Moon-like Facies by Glucocorticoid Is Associated With the Development of Diabetes and Body Image Disturbance

Abstract Context Moon-like facies (MLF) are a typical side effect of glucocorticoid (GC) therapy; however, its predisposing factors, relationship with GC-induced complications, and effects on body image are not well understood. Objective This study aimed to determine the predisposing factors for MLF during GC therapy; its association with GC-induced diabetes, hypertension, and dyslipidemia; and its effects on body image. Methods This prospective observational study spanned 24 weeks and targeted patients who received GC therapy at the University of Yamanashi Hospital from June 2020 to August 2022. The MLF was defined based on the following 3 factors: (1) an increase in facial measurement lengths, (2) subjective facial changes by patients’ self-assessment using a visual analog scale; (3) objective and qualitative facial changes assessed by physicians. We examined the predisposing factors for MLF and the association of MLF with GC-induced diabetes, hypertension, dyslipidemia, and body image. Results The cumulative incidence rate of MLF at 24 weeks was 37.6%. Predisposing factors for MLF were an initial oral prednisolone dosage of ≥ 30 mg/day [odds ratio (OR) 63.91, 95% confidence interval (CI) 5.82-701.81] and female (OR 6.66, 95% CI 1.35-32.79). MLF showed a significant association with the onset of GC-induced diabetes (OR 6.58, 95% CI 1.25-34.74). MLF was also an independent factor contributing to body image disturbance (β = −18.94, P = .01). Conclusion MLF contributes to body image disturbance and is associated with the development of GC-induced diabetes; therefore, it is clinically important as a physical manifestation of GC therapy.

Moon-like facies (MLF) is a well-known side effect of glucocorticoid (GC) therapy, but very little is known about the predisposing factors for MLF and its impact on GC-induced diabetes, hypertension, and dyslipidemia.Fardet et al conducted a series of observational studies on corticosteroidinduced lipodystrophy (CIL), which manifested as MLF and/or a buffalo hump in French patients treated with GC [1][2][3][4][5].They found that CIL was more likely to occur in women and young adults and was associated with increased energy intake and subsequent weight gain [2].Additionally, CIL is reportedly associated with the onset of hyperglycemia, hypertension, and neuropsychiatric disorders [2][3][4][5].However, no studies have focused on MLF among the complications of GC.Therefore, the factors associated with MLF and whether the onset of MLF is associated with the risk of developing GC-induced diabetes, hypertension, and dyslipidemia remain unclear.
In general, disturbances in body image due to the side effects of treatment can deteriorate the quality of life of patients [6][7][8] and affect their medication compliance and treatment continuation rates [8][9][10].In an interview-based survey, Hale et al reported that MLF negatively affects the body image of patients with systemic lupus erythematosus and causes psychological distress, although no statistical evaluation has been conducted to date [11].
This study aimed to clarify the predisposing factors for MLF and the association of MLF with GC-induced diabetes, hypertension, dyslipidemia, and body image disturbance by performing facial measurements.

Participants
This study included 90 Japanese patients aged 16 years or older who were newly prescribed oral prednisolone at the Department of Rheumatology or Nephrology at the University of Yamanashi Hospital (Yamanashi, Japan), either during hospitalization or as outpatients, between June 2020 and August 2022.All included patients provided written informed consent.This study was approved by the Clinical Trial Review Board of the University of Yamanashi Hospital (Approval No. 2246).All the research methods and procedures were conducted in accordance with the ethical principles of the Declaration of Helsinki.
We excluded patients with a history of GC use within 1 year before the start of the study, including oral, intravenous, transdermal, intranasal, or intra-articular GC use.We also excluded patients who were diagnosed with generalized edema due to renal failure, heart failure, liver failure, hypoalbuminemia, or endocrine disorders by the attending physician during the observation period; those who died during the observation period; those who discontinued regular visits; those with a history of Cushing's syndrome or its comorbidities; and those who refused to participate in the study.The use of other immunosuppressive drugs was permitted.

Study Protocol
We conducted a prospective investigation of patients who were initiated on GC therapy.Facial measurements and body image evaluations were performed weekly during hospitalization and monthly during outpatient visits.Clinical data were obtained from the medical records, including age, sex, height, body weight, blood pressure, blood glucose levels, hemoglobin A1c (HbA1c) levels, cholesterol levels, oral prednisolone dose (the daily dose at the start and the cumulative dose) during the observation period, medical history, and concurrent illnesses.The observation period was 24 weeks.If the underlying disease improved within 24 weeks and prednisolone treatment was discontinued, the time to discontinuation was included in the analysis.

Criteria for MLF
The MLF was defined based on the following 3 factors: (1) an increase in facial measurement lengths, (2) subjective facial changes by patients' self-assessment using a visual analog scale (VAS), and (3) objective and qualitative facial changes assessed by physicians ("Yes" or "No").Fulfilling conditions 1 and 2; 1 and 3; or 1, 2, and 3 were defined as MLF.
The faces of the patients were measured at 5 locations by either the investigator (T.T.) or 2 specific nurses: the length between the ends of the eyebrows, the length from the nose to the upper end of the left and right auricles, and the length from under the nose to the angle of the mandible on both sides (Fig. 1).Before the start of the study, the examiners underwent multiple practice sessions to ensure consistency.To evaluate relative reliability for the facial measurement, we calculated the intraclass correlation coefficient (ICC) and its 95% confidence interval (CI) for the interrater analysis [12,13]: the intraclass correlation coefficient (95% CI) among the 3 examiners was .99 (.96-1.00),suggesting that the internal consistency of measurement between these 3 examiners is statistically maintained (data not shown).
Measurements were taken with the same ruler and recorded in increments of .5 cm.We considered the facial measurement as significantly increased if it was increased by ≥ .5 cm after GC treatment.We used a KA-15 measure (AS ONE Co. Ltd., Japan, product #25950137) as the instrument for facial measurements as majors, which is certified by the Japanese Industrial Standards.
A VAS was used for the self-assessment of MLF by patients.As an answer to the question "How much do you think your face has changed compared to before starting the steroid treatment?", the patients indicated their responses on the VAS consisting of a 100-mm linear scale ranging from "not changed" (0 mm) to "apparently changed" (100 mm).Fifty mm or more is defined as MLF.For the objective facial changes assessed by physicians, the physician assessed whether the patient developed MLF ("Yes" or "No") at the same time as medical examination.
The process for evaluating and determining MLF is as follows: 1 of the 3 examiners measured the patient's face; then the patient answered a VAS before medical examination by a physician.Next, the physician subjectively assessed whether the patient developed MLF and informed the judgment to the investigator.Finally, the investigator decided whether the patient developed MLF according to the criteria.Therefore, no authority could affect the evaluation process and final decision regarding MLF development.
Criteria for GC-induced Diabetes, Hypertension, and Dyslipidemia GC-induced diabetes, hypertension, and dyslipidemia were evaluated after the initiation of GC therapy.GC-induced diabetes was diagnosed if the HbA1c level was 6.5% or higher, if the fasting blood glucose level was 126 mg/dL or higher, if the random blood glucose level was 200 mg/dL or higher, if the fasting blood glucose level was confirmed to be 126 mg/dL or higher on 2 separate occasions, or if medications for the treatment of diabetes (oral hypoglycemic agents, insulin, and GLP-1 receptor agonists) were started.Hypertension was diagnosed if the blood pressure at the time of examination was 140/90 mmHg or higher on 2 separate occasions or if antihypertensive medications were initiated.Dyslipidemia was diagnosed if the fasting low-density lipoprotein cholesterol level was 140 mg/dL or higher, if the high-density lipoprotein cholesterol level was less than 40 mg/dL, if the triglyceride level was 150 mg/dL or higher, or if lipid-lowering medications were initiated.

Body Image Evaluation
To evaluate body image, the Body Image in Lupus Screen Scale (BILSv1.2),which has been shown to have positive correlations with LupusQoL, 36-item Short Form Survey, EQ5D, and BIOLI in patients with systemic lupus erythematosus [14], was used.The questionnaire was administered to the patients simultaneously with the facial measurements.

Statistical Analysis
Statistical analyses were performed using Stata (version 17.0; Stata Corp., College Station, TX, USA).The normality of the data was confirmed using the Shapiro-Wilk test, and data that followed a normal distribution were presented as mean ± SD, while data that did not follow a normal distribution were presented as median [25th percentile value-75th percentile value].Group comparisons were performed using the t-test, chi-squared test, or Wilcoxon rank-sum test.The cumulative incidence rates of MLF and GC-induced diabetes, hypertension, and dyslipidemia were evaluated using the Kaplan-Meier method.Factors associated with MLF were identified using multiple logistic regression analyses.Based on previous studies, the variable selection was performed using a backward selection procedure, and a P-value of less than .2was used to retain the variables in the model [2].Factors associated with GC-induced diabetes were also identified using multiple logistic regression analysis, and variables were selected based on the risk factors for GC-induced diabetes identified in previous studies [15][16][17][18][19]. Odds ratios (ORs) were used to evaluate the effects of the identified variables.Factors associated with scores of the BILSv1.2 were identified using linear regression analysis.Statistical significance was set at P < .05.

Patient Characteristics
Among the 90 enrolled patients, 1 died due to an underlying disease, 8 refused to continue participating in the study, and 9 were excluded from the analysis because their visits to the hospital were interrupted during the observation period.A total of 72 patients were analyzed, including 69 who completed the observation period of 24 weeks, 2 who discontinued prednisolone after 20 weeks, and 1 who discontinued prednisolone after 16 weeks due to improvement in the underlying disease.
Table 1 shows the characteristics of 72 patients.The mean initial oral prednisolone dosage was 31.8 ± 13.7 mg/day.More than half of the patients (48 of 72) received other immunosuppressive therapies.
In contrast, there was no significant relationship between criteria 1 and 2 (P = .372)(data not shown).
Table 2 shows the characteristics of the patients who developed GC-induced diabetes.Patients who developed GC-induced diabetes had a significantly higher age (67.5 ± 17.0 years vs 52.8 ± 18.6 years, P = .009),HbA1c levels (6.0 ± .42%vs 5.7 ± 0.41%, P = .016),and cumulative oral prednisolone dosage (57. 5 [44.5-67.6]mg/kg vs 43.3 [30.5-60.6]mg/kg, P = .046)compared to those who did not develop this condition.MLF was also more frequent in patients who developed GC-induced diabetes compared to those who did not (80.0%vs 30.2%,P = .001).Additionally, patients who developed GC-induced diabetes showed a substantial increase in facial measurements (80.0%vs 46.5%, P = .025),particularly within the first month of treatment (53.3% vs 14.0%, P = .002).These findings suggest that an initial increase in facial measurements was associated with the onset of GC-induced diabetes.
Figure 6 shows the results of the logistic regression analysis with GC-induced diabetes as the dependent variable and MLF and the risk factors previously reported for GC-induced diabetes as the independent variables [15][16][17][18][19].In univariate analysis, MLF (OR 9.23, 95% CI 2.23-38.29,P = .002)and age (OR 1.05, 95% CI 1.01-1.09,P = .015)were found to be the significant factors.In multivariate analysis, MLF (OR 6.58, 95% CI 1.25-34.74,P = .026)and age were still significant (OR 1.06, 95% CI 1.01-1.12,P = .017),and MLF  was found to be an independent factor associated with GC-induced diabetes.In 8 (66.7%) out of 12 patients who had both MLF and GC-induced diabetes, MLF preceded or occurred simultaneously with GC-induced diabetes.Additionally, the average onset time for GC-induced diabetes was 7.3 ± 6.9 weeks, while it was 4.9 ± 4.6 weeks for MLF, indicating that MLF preceded the onset of GC-induced diabetes (data not shown).

Discussion
The present study shows that the onset of MLF during GC treatment is associated with the development of GC-induced diabetes and hypertension.Notably, the onset of MLF was found to be an independent factor associated with GC-induced diabetes.
In previous studies on high-dose prednisolone administration (≥ 20 mg/day), CIL was found to be associated with hyperglycemia, hypertension, and cardiovascular events [3].Our study uses a novel approach as we distinguished MLF from CIL and included patients treated with a broad range of prednisolone doses (10-50 mg/day) compared to those in previous studies [1][2][3][4][5].
There has been no report showing biometric or quantitative assessment for MLF.A previous report has characterized facial features of healthy subjects with high salivary cortisol levels using computer image analysis [20]: high salivary cortisol levels are associated with facial features such as a broadened perimeter of the face, a rounded facial contour, and swollen, slit-like eyes with increased extracellular plasma volume.In the present study, locations 2, 3, 4, and 5 can reflect a broadened perimeter of the face and a rounded facial contour, and location 1 can reflect swollen, slit-like eyes.Therefore, we considered these measurement locations in the present study to reflect the facial features of GC-associated facial changes.
The pathophysiology of MLF and central obesity is not completely clear, partly due to the lack of available animal models [21,22].However, the following mechanisms have been proposed for GC-induced central obesity: The glucocorticoid receptor (GR) is more abundant in visceral adipose tissue than in subcutaneous adipose tissue [23,24].Therefore, GC may have a greater effect on lipoprotein lipase activity in visceral adipose tissue, resulting in visceral fat accumulation, because GR promotes lipoprotein lipase activity [24][25][26].Additionally, GC promotes ectopic fat accumulation.Ectopic lipid accumulation in the liver and muscles caused by GC causes insulin resistance and promotes hyperglycemia [19].
In this study, an initial oral prednisolone dosage ≥30 mg/ day and female sex were identified as the predisposing factors for MLF.This is consistent with the findings of previous studies in which GC dosage was correlated with the incidence of GC-induced diabetes [15][16][17][18][19]. GC regulates glucose intolerance and adiposity.Therefore, during the pathogenesis of GC-induced diabetes, it is thought that GC dose-dependently increases the expression of glycogenic genes via the GR, such as phosphoenolpyruvate carboxykinase and glucose-6phosphatase [19].Furthermore, the GR reportedly upregulates gene expression of fatty acid synthase and acetyl-CoA carboxylase, which stimulate adipogenesis [22].Thus, we hypothesized that dose-dependent facial adipogenesis induced by GC is involved in the pathogenesis of MLF.Similar to the present study, previous studies have shown that female sex is a predisposing factor for CIL [2], although the underlying mechanism has not yet been elucidated.lipoprotein lipase activity in human subcutaneous fat is known to be higher in females than in males [27].Therefore, this may promote the accumulation of facial subcutaneous fat in females.
Patients who developed GC-induced diabetes had significantly increased facial measurements but did not experience weight gain.These patients were older than those without GC-induced diabetes.We speculate that they gained fat instead of losing muscle.Previous studies have shown that male hormones inhibit GC-induced skeletal muscle atrophy [28] and that women are more susceptible to muscle weakness in subclinical hypercortisolism [29].Additionally, GC-induced skeletal muscle atrophy is more likely to advance in female mice than in males [30].Although our study did not specifically evaluate the muscles, the loss of muscle mass may play a role in the pathogenesis of GC-induced diabetes.
In this study, patients who developed MLF experienced body image disturbances.Furthermore, patients with MLF exhibited a significant increase in body weight compared to those without MLF.In a previous study, the main reason why patients with rheumatoid arthritis refused GC therapy was an increase in body weight [31,32].In multivariate analysis adjusted for body weight, MLF was found to be an independent factor for body image disturbance.Therefore, healthcare professionals should provide psychological support to patients with MLF and explain the potential development of MLF in patients with risk factors before initiating treatment.
The limitations of this study include its small sample size, single-center design, and being restricted to Asian individuals.Furthermore, facial measurement as an evaluation method for the MLF is a new approach, and the optimal facial measurement locations for assessing the MLF have not yet been established.Therefore, it cannot be claimed that the 5 facial areas measured in this study were optimal for evaluating MLF.
In conclusion, we identified the predisposing factors for MLF and its association with GC-induced diabetes and body image disturbance by utilizing a combination of facial measurements and subjective assessments by individuals and physicians.However, further research is needed to elucidate the pathophysiology of MLF and determine the appropriate facial measurement methods for its objective assessment.

Figure 1 .
Figure 1.The measurement locations in this study.

Figure 3 .
Figure 3. Logistic regression analysis of risk factors in patients with and without moon-like facies.Risk assessment model for moon-like facies.The black square represents the odds ratio after adjusting for these factors in the logistic regression model, and the horizontal line indicates the 95% confidence interval.Variables that showed statistical significance at the end of the logistic regression analysis are presented.

Figure 4 .
Figure 4. Characteristics of patients with and without MLF after glucocorticoid therapy.(A) Cumulative oral prednisolone dosage.(B) Cumulative oral prednisolone dosage corrected for body weight.(C) Sum of facial measurements.(D) Δ Sum of facial measurements.(E) Body weight at the end of the observation period.(F) Δ Body weight.(A) Score of the BILSv1.2.We excluded 10 patients who did not provide informed consent.The results were compared between the lowest scores.(C, D) Results were compared at the time of MLF determination for patients with MLF and at the end of the observation period for those without MLF.Abbreviations: BILSv1.2,Bilateral Inverted Locus of the Skin; MLF, moon-like facies.

Figure 5 .
Figure 5. (A-C) Cumulative incidence rates of GC-induced diabetes, hypertension, and dyslipidemia.The variables with statistical significance in the log-rank test are presented.(D-F) Number of patients with GC-induced diabetes, hypertension, and dyslipidemia with or without MLF.The variables with statistical significance in the chi-square test are shown.(A, D) A total of 14 patients with diabetes at the start of the study were excluded.(B, E) A total of 27 patients with hypertension at the start of the study were excluded.(C, F) A total of 33 patients with dyslipidemia at the start of the study were excluded.Abbreviations: GC, glucocorticoid; MLF, moon-like facies.

Figure 6 . 6
Figure 6.Logistic regression analysis of risk factors with GC-induced diabetes as the dependent variable.Risk assessment model for GC-induced diabetes.The black square represents the odds ratio in the univariate and multivariate analysis, and the horizontal line indicates the 95% confidence interval.Variables that showed statistical significance at the end of the logistic regression analysis are presented.Abbreviation: GC, glucocorticoid.

Figure 7 .
Figure 7. Linear regression analysis with scores of Bilateral Inverted Locus of the Skin as the dependent variable.The black square represents the β coefficient after adjusting for these factors in the linear regression analysis, and the horizontal line indicates the 95% confidence interval.Variables that showed statistical significance at the end of the logistic regression analysis are presented.Ten patients who did not provide informed consent were excluded.

Table 2 . Characteristics of patients with GC-induced diabetes GC-induced diabetes
Abbreviations: BMI, body mass index; GC, glucocorticoid; HbA1c, hemoglobin A1c; MLF, moon-like facies.We excluded 14 patients with diabetes at the start of the study.Values are shown as mean ± SD and median [25th percentile value-75th percentile value].