Multivariable Model to Predict an ACTH Stimulation Test to Diagnose Adrenal Insufficiency Using Previous Test Results

Clinicians frequently repeat adrenal function assessments to track recovery in patients with suspected hormonal deficiencies. No prior work had resolved whether historical data could improve the accuracy of these follow-up evaluations. That uncertainty drove the need for a predictive framework incorporating longitudinal patient records. It was already known that baseline cortisol levels provide some diagnostic utility during initial screenings. However, relying solely on single-point measurements often ignores the valuable context provided by previous clinical encounters. This gap motivated researchers to investigate if combining past results with current hormone levels enhances predictive precision. Prior research has shown that the hypothalamo-pituitary-adrenal axis exhibits complex recovery patterns over time. Understanding these dynamics requires more than just isolated snapshots of endocrine function.

The study aimed to develop and validate a predictive model for estimating the results of follow-up adrenal function assessments. Researchers sought to determine if incorporating historical test data could improve diagnostic accuracy compared to current methods. Many patients undergo repeated testing to monitor the recovery of their hypothalamo-pituitary-adrenal axis, creating a need for more efficient evaluation strategies. This project addressed the challenge of predicting future hormonal responses without relying solely on new, isolated baseline measurements. By leveraging existing patient records, the team hoped to provide clinicians with a more nuanced tool for interpreting follow-up results. The motivation stemmed from the desire to reduce the burden of frequent, repetitive testing on both patients and healthcare systems. This investigation focused on identifying the most informative variables from previous clinical encounters to enhance predictive power. Ultimately, the authors intended to provide a validated, accessible resource for improving the management of adrenal health in clinical settings.

Review approach involved a retrospective, longitudinal cohort design conducted at a specialized adult endocrinology facility. Investigators examined 258 paired assessments from 175 participants to establish the initial predictive framework. They subsequently validated these findings using a distinct dataset consisting of 111 patient tests collected over one year. The team employed polynomial regression techniques combined with backwards variable selection to refine the model. Eight candidate predictors were evaluated, including previous hormone baselines and calculated cortisol/ACTH ratios. The primary outcome focused on measuring cortisol levels exactly 30 minutes after the administration of Synacthen. Data collection occurred systematically to ensure the model reflected real-world clinical practice patterns. This rigorous methodology allowed for the identification of the most influential variables for predicting future hormonal responses.

Key findings from the literature demonstrate that the multivariable model significantly outperforms single-measurement approaches for predicting adrenal function. The combination of previous baseline cortisol, previous 30-minute cortisol, and new baseline cortisol achieved an R-squared value of 0.71. In contrast, using new baseline cortisol alone resulted in a lower R-squared value of 0.53. The model showed high diagnostic accuracy with an area under the curve of 0.97. This performance metric exceeded the 0.88 area under the curve observed when relying on new baseline cortisol alone. These results indicate that historical test data provides a robust foundation for estimating future clinical outcomes. The findings remained consistent during the validation phase using the independent cohort of 111 patients. This evidence confirms that integrating longitudinal records enhances the precision of follow-up endocrine evaluations.

Synthesis and implications suggest that integrating historical data significantly improves the accuracy of follow-up adrenal function assessments. The authors propose that clinicians utilize this model to refine their monitoring strategies for patients with suspected hormonal insufficiency. This approach offers a superior alternative to relying exclusively on current morning hormone measurements. The findings indicate that previous test outcomes provide reliable indicators for future physiological responses. Researchers highlight the availability of an online tool to facilitate broader external validation of these predictive calculations. The evidence supports a shift toward more personalized diagnostic pathways in endocrinology clinics. By leveraging existing patient records, medical teams can potentially reduce the frequency of unnecessary invasive procedures. These results demonstrate that longitudinal data synthesis enhances clinical decision-making processes for adrenal health.