Association between the WHO 2017 and 2022 classification updates and ICD-10 code accuracy in patients with PitNETs: a...

This study was conducted to evaluate the association between the implementation of the WHO 2017 and 2022 classification updates for pituitary tumours and the validity of ICD-10 codes in identifying nonfunctioning and functioning subtypes of pituitary neuroendocrine tumours (PitNETs) in a real-world surgical cohort.

We analysed data from 1,096 surgically treated PitNET patients at a major Chinese medical centre between January 2020 and April 2024. The validity of the ICD-10 codes for identifying nonfunctioning and functioning PitNET subtypes was assessed using prepathological discharge diagnoses as the reference standard. The performance metrics, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), Youden index, F1 score, and kappa statistic, were calculated for each subtype. These validity measures were then compared between cases classified according to the 2017/2022 WHO criteria and those classified using the pre-2017 criteria in pathological reports.

ICD-10 code sensitivity was lowest for nonfunctioning PitNETs (72.9%; 95% CI: 69.1-76.3), followed by functioning gonadotrophs (83.3%; 95% CI: 36.5-99.1) and corticotrophs (84.0%; 95% CI: 77.4-89.0), while other subtypes maintained high sensitivity (97.9%-100%). Most subtypes had low PPVs (6.3-75.0%), except for nonfunctioning and somatotroph PitNETs. The nonfunctioning PitNET code also had a low NPV (73.8%; 95% CI: 70.1-77.2). When cases were pathologically classified using the updated 2017/2022 WHO criteria, a substantial decrease in sensitivity for nonfunctioning PitNETs was observed (94.7% to 63.0%, p < 0.001), which coincided with a reduced NPV, F1 score, and kappa, despite increased specificity (95.8% vs. 86.8%, p < 0.01). Similarly, under the new classification, corticotroph PitNETs had decreased specificity (100.0% to 92.8%, p < 0.001), PPV (94.8% to 64.6%, p < 0.001), and composite metrics. Notably, low PPVs persisted for lactotroph (59.2%) and gonadotroph (5.1%) PitNETs. Coding discrepancies primarily involved confusion with new pathological terminology and misapplication of endocrine codes for pituitary stalk compression effects.

The updated WHO classifications enhance pathological detail but should not guide clinical ICD coding, as misapplication reduces coding accuracy in surgically managed PitNETs. Future frameworks must align pathological nomenclature with clinical endocrine function in the ICD, maintaining a clear distinction between these domains. Multidisciplinary collaboration and standardized coding protocols are essential for improving accuracy.

Not applicable.