Low measurement reliability of emergency department radiographs of patellar dislocation

Primary patellar dislocation most commonly occurs in adolescents and young adults, with an incidence estimated at 29-43 per 100,000 person-years, particularly affecting adolescents [1].

Following acute dislocation, ED X-rays are routinely obtained to evaluate predisposing anatomical factors. Among the most critical markers associated with recurrent patellar instability are trochlear dysplasia and patella alta [2, 3]. Trochlear dysplasia is characterised by a shallow or flattened trochlear groove that reduces osseous containment and facilitates lateral patellar displacement [2]. Patella alta refers to an abnormally high patella, often due to patellar tendon elongation, which delays patellofemoral engagement during flexion and increases the risk of instability [3, 4].

Patellar height is commonly assessed using the Insall-Salvati [5], Blackburne-Peel [6] and Caton-Deschamps [7] indices on lateral knee X-ray. These measurements are used to identify patellofemoral malalignment [3]. However, their reproducibility has been questioned due to variability in anatomical landmarks and X-ray positioning [5]. The Dejour classification depends on subtle X-ray features and may show limited reproducibility [2, 8].

These X-ray measurements were developed and validated on true lateral X-rays, where femoral condyles are superimposed. However, in routine clinical practice, particularly in the ED, X-rays are often obtained under suboptimal conditions without strict rotational alignment control. Even minor deviations from a true lateral projection may alter the appearance of anatomical landmarks and affect measurement accuracy.

Despite the clinical relevance of these measurements, there is limited evidence evaluating their reproducibility in patients with primary patellar dislocation. Previous studies have reported limitations of X-ray assessment of patella height and trochlear dysplasia, largely based on standardised imaging protocols or specialised clinical settings [2, 5, 8]. In contrast, early evaluation after primary patellar dislocation is often based on routine ED X-rays, where data on measurement reproducibility remain limited.

This study aimed to assess the intra- and inter-rater reproducibility of commonly used X-ray measurements for trochlear dysplasia and patellar height in patients with primary patellar dislocation when applied to routine ED knee X-rays with non-true lateral positioning.

Methods

Design

This single-centre reliability study of X-ray measurements in adult patients with primary patellar dislocation was conducted at the Department of Orthopaedic Surgery and Traumatology, Odense University Hospital, Denmark. The Danish Data Agency approved the study (case number 14/15156), and reporting followed the Guidelines for Reporting Reliability and Agreement Studies [9]. All procedures involving human participants were conducted in accordance with the ethical standards of the Danish Data Protection Agency and with the 1964 Helsinki Declaration and its subsequent amendments. As this was a retrospective analysis of anonymised X-rays, informed consent was waived by the regional ethics committee.

Patients

A retrospective search of the hospital administrative system was performed for patients with patellar dislocation treated in the ED at Odense University Hospital using the International Statistical Classification of Diseases and Related Health Problems, tenth revision (ICD-10) diagnosis code for patellar dislocation (DS830) over a three-year period. Health records and X-rays were reviewed for inclusion. A total of 710 patients with patellar dislocations were identified; 420 patients were excluded based on health records, primarily due to secondary dislocation (n = 318), miscoding (n = 87), insufficient information (n = 9), or foreign residency (n = 6). Of the remaining 290 patients with primary patellar dislocation, 176 were excluded because no X-ray was available (n = 82) or the lateral radiograph was too oblique for assessment (n = 94). The final cohort consisted of 114 patients (Figure 1).

Measurements

All X-rays were obtained as part of routine ED evaluation following acute patellar dislocation. Standard anteroposterior and lateral knee X-rays were acquired according to local protocols. These protocols do not include fluoroscopic guidance or dedicated patellofemoral positioning to ensure strict rotational alignment. Consequently, greater variation in knee rotation may occur compared with standardised patellofemoral imaging protocols used in specialised outpatient settings.

The measurements were performed by four raters at different stages of training: one medical student, two residents and one consultant in knee arthroscopy. Each performed two measurements ≥ 14 days apart, and the measurements were conducted independently by the raters. The raters were aware that their measurements would be compared with those of the other raters. All measurements were assessed in the Sectra Web Picture Archiving and Communication System (version 10-02-005).

Trochlear dysplasia was assessed by the Dejour classification [10], Anterior translation [11] and Trochlear Depth [12]. The Dejour classification [10] was assessed on lateral examination of the X-ray and categorised into Grade A (crossing sign), Grade B (A + spur), Grade C (A + double contour) and Grade D (A + B + C) (Figure 2). Anterior translation and Trochlear depth were dichotomised as dysplasia present or absent to reflect their potential clinical use as screening measurements in the ED.

Anterior translation [11] was assessed by depicting a line along the anterior femur. If the trochlea crossed that line, anterior translation [11] was positive (Figure 3 A).

Trochlear depth [12] was evaluated by placing a line (1) along the posterior femur, then placing a line (2) perpendicular to line 1, running to the anterior-proximal part of the trochlea. A final line (3) was then placed 15 degrees inferior to line 2, and trochlear depth was measured along line 3 (Figure 3 B). Trochlear dysplasia was defined as a depth of less than four millimetres.

Patella height was assessed by Insall-Salvati [13], Blackburne-Peel [6] and Caton-Deschamps [7]. The Insall-Salvati [13] ratio was measured as the length of the patellar tendon divided by the length of the patella (Figure 3 C). The Blackburne-Peel index [6] was measured as the distance from a horizontal line drawn at the level of the proximal tibial articular surface to the inferior aspect of the articular surface of the patella divided by the length of the articular surface of the patella (Figure 3 C).

The Caton-Deschamps index was measured as the distance from the inferior border of the patellar articular surface to the anterior edge of the tibial plateau divided by the length of the articular surface of the patella (Figure 3 C).

Patella alta was defined as greater than 1.2 for the Insall-Salvati ratio and Caton-Deschamps index, and greater than 1.0 for the Blackburne-Peel index. Patella baja was defined as less than 0.8 for the Insall-Salvati ratio, less than 0.5 for the Blackburne-Peel index [6] and less than 0.6 for the Caton-Deschamps index [7].

Statistics

EpiData software was used for data entry and management.

For statistical purposes, the Dejour classification was grouped as low (Grade A) or high (Grade B-D), since there were very few patients in grades B to D. A sample size calculation performed using an expected kappa of 0.6, precision ± 0.15, a proportion of outcome 0.5, a confidence level of 95% and an expected drop-out rate of zero per cent yielded a minimum of 110 participants [14].

Statistical analyses were performed using unweighted kappa statistics for both intra- and inter-rater measurements, as all measurements were categorised: Dejour classification as grade A versus B–D, anterior translation positive/negative, trochlear dysplasia by depth positive/negative and all patella alta measurements as yes/no. The dichotomisation was performed for clinical relevance. However, dichotomising continuous variables may reduce statistical precision and lead to information loss, which should be considered when interpreting the findings.

Results

Trochlear dysplasia

The intra-rater reliability for the Dejour classification ranged from 0.26 to 0.53; –0.01 to 0.70 for anterior translation; and from 0.29 to 0.85 for trochlear depth (Table 1). The combined kappa rating was 0.15 for the Dejour classification, 0.23 for anterior translation and 0.26 for trochlear depth. For the medical student, the intra-rater measurements ranged from 0.26 to 0.85; –0.01 to 0.74 for the residents, and 0.3 to 0.46 for the consultant.

Patella alta

The intra-rater reliability for the Insall-Salvati index ranged from 0.63 to 0.85; for the Caton-Deschamps index, from 0.12 to 0.70; and for the Blackburn-Peel index, from 0.24 to 0.57 (Table 1). The combined kappa ratings were 0.66 for Insall-Salvati, 0.24 for Caton-Deschamps and 0.19 for Blackburn-Peel.

Discussion

In the present study, only the Insall-Salvati index demonstrated acceptable reproducibility when applied to routine ED X-rays. The reduced reproducibility observed for the remaining measurements likely reflects the challenges of applying them to X-rays with variable rotational malalignment, rather than inherent shortcomings of the measurement methods themselves. Routine ED X-rays are not optimised for rotational alignment, and variations in projection may alter anatomical landmarks, particularly in measurements of trochlear morphology.

These findings are consistent with prior literature.

Measurements of trochlear dysplasia on lateral X-rays have also been criticised for their limited reliability. The Dejour classification depends on the presence of subtle X-ray features such as the crossing sign, supratrochlear spur and double contour, which are all subject to interpretation and variability in X-ray technique [2, 10]. This is reflected in our data, which showed poor agreement for the Dejour classification and related parameters, echoing findings from Diederichs et al. and Jibri et al., who both emphasised the role of MRI over X-rays for accurate assessment of trochlear morphology [8, 12].

Increased rater seniority was not consistently associated with improved agreement. Instead, measurement complexity and the constraints of routine ED X-rays, where optimal rotational alignment is rarely achieved, likely contributed to variability. Consequently, methodological and imaging-related limitations, rather than clinical expertise per se, appear to drive measurement inconsistency, highlighting the need for assessment methods based on clearly defined and reproducible anatomical landmarks.

Our findings suggest that the Insall-Salvati ratio may be a more reproducible option when applied to routine ED X-rays. Conversely, measurements such as the Blackburne-Peel, Caton-Deschamps, and Dejour classification should be interpreted with caution. Where uncertainty remains, careful interpretation and consideration of X-ray limitations are essential, particularly when images deviate from true lateral positioning [8, 17].

This study has several limitations. X-rays were obtained as part of routine ED care without standardised rotational alignment, and the degree of rotational malrotation was not quantified. Although reflecting clinical practice, it introduces variability that may affect measurement reliability. The evaluated measurements were originally developed for true lateral X-rays, and their application to non-true lateral projections represents a deviation from their intended use. Finally, the retrospective single-centre design may limit generalisability.

Conclusions

When applied to routine ED knee X-rays, the Insall-Salvati index demonstrated the highest and most consistent reliability. In contrast, the Blackburne-Peel index, Caton-Deschamps index and trochlear dysplasia measures showed poor reproducibility, reflecting the challenges of applying X-ray indices to images with variable rotational malalignment. These findings emphasise the importance of adopting an appropriate X-ray technique and of careful interpretation when using such measurements in clinical practice.