery shares the same characteristics that are associated with failure of SRL therapy to normalize IGF-1 levels, namely invasion of the cavernous sinus, a large tumor size, and a high baseline GH levels. However, the multivariate analysis confirms the independent value of this association. The odds ratio of 3.29 suggests that the lack of IGF-1 normalization during SRL therapy has a truly significant association with surgical outcome, albeit less marked than that associated with invasion of the cavernous sinus (odds ratio 10.71). Moreover, the favorable association between the IGF-1 normalization under SRL therapy and the surgical outcome is not affected by the reduction of tumor size, which occurred in similar proportions in responsive and not responsive patients. One likely explanation of our findings is that normalization of IGF-1 during SRL treatment identifies a subgroup of GH-secreting adenomas with favorable biological characteristics beyond those already considered. To support this hypothesis, we performed a secondary analysis by including also the patients who had undergone adequate SRL therapy but had stopped it well before surgery. The results of this analysis confirmed the results of the primary analysis.

Could this association be mediated by other characteristics not included in our analysis? This is certainly possible. Previous research has demonstrated that the granulation pattern of the adenoma (densely granulated vs sparsely granulated), cellular expression of somatostatin receptors (in particular, subtype 2), and proliferation index of the tumor may affect the response to SRL treatment (26-30). However, there is less evidence that the same variables may also predict surgical outcome. The pattern of tumor granularity and the proliferation index were not associated with surgical outcome in two small studies (31, 32), while Kiseljak-Vassiliades and coworkers (30) found that patients with densely granulated tumors had higher rates of surgical remission than patients with sparsely granulated adenomas (65.7% vs 14.3%, P < .0001). The granular pattern of the GH-secreting adenoma has also been correlated with the MRI signal on T2-weighted images and the response to SRL therapy (33). However, the pattern of T2 signal intensity on MRI did not correlate with the surgical outcome (32).

Our study has unavoidable limitations. First, the study is observational and retrospective, spanning almost 30 years of neurosurgical activity in our center. The resulting heterogeneity of clinical and hormonal data is counteracted by the large sample size of SRL-pretreated patients, which allowed to perform a meaningful statistical analysis. Nevertheless, surgical technique has not changed and no difference in surgical results according to the year of surgery was observed in the present study, which is in keeping with the overall results previously published from our center (6, 34). The criteria of cure for patients with acromegaly have changed over the years, reflecting the evolution of the methodology to measure GH and IGF-1 serum levels. To avoid misclassification of surgical results, we considered the criteria for the methodology in use at the time of the evaluation. Overestimation of the surgical rate of cure is unlikely, since the recurrence rate of GH-secreting adenomas is, in our experience, even lower than that occurring in the other types of pituitary adenoma (6). The possible bias arising from the selective referral to surgery of patients who did not normalized IGF-1 levels during SRL therapy may well affect our overall estimate of the success rate of medical treatment but should not modify our comparison between IGF-I–normalized and not normalized patients. The different SRLs used during the time frame of this study might also introduce a potential bias. However, IGF-1 normalization occurred in the same proportion of patients independently of the drug that was used. No patient in the present series received Pasireotide as a first-line medical treatment. Hence, our results apply to first-generation SRLs only. Second, data on tumor shrinkage could not be collected in all patients and we could not calculate the exact volumetric changes by the appropriate technology, which was unavailable to us until the last few years. Therefore, we cannot completely exclude that tumor reduction may affect the probability of surgical remission when imputed as a continuous rather than a categorical variable. Nevertheless, univariate analysis showed no difference in the surgical remission rate among patients with tumor shrinkage as compared with patients without tumor shrinkage. Unfortunately, we did not prospectively collect the actual dimensions of the tumor at surgery and, hence, our analysis does not include this information. Again, the similar occurrence of tumor reduction in patients who had IGF-1 normalization versus those without IGF-1 normalization and also in patients cured by surgery versus those not cured may lessen the probability that the degree of tumor shrinkage is an important characteristic. Third, we do not have information about particular histological characteristics of the tumor and, hence, we cannot exclude that our findings are entirely or partially biased by an unmeasured association between some histological characteristics and responsiveness to medical therapy. Nevertheless, information on histological characteristics can be obtained only after surgery and, hence, cannot be used for the definition of a preoperative therapeutic strategy. On the contrary, the response to SRL can provide predictive information on the patient’s outcome before surgery and may aid in the planning of a therapeutic strategy.

In conclusion, our study demonstrates for the first time that normalization of IGF-1 levels during treatment with SRLs is an independent predictive factor of a favorable surgical outcome. The underlying mechanisms remain unclear, but an optimal response to medical therapy may be a characteristic of less aggressive tumors that are more likely to be entirely removed by surgery.

Abbreviations

Abbreviations
 
• GH

  growth hormone

 
• IGF

  insulin-like growth factor

 
• IQR

  interquartile range

 
• LAR

  long acting release

 
• MRI

  magnetic resonance imaging

 
• mULN

  multiple of the upper normal limit

 
• SRL

  somatostatin receptor ligand

 
• TSS

  transsphenoidal surgery

Additional Information

Disclosure Summary: The authors declare that they have no conflict of interest.

Data Availability: The datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.

References