Acromegaly: Is earlier diagnosis possible? Exploration of a screening algorithm to select high-risk patients

Abstract

Purpose

Given the ongoing need to reduce the delay in diagnosis of acromegaly, the aim of the current study was to develop a screening algorithm that: finds as many patients as possible who have already been diagnosed with acromegaly (=sensitivity), while identifying a limited number of additional patients who are at risk for acromegaly as a proxy for specificity.

Methods

CTcue data mining software (version 4.7, IQVIA Inc., USA) was used to develop a screening algorithm. Data were extracted from the electronic health records of a tertiary centre. Multiple sources (literature review, real-world evidence and expert opinion) were used to identify predictors such as clinical manifestations, comorbidities and frequently consulted specialists, for inclusion in the algorithm. Based on different combinations of predictors, several exploratory search strategies were constructed in CTcue and algorithms with the highest sensitivity were further adapted.

Results

A total of 68 predictors were identified and grouped into 5 categories. After exploratory analysis and fine-tuning, an algorithm combining pathognomonic changes in clinical appearance with manifestations considered by medical experts to be most characteristic of acromegaly (hyperhidrosis, sleep apnoea, arthralgia, headache and type 2 diabetes mellitus) detected 48/90 of patients with previously diagnosed acromegaly, while identifying an additional 1844/1,7 million of possible at-risk patients.

Conclusion

We found that our best algorithm led to the detection of more than half of patients with previously diagnosed acromegaly. The same algorithm identified 0.1 % of the hospital population as potentially having acromegaly, which is approximately 10 times higher than the estimated worldwide prevalence. These results seem promising but need further improvement and validation.