Evaluation of spinal curvature in the sagittal plane plays a central role in the diagnosis and management of spinal diseases.⁽¹⁾ Numerous studies have demonstrated an association between sagittal alignment and favorable clinical outcomes.⁽²⁾ Traditionally, sagittal spinal curves are assessed using angular measurements obtained with the Cobb method and anatomical references based on the vertebral segment evaluated—cervical, thoracic, or lumbar. Thoracic kyphosis is measured as the angle between the superior endplate of T1 and the inferior endplate of T12, whereas lumbar lordosis is commonly measured as the angle between L1 and S1.^(1-4)

However, these anatomical landmarks do not always reflect the true biomechanical transition of spinal curves. To address this limitation, Berthonnaud proposed a functional assessment of sagittal spinal curvature that prioritizes the inflection point—defined as the segment where the spinal curve changes direction—rather than fixed anatomical landmarks. Recent studies have shown that the inflection point rarely coincides with classical anatomical segments and may influence how spinal curves are classified and interpreted in clinical practice.⁵ This functional approach may have important implications for surgical planning, assessment of spinal balance, and prognosis, because improper segmentation may increase the risk of postoperative complications and unsatisfactory clinical outcomes.⁽⁷⁾

Despite its theoretical advantages, the use of the inflection point depends on adequate reproducibility and reliability among different observers. Previous studies have reported varying levels of agreement, often in populations that included more specialized raters and larger samples.⁶ However, limited data exist on determination of the spinal inflection point in routine clinical settings, particularly when evaluations involve non-specialist raters.

To assess the intra- and inter-observer reliability of spinal inflection point localization on lateral radiographs of asymptomatic adults using a standardized measurement protocol, given its important role in sagittal alignment analysis and its potential influence on biomechanical interpretation and surgical planning.

A total of 10 adults (five males and five females; age range 32–50 years) met the inclusion criteria and were evaluated (Table 1). All participants were asymptomatic and had no history of spinal disease, congenital anomaly, deformity, or prior surgery.

Table 1

Patient	Sex	Age (years)
1	M	32
2	F	34
3	M	36
4	F	37
5	M	39
6	F	42
7	M	44
8	F	45
9	M	48
10	F	50

Total n=10 (five males and five females). Age range 32–50 years.

Each of the three observers identified the cervicothoracic and thoracolumbar inflection points in three separate rounds, resulting in a total of 90 assessments (10 patients × 3 observers × 3 rounds).

The pooled distribution of inflection point determinations at both transitions, including all observers and rounds, appears in table 2. The segment C7–T1 represented the most frequently identified cervicothoracic inflection point, whereas T12–L1 represented the most common thoracolumbar inflection point.

Table 2

Transition	Segment	Frequency (n)	Percentage (%)
Cervicothoracic	C6–C7	15	16.7
	C7–T1	28	31.1
	T1–T2	27	30.0
	T2–T3	7	7.8
	C5–C6	3	3.3
	C5–C7	2	2.2
	C4–C5	1	1.1
Thoracolumbar	T10–T11	3	3.3
	T11–T12	19	21.1
	T12–L1	36	40.0
	L1–L2	10	11.1
	L2–L3	4	4.4
	T9–T10	1	1.1
	T2–T11	1	1.1
	T10–T11	1	1.1
	T10–T11	1	1.1
	T10–T11	1	1.1

Pairwise inter-observer agreement for localization of the inflection point appears in table 3. For the cervicothoracic transition, kappa coefficients ranged from 0.19 (95%CI= –0.15 to 0.55) to 0.41 (95%CI= 0.05 to 0.75), indicating slight to moderate agreement. For the thoracolumbar transition, agreement was slight or absent, with kappa values ranging from –0.03 (95%CI= –0.20 to 0.14) to 0.22 (95%CI= 0.01 to 0.44). 95%CIs were wide for all comparisons, reflecting considerable variability and imprecision among observers.

Table 3

Pair of examiners	Transition	Kappa	95% confidence interval	Strength of agreement
1 versus 2	Cervicothoracic	0.41	0.05 to 0.75	Moderate
1 versus 3	Cervicothoracic	0.19	–0.15 to 0.55	Slight
2 versus 3	Cervicothoracic	0.22	–0.06 to 0.50	Fair
1 versus 2	Thoracolumbar	–0.03	–0.20 to 0.14	None
1 versus 3	Thoracolumbar	0.22	0.01 to 0.44	Fair
2 versus 3	Thoracolumbar	0.16	–0.03 to 0.34	Slight

Intra-observer reliability, calculated using the ICC, appears in table 4. For both transitions, ICC values for single measures were low (cervicothoracic: 0.30, 95%CI= 0.07–0.54; thoracolumbar: 0.29, 95%CI= 0.08–0.53). When the mean of the measurements was considered, ICC increased to 0.56 for both transitions, indicating moderate intra-observer consistency.

Table 4

Transition	ICC (single measure)	95% CI	ICC (mean of measures)	95% CI
Cervicothoracic	0.30	0.07–0.54	0.56	0.19–0.77
Thoracolumbar	0.29	0.08–0.53	0.56	0.22–0.77

95%CI: 95% confidence interval.

Repeatability of inflection point localization for the cervicothoracic transition, stratified by patient and considering all raters and all rounds, appears in table 5. Percent agreement ranged from 66.6% to 100% across patients. Similar patterns were observed for the thoracolumbar transition and are summarized textually.

Table 5

Patient	Agreement across rounds (%)	Most frequent segment
1	66.7	T1–T2/C7–T1
2	66.7	C7–T1
3	66.7	C7–T1/C6–C7
4	100	C7–T1
5	100	T1–T2
6	66.7	T1–T2/C7–T1
7	100	C6–C7/C7–T1
8	66.7	C7–T1
9	66.7	C6–C7/C7–T1
10	100	C6–C7

Overall, both inter- and intra-observer agreement for inflection point localization ranged from low to moderate, with most discordance occurring near anatomical transition segments. The findings indicate substantial variation both between observers and within individual observers, regardless of the anatomical region evaluated. Overall, the consistently low agreement across both spinal transitions highlights the limited reliability of inflection point localization using the current methodology.

This study demonstrated low intra- and inter-observer agreement in the determination of inflection points at both the cervicothoracic and thoracolumbar transitions, supporting prior evidence that these functional landmarks often do not align with classical anatomical references.⁶ This pattern of divergence highlights the challenges associated with applying functional segmentation of the spine in the clinical assessment of sagittal balance and disease management.^(11,12)

Our data identified C7–T1 as the most frequent inflection point at the cervicothoracic transition and T12–L1 at the thoracolumbar transition. Malka et al.⁽¹⁰⁾ similarly reported the highest frequency of inflection points at C7–T1, although the distribution ranged from T8 to L4, with 44% of cases located between C6–C7 and C7–T1. Park et al.⁽¹¹⁾ observed a predilection for L2 and T1 levels, with age-associated distalization, whereas Vacari et al.⁽⁶⁾ identified T12–L1, particularly T12, as the most common thoracolumbar inflection point. Collectively, these studies and the present results indicate that inflection points rarely coincide with fixed anatomical landmarks.^(13,14)

The shift from anatomical to functional spinal segmentation arises from the clinical importance of sagittal alignment parameters. As introduced by Berthonnaud et al.,⁽⁴⁾ functional segmentation defines lordosis as spinal segments in extension, kyphosis as segments in flexion, and the inflection point as the transition between these regions. Subsequent studies have reinforced this concept by demonstrating associations with overall sagittal alignment. This approach parallels determination of terminal vertebrae in the coronal plane, a method frequently used in planning deformity surgery.

Despite this conceptual advancement, the present findings, along with those of previous studies, consistently demonstrate limited reproducibility and precision in functional inflection point determination, particularly when performed by non-specialist raters with only basic radiographic training. The wide 95%CIs observed for kappa and ICC in this study further highlight substantial inter- and intra-observer variability, which represents a significant barrier to clinical adoption. Although the angular measurements themselves are not inherently complex, consistent identification of the inflection point may require not only clearly defined protocols but also improved training and possibly technological support.

Limitations of this study include the small sample size, the single-center convenience sample, and the limited experience of the observers. However, these characteristics reflect real-world settings in which such measurements are often performed by residents or physicians in training, thereby highlighting the practical challenges of incorporating functional segmentation into routine clinical workflows. Additional limitations include the absence of expert raters for comparison and the lack of correlation with clinical outcomes. Nonetheless, these factors represent relevant clinical scenarios and further emphasize the need for structured training and standardized protocols.

Future studies should expand these findings by including larger multicenter cohorts, incorporating experienced spine specialists as comparators, and exploring clinical correlations between inflection point determination and patient-reported outcomes. The application of automated image analysis and artificial intelligence, as suggested in recent literature, may further improve reproducibility and reduce observer dependence, thereby providing a more robust framework for integrating functional segmentation into both clinical care and research.

From a clinical perspective, variability in inflection point localization may directly influence sagittal alignment assessment and the interpretation of spinal biomechanics. Inconsistent identification of transition zones could impact surgical planning, particularly in defining fusion levels and alignment targets, potentially affecting postoperative outcomes. Therefore, improving the reliability of this parameter is essential before its incorporation into routine.

In summary, the results emphasize the discordance between anatomical and functional segmentation in inflection point determination and highlight the need for standardized, training-oriented, and potentially technology-assisted protocols before functional segmentation can be reliably incorporated into clinical or research workflows.

Despite limitations related to sample size and observer experience, this study demonstrated poor agreement between anatomical and functional (dynamic) references in the determination of the spinal inflection point. In addition, intra- and inter-observer agreement for inflection point localization remained consistently low across methods and spinal transitions. These findings highlight the need for more standardized protocols and advanced measurement strategies before routine clinical or research use of functional spinal segmentation can be recommended. Reliable identification of the spinal inflection point is clinically relevant, as it may impact sagittal alignment assessment, biomechanical interpretation and surgical decision.