Development of a Screening Tool Differentiating Patient Symptoms in Cushing's, Polycystic Ovarian and Metabolic Syndromes

Background:

The first point of contact and gate-keeping for referrals to specialty care in many medical systems, is through a primary care practitioner.  However, the difficulty in identifying patients with Cushing’s syndrome in a primary care setting often leads to a significant delay or in missed diagnosis for this disease.

Hypercortisolism, including Cushing’s syndrome (CS) and disease (CD) is considered rare, with an estimated incidence of 0.2–5.0 per million people per year and a prevalence of 39–79 per million in various populations.  Women have a higher prevalence than men 3:1 with a median age of 41·4 years at diagnosis (Lacroix, Feelders, Stratakis, & Nieman, 2015).  Cushing’s disease (CD) represents 80% of the cases of Cushing’s syndrome and is caused by an ACTH-secreting pituitary adenoma with an estimated prevalence of  2.4 cases per million inhabitants (Acebes, Martino, Masuet, Montanya, & Soler, 2007). CS confers a significant mortality and morbidity risk and lower quality of life (QoL) from multiple comorbidities such as cardiovascular disease, diabetes and osteoporosis.  Early diagnosis is paramount to ameliorate these factors (Nieman, 2015).   The associated delay in diagnosis has been estimated to be between 6 months and 10 years and not only contributes to morbidity and mortality, but also likely results in an underestimation of prevalence (De, Evans, Scanlon, & Davies 2003).  

Presenting symptoms are similar to diseases such as metabolic syndrome and polycystic ovarian disease (Brzana et al., 2014).  Hence, patients report frustration and humiliation in a medical system where they have often sought multiple consultations prior to diagnosis.    Once CS is suspected, patients are usually referred to an endocrinologist for a full workup and formal diagnosis. However, informed by the internet, many patients self-diagnose and request a specialty referral based on symptomology. In the time provided in a typical medical consultation both patients may have difficulty succinctly presenting key symptoms.  Coupled with the provider lack of time and knowledge regarding this rare disease associating symptoms with more uncommon diagnoses may be an insurmountable challenge.  Thus, identification of  symptoms most commonly described by CS patients may provide a guide to primary care providers that specialty referral and work up is indeed warranted.

Aim:

To identify sentinel patient reported symptoms and/or functional limitations that have a high specificity for CS and can differentiate CS patients from those with metabolic syndrome (MS) or PCOS. Secondarily to develop a valid and reliable screening tool and scoring guide for use in primary care settings indicating the need for specialty referral.

Method:

A prospectively administered questionnaire was completed by 139 patients (42 male/97 Female) presenting with symptoms of pituitary dysfunction or with pituitary tumors at a single institution pituitary center from 2011-2015. Patients with unstable co-morbidities and /or a severe life stressor within 12 months of presentation were excluded.   All subjects completed a 205 item questionnaire evaluating their presenting symptoms and perceived dysfunction and associated severity on a 6 point Likert scale from:  not symptomatic (0);  to most severe (5).  The questionnaire was developed using modified scales, such as the Beck Depression Inventory, Eysneck Personality scale, Epworth Sleepiness Scale, Krupp Fatigue Severity Scale, Functional Assessment Rating Scale and symptoms of pituitary diseases derived from review of literature and patient interviews. All patients completed the questionnaire prior to disease work up and diagnosis confirmation.

Item reliability analysis was performed using SPSS 18 and poorly discriminating items were removed.  Patients were then separated by diagnosis. Only those patients with biochemically and pathology confirmed CD (14) and non-function pituitary adenomas (NFA) (52) were included in the final analysis to select screening tool items.  The remaining items were re-examined using independent T test.  Items were selected that demonstrated a significant difference (p=<0.05) between the 2 diagnostic groups that were CS specific.  Collectively they formed the Cushing’s Syndrome Screening Tool (CSST).

The tool was re-examined for item reliability, sensitivity and specificity. To evaluate the sensitivity and specificity of the tool, the respondent pool was re-examined and responses divided into 3 groups:  Group 1 with confirmed Cushing’s syndrome or disease; Group 2 patients meeting criteria for metabolic syndrome (MS); Group 3 patients with a diagnosis of PCOS; Group 4 comprised of healthy controls was solicited from the community.  

Metabolic syndrome was defined by most commonly agreed upon criteria as any three of five risk factors:  Hyperglycemia fasting glucose > 100mg/dL; obesity BMI  ≥ 30.00; elevated triglycerides >150 mg/dL,;  HDL <40-50mg/dL;  or hypertension (Alberti et al., 2009)

Polycystic ovarian disease was defined as an endocrine disorder involving infertility, hyperandrogenism, and insulin resistance in women with or without the presence of polycystic ovaries on imaging (Legro et al., 2013; Dunaif et al., 1992 ).

ANOVA with Tamhane’s post hoc analysis and ROC analysis were performed to evaluate a significant difference between groups and to establish the tool sensitivity and specificity using PSAW 18. This study was approved by the OHSU Internal Review Board.

Results:

After item analysis, a total of 56 items were retained to achieve a Cronbach’s alpha of .97 for the primary instrument.   Independent T test revealed 10 questions demonstrating a significant difference between CS and other diagnostic groups and specificity for CS.   Cronbach’s alpha for this screening tool was .95. 

The screening tool was applied to 56 subjects:  Group 1, 14 CS (3Male/11Female) ; Group 2,  10 MS (3Male /7Female);  Group 3, 11 PCOS (11Females); Group 4  21 Controls (8 Males/13 Females).   Mean age was similar for the control group and patients with CS (40 vs 42years), but patients with PCOS were significantly younger that those with CS (p=0.05) and MS (p=0.004).  Gender distribution was only different between controls and patients with PCOS (all females) as expected (P=0.01). BMI was only different between controls and those with MS (p=0.000).  Also as expected patients with MS had significantly more hyperlipidemia, diabetes mellitus and hypertension (p=>0.001) than all other groups.   The mean severity scores of a possible of 50 were: Gr1 (CS), 38.07 (range 14-58); Gr 2 (MS), 15.1 (range 5-25); Gr3 (PCOS), 13.27 (range 7-20); Gr 4 Controls, 3.67 (range 0-10).  Mean total scores for patients with confirmed CS were significantly higher than for all other groups (p=>0.001).  There was no significant difference for mean total scores between patients with MS and PCOS using this tool. All control patients scored < 10.

The sensitivity of the tool for CS was 85.7% when the specificity was 97% (AUC= 0.965).  At a mean severity score of >30, the positive predictive value of this tool for CS was .99 with the negative predictive value of .93.

Conclusion:

Although further validation with larger a population of patients with CS is required, the Cushing Syndrome Screening Tool demonstrates high item reliability, sensitivity and specificity for CS.  As a clinical tool, it may be useful for differentiating patient with Cushing’s syndrome from those with both metabolic syndrome and polycystic ovarian disease. This allows for the identification and referral of patients who may benefit from further evaluation for CS or CD.