Cephalometric measurement error: interpreting ANB and Wits under uncertainty

Simplified Orthodontics Academy · Draft prepared for review 30 October 2026

Measurement error is part of the result

A cephalometric number is not a direct readout of anatomy. It is the end of a chain: an image is acquired, structures are represented on that image, landmarks are placed, lines or planes are constructed, and a calculation is made. Each link can add variation. Projection, contrast, superimposition, head position, magnification, examiner decisions and software settings can all affect the final value. The existing cephalometric landmarks guide explains what the named points and reference lines mean; this article addresses the uncertainty that remains after those definitions are understood.

That distinction matters when two tracings differ. A change on the screen may reflect a different point placement or a different construction, rather than a change in anatomy. The relevant question is not simply whether a value moved, but whether the movement is larger than the uncertainty expected from that particular method, image and workflow.

Repeatability and reproducibility are different questions

Repeatability describes consistency under closely matched conditions: the same examiner, method, image type, software and protocol, repeated over a short interval. It is an answer to “Can this workflow produce a similar number again?” Reproducibility is broader. It asks whether the result remains comparable when an examiner changes, when a method is compared with another method, or when the protocol is implemented in another setting. A method can be highly repeatable and still disagree systematically with another method.

Trpkova and colleagues’ meta-analysis of 15 landmark locations reported standard mean errors and confidence intervals for repeatability and reproducibility. The important lesson is not one universal error value. It is that error differs by landmark and by the question being asked. Within-method stability should therefore not be used as proof that a value is interchangeable with a value obtained by a different tracing process [1].

Tsolakis and colleagues illustrate the same issue in a comparison of automated and manual tracing. In their 100 lateral cephalograms, within-method ICCs were above 0.97, while between-method ICCs ranged from 0.70 to 0.92. ANB had no significant mean difference under that study’s adjusted criterion, but other measurements did differ. Those findings belong to the tested images, software and protocol; they do not establish a universal agreement level for every system [4].

Why image and operator effects remain important

Image quality changes how confidently a landmark can be located. Blur, noise, overlap and positioning can make a boundary broad rather than point-like. The operational definition also matters: two observers may use the same label while applying slightly different rules to a difficult structure. A CBCT study reported that disagreement was partly related to differing definitions of Condylion and Gonion, and noted that there is no definitive in-vivo gold standard for the exact landmark position [2].

Automation changes who or what places the initial point, but it does not remove the need to inspect the tracing. Jung and colleagues studied one Invivo7 CBCT system and 76 volumes. Their automated landmark mean absolute error was 0.94 ± 0.99 mm; 87% of landmark MAEs were below 2 mm; linear measurement MAE was 1.35 ± 1.33 mm and angular MAE was 0.89 ± 0.89°. These are performance results for one system, sample and protocol. They are not acceptance limits, treatment thresholds or evidence that another system will behave the same way [2]. An umbrella review likewise criticised using 2 mm as a universal clinical cutoff and concluded that landmark accuracy differs by landmark, with expert supervision still needed [3].

ANB: a small point change can alter the angle

ANB is a geometric relationship built from the directions of lines joining the relevant landmarks. If one of those points is placed differently, the direction of a line changes. The size and direction of the angular effect depend on where the displacement occurs relative to the other points and on the geometry of the image. The same apparent numerical movement can therefore have different explanations in different tracings.

This is why the definition of a measurement and its uncertainty should be kept separate. The companion ANB and Wits appraisal explainer covers what the two appraisals measure and why they can disagree. Here, the narrower point is that even a correctly defined measurement can vary when the points or construction are not located identically. A difference should be described as compatible with a change only after the tracing conditions and method error have been examined; it should not be treated as standalone proof of a biological or diagnostic change.

Why Wits has geometric sensitivity

Wits introduces an additional constructed object: the selected occlusal plane. The projected positions of A and B depend on that plane’s orientation as well as on the locations of A and B. If the plane is drawn through a different set of contacts, or its endpoints are judged differently, the projected AO–BO distance can change even when the underlying jaw anatomy has not.

Haynes and Chau’s study specifically identified construction of the functional occlusal plane as an error source. Repeatability within observers was more satisfactory than reproducibility between observers, and the highest single differences between tracing series were 3.1 mm for one observer and 5.8 mm for another. These are findings from that study, not clinical limits or advice about what amount of change should be accepted [5].

Del Santo examined 122 finished orthodontic cases and reported a tendency for ANB and Wits to be inconsistent more often in the group with a high occlusal-plane angle. Again, this is a study-specific association, not a diagnostic rule. It reinforces the need to record how the plane was constructed and to interpret a result in the context of the method that produced it [6].

A method-aware verification sequence

1. Check that the images being compared are comparable in positioning, magnification, exposure and visible anatomy.
2. Review the operational definitions and placement of the relevant landmarks. If a point is unclear, repeat the placement rather than implying false precision.
3. For Wits, document which teeth or contacts defined the occlusal plane and whether that construction can be reproduced.
4. Repeat the tracing with the same settings, or have a calibrated second examiner repeat it, while preserving the method details.
5. Report the observed difference alongside the method used to estimate error or agreement. Do not substitute a remembered universal cutoff.
6. Use the measurement with the rest of the available record. One borderline number should not carry the whole interpretation.

If a difference is not clearly larger than the uncertainty of the method, it is not independent evidence of biological change. It may be compatible with change, but the claim needs other supporting information.

Limits of the evidence

The evidence supports a general principle rather than one universal number: cephalometric measurements include landmark, image, operator and geometric uncertainty. Error estimates vary with the landmark, radiographic quality, population, software, examiner and tracing protocol. The reported 0.94 mm, 0.89°, 3.1 mm and 5.8 mm figures above should be read as study-specific results. They must not be converted into clinical thresholds, diagnosis or treatment advice. A consistent workflow can make measurements more auditable, but it cannot make uncertainty disappear.

Educational content for clinicians and students. This draft is prepared for review by Dr Salama before publication. It does not provide diagnosis, treatment, prognosis or patient-specific advice.

References

1. Trpkova B, Major P, Prasad N, Nebbe B. Cephalometric landmarks identification and reproducibility: a meta-analysis. Am J Orthod Dentofacial Orthop. 1997;112(2):165-70. PubMed
2. Jung YE, Suh H, Park J, Oh H. Accuracy and reliability of automated landmark identification and cephalometric measurements on cone beam computed tomography using Invivo software. Angle Orthod. 2025;95(4):362-370. doi: 10.2319/122324-1049.1
3. Polizzi A, Leonardi R. Automatic cephalometric landmark identification with artificial intelligence: an umbrella review of systematic reviews. J Dent. 2024;146:105056. doi: 10.1016/j.jdent.2024.105056
4. Tsolakis IA et al. Comparing a Fully Automated Cephalometric Tracing Method to a Manual Tracing Method for Orthodontic Diagnosis. J Clin Med. 2022;11(22):6854. doi: 10.3390/jcm11226854
5. Haynes S, Chau MN. The reproducibility and repeatability of the Wits analysis. Am J Orthod Dentofacial Orthop. 1995. doi: 10.1016/S0889-5406(95)70108-7
6. Del Santo M. Influence of occlusal plane inclination on ANB and Wits assessments of anteroposterior jaw relationships. Am J Orthod Dentofacial Orthop. 2006. doi: 10.1016/j.ajodo.2005.09.025

More to read

• Cephalometric landmarks explainedDefinitions and reference lines
• ANB and Wits appraisal explainedWhat the two measurements describe
• All articlesOrthodontic analyses and methods
• خطأ القياس السيفالومتريThe same article in Arabic
• Arabic article indexالمقالات العربية

This article is for education. Clinical decisions remain the responsibility of the treating clinician.