Conventional radiography cannot replace CT scanning in detecting tibial tubercle lateralisation

1. Introduction 

Malalignment of the extensor mechanism can cause patellar instability and is frequently encountered in knee surgery clinics. Factors contributing to instability include patella alta, trochlear dysplasia, medial patellofemoral ligament rupture, muscular imbalance, and an excessive lateral distance between the tibial tubercle and the trochlear groove (TT–TG) [1], [2], [3], [4]. Recently, Shakespeare and Fick [5] concluded that visual inspection of the TT–TG was unreliable. Another recent study, comparing CT and MRI, reported on MRI being a reliable alternative to detect the TT–TG [6]. However, the general consensus is that routine conventional radiography (CR) must be supplemented by computer tomographical (CT) analysis to assess the TT–TG adequately [1], [2], [3], [7], [8], [9], [10], [11]. CT analysis is time-consuming, not always available, and has considerably higher costs compared to axial CR.

We set out to determine the relationship between the TT–TG on CT and an axial CR with external determination of the tibial tubercle and femoral epicondyles using lead markers. We devised a new technique to measure the TT–TG on axial CR based upon the principles of the TT–TG in reformatted CT analysis. We hypothesised that standardised, axial conventional radiography using lead markers would produce the same projection as reformatted CT and would be a reproducible and reliable alternative to CT.

3. Results 

No patellofemoral arthrosis, patella alta, or trochlear dysplasia on any of the axial CR or CT was seen. Physical examination revealed no signs of medial patellofemoral ligament rupture or muscular imbalance. The measurements of the TT–TG on CT and axial CR are presented in Table 1, whereas Table 2 presents the ICC and 95% confidence intervals (CI95).

Table 1. Measurements of conventional radiography and CT analysis

	Knee	Conventional radiography (CR)a								Reformatted CTb
		Observer 1				Observer 2				Observer 1		Observer 2
		CR 1A	CR 1B	CR 2A	CR 2B	CR1A	CR 1B	CR 2A	CR 2B	A	B	A	B
	1	19	17	22	22	18	20	22	21	16	16	16	16
	2	6	1	7	9	3	2	7	8	16	15	18	17
	3	21	24	24	20	23	24	21	22	11	9	13	12
	4	23	24	8	8	23	25	9	9	20	22	21	33
	5	12	11	17	18	11	11	15	17	22	24	24	26
	6	18	16	16	17	13	15	18	18	16	12	14	13
	7	8	8	12	12	7	7	12	11	16	15	17	17

a Two axial radiographs were taken (CR 1 and 2). Each CR was assessed twice (A and B) by each observer. b The reformatted CT was assessed twice (A and B) by each observer.

Table 2. Intraclass correlation coefficients (ICC) and 95% confidence interval for the CR and CT analysis

			ICC	95% confidence interval
	Conventional radiography	Intraobserver reliability observer 1	0.862	0.537–0.995
		Intraobserver reliability observer 2	0.891	0.640–0.979
		Interobserver reliability	0.948	0.855–0.990
	Reformatted CT	Intraobserver reliability observer 1	0.943	0.693–0.990
		Intraobserver reliability observer 2	0.980	0.886–0.997
		Interobserver reliability	0.976	0.924–0.995

The ICC for CT was high, varying between 0.943 and 0.980 (CI95 0.693–0.997). The ICC for axial CR was good-to-high varying between 0.862 and 0.948 (CI95 0.537–0.995). Note the broader range in the CI95 for the axial CR as compared to the CT measurements. The limits of reproducibility for CT was 4 mm while it was 18 mm for axial CR. Fig. 2, Fig. 3 present an example of the measurements for CT and axial CR by the same observer. Note the difference of 5 mm between the two axial CR, and the 10 mm difference between CT and axial CR.

4. Discussion 

In the present study the reliability (ICC) of the axial CR method was comparable to CT, but the variability as shown by the limits of reproducibility of the axial CR method (18 mm) was more than four times that for CT (4 mm). The limits of reproducibility quantify the reliability of the measurement technique as it determines the measurement error from all sources (measuring technique and inter- and intra-variability from the observers). Thus, this value indicates how great a difference must be before one can speak of a real difference that is greater than the measurement error [17]. In the present study, the actual value or the TT–TG on the axial CR could be up to 18 mm lower or 18 mm greater than the value measured, and the difference between two measurements made by different observers must be greater than 18 mm to be classified as being a real difference. For the CT this is 4 mm, i.e. a factor four smaller. Fig. 3 exemplifies the large measurement error and therefore variability of the axial CR method, in this case a difference of 5 mm between the two axial CR. Even greater differences were found in other knees (Table 1). The large difference compared to CT can also be seen (Fig. 2 and Table 1). The relationship between the tibial tubercle and the femoral epicondyles, and the relationship between the tibial tubercle and the patellar groove on a 30° axial view could be visualised using lead markers, which is analogous to others [18].

Literature reports the difficulties in the assessment and comparison of CT with axial radiography using conventional radiographic parameters [7], [19]. Nagamine et al. [18] used a marker for the visualisation of the tibial tubercle as possible screening method and concluded that a marking wire may demonstrate an abnormal lateral position of the tibial tubercle in patellofemoral osteoarthritis at 30° of flexion. They determined this by two self-developed, reproducible, methods using the line between the highest points of the medial and lateral condyle as the reference axis, thus not measuring the TT–TG. However there was no significant correlation between the CR and CT. We, on the other hand, used the clinically determined epicondylar line as reference axis, analogous to the TT–TG in reformatted CT. Other factors can also account for the large measurement error of the axial CR. Positioning of the markers, in particular the tibial tubercle marker, is not always clinically straightforward. Secondly, the positioning of the patient probably plays an even greater role. Rotation of the tibia and/or hip [9] can influence the tibial marker position in the horizontal plane as well as the unintended movements by the patient. Movement of the skin could also affect the position of the tibial tubercle although every effort was made to prevent knee movement after the markers were fixed. Furthermore, the tibial tubercle and/or epicondylar points could not always be palpated with the knee positioned in 30° flexion, resulting in the knee flexed slightly more while positioning the markers and thus potentially influencing the marker position. An exact 90° angulation of the X-ray cassette can also be difficult to achieve and maintain by the patient, hence influencing the projection of the tibial marker on the X-ray cassette. We believe that even the Merchants view [20] cannot reduce these variations.

An adequate, reproducible clinical research method must have a high reliability and low measurement error, i.e. low variability. Axial CR can, in theory, be used to identify abnormal positioning of the tibial tubercle if this tubercle can be detected. While we could detect the tibial tubercle using a simple lead marker and determine the TT–TG, in this study the 95% prediction limit of the measurement error was too high (more than four times that of CT) even though the qualitative reliability, as shown by the ICC, was comparable to that of CT. The large measurement error makes the CR method unusable to detect (within the 95% prediction limit) a TT–TG distance under 18 mm, which is unacceptable for clinical practice. Factors related to the set-up of the axial CR probably account for this large measurement error. For that reason the study was terminated after analysis of the seven knees. In conclusion, the axial CR method using lead markers was not reproducible due to the large measurement error and can therefore not be used as an alternative to CT scanning to detect the TT–TG in patients with patellofemoral complaints without apparent abnormalities.