MRI measures of hypothalamic injury are associated with GLP-1 receptor agonist treatment response in people with hypothalamic obesity

Abstract

Aims

To evaluate whether neuroimaging-delineated regions of hypothalamic injury are associated with a differential treatment response to glucagon-like peptide receptor agonist (GLP-1RA) in patients with hypothalamic obesity (HO).

Materials and Methods

We performed a prespecified secondary analysis of a randomized, multicenter, double-blind, placebo-controlled trial of 10–25-year-old people with hypothalamic injury and HO randomized to GLP-1RA exenatide (ExQW) or placebo for 36-weeks. Subjects underwent MRI prior to enrollment and degree of hypothalamic damage was assessed using an integrative hypothalamic lesion score (HLS). Mammillary body (MB) damage was specifically determined. The main clinical endpoints were % change in BMI and change in % body fat. Nested ANCOVA models including a treatment x imaging measure interaction were compared using partial F-tests to assess whether the effect of ExQW treatment differed by severity of hypothalamic damage.

Results

Complete data were available in 35/42 randomized participants (placebo n=15; ExQW n=20). ExQW-treated patients with worse HLS or bilateral MB damage had greater reductions in % body fat at 36-weeks (interaction coefficient estimates for HLS: –0.9%, 95% CI –1.6% to 0.2%, P=0.02; for MB damage: –7.4%, 95% CI –10.1% to –4.7%, P<0.001, respectively) but not for BMI % change. Similarly, patients with more damaged and smaller MB cross sectional areas had greater reductions in % body fat following ExQW (interaction coefficient estimate 0.3%, 95% CI 0.2% to 0.4%, P<0.001).

Conclusions

In people with HO, greater hypothalamic damage as determined by MRI, in particular MB injury, is associated with greater reductions in adiposity following GLP-1RA treatment.

INTRODUCTION

Craniopharyngioma is the most common pediatric suprasellar tumor of nonglial origin but also occurs in adults.¹ Although histologically benign, damage to the hypothalamus from tumor mass effect or treatment complications often causes morbidity, including hypothalamic obesity (HO),² which occurs in half of craniopharyngioma survivors.³ HO can also occur following hypothalamic damage due to numerous other causes. Structural damage to the hypothalamus can disrupt neuronal pathways critical for regulating feeding behavior and energy homeostasis leading to hyperphagia, rapid weight gain, central insulin and leptin resistance, decreased energy expenditure and increased energy storage in adipose tissue.⁴ As a result, safe and effective pharmacological treatment for HO is very challenging.^5–11

We recently reported promising results to stabilize or reduce adiposity in HO following 36-week treatment with exenatide once-weekly (ExQW), a glucagon-like peptide-1 receptor agonist (GLP-1RA).¹² GLP-1 is an incretin that functions as a satiety hormone. GLP-1 binds to receptors in the vagus nerve, hindbrain (solitary tract nucleus and area postrema), hypothalamus (arcuate and dorsomedial nuclei), hippocampus, and mesolimbic reward pathways.¹³ GLP1 modulates neuronal activity in appetite- and reward-related neuronal pathways^14,15 and promotes decreased food intake.

ExQW and other GLP-1RA agents are novel and encouraging therapies for HO because the mechanism of action includes targets outside of the damaged hypothalamus. In our phase 3, randomized, clinical trial evaluating the safety and efficacy of GLP-1RA treatment in HO,¹² half of the subjects receiving ExQW responded with reduction in adiposity. Similar heterogeneous treatment responses to GLP-1RA have been reported in other patient populations.¹⁶ Determining why certain subjects demonstrate a treatment response could lead to the identification of biomarkers to guide therapy and predict treatment response.

Several neuroimaging scoring systems have been described that rely on MRI features of hypothalamic injury as biomarkers to predict risk of developing complications including HO in patients diagnosed with craniopharyngioma with the hope of providing an opportunity for early treatment intervention.^17–21 Damage in the posterior hypothalamus and mammillary bodies (MB) is associated with increased risk for HO.^19,21 Similarly, such scoring systems might also be useful in predicting treatment response.

We therefore evaluated whether MRI measures of extent and severity of hypothalamic injury were associated with ExQW treatment response in HO.

MATERIALS AND METHODS

Trial Design and Participants

We performed a prespecified secondary analysis of the Energy Balance & Weight Loss in Craniopharyngioma-related or Other Hypothalamic Tumors in Hypothalamic Obesity (ECHO) trial, an investigator-initiated, multi-site, 36-week double-blind, placebo-controlled randomized trial evaluating the efficacy and safety of ExQW for BMI reduction in adolescents and young adults with HO.¹²

Participants 10–25y were recruited at three clinical sites (Children’s Minnesota, St. Paul, MN; Seattle Children’s Hospital, Seattle, WA; and Vanderbilt University Medical Center, Nashville, TN). A detailed description of the clinical trial was previously published.¹² Briefly, all participants had a clinical diagnosis of HO due to a suprasellar tumor and had completed surgery or radiation treatment at least 6mo before enrollment. Participants were randomized 1:1 to placebo or once-weekly subcutaneous injections of exenatide 2mg (ExQW) for 36-weeks.

The clinical trial was approved by the Institutional Review Board at each site and conducted in accordance with the Declaration of Helsinki and human use guidelines for Good Clinical Practice. The study was registered at ClinicalTrials.gov (NCT02664441). Parents of participating children provided consent and children provided assent.

Weight and height were assessed using calibrated instruments. Dual energy x-ray absorptiometry (DXA) was performed (GE Lunar Prodigy Advance; GE Lunar iDXA; and Hologic Discovery and Horizon) to estimate fat mass, fat free mass, and percentage body fat (BF%).

Neuroimaging and measures of hypothalamic damage

A brain MRI performed within 8mo before each participant’s enrollment was reviewed by a central study neuroradiologist to confirm evidence of hypothalamic injury. Patients without a recent brain MRI or with technically inadequate MRI studies underwent a research MRI (n = 4) including volumetric 3D T1 with 1 mm slice thickness. Hypothalamic damage was qualitatively assessed using the hypothalamic lesion score (HLS) as previously described.²¹ In brief, the scoring system ranges from 0 to 7, with higher scores indicating more severe injury, and assesses the presence of hypothalamic injury in the floor of third ventricle; anterior, medial, and posterior hypothalamus; and MB. Additional score components include the presence of lateral or third ventriculomegaly. Scoring was independently performed by a neuroradiologist and the pediatric endocrinologist who developed the scoring system. All neuroimaging analyses were conducted without knowledge of treatment assignment or response. Discordant scores were resolved by consensus review. Based on prior results,²¹ the primary prespecified imaging measures were total HLS, posterior hypothalamus injury subscore, and MB injury subscore.

As part of a confirmatory secondary investigation, MB sizes were measured using the highest spatial resolution images available for each patient which were typically volumetric, 1 mm thin, T1- weighted sequences. Using OsiriX MD imaging software (version 11.0.4; Pixmeo, Geneva, Switzerland) the maximum MB cross sectional areas for both MB in each patient were measured on coronal, axial, and sagittal images by the study neuroradiologist. Maximum bidimensional, perpendicular diameters of the MB on coronal images and maximum anterior-posterior diameter on axial images were also measured. The prespecified primary MB size measurement was the mean of the maximum cross- sectional areas determined on each imaging plane. Several simpler and easier to perform secondary measures of total MB sizes were also assessed as follows: maximum coronal cross sectional area, estimated ellipsoid volume using the three longest orthogonal diameters,²² and product of maximum coronal perpendicular, bidimensional diameters.

Clinical outcome measures

The prespecified main clinical endpoint was % change in BMI (ΔBMI%). Change in % body fat (ΔBF%) was prespecified as an additional primary clinical endpoint based on treatment-related differences identified in the ECHO trial.¹² Secondary clinical outcome measures included changes in derivative measures of adiposity (including changes in total body fat mass and fat mass index) and % lean body mass.

Statistical Analysis

Analysis was based on treatment assignment using all available data at baseline and 36 weeks without imputation and regardless of treatment compliance. The main statistical analysis of interest was whether the effect of ExQW treatment on ΔBMI% and ΔBF% was different depending on the severity of hypothalamic damage as assessed on MRI. This was determined using nested ANCOVA models with the change in the clinical outcome measure at 36 weeks as the dependent variable and the main predictor variable of treatment group. Model covariates included the clinical outcome variable at baseline, study site, age group, and sex. To test for a modification of treatment effect based on MRI measures of hypothalamic damage, a treatment group X imaging measure interaction term was added to the base model and improvement in the model fit was assessed with a partial F-test. Analysis was performed using R statistical programming language (version 4.0.2²³). Two-sided alternatives were assumed, with α=0.05. P-values are reported without adjustments for multiple comparisons.

RESULTS

Patients and baseline measures

Detailed reporting of patient dispositions and a CONSORT flow diagram for the ECHO trial have been previously reported.^12,12 In summary, 43 patients were screened, and patients were enrolled. All 23 patients randomized to ExQW, and 18 of 19 patients randomized to placebo, received at least one dose of study medication (Table 1). For the clinical endpoints at 36 weeks, ΔBMI% data were available for 36 (placebo n=15; ExQW n=21) and ΔBF% for 35 (placebo n=15; ExQW n = 20) patients.

Table 1.

Baseline patient and imaging characteristics

	Placebo (n =18)	ExQW (n =23)	P-value
Age, y	16.9 (4.7)	16.9 (4.3)	>0.99
Sex, n (%)
Male	7 (39%)	9 (39%)	0.99
Female	11 (61%)	14 (61%)
Reported GI-related adverse events	12 (67%)	15 (65%)	>0.99
Anthropometries and body composition
Height, cm	161.3 (13.3)	160.9 (15.6)	0.94
Weight, kg	103.7 (30.0)	95.1 (27.9)	0.35
Waist circumference, cm	119.7 (16.3)	112.8 (17.4)	0.20
Body mass index (BMI), kg/m2	39.2 (7.3)	35.2 (6.1)	0.07
DXA
% body fat	49.0% (4.5%)	47.4% (4.9%)	0.27
% lean body mass	48.8% (4.5%)	50.4% (4.9%)	0.31
Lean body mass index (LBMI)	19.1 (4.2)	17.7 (3.0)	0.23
Fat body mass index (FMI)	19.3 (4.4)	16.9 (3.8)	0.07
Neuroimaging
Pre-enrollment MRI (days performed from enrollment)	75.7 (70.8)	91.3 (72.7)	0.49
MRI slice thickness, mm	1.4 (1.0)	1.6 (1.2)	0.56
Hypothalamus lesion score total	5.3 (1.6)	4.8 (1.7)	0.30
Floor of third ventricle, n (%)
Intact	3 (17%)	3 (13%)	>0.99
Damaged	15 (87%)	20 (83%)
Anterior hypothalamus, n (%)
Intact	0 (0%)	0 (0%)	0.62
Unilateral damage	1 (6%)	3 (13%)
Bilateral damage	17 (94%)	20 (87%)
Medial hypothalamus, n (%)
Intact	1 (6%)	1 (4%)	0.15
Unilateral damage	0 (0%)	4 (17%)
Bilateral damage	17 (94%)	18 (78%)
Posterior hypothalamus, n (%)
Intact	5 (28%)	7 (30%)	>0.99
Unilateral damage	1 (6%)	2 (9%)
Bilateral damage	12 (67%)	14 (61%)
Mammillary bodies, n (%)
Intact	4 (22%)	5 (22%)	0.92
Unilateral damage	9 (50%)	10 (43%)
Bilateral damage	5 (28%)	8 (35%)
Third ventriculomegaly, n (%)
Absent	2 (11%)	9 (39%)	0.08
Present	16 (89%)	14 (61%)
Lateral ventriculomegaly, n (%)
Absent	8 (44%)	15 (65%)	0.37
Unilateral	4 (22%)	4 (17%)
Bilateral	6 (33%)	4 (17%)
Mammillary body size measures
Mean of max areas (mm2)	25.4 (7.7)	20.8 (7.8)	0.07
Coronal max area (mm2)	23.8 (8.1)	19.8 (7.0)	0.10
Estimated ellipsoid volume (mm3)	79.2 (38.7)	58.0 (31.6)	0.07
Product of max coronal bidimensional diameters (mm2)	30.3 (10.8)	24.8 (9.1)	0.09

Abbreviations: DXA, dual energy x-ray absorptiometry; ExQW, exenatide once weekly.

Data are expressed as Means (SD) or count (% of group total). Statistical analyses were performed using Student’s t-test or Fisher’s exact test. GI-related adverse events include abdominal pain, nausea, vomiting, diarrhea, or constipation.

There were no differences in baseline clinical measures, including reported GI-related adverse events, or MRI hypothalamic damage measures between placebo and ExQW-treated groups (Table 1). There was substantial inter-rater reliability for the primary imaging measure of total HLS (weighted kappa coefficient = 0.80). There were no associations between baseline BMI and MRI hypothalamic damage scores or MB sizes; for example, total HLS effect estimate for baseline BMI was 1.0 (95% CI –0.1 to 2.2, P=0.08). However, there was a statistically significant estimated 0.9% greater baseline BF% for each 1-point increase in total HLS (95% CI 0.2% to 1.6%, P=0.02). Moreover, patients with bilateral MB damage had a 1.1% greater adjusted mean baseline BF% (95% CI –7.1% to –0.7%, P=0.02) compared to those with no damage, and an estimated 0.2% greater BF% for each mm² decrease in mean of the maximum MB areas (95% CI 0.3 to 0.0, P=0.04). Patients with bilateral MB damage had a greater likelihood of also having bilateral posterior hypothalamic damage compared to those with intact MB (relative risk 6.9, 95% CI 1.1 to 43.4, P<0.001).

Effect modification of qualitative imaging measures

Partial F-tests were used to compare the relative goodness-of-fit for nested ANCOVA models without and with an imaging measure x treatment interaction term (Table 2). The interaction coefficients were calculated to estimate differences in treatment effects.

Table 2.

Treatment effect modification by imaging measures of hypothalamic damage

	% Change inBMI				Change in %body fat
Imaging measure	n	Treatment Adjusted R2	Imaging measure × Treatment Adjusted R2	P-value	n	Treatment Adjusted R2	Imaging measure × Treatment Adjusted R2	P-value
Hypothalamic lesion score
Total	36	0.18	0.12	0.93	35	0.09	0.33	0.006
Floor of third ventricle	36	0.18	0.17	0.46	35	0.09	0.09	0.34
Anterior hypothalamus	36	0.18	0.15	0.59	35	0.09	0.05	0.60
Medial hypothalamus	36	0.18	0.12	0.82	35	0.09	0.20	0.92
Posterior hypothalamus	36	0.18	0.15	0.58	35	0.09	0.26	0.06
Mammillary bodies	36	0.18	0.13	0.67	35	0.09	0.65	<0.001
Third ventriculomegaly	36	0.18	0.13	0.88	35	0.09	0.21	0.06
Lateral ventriculomegaly	36	0.18	0.20	0.36	35	0.09	0.00	0.82
Mammillary body size measures
Mean of max areas	35	0.17	0.16	0.45	34	0.08	0.48	<0.001
Coronal max area	35	0.17	0.24	0.12	34	0.08	0.48	<0.001
Estimated ellipsoid volume	34	0.17	0.28	0.07	33	0.08	0.43	0.001
Product of max coronal bidimensional diameters	35	0.17	0.28	0.06	34	0.08	0.48	<0.001

Comparison of nested ANCOVA models for the primary dependent clinical endpoints without and with an imaging measure × treatment interaction term. Adjusted R² values indicate goodness-of-fit for the models which were compared using partial F-tests (P-values). Statistically significant interaction coefficient estimates are reported in the Results text.

For ΔBMI% at 36 weeks, the imaging measure x treatment interaction terms did not significantly improve model fits for any imaging measure (Table 2 and Figure 1A–C).

Figure 1.

Plots of BMI % change at 36 weeks and imaging measures of hypothalamic damage by treatment group (circles = placebo; triangles = exenatide once-weekly, ExQW). A and D, scatterplots of total hypothalamic lesion score (HLS) or mammillary body (MB) mean of max areas with linear regression fits (dashed line = placebo; solid line = ExQW) and associated 95% CI bands (light gray = placebo; dark gray = ExQW). There were no statistically significant interactions between treatment and total HLS or MB mean area. B and C, box and whisker plots of hypothalamus or MB damage subscores by treatment group (horizontal line = median; box = interquartile range; whiskers = min and max). There were no statistically significant interactions between treatment and degree of posterior hypothalamus damage or MB damage.

However, for ΔBF% at 36 weeks there was evidence for a differential response to treatment based on severity of total HLS, with an estimated adjusted mean 0.9% decrease in %BF for each 1 point increase in lesion score among ExQW-treated patients (95% CI 0.2% to 1.6% decrease, P=0.02; Table 2 and Figure 2A) taking into account baseline BF%, study site, age group, and sex. Among the subscore imaging measures, there was no statistically significant interaction for posterior hypothalamus damage x treatment (P=0.06, Figure 2B), but there was for MB damage (interaction effect estimate for unilateral damage –2.8%, –5.3% to –0.3%, P=0.028; bilateral damage –7.4%, 95% CI –10.1% to – 4.7%, P<0.001; Table 2, Figure 2C, and Figure 3). The adjusted mean ΔBF% at 36 weeks for ExQW-treated patients with intact, unilateral damaged, and bilateral damaged MB was 2.9% (95% CI 1.4% to 4.4%), 0.5% (95% CI – 0.4% to 1.3%), and –2.1% (–3.1% to –1.2%), respectively. Patients with worse MB damage scores had improved ΔBF% at 36 weeks following ExQW treatment. The adjusted mean difference in ΔBF% for ExQW-treated patients with unilateral MB damage compared to intact was –2.5% (95% CI, –4.1% to – 0.8%, P=0.006) and for bilateral MB damage compared to intact was –5.0% (95% CI, –6.7% to –3.3%, P<0.001).

Figure 2.

Plots of change in % body fat at 36 weeks and imaging measures of hypothalamic damage by treatment group (circles = placebo; triangles = exenatide once-weekly, ExQW). A and D, scatterplots of total hypothalamic lesion score (HLS) or mammillary body (MB) mean of max areas with linear regression fits (dashed line = placebo; solid line = ExQW) and associated 95% CI bands (light gray = placebo; dark gray = ExQW). A, There were statistically significant interactions for treatment x total MRI HLS (interaction coefficient estimate −0.9%, 95% CI −1.6% to −0.2%, P = 0.02) and, D, treatment x MB mean area (interaction coefficient estimate 0.3%, 95% CI 0.2 to 0.4, P < 0.001). ExQW-treated patients with worse total HLSs or smaller MB sizes were associated with improved % body fat reduction at 36 weeks. B and C, box and whisker plots of hypothalamus or MB damage subscores by treatment group (horizontal line = median; box = interquartile range; whiskers = min and max). There was a statistically significant interaction between treatment and MB damage (P < 0.001) but not for posterior hypothalamus damage (P = 0.06). ExQW-treated patients with bilateral MB damage had statistically significant reduction in % body fat compared to those with intact or unilateral MB damage (* <0.05; ** < 0.01, *** < 0.001).

Figure 3.

Representative coronal (A and C) and axial (B and D) T1-weighted MRI images from a patient with intact mammillary bodies (arrows in A and B) and another patient with bilateral damaged mammillary bodies (arrows in C and D).

Among placebo-treated patients, the adjusted mean ΔBF% at 36 weeks with intact, unilateral damaged, and bilateral damaged MB was 0.7% (95% CI –0.9% to 2.3%), 1.1% (95% CI 0.2% to 1.2%), and 3.1% (1.6% to 4.6%), respectively. Placebo-treated patients with bilateral MB damage had an adjusted absolute 2.6% greater increase in ΔBF% compared to those with unilateral MB damage (95% CI 1.3% to 3.9%, P=0.02; Figure 2C). Patients with bilateral MB damage treated with ExQW had an adjusted absolute mean –5.2% difference in ΔBF% at 36 weeks compared to placebo treated (95% CI, –7.0% to – 3.5%, P<0.001; Figure 2C).

There were also statistically significant interactions between change in % lean body mass and total HLS and MB damage subscores with greater increases in % lean body mass among ExQW-treated patients with worse hypothalamic and MB damage (interaction effect estimate for total HLS 0.9%, 0.2% to 1.6%, P = 0.015; unilateral mamillary body damage 2.8%, 0.3% to 5.3%, P = 0.03; bilateral damage 7.3%, 95% CI 4.6% to 10.0%, P <0.001).

Effect modification of imaging measures of mammillary body size

A secondary investigation measuring MB sizes on MRI was performed to validate the main finding of decreased ΔBF% in ExQW-treated patients with qualitatively more severe MB damage. Mean of the maximum MB areas could be measured in 40/41 subjects; the MB could not be reliably identified in one placebo-treated patient with severe bilateral MB injury and suboptimal MRI spatial resolution. Among all patients, mean of the maximum MB areas was statistically significantly smaller in bilateral damaged (15.7 mm², 95% CI 12.4 to 19.1 mm², P<0.001) and unilateral damaged (24.5 mm², 95% CI 21.6 to 27.3 mm², P<0.001) groups compared to those with intact MB (30.1 mm², 95% CI 26.0 to 34.1 mm²). Patients with qualitative bilateral posterior hypothalamus damage had smaller MB compared to those with intact posterior hypothalamus (mean difference 6.3 mm², 95% CI 0.7 to 11.9 mm², P<0.001).

There was evidence of a differential response to treatment, based on ΔBF%, depending on MB sizes using the primary method of mean of the maximum areas (Table 2 and Figure 3D). Among ExQW- treated patients, each 1 mm² decrease in MB size was associated with an adjusted mean decrease in ΔBF% at 36 weeks of 0.3% (95% CI 0.2% to 0.4%, P < 0.001). Results were similar using simpler MB size measures including maximum coronal area, estimated ellipsoid volume, and product of maximum coronal bidimensional areas (Table 2).

ExQW responders versus non-responders

Exploratory analyses were performed to assess for differences in imaging measures of hypothalamic injury and MB sizes for ExQW-treated patients with the prespecified definitions of treatment response as decreased BMI or %BF at 36 weeks (Supplemental Table 1). There were no differences in imaging measures for ExQW-treated patients with decreased BMI. However, ExQW- treated patients with a decrease in %BF at 36-weeks were more likely to have bilateral damaged MB (86% increased likelihood of %BF reduction with bilateral MB damage compared to none, 95% CI 60% to 112%, P=0.01) and smaller MB sizes (mean difference in MB mean area 10.0 mm², 95% CI 5.4 to 16.7 mm², Supplemental Table 1). There were no differences in reported GI-related adverse events between ExQW responders and nonresponders (Supplemental Table 1).

DISCUSSION

We found that worse total HLS was associated with improved reductions in adiposity following 36-week treatment with ExQW. ExQW-treated subjects with qualitatively more severe MB damage subscores had greater reduction in %BF. We found no association between ΔBMI% and imaging measures of hypothalamic damage; as has been previously discussed,¹² BMI change may not be as reliable an outcome measure in growing children or the effect of GLP-1RA may be greater in altering body fat composition rather than body mass.

Although not the primary objective of this study, we validated the utility of the MRI-based HLS previously found to be associated with development of HO in a separate patient cohort following treatment for craniopharyngioma.²¹ At baseline prior to treatment, HO patients with worse total HLS had higher adiposity. We did not confirm a relationship with BMI which was previously reported;²¹ however, the patients in this study cohort all had a diagnosis of HO with very high BMI which, given the sampling bias, could limit the ability to detect a relationship.

We studied MB injury scores as surrogate indicators for damage to the posterior hypothalamic nuclei given their general proximity. Moreover, damage to the posterior hypothalamus and MB have both been associated with robust weight gain in HO, and MB can be recognized as landmarks on MRI most of the time even with large tumors.^19,21 Nevertheless, qualitative evaluation for the presence of MB damage relies on subjective evaluation of MB sizes, configurations, and MRI signal characteristics. For similar reasons, existing qualitative neuroimaging scoring systems of hypothalamic involvement for craniopharyngiomas can have poor reliability.²⁴ Therefore, we conducted a secondary confirmation study to quantitatively, and more objectively, assess MB sizes as a proxy for MB damage. As expected, patients with worse qualitative MB damage subscores had smaller MB sizes. In addition to measures of MB maximum cross-sectional areas, we evaluated simpler measures of MB sizes including the maximum bidimensional diameter of each MB on coronal images. These linear diameter measurements are practical because they can be performed easily and quickly on routine brain MRI studies without specialized MRI protocols or image post-processing. Although these MB measurements are likely to be more reliable than qualitative MB damage assessment, future work is needed to validate these findings and assess reliability.

Our results suggest that extent of hypothalamic damage on MRI is not only associated with risk of developing HO but also response to GLP-1RA therapy. Our hypothesis was that patients with more extensive hypothalamic injury would have a diminished response to GLP-1RA therapy due to disruption of critical neural circuitry regulating metabolism and feeding behavior. On the other hand, the rationale for GLP-1RA therapy in HO is that the effect of these agents may not solely depend on intact hypothalamic structures. GLP-1 receptors are present in the vagus nerve and widely expressed throughout the brain including appetite-related sites in the hindbrain (nucleus of the solitary tract, area postrema) and the hypothalamus (arcuate and dorsomedial nuclei).²⁵ By binding to these receptors, GLP-1 modulates activity in appetite- and reward-related brain areas,^14,15 and functions as a satiety hormone, promoting reduced food intake and meal termination.^26–28 Surprisingly, patients with more extensive hypothalamic and MB damage not only responded but had the greatest reductions in %BF in response to ExQW. One possible explanation for this finding is that destruction of endogenous ligand sites of action in the hypothalamus heightens responsiveness and sensitivity of extra-hypothalamic sites of action to exogenous ligands. It furthermore suggests that disruption of hypothalamic pathways involved in appetite and energy homeostasis may result in alterations in other pathways such as GLP-1- mediated signaling in the brainstem, which remain intact in HO patients.

The association of MB damage with greater reduction in %BF among subjects treated with ExQW does not necessarily directly implicate the role of MB in the mechanism for this response. MB damage may be a proxy for other injury sites responsible for ExQW-treatment response. Patients with bilateral MB injury or small MB were more likely to also have posterior hypothalamus injury. In particular, the supramammillary nucleus, located between the lateral hypothalamus and ventral tegmental area, contains GLP-1 receptors and in a rodent model may contribute to the observed reduction of food intake and food-motivated behavior following GLP-1RA microinjection.^29,30 Alternatively, MB damage and small MB could reflect secondary injury to other remote brain structures responsible for GLP-1RA response.

Finally, although there is currently little evidence to suggest a role of MB in control of feeding behavior or energy homeostasis,³¹ our findings may indicate a role in modulating the neuronal circuitry underlying GLP-1RA and body fat regulation. The MB are critically involved in learning and memory and thought to serve as a hippocampal relay or coordinating activity between anterior thalamic nuclei and cingulate cortex.^32–35 In Macaques, projections from the hippocampal formation to the posterior hypothalamus are confined to the medial and lateral MB.³⁶ There is growing evidence that the hippocampus contributes to control of feeding behavior,^13,37 perhaps modulated by MB inputs. GLP-1 receptors are expressed in hippocampal neurons³⁸ and GLP-1 has neuroprotective effects and modifies synaptic plasticity.³⁹ One possibility for increased sensitivity to GLP-1RA therapy in HO patients with MB damage could be altered regulation and increased sensitivity of hippocampal GLP-1 pathways due to loss of MB input. However, the potential role of the MB in the regulation of energy homeostasis and satiety remains poorly understood and future studies are necessary including in preclinical models.

Our study has several limitations. First, the study was a secondary analysis of a relatively small but effectively designed randomized, double-blinded, randomized controlled trial. Second, most of the MRI studies in this study were obtained for clinical purposes with varying techniques from multiple sites, and a standardized, prospective research MRI protocol to optimize imaging of the MB was not performed. Volumetric analysis of the MB typically requires specialized research MRI sequences;⁴⁰ therefore, we did not measure MB volume but relied on measures of cross-sectional area and maximum linear measures which are more practical for future clinical application and have been previously employed.²² Third, reliably identifying residual tissue from severely damaged MB can be challenging; indeed, MB size of one placebo-treated patient with severe bilateral MB damage could not be reliably measured. Future work is necessary to further validate quantitative MB size assessment; nevertheless, we found an association between GLP-1RA treatment response and the prespecified imaging measures of qualitative HLS, which has substantial inter-rater reliability, and MB damage.

In conclusion, MRI measures of hypothalamic damage are promising biomarkers to predict response to GLP-1RA therapy in HO. Patients with more severe hypothalamic damage, particularly damage to the MB, demonstrated improved response to GLP-1RA with greater decreases in adiposity. This suggests that GLP1RA treatment supplements a pathway that is deficient due to hypothalamic damage, while sensitivity to treatment is maintained. Although these novel findings need to be confirmed in a prospective trial, they should be considered when planning future clinical trials evaluating efficacy for HO treatments.