Introduction

Insall–Salvati, Grelsamer–Meadows (the modified Insall–Salvati), Caton–Deschamps, and Blackburne–Peel ratio are, among a multitude of others [1], the most common patellar height ratios [1, 2] used in diagnosing a high-riding patella (patella alta) on routine digital lateral radiographs of the knee. These measurement methods, however, may be somewhat complicated and time-consuming [3]. Therefore, an actual need exists for an easier alternative, especially in the emergency radiology setting.

The objective of our study was to propose a new reference line in diagnosing a high-riding patella on routine digital lateral radiographs of the knee.

Materials and methods

Materials

A retrospective cross-sectional preliminary study was conducted in November of 2012. Institutional review board approval was obtained. Given the retrospective design, an informed patients’ consent was not deemed necessary. The Declaration of Helsinki as revised at the 59th World Medical Association General Assembly in Seoul 2008 was closely followed where applicable.

Out of 224 consecutive routine digital lateral radiographs of the knee obtained during September and October 2012, 64 (64/224, 28.6 %) were included in our study. Inclusion criteria were as follows: (1) no bone (patellar, femoral, tibial or fibular) pathology, and (2) no bone surgery (e.g., total knee arthroplasty).

The demographics of participants are shown in Table 1.

Table 1 Demographics of participants
Full size table

Methods

The measurements were performed on routine digital lateral radiographs of the knee in Ginkgo CADx (version 2.14.0.4972.; MetaEmotion Healthcare, MetaEmotion SL, Valladolid, Spain) separately by two well-trained observers (the first and the second author) at 2-week intervals in a blinded manner.

Routine digital lateral radiography of the knee

Routine digital lateral radiography of the knee (Fig. 1) was performed with the patient lying on the affected side [4–6]. The patient’s hands were positioned behind the head or extended above it [4]. The affected leg was slightly flexed at the knee joint [4, 5, 7] (ideally at 30°) with a support for the ankle and foot [4–6]. The unaffected leg was positioned either in front of or behind the affected leg [4–6]. The central X-ray beam passed vertically through the center of the knee [4].

Fig. 1
figure 1

Routine digital lateral radiography of the knee. The patient was lying on the affected side. The affected leg was slightly flexed at the knee joint (34.4° ± 16.6°) with a support for the ankle and foot. The unaffected leg was positioned either in front of or behind the affected leg. The central X-ray beam (1) passed vertically through the center of the knee. In the routine analogue (classic) lateral radiography of the knee, an X-ray cassette (2) would have been positioned under the knee of the affected leg

Full size image

Subsequently, on the routine digital lateral radiographs of the knee: (1) the bones (the lateral aspect of the patella, the distal portion of the femur, and the proximal portion of the tibia and the fibula) were demonstrated, (2) the joint spaces (of the patellofemoral and the femorotibial joint) were open, and (3) the femoral condyles were largely superimposed [4, 6].

The most common patellar height ratios

The Insall–Salvati ratio (Fig. 2a) was calculated by dividing the length of the patellar tendon with the longest diagonal distance between the lower and the upper end of the patella [1, 3, 8–15]. Values higher than 1.20 indicated a high-riding patella [1, 8–10, 12, 14, 15] and values lower than 0.80 indicated a low-riding patella (patella infera or patella baja) [1, 8–10, 12, 14, 15].

Fig. 2
figure 2

The most common patellar height ratios in diagnosing a high-riding patella on routine digital lateral radiographs of the knee. a The Insall–Salvati ratio. a the length of the patellar tendon, b the longest diagonal distance between the lower and the upper end of the patella. b The Grelsamer–Meadows (the modified Insall–Salvati) ratio. a the distance between the patellar tendon insertion and the lower end of the articular surface of the patella, b the length of the articular surface of the patella. c The Caton–Deschamps ratio. a the distance between the peak of the visualized anterosuperior angle of the tibial plateau (dotted) and the lower end of the articular surface of the patella, b the length of the articular surface of the patella. d The Blackburne–Peel ratio. a the vertical distance between the horizontal line (dotted) projected along the tibial plateau and the lower end of the articular surface of the patella, b the length of the articular surface of the patella

Full size image

The Grelsamer–Meadows (the modified Insall–Salvati) ratio (Fig. 2b) was calculated by dividing the distance between the patellar tendon insertion and the lower end of the articular surface of the patella with the length of the articular surface of the patella [1, 10–12, 14, 16]. Values higher than 2.00 indicated a high-riding patella [1, 10, 12, 14, 16].

The Caton–Deschamps ratio (Fig. 2c) was calculated by dividing the distance between the peak of the visualized anterosuperior angle of the tibial plateau and the lower end of the articular surface of the patella with the length of the articular surface of the patella [1, 3, 11, 13, 14, 17]. Values higher than 1.30 indicated a high-riding patella [1, 14, 17] and values lower than 0.60 indicated a low-riding patella [1, 13, 14, 17, 18].

The Blackburne–Peel ratio (Fig. 2d) was calculated by dividing the vertical distance between the horizontal line projected along the tibial plateau and the lower end of the articular surface of the patella with the length of the articular surface of the patella [1, 3, 9, 11, 14, 19]. Values higher than 1.00 indicated a high-riding patella [1, 14, 19] and values lower than 0.54 indicated a low-riding patella [10, 12, 19].

The new reference line

On routine digital lateral radiographs of the knee, the long axis of the superimposed egg-shaped femoral condyles was drawn and moved upward to pass through the point where the posterior contours of the femoral diaphysis and the femoral condyles meet (the posterior reference point) to become a new reference line (Figs. 3 and 4). Any part of the patella crossing the new reference line indicated a high-riding patella.

Fig. 3
figure 3

The new reference line in diagnosing a high-riding patella on routine digital lateral radiographs of the knee. a An egg has a short axis (SA) and a long axis (LA). b On routine digital lateral radiographs of the knee, the superimposed femoral condyles appeared as an egg-shaped structure with its long axis (LA). The point where the posterior contours of the femoral diaphysis (F) and the femoral condyles meet was the posterior reference point (PRP). c Sequentially, the long axis (LA) of the superimposed femoral condyles was drawn and moved upward (arrows) to pass through the posterior reference point (PRP) and to become a new reference line (RL). d Non-high-riding patella did not cross the new reference line (RL). e High-riding patella crossed the new reference line (RL)

Full size image
Fig. 4
figure 4

The new reference line in diagnosing a high-riding patella on routine digital lateral radiographs of the knee. a Non-high-riding patella. b High-riding patella. L left knee, LA long axis of the superimposed egg-shaped femoral condyles, PRP posterior reference point where the posterior contours of the femoral diaphysis and the femoral condyles meet, R right knee, RL the new reference line

Full size image

To the extent of our knowledge, a similar reference line has not been described in the pertinent literature.

Statistical analysis

Statistical analysis was performed in MedCalc (version 11.3.0.0.; MedCalc Software bvba, Ostend, Belgium). A p value <0.05 was considered statistically significant.

The time needed to draw the new reference line as well as the lines of the most common patellar height ratios—the drawing time—was measured by a stopwatch. The average drawing time with the standard deviation (SD) was calculated.

The statistical significance of differences between the binary interpretations of the measurements for the new reference line and the binary interpretations of the measurements for the most common patellar height ratios was assessed using McNemar’s test [20]. Two-tailed p values with the 95 % confidence intervals (CI) were calculated.

The intra- and interobserver agreement of the most common patellar height ratios was expressed as concordance correlation coefficient [21, 22] with the 95 % CI.

The intra- and interobserver agreement of the new reference line was expressed as percent agreement [23] of the binary interpretations of the measurements for the new reference line.

Results

The average drawing time of the new reference line was 2.7 ± 0.6 s. The average mean values and the average drawing time of the most common patellar height ratios are shown in Table 2.

Table 2 Average mean values and the average drawing time of the most common patellar height ratios
Full size table

The new reference line required approximately one-half of the drawing time compared to Insall–Salvati, Grelsamer–Meadows (the modified Insall–Salvati), and Caton–Deschamps ratio, and approximately one-third of the drawing time compared to Blackburne–Peel ratio.

There were no statistically significant differences between the binary interpretations of the measurements for the new reference line and the binary interpretations of the measurements for the most common patellar height ratios (Table 3).

Table 3 There were no statistically significant differences between the binary interpretations of the measurements for the new reference line and the binary interpretations of the measurements for the most common patellar height ratios when cross-compared using McNemar’s test and the two-tailed p values with the 95 % confidence intervals
Full size table

McBride [24] classified the strength of agreement using the concordance correlation coefficient as poor (<0.90), moderate (0.90–0.95), substantial (0.95–0.99), and almost perfect (>0.99). The range of intraobserver agreement of the most common patellar height ratios was 0.87–0.95 (Table 4). The range of interobserver agreement of the most common patellar height ratios was 0.85–0.95 (Table 5). Therefore, the intra- and interobserver agreement of the most common patellar height ratios ranged from poor to moderate.

Table 4 Intraobserver agreement of the most common patellar height ratios. The observers are written in brackets, and measurements as indexes. (1) first observer, (2) second observer, 1 first measurement, 2 second measurement
Full size table
Table 5 Interobserver agreement of the most common patellar height ratios. The observers are written in brackets, and measurements as indexes. (1) first observer, (2) second observer, 1 first measurement, 2 second measurement
Full size table

The intra- and interobserver agreement of the new reference line were 100 %. The new reference line repeatedly classified the same 12 (12/64, 18.8 %) patellae into the high-riding group and the other 52 (52/64, 81.3 %) patellae into the non-high-riding group.

Discussion

Although many different methods of measuring the patellar height exist [1], none is perceived as the gold standard [2, 12]. Therefore, all we could do was to compare the new reference line with the most common patellar height ratios: Insall–Salvati, Grelsamer–Meadows (the modified Insall–Salvati), Caton–Deschamps, and Blackburne–Peel [1, 2].

An ideal method of measuring the patellar height implies: (1) a good correlation with clinical symptoms, (2) simplicity, (3) a reliable distal femoral reference point, (4) accuracy in a routine radiographic setting (regardless of age, sex, ethnicity, radiographic magnification, physical size of the knee, or the angle of knee flexion), and (5) a good intra- and interobserver agreement [2, 3]. Currently, no method of measuring the patellar height meets all of the aforementioned criteria [2].

In daily work, the most common patellar height ratios may be time-consuming because each requires two measurements and a calculation within the normal ranges that are difficult to remember [3]. Furthermore, they all measure the patellar height using the proximal tibial reference points [2, 11, 12, 25]. The values of the most common patellar height ratios established in adults cannot be transposed to pediatric population [1]. Aparicio et al. [26] reported that the Caton–Deschamps ratio was the least affected by skeletal maturation compared to Blackburne–Peel and Koshino. The angle of knee flexion is no less important [12]—except for the Insall–Salvati ratio [9, 15, 16]—which, on the other hand, appears dependent on sex and ethnicity, although the data are conflicting [1, 8, 15, 27]. What is more, the available research on the interobserver agreement of the most common patellar height ratios is scarce [11], with results varying from poor [10–12] to excellent [2, 3, 28], the latter being ascribed to more scrutinized methodology [2] and training [28]. In our study, the intra- and interobserver agreement of the most common patellar height ratios ranged from poor to moderate. Our results could be corroborated by the difficulties in identifying the required references, namely the patellar tendon insertion relative to the Insall–Salvati as well as the Grelsamer–Meadows (the modified Insall–Salvati) ratio [3, 11], the peak of the visualized anterosuperior angle of the tibial plateau relative to the Caton–Deschamps ratio [3], and the projection of the horizontal line along the tibial plateau relative to the Blackburne–Peel ratio [3, 11, 29]. In our study, the best intra- and interobserver agreement was obtained for the Insall–Salvati ratio. This was, perhaps, assisted by the absence of the certain conditions that might interfere with the measurement of the Insall–Salvati ratio, such as Osgood–Schlatter’s and Sinding–Larsen–Johansson’s disease on one side, as well as better imaging of soft tissues on routine digital lateral radiography of the knee, namely the patellar tendon [14], on the other.

A high-riding patella is frequently associated with a plethora of diverse clinical entities, ranging from anterior knee pain [1, 2, 9, 18, 25, 30, 31]—undoubtedly a very recurrent symptom [32]—to chondromalacia patellae [3, 11, 14, 15, 25, 33], chronic tendinopathy in athletes [27], joint effusion [25], neurological disorders [9, 18, 30], Osgood–Schlatter’s disease [1, 9, 15], patellar subluxation [15, 18, 33] and dislocation [3, 14, 15, 18, 25, 27, 30, 33], patellofemoral instability [1, 2, 9, 11, 18, 30], Sinding–Larsen–Johansson’s disease [9, 15], tendonitis of the patellar and the quadriceps tendon [11, 15], trauma (rupture of the patellar tendon) [9, 15, 18], all the way to idiopathic cases [30]. Thus, the abnormal patellar height in the sagittal plane certainly seems of practical clinical significance [25].

The new reference line required approximately one-third to one-half of the drawing time compared to the most common patellar height ratios. No measurement or calculation was needed. In addition, when only hardcopy lateral radiographs of the knee are at hand or—for any reason—it is impossible to draw the new reference line (e.g., the necessary software tools are not available), the position of the new reference line could be emulated with a pencil or a paper margin. Any part of the patella crossing the pencil or the paper margin would indicate a high-riding patella (Fig. 5). Therefore, we consider the new reference line a very simple tool.

Fig. 5
figure 5

The position of the new reference line could be emulated with a pencil or a paper margin. First, the long axis of the superimposed egg-shaped femoral condyles could be emulated with the pencil (a) or the paper margin (c). The point where the posterior contours of the femoral diaphysis and the femoral condyles meet is the posterior reference point (dot). Sequentially, the pencil or the paper margin should be moved upward to pass through the posterior reference point. Any part of the patella crossing the pencil (b) or the paper margin (d) would indicate a high-riding patella

Full size image

In general, the patellofemoral joint is defined by the patellar height with regard to the distal portion of the femur [2]. Therefore, the patellar height should be measured using a distal femoral reference point [1, 2, 12, 25]. In our study, the point where the posterior contours of the femoral diaphysis and the femoral condyles meet—the posterior reference point—proved a reliable distal femoral reference point because it was easily identifiable on every routine digital lateral radiograph of the knee.

We have also shown that there were no statistically significant differences between the binary interpretations of the measurements for the new reference line and the binary interpretations of the measurements for the most common patellar height ratios. Therefore, the new reference line and the most common patellar height ratios were equally accurate in diagnosing a high-riding patella. In our study, the mean age of participants was 53 years (range: 16–85 years). The number of females and males was almost equal, with a slight female preponderance (33:31). We, therefore, believe that—in adults—the new reference line should not be sex-dependent. Pursuant to the regional demographics, though, we could only include Caucasian participants. The radiographic magnification and the physical size of the knee seemed to have no bearing on the new method.

As previously stated, the new reference line was drawn on routine digital lateral radiographs of the knee. In a standardized lateral radiography of the knee, the angle of knee flexion is usually around 20–30° to ensure the tightness of the patellar tendon [30]. In practice, however, it is prone to considerable variation [12]. The methods of measuring the patellar height using a distal femoral reference point seem to depend crucially on the angle of knee flexion [2]. Unfortunately, due to our limited resources, we did not analyze the new reference line over the full range of knee flexion. It is important to note, however, that the angle of knee flexion was not a criterion in our study. Instead, we measured the angle after inclusion criteria were met and obtained the mean value with the standard deviation and range (34.4° ± 16.6°; 1.6–79.1°). Seyahi et al. [12] reported that the majority of routine lateral radiographs of the knee at their institution had been performed over the range of knee flexion between 20° and 45°. In our study, 57.8 % (37/64) of data were in that range (Fig. 6). Therefore, we believe that the new reference line should certainly prove stable within these limits (Fig. 7). On the other hand, Caton [13] suggests that a good method for measuring the patellar height should be valid over the range of knee flexion between 10° and 80°. Since our results are valid for the 1.6–79.1° range, this criterion appears satisfied, too. Nevertheless, we emphasize that any excessive flexion of the knee during routine digital lateral radiography increases the possibility of unnecessary damage (e.g., separation of the patellar fragments) as well as hindering of important diagnostic information (e.g., effusion). The pain itself may often restrict the movement.

Fig. 6
figure 6

Box-and-whisker plot showing the values of the angle of knee flexion spreading across the range between 1.6° and 79.1°

Full size image
Fig. 7
figure 7

The new reference line (full line) indicating a non-high-riding (a) and a high-riding (b) patella at different angles of knee flexion is parallel to the long axis of the superimposed egg-shaped femoral condyles (dotted line)

Full size image

The intra- and interobserver agreement of the new reference line were 100 %. In statistics, when two well-trained observers rate a variable with only two possible conditions that are clearly differentiated, a high agreement should be expected [23]. Defining whether any part of the patella crossed the new reference line was rather straightforward. Therefore, the possibility of guessing the outcome was rather low. In that regard, other statistical tests such as Cohen’s kappa were not necessary [23]. It could be theorized, though, that the less-trained observers would have achieved a less favorable agreement.

There were several limitations to our study, apart from the retrospective design. The sample was moderate in size [34] and formed solely of Caucasian adults. Furthermore, the new reference line was not designed to indicate a low-riding patella—albeit a serious [35], yet relatively rare condition [35, 36] associated with achondroplasia [9, 11, 15, 18, 33], anterior knee pain [1], juvenile rheumatoid arthritis [33], limited range of knee motion [1, 2, 11, 14], neurological disorders [11, 15, 33], Osgood–Schlatter’s disease [2, 11], patellofemoral osteoarthritis [2, 3, 11, 14], and trauma [9, 11, 15, 18, 35]. In addition, a low-riding patella is often a consequence of various surgical procedures [11, 35] such as anterior cruciate ligament reconstruction [9, 18, 30], knee arthroscopy [18], retrograde femoral nailing [35], tibial tubercle osteotomy [9, 15, 18, 30, 33], or total and unicompartmental knee replacement [30]. In fact, it usually occurs in postoperative knees [18]. The same limitation is also inherent to some of the other direct methods of measuring the patellar height, like that of Labelle et al. and Laurin [1]. In our study, the total number of measurements indicating a low-riding patella according to the most common patellar height ratios was very low: 2.3 % (6/256; Insall–Salvati), 0.0 % (0/256; Caton–Deschamps), and 0.4 % (1/256; Blackburne–Peel). At length, given that the lateral femoral condyle has a greater anteroposterior dimension [1], when the two are not largely superimposed, this could open a question of which should be regarded as a reference.

Conclusions

The new reference line proved to be a very simple, accurate and reproducible tool in diagnosing a high-riding patella. We believe that our preliminary results are an encouraging impetus to the more complex research.