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

Abstract

Background

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.

Methods

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.

Results

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.

Conclusion

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.

Clinical trial number

Not applicable.

Supplementary information

The online version contains supplementary material available at 10.1186/s12902-025-02121-w.

Background

Pituitary neuroendocrine tumours (PitNETs) represent a clinically diverse group of neoplasms for which precise classification is critical for diagnosis, treatment, and epidemiological research [1]. The World Health Organization (WHO) has substantially revised the pathological criteria for these tumours, moving from a primarily histological classification to one based on cell lineage and transcriptional profiling in its 2017 and 2022 updates [2–4]. A pivotal change in the 2022 classification was the redefinition of “pituitary adenomas” as PitNETs, emphasizing their neuroendocrine characteristics [4].

Despite these advances in pathological nosology, the clinical utility and implications of the new classification system remain debated [5–11]; a significant yet underinvestigated aspect is its association with the International Classification of Diseases (ICD) coding system. The ICD is the cornerstone for healthcare data management and directly influences resource allocation, clinical decision-making, and the validity of epidemiological research. Crucially, the current ICD-10 framework classifies these tumours as benign neoplasms (D35.2), with their endocrine function (e.g., acromegaly, Cushing’s disease) separately coded within the chapter on endocrine diseases. This creates a potential disconnect with the modern WHO system, which prioritizes cell lineage over clinical secretory status.

While ICD-10 codes have been used to identify pituitary adenomas in epidemiological studies [12–15], their accuracy in the context of the updated WHO classifications is unclear. Specifically, it is unknown how the shift in pathological diagnostic criteria relates to the accuracy of ICD-10 codes in capturing the distinction between clinically functioning and nonfunctioning (NF) PitNETs—a distinction retained as a primary axis of the pituitary in ICD-11 [16]. This gap is critical because coding inaccuracy can lead to significant biases in clinical databases. Therefore, to bridge this knowledge gap, we conducted a study utilizing a comprehensive single-centre surgical cohort of PitNET patients. We aimed to (1) evaluate the accuracy of ICD-10 codes in identifying clinically functioning and nonfunctioning PitNETs, using clinical discharge diagnosis as the reference standard, and (2) assess the association between the adoption of the WHO 2017/2022 classifications and ICD-10 coding performance. Our findings seek to inform standardized coding protocols and improve the reliability of healthcare data for PitNET management.

Materials and methods

Design and study setting

This retrospective study was conducted at Peking Union Medical College Hospital (PUMCH), a tertiary academic medical centre with 2023 inpatient beds. As a national pioneer in pituitary disorder management, PUMCH has had a dedicated multidisciplinary pituitary team since 1979, functioning as China’s primary referral centre for complex pituitary-related cases [17–21]. The institution also leads the Chinese Pituitary Adenoma Cooperative Group. PUMCH established the Pituitary Disease Innovation and Treatment Center in 2022 to further advance clinical services. This centre implements an integrated “one-stop service” model, delivering coordinated multidisciplinary care for patients with pituitary and hypothalamic disorders.

Data sources

This retrospective cohort study included consecutive patients who underwent surgical intervention for PitNETs between January 1, 2020, and April 30, 2024. The inclusion criteria were the availability of comprehensive pathological reports with complete immunohistochemical staining profiles for lineage-specific transcription factors (pituitary-specific positive transcription factor 1 [PIT1], T-box transcription factor [TPIT], and steroidogenic factor 1 [SF1]) and anterior pituitary hormones (growth hormone [GH], prolactin [PRL], adrenocorticotropic hormone [ACTH], thyroid-stimulating hormone [TSH], follicle-stimulating hormone [FSH], and luteinizing hormone [LH]). Patients with incomplete immunohistochemical profiles or pathological diagnoses of pituitary hyperplasia or nonneoplastic pituitary lesions were excluded from the analysis. Patients were identified through the hospital’s discharge abstract database managed by the Department of Medical Records using the following ICD-10 code as either a primary or secondary diagnosis: D35.2 (benign neoplasm of the pituitary gland). A standardized protocol was followed for the data extraction process. We collected the following data for each eligible patient: (1) medical record number as a unique patient identifier; (2) primary and secondary diagnostic statements (up to 10 entries per patient); (3) corresponding ICD-10 codes and classification terminologies; and (4) detailed pathological reports, including pathological diagnoses and immunohistochemical staining results, with a particular focus on three pituitary-specific transcription factors (PIT1, TPIT, and SF1) and six anterior pituitary hormones (GH, PRL, ACTH, TSH, FSH, and LH).

For patients who underwent more than one surgical procedure during the study period, only the data from their first admission were included.

To ensure consistency with current diagnostic standards, all pathological classifications were reviewed according to the 2022 WHO Classification of Pituitary Tumours [5]. Importantly, while pathological reports from the entire study period (2020–2024) were available, the institutional adoption of updated classification standards demonstrated a transitional pattern. Specifically, although the 2017 or 2022 WHO criteria were formally introduced during the study period, a substantial proportion of cases continued to utilize pre-2017 pathological formats and nomenclature in their original reports. This phenomenon primarily reflected institutional operational inertia in fully implementing updated documentation templates rather than any fundamental disagreement with the new classification system. For patients who were originally classified using the 2017 or pre-2017 WHO classification standard, a senior neuropathologist with more than 8 years of experience in pituitary pathology was invited to perform reclassification according to the updated 2022 standard where necessary. This reclassification process involved a comprehensive review of histological features and complete immunohistochemical profiles.

Coding mode

The Department of Medical Records at PUMCH employed a stable team of certified coding professionals responsible for assigning diagnostic coding for all hospital admissions throughout the study period. Notably, during the entire study timeframe, there were no major changes to the electronic medical records (EMR) system and no significant turnover in coding personnel, thereby minimizing potential temporal confounders related to administrative or technological shifts. These coders, who maintained a generalist approach without specializing in specific disease entities or clinical departments, operated without specific training in pituitary tumour coding. The coding process was systematically conducted within 15 days post-discharge to meet mandatory reporting deadlines for morbidity statistics, healthcare quality monitoring, and hospital performance evaluation. In cases where a tumour was clinically suspected, the coders initially assigned provisional ICD-10 codes based on discharge diagnoses. Following the release of final pathology reports, the coders then systematically reviewed and updated the tumour-related ICD-10 codes accordingly.

Algorithms for the identification of patients with nonfunctioning or functioning PitNETs using ICD-10 codes

The types of PitNETs and their corresponding ICD-10 codes used in daily coding practice and descriptions are systematically presented in Table 1. Please see Supplementary Table S1 for a complete mapping of the local extended ICD-10 codes to the standard ICD-10 framework.

Table 1.

Clinically nonfunctioning and functioning PitNET subtypes with the corresponding ICD-10 codes and clinical descriptions

PitNET subtypes	ICD code combinations	Code description
Nonfunctioning PitNETs	D35.202* or D35.2 (without E codes)	D35.2: Benign neoplasm of the pituitary gland
		D35.202*: Nonfunctioning pituitary tumour
Functioning PitNETs subtypes
Somatotroph	D35.2 + E22.0	E22.0: Acromegaly or gigantism
Lactotroph	D35.2 + E22.101*	E22.101*: Hyperprolactinemia
	D35.2 + E22.102*	E22.102*: Prolactinoma
Corticotroph	D35.2 + E24.0	E24.0: Pituitary-dependent Cushing’s disease
Thyrotroph	D35.2 + E05.802*	E05.802*: Thyrotropinoma/TSHoma
	D35.2 + E05.803*	E05.803*: TSH-dependent hyperthyroidism
	D35.2 + E05.805*	E05.805*: Syndrome of inappropriate TSH secretion
Gonadotroph	D35.2 + E22.804*	E22.804*:Gonadotropin-secreting pituitary tumour
Plurihormonal	D35.2 + E22.801* or D35.2 + ≥2 E codes	E22.801*: Plurihormonal pituitary tumour

* Modified version of ICD-10 codes in the hospital setting

Diagnostic criteria for clinically nonfunctioning PitNET and functioning PitNET subtypes

Nonfunctioning PitNETs (NF-PitNETs) are tumours that do not produce hormones and may present with mass effect symptoms (visual field defects, headache, and anterior hypopituitarism) [22]. Functioning PitNETs are defined by the hypersecretion of pituitary hormones. The primary subtypes, classified based on the hormone produced, are prolactotroph, somatotroph, corticotroph, and thyrotroph PitNETs, corresponding to the excess release of prolactin, growth hormone, ACTH, and TSH, respectively.

The diagnostic criteria for functioning somatotroph PitNETs included (1) clinical signs and symptoms related to excessive GH and insulin-like growth factor-1 (IGF-1); (2) a nadir GH exceeding 1.0 ng/ml with a 75 g oral glucose tolerance test based on the updated Chinese consensus (2021 edition) for the diagnosis and treatment of acromegaly [23] and supranormal IGF-1 levels adjusted for age and sex [24]; and (3) evidence of a pituitary mass on magnetic resonance imaging (MRI).

The diagnosis of functioning lactotroph PitNETs required (1) persistent hyperprolactinemia, typically > 200 ng/mL for macroadenomas after excluding secondary causes; (2) radiographic confirmation of a pituitary adenoma on MRI; and (3) concordance between the prolactin level and tumour size. Supportive clinical features included hypogonadal symptoms (e.g., amenorrhea and galactorhea in women and decreased libido or erectile dysfunction in men) or manifestations of mass effects (e.g., visual field defects) [25].

Functioning corticotroph PitNETs were diagnosed using a standardized protocol. Biochemical criteria included elevated 24-hour urinary free cortisol and serum ACTH levels and a positive dexamethasone suppression test (nonsuppression with low-dose, > 50% suppression with high-dose). Tumour localization was achieved via pituitary MRI (microadenoma: < 10 mm; macroadenoma: ≥10 mm). For equivocal cases, inferior petrosal sinus sampling (IPSS) with desmopressin stimulation was performed as previously described by Feng [26], with a central-to-peripheral ACTH ratio ≥2.0 (basal) or ≥3.0 (poststimulation), confirming the diagnosis.

The suspicion of functioning thyrotroph PitNETs was based on two primary findings: the presence of elevated circulating free thyroid hormone levels alongside nonsuppressed TSH levels and the identification of a pituitary mass on contrast-enhanced MRI [27, 28]. Confirmatory dynamic tests, namely, the TRH stimulation test and T3 suppression test, could not be performed due to the unavailability of specific test reagents in our clinical setting. Consequently, the critical differential diagnosis from resistance to thyroid hormone (RTH) relies on alternative approaches, including somatostatin analogue tests and the evaluation of peripheral bone and liver markers. When needed, genetic testing for thyroid hormone receptor mutations was also utilized to achieve a definitive distinction. For hormone quantification, serum levels of TSH, FT3, FT4, T3, and T4 were measured using the direct chemiluminescence method (ADVIA Centaur, Siemens, USA), with the following reference intervals: 0.38–4.34 μg/mL TSH, 1.80–4.10 pg/mL FT3, 0.81–1.89 ng/dL FT4, 0.66–1.92 ng/mL T3, and 4.30–12.50 μg/dL T4 [29].

The diagnosis of functioning gonadotroph PitNETs was based on the following integrated criteria: (1) characteristic clinical presentations, including menstrual disorders, recurrent large ovarian cysts ( > 1 cm), or spontaneous ovarian hyperstimulation syndrome in women; testicular enlargement in men; or isosexual precocious puberty in children; (2) documented abnormalities in sex hormone profiles, where females exhibited either elevated oestradiol (E2) with increased/nonsuppressed FSH and suppressed/inappropriately normal LH or elevated LH with suppressed/inappropriately normal FSH, while males showed either reduced testosterone (T) with elevated/nonsuppressed FSH and suppressed/inappropriately normal LH or elevated T with elevated/nonsuppressed LH and suppressed/inappropriately normal FSH; and (3) radiologic confirmation of a pituitary adenoma on MRI [30].

Functioning plurihormonal PitNETs were diagnosed when a patient’s pituitary tumour met the biochemical and clinical diagnostic criteria for two or more of the distinct functioning PitNET subtypes described above. The diagnosis was primarily based on the biochemical evidence of hypersecretion of multiple hormones, supported by corresponding clinical syndromes.

Adjudication process for true and false cases of clinically nonfunctioning and functioning PitNETs

The reference standard for this study was the final clinical discharge diagnosis documented in the discharge summary, as established by a treating neurosurgeon or endocrinologist at the time of patient discharge. This diagnosis integrated preoperative and intraoperative findings, including clinical manifestations, biochemical endocrine profiles, imaging characteristics, and surgical observations. Importantly, this clinical diagnosis was formulated independently of the postoperative pathology report, which typically became available approximately two weeks after discharge. Once the pathology report was issued, the medical records were already archived, and clinicians generally did not revise the original discharge diagnosis based on pathological findings. To ensure accurate extraction of clinical discharge diagnoses, an expert panel comprising two neurosurgeons and two endocrinologists revalidated the diagnoses through a comprehensive review of electronic medical records (EMRs), focusing specifically on distinguishing between nonfunctioning and functioning PitNET subtypes. For cases where the diagnosis was uncertain, a senior pituitary endocrinologist made the final determination. Initial ICD-10 codes were assigned by medical coders exclusively based on the clinical discharge diagnosis. After the pathology report became available, the coders evaluated whether the pathological findings aligned with the original clinical diagnosis. Based on this assessment, they either retained the initial codes or updated the diagnostic codes according to the pathological classification. The accuracy metrics (based on true positive, false positive, true negative, and false negative cases) used in this study were derived by comparing the final ICD-10 codes against the original pathology-independent discharge diagnoses. Cases were classified according to a standardized protocol: cases showing concordance between hormone-secretory profiles at discharge and the corresponding ICD-10 codes were classified as true positives for their respective subtypes, whereas discordant cases were designated as false positives. Cases lacking functioning PitNET-specific ICD-10 codes underwent further categorization: accurately coded cases were classified as true negatives for functioning PitNET subtypes and as true positives for nonfunctioning PitNETs; cases with coding omissions were considered false negatives for functioning subtypes. The primary unit of analysis for all validation metrics was the unique patient. However, to verify the accuracy of the ICD-10 codes, each assigned code was validated independently against the discharge diagnoses as a reference standard. Special consideration was given to patients with clinically diagnosed plurihormonal PitNETs. These patients naturally receive multiple ICD-10 codes (e.g., one for each hormonally active subtype) during a single admission. Each of these codes was evaluated separately. Consequently, a single patient with a plurihormonal tumour could contribute more than one count to the true positive (TP) total across different coding categories.

Comparison of the hormone profiles reflected in the three-dimensional diagnostic approach

PAs/PitNETs were characterized using three diagnostic methods: (1) clinical diagnosis, established by physicians on the basis of a comprehensive evaluation of tumour characteristics and endocrine function at the time of discharge, prior to the availability of pathological confirmation; (2) standardized classification on the basis of ICD coding systems; and (3) definitive pathological subtyping according to the 2017 or 2022 WHO criteria, incorporating transcription factor expression and hormone immunohistochemical profiles. For comparative analysis, the patients were reclassified into seven hormonal profile groups (nonfunctioning/no hormone, PRL, GH, ACTH, TSH, multihormonal (MULTI), and LH/FSH) across all three diagnostic approaches. A Venn diagram was constructed to visualize the concordance of hormone secretion patterns in both functioning and nonfunctioning PitNETs among clinical diagnoses documented at discharge, final ICD-10 coding assignments, and pathological reports. The pathological diagnoses were realigned to the seven groups to ensure consistency: null cell tumours were categorized as nonfunctioning, while mammosomatotroph tumours, mature/immature plurihormonal PIT1-lineage tumours, mixed somatotroph-lactotroph tumours, and plurihormonal tumours without distinct lineages were consolidated into the multihormonal group. Similarly, mixed-lineage cases were assigned to the multihormonal group. Notably, three acidophil stem cell tumour cases did not conform to any of the seven predefined groups, whereas all the other pathological diagnoses could be mapped to one of these seven hormonal profile groups.

Statistical analysis

All the statistical analyses were performed using R software (version 4.3.0). Using the final clinical diagnosis from the discharge summary as the gold standard, we calculated diagnostic accuracy metrics—including sensitivity, specificity, and positive and negative predictive values (PPVs and NPVs) with 95% confidence intervals (CIs)—for the ICD-10 code combinations assigned to identify nonfunctioning and functioning PitNETs listed in Table 1. The Youden index and F1 score were also computed to provide a more comprehensive assessment of each coding strategy. Interrater agreement between clinical diagnoses and ICD-10 codes was evaluated using Cohen’s kappa (κ) statistic.

To examine the association between evolving pathological classification standards and ICD-10 coding accuracy, diagnostic accuracy metrics were analysed separately according to two pathology reporting styles influenced by the updated WHO classification for pituitary tumours. Sankey diagrams were used to visualize changes in diagnostic trajectories, specifically in hormonal patterns from clinical diagnosis through pathological diagnosis to the final ICD-10 code. Group comparisons were performed using chi-square tests (with and without continuity correction) and Fisher’s exact test to assess differences in code performance metrics. Furthermore, we conducted a comprehensive analysis of diagnostic discrepancies by categorizing and quantifying the causes of false-positive and false-negative classifications through a meticulous review of original medical records, with particular attention given to the frequency, proportion, and nature of coding inaccuracies.

Results

Basic information

A total of 1,096 unique patients with surgically treated PitNETs were included in the final analysis, with a female predominance (656 females, 59.9%) over a male predominance (440, 40.1%). The mean age at admission was 46.2 ± 14.0 years, with a significant difference in age between the sexes (males: 49.0 ± 14.0 years vs. females: 44.0 ± 14.0 years). The pathological diagnoses were made according to the 2017/2022 WHO criteria for PitNETs in 816 cases (74.5%), whereas the diagnoses for the remaining 280 cases (25.5%) were based on the pre-2017 classification systems.

Distribution of the hormone patterns of PitNETs in patients identified on the basis of clinical diagnosis, ICD-10 codes and pathological classification

Among the 605 patients with clinically NF-PitNETs, the SF1 lineage (60.3%) and TPIT lineage (28.9%) were predominant; among the 209 patients with clinically functioning somatotroph PitNETs, somatotroph tumours accounted for 50.7%, followed by mammosomatotroph tumours (24.9%) and mature plurihormonal PIT1-lineage tumours (15.3%); TPIT-lineage predominance (100.0%) was observed for functioning corticotroph PitNETs (n = 169); the PIT1 lineage (97.8%) was predominant among patients with functioning lactotroph PitNETs (n = 90); and functioning thyrotroph PitNETs (n = 11) demonstrated an equal distribution between patients with mature plurihormonal PIT1 lineage and those with immature PIT1 lineage tumours (45.5% each); functioning gonadotroph PitNETs (n = 6) exhibited 71.4% concordance between clinical and pathological hormone secretion profiles; and functioning plurihormonal PitNETs (n = 6; 5 with cosecretion of GH and PRL, and 1 with GH and TSH were pathologically confirmed to be positive for multiple hormones in all the patients. The comprehensive correlation between the clinically defined endocrine function subtypes of PitNETs and their corresponding pathological classifications is detailed in Table 2.

Table 2.

Distribution of the clinically nonfunctioning and functioning PitNET subtypes and their corresponding pathological classification according to 2022 WHO classification

PitNET type reflected in pathology report	Clinically nonfunctioning and functioning PitNET subtypes reflected in discharge diagnosis (n = 1096)
	NF (%)	GH (%)	ACTH (%)	PRL (%)	TSH (%)	FSH/LH (%)	Multihormonal (%)	Total
PIT1-lineage PitNETs
Somatotroph tumour	5(0.8)	115(55.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	120(10.9)
Lactotroph tumour	13(2.1)	0(0.0)	0(0.0)	88(97.8)	0(0.0)	0(0.0)	0(0.0)	101(9.2)
Mammosomatotroph tumour	2(0.3)	52(24.9)	0(0.0)	1(1.1)	0(0.0)	0(0.0)	2(33.3)	57(5.2)
Thyrotroph tumour	3(0.5)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	3(0.3)
Mature plurihormonal PIT1-lineage tumour	6(1.0)	32(15.3)	0(0.0)	0(0.0)	5(45.5)	0(0.0)	2(33.3)	45(4.1)
Immature PIT1-lineage tumour	6(1.0)	4(1.9)	0(0.0)	0(0.0)	5(45.5)	0(0.0)	0(0.0)	15(1.4)
Acidophil stem cell tumour	3(0.5)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	3(0.3)
Mixed somatotroph and lactotroph tumour	1(0.2)	6(2.9)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	2(33.3)	9(0.8)
TPIT-lineage PitNETs
Corticotroph tumour	175(28.9)	0(0.0)	169(100.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	344(31.4)
SF1-lineage PitNETs
Gonadotroph tumour	365(60.3)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	5(83.3)	0(0.0)	370(33.8)
Mixed lineages
TPIT-lineage/PIT1-lineage	3(0.5)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	3(0.3)
SF1-lineage/PIT1-lineage	0(0.0)	0(0.0)	0(0.0)	1(1.1)	0(0.0)	0(0.0)	0(0.0)	1(0.1)
Mature plurihormonal PIT1-lineage tumour/SF1-lineage	0(0.0)	0(0.0)	0(0.0)	0(0.0)	1(9.1)	0(0.0)	0(0.0)	1(0.1)
Immature PIT1-lineage tumour/SF1-lineage	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	1(16.7)	0(0.0)	1(0.1)
PitNETs without distinct cell lineage
Plurihormonal tumour	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)
Null cell tumour	23(3.8)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	0(0.0)	23(2.1)
Total	605(100.0)	209(100.0)	169(100.0)	90(100.0)	11(100.0)	6(100.0)	6(100.0)	1096(100.0)

The concordance in PitNET hormone profiles among clinical discharge diagnoses, ICD-10 codes, and pathological reports is shown in Fig. 1. Consistency between the ICD-10 codes and clinical diagnoses was observed for 869 patients. Additionally, the ICD-10 codes of 516 patients (364 + 152) were consistent with those of the pathological reports, and those of 406 patients (42 + 364) were consistent between the pathological reports and clinical diagnoses. Complete concordance across all three criteria was found in 364 patients, whereas only 33 patients were entirely inconsistent across all the criteria. The distribution of hormone profiles for each subtype according to the three documentation methods is shown in Fig. 2.

Fig. 1.

Concordance among the hormone profiles of PitNETs as documented by clinical discharge diagnoses, ICD-10 codes, and pathological reports. Legend: agreement in PitNET hormonal subtyping across three documentation methods is illustrated in the venn diagram. The ellipses represent clinical diagnoses (yellow), ICD-10 codes (light purple), and pathological reports (pink). Each segment indicates the number of patients with a specific agreement pattern. The total cohort of 1,096 patients was categorised into five subgroups (n = 364 + 152 + 505 + 42 + 33) based on these patterns. Case distribution shows complete agreement (P&C&ICD) in 364 cases, with other notable patterns being C&ICD (505), P&ICD (152), P&C (42), and cases exclusive to a single method (33)

Fig. 2.

Concordance among the hormone profiles of PitNETs as documented by clinical discharge diagnoses, ICD-10 codes, and pathological reports, stratified by hormonal subtype. Legend: this series of venn diagrams illustrates the agreement in the hormonal subtyping of PitNETs across three diagnostic dimensions—prepathological clinical diagnosis, ICD-10 codes, and pathological reports—stratified by hormonal subtype. Each panel includes patients whose subtypes were documented by at least one method, with overlapping areas indicating concordance and exclusive areas representing unique identification by a single dimension. Key numerical findings by subtype are as follows: NF (n = 634) was characterised by a predominant C&ICD agreement pattern (n = 421, 66.4%) and a low level of complete concordance (P&C&ICD = 20, 3.2%). PRL (n = 132) demonstrated substantial complete concordance (P&C&ICD = 86, 65.2%). GH (n = 214) exhibited moderate complete concordance (P&C&ICD = 107, 50.0%) and a predominant C&ICD overlap (n = 178, 83.2%). ACTH (n = 344) showed a moderate level of triple agreement (P&C&ICD = 141, 41.0%). In contrast, LH/FSH (n = 371) had minimal complete concordance (P&C&ICD = 4, 1.1%), with the vast majority of cases (n = 296, 79.8%) being exclusive to a single method. For TSH (n = 15), the majority were identified by C&ICD overlap (n = 10, 66.7%). Finally, the multi (n = 142) subgroup had limited triple agreement (P&C&ICD = 6, 4.2%), with the vast majority of its cases (n = 112, 78.9%) also being exclusive to a single method

Validity of the ICD-10 codes in identifying clinically nonfunctioning and functioning PitNET subtypes

The diagnostic validity of ICD-10 codes for identifying NF-PitNETs and functioning PitNETs varied considerably across different subtypes. For NF-PitNETs, the sensitivity was 72.9% (95% CI: 69.1–76.3), with a specificity of 94.1% (95% CI: 91.5–95.9). The PPV for NF-PitNETs was 93.8% (95% CI: 91.2–95.8), while the NPV was 73.8% (95% CI: 70.1–77.2). Among clinically functioning PitNETs, somatotroph PitNETs demonstrated perfect discrimination, with a sensitivity, specificity, PPV, and NPV of 100% (95% CIs: 97.8–100.0, 99.5–100.0, 97.8–100.0, and 99.5–100.0, respectively). Lactotroph PitNETs showed high sensitivity (97.9%; 95% CI: 91.9–99.6) and specificity (92.4%; 95% CI: 90.5–93.9) but low PPV (55.0%; 95% CI: 47.2–62.6). Corticotroph PitNETs exhibited balanced performance, with a sensitivity of 84.0% (95% CI: 77.4–89.0), a specificity of 94.5% (95% CI: 92.8–95.8), and a PPV of 73.6% (95% CI: 66.7–99.5). For less common subtypes, thyrotroph PitNETs showed perfect sensitivity (100.0%; 95% CI: 69.9–100.0) and high specificity (99.6%; 95% CI: 99.0–99.9) but limited PPV (75.0%; 95% CI: 47.4–91.7). The PPV of gonadotroph PitNETs was notably low (6.3%; 95% CI: 2.3–14.8), despite their reasonable sensitivity (83.3%; 95% CI: 36.5–99.1). Plurihormonal PitNETs also had a low PPV (13.0%; 95% CI: 5.4–27.0) despite perfect sensitivity (100.0%; 95% CI: 51.7–100.0). Agreement measures reinforced these patterns, with excellent performance for somatotroph PitNETs (κ = 1.000; Youden index = 1.000; F1 score = 1.000) and substantial agreement for lactotrophs (κ = 0.668; Youden index = 0.903; F1 score = 0.705) and corticotroph PitNETs (κ = 0.742; Youden index = 0.785; F1 score = 0.785). In contrast, gonadotroph (κ = 0.109; Youden index = 0.765; F1 score = 0.118) and plurihormonal PitNETs (κ = 0.223; Youden index = 0.963; F1 score = 0.231) showed poor agreement metrics despite high Youden indices, reflecting the challenge in correctly identifying these rare subtypes. The complete diagnostic accuracy metrics are presented in Table 3.

Table 3.

Validity of the ICD codes in identifying clinically nonfunctioning and functioning PitNETs subtypes (n = 1096)

Subtypes	TP	FN	FP	TN	Sensitivity	Specificity	PPV	NPV	Youden_Index	F1_Score	Kappa
NF	441	164	29	462	72.9	94.1	93.8	73.8	0.670	0.820	0.653
Somatotroph*	215	0	0	881	100.0	100.0	100.0	100.0	1.000	1.000	1.000
Lactotroph#	93	2	76	925	97.9	92.4	55.0	99.8	0.903	0.705	0.668
Corticotroph	142	27	51	876	84.0	94.5	73.6	97.0	0.785	0.785	0.742
Thyrotroph△	12	0	4	1080	100.0	99.6	75.0	100.0	0.996	0.857	0.855
Gonadotroph	5	1	74	1016	83.3	93.2	6.3	99.9	0.765	0.118	0.109
Plurihormonal/Multihormonal	6	0	40	1050	100.0	96.3	13.0	100.0	0.963	0.231	0.223

TP, true positive; FP, false positive; TN, true negative; FN, false negative. The complete 95% confidence intervals for all metrics are provided in Supplementary Table S2 (for Table 3). Note that the sum of true positives and false negatives (n = 1108) exceeds the total number of patients (n = 1096) because patients with clinically functioning plurihormonal tumours contributed multiple codes, each of which was validated independently. * Including 6 cases of functioningplurihormonal PitNETs with a GH-positive component; ^# Including 5 cases of functioningplurihormonal PitNETs with a PRL-positive component; ^△ Including 1 case of functioningplurihormonal PitNETs with a TSH-positive component

Association between the updated WHO classification and the validity of ICD codes for identifying nonfunctioning and functioning PitNET subtypes

A substantial decrease in the sensitivity of ICD codes for nonfunctioning PitNETs was observed in association with the adoption of the WHO 2017 or 2022 classification, declining from 94.7% to 63.0% (p < 0.001). A corresponding reduction in NPV was also noted, from 88.8% to 72.1% (p < 0.001). In contrast, the sensitivity for identifying lactotroph PitNETs under the new classification was 97.9%; however, a pre-2017 comparison was not feasible due to the absence of true positive cases during that period. With respect to specificity, the use of the new classification was associated with a significant increase only for nonfunctioning PitNETs (95.8% vs. 86.8%, p < 0.01). Conversely, the specificity significantly decreased for several functioning PitNET subtypes according to the new classification: corticotroph PitNETs (100.0% vs. 92.8%, p < 0.001), gonadotroph PitNETs (100.0% vs. 90.9%, p < 0.001), and multihormonal PitNETs (99.3% vs. 95.3%, p < 0.01). Specificity remained consistently high (≥99.6%) for thyrotroph PitNETs under both classification systems. Marked reductions in PPV were observed under the new classification for corticotroph PitNETs (94.8% to 64.6%, p < 0.001) and lactotroph PitNETs (where the baseline PPV was undefined due to zero true positives, declining to 59.2% under the new classification). The PPV was also notably low for the gonadotroph (5.1%) and plurihormonal (13.6%) subtypes under the updated system. These changes in performance metrics were further reflected in composite measures. The Youden index, F1 score, and kappa statistic all substantially decreased for nonfunctioning PitNETs under the new classification. A concurrent, albeit less pronounced, reduction in these composite metrics was also observed for corticotroph PitNETs. Table 4 provides detailed information on the association between the updated WHO classification and the validity of ICD-10 codes in identifying the clinical endocrine function subtypes of PitNETs. The diagnostic trajectory changes in hormonal patterns from clinical diagnosis to the final ICD-10 code, as well as the mediating effect of pathological diagnosis between these stages via the pre-2017 WHO classification and the 2017/2022 WHO classification of PitNETs, are visualized via Sankey diagrams in Fig. 3.

Table 4.

Association between the updated WHO classification and the validity of ICD codes for identifying nonfunctioning and functioning PitNET subtypes

Endocrine function reflected by ICD codes	TP	FN	FP	TN	Sensitivity	Specificity	PPV	NPV	Youden Index	F1-score	Kappa
NF
2017/2022 WHO classification	262	154	17	383	63.0	95.8	93.9	72.1	0.587	0.754	0.583
Pre-2017 WHO classification	179	10	12	79	94.7***	86.8**	93.7	88.8***	0.815	0.942	0.820
Somatotroph*
2017/2022 WHO classification	187	0	0	629	100.0	100.0	100.0	100.0	1.000	1.000	1.000
Pre-2017 WHO classification	28	0	0	252	100.0	100.0	100.0	100.0	1.000	1.000	1.000
Lactotroph#
2017/2022 WHO classification	93	2	64	657	97.9	91.1	59.2	99.7	0.890	0.738	0.694
Pre-2017 WHO classification	0	0	12	268	/	95.7	/	100.0	/	/	0.000
Corticotroph
2017/2022 WHO classification	93	15	51	657	86.1	92.8	64.6	97.8	0.789	0.738	0.691
Pre-2017 WHO classification	49	12	0	219	80.3	100.0***	94.8***	94.8	0.803	0.891	0.865
Thyrotroph△
2017/2022 WHO classification	12	0	3	801	100.0	99.6	80.0	100.0	0.996	0.889	0.887
Pre-2017 WHO classification	0	0	1	279	/	99.6	/	100.0	/	/	0.000
Gonadotroph
2017/2022 WHO classification	4	0	74	738	100.0	90.9	5.1	100.0	0.909	0.098	0.089
Pre-2017 WHO classification	1	1	0	278	50.0	100.0***	100.0	99.6	0.500	0.667	0.665
Plurihormonal/Multihormonal
2017/2022 WHO classification	6	0	38	772	100.0	95.3	13.6	100.0	0.953	0.240	0.230
Pre-2017 WHO classification	0	0	2	278	/	99.3**	/	100.0	/	/	0.000

TP, true positive; FP, false positive; TN, true negative; FN, false negative. *p < 0.05; **p < 0.01; ***p < 0.001. Data are presented as point estimates. The complete 95% confidence intervals for all metrics are provided in Supplementary Table S3 (for Table 4). Note that the sum of true positives and false negatives (n = 1108) exceeds the total number of patients (n = 1096) because patients with clinically functioning plurihormonal tumours contributed multiple codes, each of which was validated independently. * Including 6 cases of functioning plurihormonal PitNETs with a GH-positive component; ^# Including 5 cases of functioning plurihormonal PitNETs with a PRL-positive component; ^△ Including 1 case of functioning plurihormonal PitNETs with a TSH-positive component

Fig. 3.

Sankey diagram illustrating the diagnostic trajectories of PitNET hormonal patterns across the clinical diagnosis, pathological classification, and ICD coding systems. Legend: left column: hormonal profiles derived from clinical discharge diagnoses; centre column: pathological classification according to the (a) pre-2017 who criteria or (b) integrated 2017/2022 who criteria; right column: final ICD-10 code assigned to each hormonal subtype. (a) The majority of clinical NF diagnoses (n = 189) were reclassified into specific subtypes pathologically (122 as LH/FSH; 54 as ACTH). However, ICD codes showed limited association with pathological findings, as most reclassified cases (e.g., 116 of 122 LH/FSH) retained the original NF code, aligning more closely with the initial clinical diagnosis. (b) Among 416 patients with clinically diagnosed NF pitNets, 154 (37%) showed discordant final ICD coding: LH/FSH (n = 69), ACTH (n = 48), PRL (n = 28), and others (n = 9). Separately, among 181 somatotroph pitNets, pathology identified plurihormonal expression in 92 cases, with 25% (n = 23) receiving MULTI-coding that diverged from the original GH diagnosis

Categories of coding errors in identifying patients with nonfunctioning and functioning PitNETs by ICD-10 codes

A comprehensive analysis of 227 patients with 467 coding errors leading to either false positive or false negative cases revealed four distinct categories of errors. The first category (12.8%, 95% CI: 10.0% − 16.2%; 60/467) comprised pure coding errors, primarily resulting from coders’ misinterpretation of hormone types released by PitNETs or the use of unspecified or ambiguous codes instead of specific codes. The second and most prevalent category (62.3%, 95% CI: 57.7% − 66.7%; 291/467) was related to the adoption of new descriptive terminology in the updated WHO classification of PitNETs used in pathological reports. The third category (22.7%, 95% CI: 19.0% − 26.8%; 106/467) consisted of false-positive errors predominantly observed in functioning lactotroph PitNETs, which occurred when coders mistakenly assigned endocrine function ICD codes to indicate hyperprolactinemia caused by pituitary stalk compression effects rather than true prolactinomas. The fourth category (2.1%, 95% CI: 1.1% − 3.9%; 10/467) resulted from the combined effects of hyperprolactinemia due to pituitary stalk compression and the descriptive terminology of the new WHO classification of PitNETs in pathological reports, primarily resulting in false-positive cases of pituitary plurihormonal PitNETs. The pathways leading to these coding inaccuracies are schematically summarized in Fig. 4. More details of the causes of coding errors are provided in Table 5.

Fig. 4.

Pathways and mechanisms of ICD-10 coding errors in PitNETs

Table 5.

Descriptions of causes of coding errors in identifying nonfunctioning and functioning PitNETs subtypes based on the corresponding ICD code combinations

Subtypes	ICD code combinations	Type of inconsistency	Frequency	Causes
NF	D35.202 or (D35.2 without any secreting function related E codes)	False-positive	29	Pure coding error (n = 29)
		False-negative	164	(1) The endocrine function ICD codes used in these cases reflected hyperprolactinemia caused by pituitary stalk compression effects rather than true prolactinomas (n = 45) (2) Influenced by the descriptive terminology of the new WHO classification of PitNETs used in pathological reports (n = 119)
Thyrotroph	E05.8	False-positive	4	(1) Pure coding error (n = 1) (2) Influenced by the descriptive terminology of the new WHO classification of PitNETs used in pathological reports (n = 2) (3) Influenced by the combined effects of pituitary stalk compression and incorrect pathological interpretation and misclassified into the pituitary plurihormonal PitNETs group (n = 1)
Lactotroph	E22.1	False-positive	76	(1) The endocrine function ICD codes used in these cases reflected hyperprolactinemia caused by pituitary stalk compression effects rather than true prolactinomas (n = 52) (2) Influenced by the descriptive terminology of the new WHO classification of PitNETs used in pathological reports (n = 24)
		False-negative	2	Pure coding error (n = 2)
Corticotroph	E24.0	False-positive	51	Influenced by the descriptive terminology used in pathological reports (n = 51)
		False-negative	27	Pure coding error (n = 27)
Gonadotroph	E22.8	False-positive	74	(1) Influenced by the descriptive terminology of the new WHO classification of PitNETs used in pathological reports (n = 70) (2) Influenced by the combined effects of pituitary stalk compression and incorrect pathological interpretation and misclassified into the pituitary plurihormonal PitNETs group (n = 4)
		False-negative	1	Pure coding error (n = 1)
Plurihormonal/Multihormonal	(D35.2 and E22.801) or (D35.2 and any two or more of the above endocrine function codes)	False-positive	40	(1) Influenced by the descriptive terminology of the new WHO classification of PitNETs used in pathological reports (n = 26) (2) The endocrine function ICD codes used in these cases reflected hyperprolactinemia caused by pituitary stalk compression effects rather than true prolactinomas (n = 9) (3) Influenced by the combined effects of pituitary stalk compression and incorrect pathological interpretation and misclassified into the pituitary plurihormonal adenoma group (n = 5)

Discussion

The discharge diagnosis of PitNETs constitutes a comprehensive clinical synthesis, integrating manifestations, imaging features, biochemical markers, and intraoperative findings. Notably, the pathological classification systems (WHO 2017/2022) are intended for diagnostic and prognostic purposes and should not be directly used to assign clinical ICD codes—a distinction that appears central to certain coding discrepancies observed in this study. Owing to inherent delays in pathological processing, pathological confirmation is generally unavailable at the time of discharge. Therefore, discharge diagnoses and their corresponding ICD codes must be based primarily on the clinical syndrome and biochemical evidence available at that time rather than on subsequent pathological findings. Accordingly, clinical features documented at discharge—including endocrine function subtype, tumour size, and invasiveness—are routinely translated into specific ICD codes for standardized data management, whereas detailed immunohistochemical results from pathology reports should not serve as the primary basis for code assignment. Among these features, the hormonal hypersecretion profile represents a cornerstone for clinical coding.

Previous studies have examined the accuracy of ICD codes for specific pituitary adenoma subtypes, such as acromegaly [13, 14] and Cushing’s disease [31]. However, a comprehensive evaluation of coding accuracy for endocrine function subtypes across diverse healthcare databases remains limited. Our findings reveal previously underrecognized limitations in the validity of ICD-10 codes for identifying both nonfunctioning and functioning PitNETs. A key observation from our analysis is that a notable source of coding discrepancy arises when the updated WHO pathological classification is inappropriately applied to clinical ICD coding. In particular, we identified a reduction in coding accuracy in cases where the 2017 or 2022 WHO classification was used, highlighting the systemic challenges that can occur when pathological nomenclature is conflated with clinical coding practices. These results underscore the need for clearer guidelines that distinguish pathological reporting from clinical code assignment to improve the accuracy of data captured in administrative databases.

Our study revealed relatively low PPVs across different PitNET subtypes, including gonadotroph, plurihormonal, lactotroph, corticotroph, and thyrotroph PitNETs. This pattern was associated with the widespread use of endocrine disease (E) codes in clinical coding practices. A striking example is the case of a functioning lactotroph PitNET, where a substantial proportion of hyperprolactinemia cases are, in fact, driven by the pituitary stalk effect—a mechanism in which the mass effect of the tumour disrupts the hypothalamic inhibitory pathway rather than the effect being caused by autonomous hormone secretion from neoplastic cells. This pathophysiological mechanism underscores the potential for false-positive diagnoses when hyperprolactinemia (E22.1) is used as a diagnostic marker for functioning lactotroph PitNETs, highlighting inherent limitations in the use of the ICD-10 classification system for accurate characterization of these tumours. Therefore, precise ICD codes are needed to determine whether hyperprolactinemia is caused by a lactotroph PitNET or results from the pituitary stalk effect associated with non-prolactin-secreting adenomas. Without such a distinction, the use of ICD codes may lead to an overestimation of the prevalence of lactotroph PitNETs in hospital administrative databases.

Our findings indicated that the introduction of the 2017/2022 WHO classifications in pathological reporting was associated with reduced sensitivity of ICD-10 codes in identifying clinically nonfunctioning PitNETs. In addition, the use of these updated classifications correlated with significantly decreased specificity across multiple functioning PitNET subtypes, such as lactotrophs, corticotrophs, gonadotrophs, and plurihormonal PitNETs. The revised WHO system emphasizes detailed pathological characterization based on cellular lineage and immunohistochemical profiles rather than clinical endocrine activity. This shift appears to have introduced challenges in coding practice: when ICD-10 codes are revised based on pathological reports, coders may misclassify histologically defined tumours in a quiescent phase as clinically functioning neoplasms [32]. Such misclassifications have been linked to increased inaccuracies in coding hyperfunctioning pituitary states, thereby affecting the validity of the ICD-based representation of both nonfunctioning and functioning pituitary tumours. These observations suggest that morphology-based coding alone is unlikely to constitute an effective identification strategy for lactotroph PitNETs. In contrast, before the widespread adoption of the 2017/2022 WHO classifications, pathological diagnoses were often limited to the nonspecific term “pituitary adenoma,” with immunohistochemistry findings reported without further subtyping—as seen, for instance, in the case of gonadotroph PitNETs. In that context, coding was more straightforward, as coders could rely on discharge diagnoses with little ambiguity. To reduce the misapplication of E codes, coders should receive training in the key clinicopathological distinctions outlined in the latest WHO classification, with emphasis on differentiating pathological lineage from clinically relevant endocrine function to improve coding accuracy and clinical utility.

Within the ICD-10 coding framework, the classification of pituitary tumours relies predominantly on tumour location codes to denote biological behaviour (e.g., benign, malignant, or atypical/borderline malignant) and endocrine chapter codes to reflect clinical endocrine function. While the ICD-11 framework has an integrated ICD-O (International Classification of Diseases for Oncology) to enable more precise coding of cellular pathological morphology for specific tumours, a comprehensive system for capturing pathological tumour lineages and molecular subtypes is still lacking. Notably, even after the 2022 revision of the WHO classification for pituitary tumours, the ICD-11 framework continues to rely on stem codes for nonfunctioning PitNETs and broad categories for functioning PitNETs [16]. The newly introduced lineage-based classification and terminology for PitNETs have yet to be directly incorporated into this update, likely owing to ongoing significant controversies surrounding the current nomenclature and classification of these tumours [6, 10].

Consequently, pituitary tumours are currently characterized through three distinct yet interrelated frameworks: ICD coding systems, clinical diagnostic evaluations, and pathological assessments. Although these classification levels exhibit partial overlap, they remain operationally independent. The ICD coding system is fundamentally designed to prioritize epidemiological surveillance by meeting statistical requirements for morbidity and mortality tracking. While this system effectively addresses the clinical characterization of tumour endocrine function, it fails to comprehensively capture complex transcriptional profiles at the molecular level. This limitation results in inadequate representation of molecular-level typing, which is critical for a deeper understanding of disease biology and behaviour.

To address these challenges, a multidimensional coding strategy that systematically integrates tumour localization, lineage-based classification, and clinical endocrine function must be rigorously incorporated into the updated ICD-11 framework. This approach would ensure more accurate and clinically relevant classification. We strongly advocate for the active involvement of pathologists, clinicians, and medical coders in the development of standardized nomenclature and classification systems, leveraging their complementary expertise. Such interdisciplinary collaboration should focus on refining current classification frameworks to ensure that international disease coding systems accurately reflect both the essential biological features and the clinical endocrine function of pituitary tumours. This integrated approach would better serve the diverse needs of healthcare practitioners, researchers, and public health professionals, ultimately enhancing disease surveillance and clinical management.

By identifying pitfalls in current coding practices, this study provides actionable recommendations for improving the ICD-11 framework. These recommendations include advocating for multidimensional coding frameworks that integrate molecular, histological, and endocrine function data—a critical step towards personalized medicine in pituitary tumour management. Our study emphasizes the need for collaborative efforts among clinicians, pathologists, and coders to align diagnostic, pathological, and coding terminologies. This approach addresses the operational independence of current classification systems and proposes strategies to harmonize clinical documentation with evolving WHO standards. This work fills a critical evidence gap as one of the first large-scale evaluations of ICD-10 code accuracy in the context of WHO updates. The findings of this study directly inform efforts to standardize coding protocols, reduce biases in administrative data, and optimize the transition to the ICD-11 for pituitary tumour management.

Limitations

This study has several limitations. First, as the data were derived from a single tertiary medical centre, the generalizability of our findings to other healthcare settings may be limited. Second, although the overall sample size was substantial, certain rare subtypes of PitNETs—such as gonadotroph and thyrotroph tumours—had relatively low case numbers, which may affect the stability of the estimates in the subgroup analyses. Moreover, the study was confined to patients who underwent surgical resection and had pathological confirmation; those managed conservatively or in outpatient settings were excluded, which may further restrict the external validity of the results. Another important limitation relates to institutional coding protocols. At our centre, the initial ICD-10 codes assigned at discharge based on clinical diagnosis are systematically replaced after pathological reports become available, and the original codes are not retained. As a result, a direct paired comparison of pre- and postpathology coding was not feasible. Although our group-level analysis evaluated the association between WHO classification updates and coding accuracy, it could not assess individual code-level modifications. Finally, the phased introduction of the updated WHO classifications during the study period raises the possibility of temporal confounding, such as concurrent changes in the electronic medical records system or coding personnel. However, we confirmed that no major EMR updates or turnover among core coding staff occurred during the study interval. While this stability supports an association between the adoption of the new pathological criteria and the observed shifts in coding accuracy, the influence of other unmeasured time-related factors cannot be entirely ruled out.

Conclusion

This study, which is based on a surgical cohort, highlights challenges in the alignment between the updated WHO classification frameworks and the accuracy of ICD-10 coding for surgically managed pituitary neuroendocrine tumours (PitNETs). While the revised classifications have improved pathological characterization, their introduction has been associated with reduced sensitivity in coding for nonfunctioning PitNETs and decreased specificity for several functioning subtypes within this surgical population. These observations underscore a critical difference between pathology reporting systems and administrative coding practices. Importantly, the WHO classifications (2022 and 2017) are intended for pathological diagnosis and should not be directly equated with clinical patient classification using ICD codes—particularly given the increased coding discrepancies observed following the 2022 update. Moving forwards, efforts to improve coding accuracy should focus on multidisciplinary collaboration, enhanced training for coders, and the development of integrated coding systems that better reconcile pathological definitions with clinical documentation. Such steps are essential for refining disease classification and supporting the reliable use of coded data in both clinical practice and research.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

PA

Pituitary adenoma

PitNET

Pituitary neuroendocrine tumour

NPV

Negative predictive value

PPV

Positive predictive value

EMR

Electronic medical record

NF

Nonfunctioning

PIT1

Pituitary-specific positive transcription factor 1

TPIT

T-box transcription factor

SF1

Steroidogenic factor 1

PRL

Prolactin

GH

Growth hormone

ACTH

Adrenocorticotropic hormone

TSH

Thyroid-stimulating hormone

LH

Luteinizing hormone

FSH

Follicle-stimulating hormone

MULTI

Multihormone

IGF-1

Insulin-like Growth Factor-1

WHO

World Health Organization

ICD-10

International Classification of Diseases, 10th Revision

ICD-11

International Classification of Diseases, 11th Revision

ICD-O

International Classification of Diseases for Oncology

Author contributions

J. Z. and X. G. were responsible for writing the paper and conducting the statistical analysis. X. G. and L. D. contributed to the adjudication process of the clinical endocrine function of PitNETs by reviewing the original medical records. J. Z. and A. L. retrieved and acquired the data. X. M. was responsible for delivering the pathological report and interpreting the results. N.L. and B.X. critically revised the entire manuscript and formulated the problem. All the authors participated in the drafting of the manuscript and approved the final version.

Funding

This work was supported by the National High Level Hospital Clinical Research Funding (2025-PUMCH-C-006).

Data availability

We have deposited the anonymized data and reproducible scripts in the Mendeley Data repository at doi: https://doi.org/10.17632/mynjx8hjvt0.1 (Zhou, Jingya, 2025, “The Three-Dimensional Hormonal Profile in Patients with PitNETs”, Mendeley Data, V1).

Declarations

Ethical approval and consent to participate

Our work was approved by the institutional ethics committee of PUMCH (No. K8661) and adhered to the 1964 Declaration of Helsinki and its subsequent amendments or comparable ethical standards. Given that this was a retrospective study, written informed consent was waived by the institutional ethics committee of PUMCH.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.