Serum Cortisol, Hepcidin and Thyroid Hormone Levels in Neonates with Late-Onset Sepsis; Methodological Issues on Diagnostic and Prognostic Studies.

Abstract

I was interested in reading the paper by Hagag A and colleagues that was published in the March 2020 edition of the Infect Disord Drug Targets [1]. Neonatal sepsis is a clinical syndrome characterized by symptoms and signs of infection in the first twenty-eight days of life. Serum thyroid, cortisol and hepcidin are affected by neonatal sepsis. The purpose of the authors was to assess the predictive value of serum thyroid hormones including free triiodothyronine (free TT3) and free tetraiodothyronine (free TT4), serum cortisol and hepcidin levels through comparison of their concentrations between normal neonates and neonates with high probable late onset sepsis. They carried out a case control study on 40 neonates with suspected high probable late onset neonatal sepsis and 40 healthy neonates matched in age and sex as a control group (group II). For patients and controls; blood culture, highly sensitive C-reactive protein (H-s CRP), serum hepcidin, serum cortisol and thyroid hormones levels including free TT3 and free TT4 were assessed. They reported a significant positive correlation between H-s CRP and serum hepcidin and cortisol in group I while there was significant negative correlation between H-s CRP and free TT3 and free TT4. ROC curve of specificity and sensitivity of H-s CRP, serum hepcidin, cortisol, free TT3 and free TT4 in prediction of neonatal sepsis shows that serum hepcidin had the highest sensitivity and specificity with 95% and 90% respectively followed by serum cortisol, H-s CRP, free TT3 and lastly free TT4. First, reproducibility and validity (accuracy) are two different methodological issues of diagnostic studies [2]. Second, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) as well as likelihood ratios [likelihood ratio positive (ranging from 1 to infinity; the higher the LR+, the more accurate the test) and likelihood ratio negative (ranging from 0 to 1; the lower the LR-, the more accurate the test)] are among the estimates to assess validity (accuracy) of a diagnostic test [2-6]. Third, receiver operative characteristic (ROC) curve is usually used to assess diagnostic accuracy (discrimination) of a diagnostic model and has nothing to do with prediction of neonatal sepsis. Finally, logistic regression analyses have nothing to do with prediction studies. For prediction, we need at least two different cohort datasets or at least one cohort dataset, splitting them for development and validation of our model [2,7,8]. They concluded that neonates with high probable sepsis had significantly higher serum cortisol and hepcidin and significantly lower free TT3 and free TT4 compared with healthy neonates. They claimed that higher serum cortisol and hepcidin and lower free TT3 and free TT4 can lead to early diagnosis and subsequently better prognosis. Such conclusion should be supported by the above-mentioned methodological issues. Otherwise, misdiagnosis and mismanagement of the patient may occur.