The main contribution of this study is that the API was proposed for assessing patellar height. The evidences suggested the intraclass correlation coefficients of this new method were excellent, comparable with those of previously reported for ISI, mISI, BPI, CDI and PPA(2, 5, 17). The API seems to have slightly better intra-observer repeatability and inter-observer reproducibility in more than 200 radiographs. Furthermore, the analyses of Pearson’s correlation coefficient demonstrated strong correlations between the API and the aforementioned ones. Overall, the API is a valid and reliable alternative for assessing patellar height in radiographically normal knees.
Currently, the five published methods mentioned in this article are the most popular. However, there are inherent methodological difficulties, which may lead to an increase in inter- and intra-observer variability(11, 17–19). The ISI remains the most popular one, however, it is imperative to standardize the tibial attachment of the patellar tendon in order to avoid a significant variation resulting from the obscure morphology and the pathological overgrowth of the tibial tubercle(18, 20). Besides, the variants of patellar shape can lead to a greatly different result, especially in a patella being characterized by a long, non-articulating inferior pole (Cyrano patella)(19–21). The mISI described by Grelsamer et al(7) was designed to overcome the above disadvantages. Nonetheless, it did not provide the cutoff value for patellar baja. Blackburne and Peel(8) devised a method which exchanged the easily misidentified the tibial attachment of the patellar tendon for the tibial plateau as a landmark. Regrettably, they did not make a precise description of the tibial plateau line(12), resulting in the measurement variability. Portner et al(2) proposed the PPA which was formulated by the patella and tibial plateau. However, it is commonly found that the posterior margin of tibial plateau was rounded and smooth, making it difficult to unify the vertex of the angle in repeatable measurements. Caton et al(9) described that the antero-superior edge of the tibia was used as the landmark. Whereas an accurate identification of the reference could be difficult in osteoarthritic knees with osteophytes(12, 14). In view of this, it is necessary to find an accurate measurement method that adopts new reference to avoid the above identification difficulties.
The tibial landmark of the API, formed by the tibial shaft axis and its perpendicular line passing through the tip of the fibular head, intrinsically owns better positional stability. Compared with the tibial plateau, the fibular head is independent of osteophytes and articular surface destruction. And the recognition of the styloid process of fibular head is accurately acquired on the standard lateral radiograph. Meanwhile, the tibial shaft axis is an independent reference with spatial constancy. In this study, it was determined by selecting two mid-diaphyseal points at 70- and 110-mm distal to the tibial plateau down the tibial shaft, in which could avoid an increase of the measurement variability, as described by previous studies(15, 22). As an intramedullary reference, the tibial shaft axis can theoretically provide a constant landmark compared to marginal counterparts, especially during tibial rotating subsequent to knee flexion. Thus, we speculated that maintaining the greater accuracy and less variability in the measurements could be based on the above procedures.
Consistent with standards of previous methods, this new method is independent of radiographic magnification, physical size of the knee, and angles of knee flexion(2, 7, 8, 12). In this study, as mentioned above, the radiologically normal knees were enrolled in an endeavour to establish a normal range of values. The results demonstrated that the average value of API is approximately 1.5 using the sample size of 649 radiographs, thereby the reference range is recommended as 1.2–1.9 based on the 95% confidence interval. When the value is greater than 1.9 or less than 1.2, surgeons should be alert to the occurrence of patella alta or infera. We also evaluated the correlations between the API and other five methods, and delighted to find that the new method was strongly correlated with the mISI, BPI and CDI. This is partially explained by the fact that these four methods share the patellar articular surface. Moreover, the results showed that the API was less correlated with ISI and PPA, and a potential explanation might be the influence of the anatomic variability of patellar shape, as illustrated by Grelsamer et al(20).
There are several limitations of the present study that warrant consideration. Firstly, this study employed a retrospective design and the case data were extracted from the hospital information system. The subjects were rated as radiologically normal, but the clinical symptoms could not be determined and injuries of meniscuses or cruciate ligaments might be included. However, these soft-tissue problems could be deemed acceptable on the assumption that the patellofemoral relationship is little affected, as mentioned in the previous studies(6, 8, 9). Secondly, variabilities in the distance between the tip of the fibular head and the joint line should not be ignored, and the reference range derived from this study ought to be treated with caution. Thirdly, the authors speculated that the API cannot be used when the fibular head is damaged by bone neoplasms, or in patients who have had the fibular head removed. Finally, the samples included in this study were radiographically normal knees, it is necessary to investigate the application of the API in pathological conditions in further studies. Despite the limitations set out above, there are still several design advantages, including a large sample size, measuring patellar height with six different methods, and simultaneously assessing the inter- and intra-observer reliability in all methods.