The Validity and Absolute Reliability of Lower Extremity Angle Values on Full-Leg Standing Radiographs Using the TraumaMeter Software

Featured Application

TraumaMeter v.873 was designed to increase the accuracy of specific measurements at the spine level. The system meets the characteristics of open-source and modular functions and has been extended to perform angular measurements on lower extremity X-ray studies with high reliability.

Abstract

To establish classifications and to obtain pre- and post-operative information on patient-specific alignments, it is necessary to measure different angular values accurately and precisely, mainly on weight-bearing, full-length anteroposterior X-rays of the lower limbs (LLRs). This study evaluated angular measurements’ validity and absolute reliability on LLRs with a self-developed, computer-aided measurement system (TraumaMeter v.873). Eight independent observers measured the preoperative mechanical hip-knee-ankle (mHKA) angle of 52 lower extremities (26 cases) in a blinded fashion on three occasions separated by two weeks. We obtained an intra-observer mean bias error (MBE) of 0.40°, a standard deviation (SD) of 0.11°, and a 95% confidence interval (CI) of 0.37°–0.43°. We also obtained an inter-observer MBE of 0.49°, an SD of 0.15°, and a 95% C of 0.45°–0.53°. The intra-observer MBE for the measurement pair between the second and the first measurement round (T2T1) was 0.43°, the SD was 0.13°, and the 95% CI was 0.39°–0.47°; the MBE between the third and the second round (T3T2) was 0.37°, with an SD of 0.10° and a 95% CI of 0.34°–0.40°; and the MBE between the third and the first round (T3T1) was 0.40°, with an SD of 0.10° and a 95% CI of 0.37°–0.43°. The interobserver MBE for the first round of measurements was 0.52°, with an SD of 0.16° and a 95% CI of 0.48°–0.56°; the MBE for the second round was 0.50°, with an SD of 0.15° and a 95% CI of 0.46°–0.54°; and the MBE for the third round was 0.46°, with an SD of 0.14° and a 95% CI of 0.42°–0.50°. There were no statistically significant differences in the inter-observer errors for the three tests. In the case of the intra-observer analysis, there were differences between T2T1 and between T3T2, but these differences were minimal, with no overlaps in the lower or upper values, respectively, of the confidence intervals. These results led us to conclude that the TraumaMeter v.873 software extension for measuring lower-limb angles in LLRs is an accurate tool with low intra- and inter-observer variability.

1. Introduction

A trend of innovative thinking has brought about enormous changes in knee replacement surgeries over the last decade. Since the introduction of kinematic alignment (KA) [1] and the concept of constitutional alignment [2], we have seen a remarkable increase in the scientific output calling for a personalised approach in total knee arthroplasty (TKA) surgeries. As a result, alternatives such as the KA and its restricted option, functional alignment, adjusted mechanical alignment, inverse kinematic alignment, and its restricted version are replacing the “one-size-fits-all” almost dogmatic approach of mechanical alignment [3,4]. These different alternatives have generated enormous interest in the information related to the different phenotypes fundamentally studied in the coronal plane [5,6,7]. It is necessary to measure different angular values reliably and accurately, basically on weight-bearing, full-length anteroposterior X-rays of the lower limbs (LLRs), to establish classifications and obtain pre- and post-operative information on the specific alignments of each patient [8,9,10,11,12]. The mechanical hip-knee-ankle (mHKA) angle measures the angulation between the femur and tibia mechanical axes in the coronal plane [13]. The mHKA helps quantify the degree of valgus or varus alignment in the knee [2,14]. Quantifying misalignment is essential as it has been implicated in the pathogenesis of knee osteoarthritis [14,15,16,17]. The mHKA is also used to assess post-operative knee alignments after total knee replacements and is necessary for angular correction surgeries and assessments of the corrections achieved [18]. Knowledge of the mHKA is also helpful in the analysis of the causes of TKA failure and the planning of revision surgery [19]. In addition, advances in digital technology in radiology in the last decade have enabled advantages such as reduced radiation exposure, more efficient image comparisons, variable contrast scales, and ease of storage. These advantages have encouraged the development of software tools for evaluating medical images [20]. Today, we have different software tools that allow us to measure medical images more easily, quickly, precisely, and accurately, with less intra- and inter-observer variability. There are even systems based on artificial intelligence which can automatically detect anatomical references on X-rays to establish determined angular measurements [21,22,23,24,25,26,27]. These advances in angle information management have not always been accompanied by the validation of accuracy and reliability, with few exceptions [28,29,30,31,32].

An open-source, computer-aided measurement system (TraumaMeter v.873, José Hurtado-Avilés and Fernando Santonja-Medina, reg. no. 08/2021/374, Murcia, Spain) has been developed. Initially, the software was designed to improve the manual measurement of variables necessary to study vertebral misalignments, such as the Cobb angle and axial vertebral rotation (AVR) on digital radiographic images, without introducing the scale of the radiographic study performed [33,34,35].

Our research aimed to evaluate the validity and absolute reliability (according to the Hopkins criteria [36]) of an extension of the initial TraumaMeter v.873l software for measuring the angular values of LLRs. This validation, which was the first to be carried out, was necessary for the usefulness of the software in clinical practice to be confirmed.

2. Materials and Methods

We conducted a prospective, observational study with 52 preoperative LLRs from the medical image repository (by random selection from 26 patients undergoing bilateral TKA surgeries at one time).

2.1. Software

The TraumaMeter v.873 software was developed in C++ language under the Microsoft Visual Studio 2019 development environment and using the OpenCV 3.4.10 artificial vision libraries and the DCMTK libraries from OFFIS (the Institute for Information Technology) to operate with the digital imaging and communication on medicine (DICOM) files. The software accepts images in .ima (DICOM), .tiff (tagged image file format), .bmp (Windows bitmap) and .jpeg (JPEG (joint photographic experts group)) formats and incorporates additional tools, such as the ability to zoom in on regions of interest and to vary the contrast (i.e., the fractional difference in optical density of the brightness between two areas of an image) of a digitalised X-ray image [25,26,27].

In different trials carried out with TraumaMeter v.873 for the study of scoliosis, we previously demonstrated its increased validity and absolute reliability in determining the AVR on digital X-rays compared to conventional manual measurements [26], lower intra-observer measurement error values when using the software (a mean bias error (MBE) of 1.8° and a standard deviation (SD) of 0.65°) compared to the manual Cobb angle measurement method (which has an MBE of 2.31° and an SD of 0.83°) [27], and a minimum detectable change (MDC95) equal to or less than 0.5° [27].

2.2. X-ray Studies

All studies on the LLRs had been conducted between 2011 and 2017. We used, similarly to other studies in our group [37], the Ysio digital X-ray machine (Siemens Healthcare GmbH, Erlangen, Germany). The patients were barefoot, bearing equal weight on both lower extremities, with the feet facing directly forward, the patellae centred on the femoral condyles, and a standard foot separation of eight centimetres (20 cm for the valgus alignments over 10°). A three-part cassette with a graduated grid was situated behind each patient and the X-ray beam source was located perpendicular to the detector at a distance of two metres. The radiographs of the hips, knees, and ankles were then digitally “stitched” together to generate a singular image, with the appropriate optical density adjustments. Despite the standardised conduct of LLRs, we did not absolutely control rotation [38,39] (by employing, for instance, the relative position of the proximal fibula to the proximal tibia as proposed by Maderbacher et al. [40]). The X-ray images were archived on the Picture Archiving and Communication System (PACS, Siemens Healthcare GmbH, Erlangen, Germany) server in the international standard DICOM format, which we provided to the researchers. Ethical approval was obtained for the analysis of joint imaging by the institutional Ethical Committee (C.P.VLM_HMM_01_2020-C.I.EST: 64/20). All patients gave their informed consent for the analysis of their limb imaging. Our study followed the ethical guidelines of the World Medical Association Declaration of Helsinki, as revised in 2013.

2.3. Study Methodology and Hopkins Criteria

We evaluated the absolute reliability using the Hopkins criteria (a minimum of thirty cases, at least six blinded observers as evaluators, and at least three evaluations per observer separated by at least two weeks) [36]. Reliability can be understood as the reproducibility of the values of a test, trial, or other measurement in repeated trials with the same individuals. Higher reliability means a higher precision in individual measurements. Systematic changes in the measurement means between successive trials represent effects such as learning, motivation, or fatigue, and they should be removed from estimates of within-subject variations. The Hopkins criteria ensured that these factors were eliminated.

The study was conducted with eight independent evaluators. One was an orthopaedic specialist dedicated to knee surgeries and with routine (daily practice) measurements of the study values (we considered him an expert). Two observers occasionally performed angle measurements of the lower limbs with LLRs and were considered semi-experts. The remaining five observers needed to gain experience in measuring the study values and were considered novices.

Several briefing sessions were held before completing the measurements, with comprehensive information on the survey and training on the software used, with images selected for training practices other than those of the study. We used audiovisual media for training in .pdf and .mp4 formats.

For each image, the observers determined the centre of the femoral head, assisted semi-automatically by the software, and the centre of the femoral intercondylar notch to trace the mechanical femoral axis. The centre of the femoral head was determined graphically. First, the area of the femoral head was zoomed in on to facilitate the measurement, and the brightness and contrast were adjusted if necessary. The software displays a horizontal line across the screen’s full width that follows the cursor’s movement. By clicking with the mouse, the observer fixes the line at the most proximal position of the femoral head. The software then displays a circle with the cursor’s position determining its horizontal position and radius and its top point anchored to the previously defined line. The operator matches the position and radius of the circle to those of the femoral head and clicks the mouse.

The observers also determined the centre of the notch of the tibial spines (marking the centre of the distance between the tips of both spines) and the centre of the ankle joint to trace the mechanical tibial axis in each image. We decided that the anatomical landmark of the ankle should be the midpoint of a straight line between the most prominent points of the medial and lateral malleoli cortex close to the joint line [11]. The software calculates the ankle centre’s anatomical reference by dividing the distance between the two marks by two. On postoperative X-rays, the observers marked the anatomical landmarks at the neo-joints (the centres of the intercondylar notches and the tibial metal trays).

To simplify our analysis of accuracy and absolute reliability but still strictly adhere to the Hopkins criteria of absolute reliability [36], we decided to use the observers’ preoperative mHKA angles (the angles between the mechanical axes of the femurs and tibias in the coronal planes) of the 52 LLRs in the three-spaced sessions (i.e., 1248 values). We decided not to use the lateral distal femoral and medial proximal tibial angles, also measured by the observers, in this study. Osteoarthritic knees show enormous variation in the overall coronal alignments of the limbs, as well as in the coronal femoral and tibial alignments [41,42]. Since osteoarthritic knees present no mirror image of one knee over the other for each patient, we could directly consider each knee as an independent case (n = 52) if the object of the analysis was to compare the angular values. However, as we were interested in analysing measurement errors, this solution was more complex from a statistical point of view. We adopted the concept of validity or accuracy as the difference between the average of the measurements obtained and the actual value. Systematic errors (the predictable ones) are what determine validity or accuracy. We adopted the concept of precision or reliability as the variability in the measurements obtained (standard deviation (SD)). Precision or reliability is determined by random errors (due to a lack of internal consistency or the temporal stability of the observer, the measuring instrument, and environmental factors).

The TraumaMeter v.873 software returns the result of the measurements on the medial aspect. It calculates two values (Figure 1) that we called the HKA′ (the angle between the femoral mechanical axis and a horizontal line to the ground) and the HKA″ (the angle between the tibial mechanical axis and a horizontal line to the ground), which were the values that the observers entered into the Microsoft Excel spreadsheet (.xlsx) that was given to them to record the results of their measurements. We calculated the mHKA angle by adding HKA′ and HKA″ together. We showed the mHKA angle, not to be confused with the arithmetic HKA (aHKA), which was the result of subtracting the value of the lateral distal femoral angle (LDFA) from the medial proximal tibial angle (MPTA).

Figure 1 shows the angles used by TraumaMeter v.873 for calculation of the mechanical hip-knee-ankle (mHKA) angle (HKA′, which is the angle between the femoral mechanical axis and a horizontal line to the ground, and HKA″, which is the angle between the tibial mechanical axis and a horizontal line to the ground). The mHKA was calculated by adding the HKA′ and HKA″ together, whilst a denotes the horizontal line to the ground.

2.4. Statistical Analysis

The statistical analysis was performed using the Statistical Package for the Social Sciences (SPSS), version 25 for Windows (SPSS, Inc., Chicago, IL, USA). The results were rounded to one decimal place in the measurements obtained with the software. For the software’s intra- and inter-group concordance analysis, we calculated the validity or degree of agreement between the mean obtained from a large set of results and the true or reference value (MBE (mean bias error)), reliability (SD), and the standard error of the mean (SEM). We derived the error distributions as t procedures for pairwise designs (each data point resulted from the difference between a pair of measurements from a different test or observer). Each error distribution was therefore comparable to that obtained by a paired samples t-test and allowed us to apply inferential methods. By identifying outliers, i.e., values below Q1 − (1.5 × IQR (interquartile range)) and above Q3 + (1.5 × IQR), we improved the distributions of the measurements. We removed outliers from each distribution based on statistical methods to avoid the effect of losing the normality of the data distributions, as the normality of the distributions was necessary for the application of the inferential statistical methods. The outliers were of scant value. Th significantly inaccurate measurements were due to the observers’ errors in writing the wrong values in the cells of the Excel spreadsheets (e.g., instead of writing 178°, an observer wrote the wrong order of numbers (187°) or did not write the three digits and instead wrote 17°, or they wrote the dot that separated the whole and fractional parts of the numbers incorrectly). We analysed whether the differences in the MBE values between each set of measurements were statistically significant using ANOVA and the Tukey’s method for multiple comparisons. We used the Shapiro–Wilk test to test the normality of the data set. As the p-value was greater than 0.05, we concluded that the data were normally distributed.

3. Results

We obtained an intra-observer error in the preoperative right knee measurements for MBE = 0.41°, SD = 0.14°, and 95% CI: 0.36°–0.46° and MBE = 0.42°, SD = 0.12°, and 95% CI: 0.38°–0.46° for the left knee. We obtained an inter-observer error in the preoperative right knee measurements for MBE = 0.5°, SD = 0.17°, and 95% CI: 0.43°–0.57° and MBE = 0.49°, SD = 0.12°, and 95% CI: 0.38°–0.53° for the left knee.

Table 1. Mean bias errors for the intra- and inter-observer errors (n = 26 and n = 26).

Right Knee Intra-Observer Error				Left Knee Intra-Observer Error
	T2T1	T3T2	T3T1		T2T1	T3T2	T3T1
MBE	0.43°	0.37°	0.43°	MBE	0.45°	0.40°	0.40°
SD	0.19°	0.10°	0.12°	SD	0.10°	0.14°	0.12°
SEM	0.04°	0.02°	0.02°	SEM	0.02°	0.03°	0.02°
95% CI	0.36°–0.50°	0.33°–0.41°	0.38°–0.48°	95% CI	0.41°–0.49°	0.35°–0.45°	0.36°–0.44°
Right Knee Inter-Observer Error				Left Knee Inter-Observer Error
	T1	T2	T3		T1	T2	T3
MBE	0.49°	0.50°	0.49°	MBE	0.54°	0.50°	0.43°
SD	0.18°	0.17°	0.16°	SD	0.12°	0.13°	0.11°
SEM	0.04°	0.03°	0.03°	SEM	0.02°	0.03°	0.02°
95% CI	0.52°–0.56°	0.43°–0.57°	0.43°–0.55°	95% CI	0.49°–0.59°	0.45°–0.55°	0.39°–0.47°

shows the different comparisons’ MBE values for intra- and inter-observer errors.

Table 1 shows the values of the mean bias error (MBE); the reliability or standard deviation (SD), which described the variability within a single sample; the sample standard error (SEM), which estimated the variability across multiple samples; and the SEM for a 95% confidence interval (CI). T2T1 corresponds to the pair of measurements used to obtain the intra-observer error between the second and the first measurement rounds, T3T2 refers to that between the third and the second, and T3T1 refers to that between the third and the first. T1 shows the inter-observer error in the first round of measurements, T2 shows that in the second, and T3 shows that in the third.

No significant differences were found when analysing the intra-observer error distributions by ANOVA (p = 0.205). When analysing the inter-observer error distributions, we also found no significant differences (p = 0.215), and so we accepted the hypothesis of the equality of the means. However, in the post hoc rank test (Tukey’s test), when comparing the differences in the inter-observer error MBE values for each pair of distributions, we obtained a significant difference (p = 0.002) at a confidence level of at least 95% in the analysis of tests 1 (T1) and 3 (T3) of the left lower extremity.

We obtained an intra-observer error in preoperative measurements for MBE = 0.40°, SD = 0.11°, and 95% CI: 0.37°–0.43°. We obtained an inter-observer error in the preoperative measurements for MBE = 0.49°, SD = 0.15°, and 95% CI: 0.45°–0.53°.

Table 2. Mean bias error for the intra- and inter-observer errors (n = 52).

Intra-Observer Error
	T2T1	T3T2	T3T1
MBE	0.43°	0.37°	0.40°
SD	0.13°	0.10°	0.10°
SEM	0.02°	0.01°	0.01°
95% CI	0.39°–0.47°	0.34°–0.40°	0.37°–0.43°
Inter-Observer Error
	T1	T2	T3
MBE	0.52°	0.50°	0.46°
SD	0.16°	0.15°	0.14°
SEM	0.02°	0.02°	0.02°
95% CI	0.48°–0.56°	0.46°–0.54°	0.42°–0.50°

shows the MBE values of the different comparisons for the intra- and inter-observer error, considering n = 52.

Table 2 shows the values of the mean bias error (MBE), the reliability or standard deviation (SD), the standard error of the mean (SEM), and the SEM for a 95% confidence interval (CI). T2T1 corresponds to the pair of measurements used to obtain the intra-observer error between the second and the first measurement rounds, T3T2 corresponds to that between the third and the second, and T3T1 corresponds to that between the third and the first. T1 shows the inter-observer error in the first round of measurements, T2 shows that in the second, and T3 shows that in the third.

Analysing the intra-observer error distribution for the n = 52 sample with ANOVA, we could not accept the hypothesis that the MBE values of the T2T1 and T3T2 distributions were equal since the value of p was 0.019. Indeed, the lower value of the 95% CI (MBE − 1.96SEM) of T2T1 (0.3974°) did not overlap with the upper value of the 95% CI (MBE + 1.96SEM) of T3T2 (0.3975°), as shown in Figure 2, but it was no less true that there was such a significant difference. No significant differences were found when analysing the inter-observer error distributions by ANOVA (p = 0.155).

Figure 2 shows the 95% confidence intervals for the mean intra- and inter-observer errors. T2T1 corresponds to the pair of measurements used to obtain the intra-observer error between the second and the first measurement rounds, T3T2 corresponds to that between the third and the second, and T3T1 corresponds to that between the third and the first. T1 shows the inter-observer error in the first round of measurements, T2 shows that in the second, and T3 shows that in the third.

4. Discussion

Our research aimed to assess the validity and absolute reliability of using the TraumaMeter v.873l software for measuring LLR angular values, which was necessary to confirm the usefulness of the software in clinical practice. The most notable findings of the present study were the high inter- and intra-observer validity and reliability of the adapted TraumaMeter v.873 tool for assessing the angular values at the lower extremities.

When analysing the measurement errors (not the angular values of the HKA, which we necessarily assumed to be different between the two knees) and after examining the 26 right and 26 left knees, we determined that the mean intra- and inter-observer measurement errors were equal in both cases. For instance, a measurement error may vary depending on how an X-ray is presented to an observer (laterality and anteroposterior or posteroanterior acquisition [35]) and whether the measurements were taken on one or the other side of a knee. This result allowed us to use the total number of medical images considered in the study (n = 52) as a sample of the population of knees that could be measured with the TraumaMeter software, further increasing the statistical power and exceeding the Hopkins criteria for calculating the absolute precision [28].

Daily clinical practice and research require the measurement of variables. Each measurement process requires a measuring instrument that must be validated. For such validation, an instrument must be reliable [43]. Our study was designed to validate a semi-automatic instrument for measuring lower limb angular values in LLRs.

In the late 1980s and early 1990s, several authors established criteria for the determination of angular values relevant to the phenotypic characterisation of the lower limbs and for the calculations required for conservative management or surgical planning [10,11,44]. It is also worth noting that it was the intention of a group of rehabilitators, led by T. Derek V. Cooke, to standardise the methodology for such measurements [12,45]. This interest in validity and reliability was heightened by the enormous variability in lower limb alignments in osteoarthritic knees and the alignment changes during the arthritic process [41,46].

Interest in coronal plane angle measurements has increased significantly over the last decade (perhaps motivated by new alignment philosophies in TKA surgeries and because of the growing interest in classifying the knees according to their constitutional alignment [2,5,6]).

The reliability of different systems for measuring the same angular values (mHKA), as we have determined, has also been evaluated by other authors. For example, Marx et al. [47] aimed to assess the intra- and inter-observer reliability of lower limb alignment measurements obtained manually from hardcopy radiographs compared with measurements obtained using the Philips Easy Vision system (Philips, Eindhoven, The Netherlands). Four raters (two radiologists and two orthopaedic surgeons) blindly measured each of the 42 hardcopy radiographs and computer images on two separate occasions, separated by at least two weeks. Intra-observer reliability for the mechanical axis data measured on the PACS ranged from an intraclass correlation coefficient (ICC) of 0.93 to 0.99, and the inter-observer reliability ranged from an ICC of 0.93 to 0.97. In a prospective study comparing alignments measured on LLRs and computer tomography with intra-operative computer navigation measurements in 40 patients, Babazadeh et al. [9] observed excellent intra-observer correlations in all cases (>0.98), with a coefficient of repeatability of <1.1°. Three observers measured the alignments on LLRs three times using a computer PACS system’s Cobb angle measure (Centricity EnterpriseWeb, version 3.0, GEMedical systems, North Miami Beach, FL, USA). For pre-operative alignments as measured by LLRs, the inter-observer correlation was >0.974 (range of 0.974–0.994, p < 0.001), with the coefficient of repeatability being 2.8° [9]. Bowman et al. [48] aimed to determine the inter- and intra-observer alignment reliability before and after TKA using LLRs. Observers of differing experience levels (including an orthopaedic consultant, a senior orthopaedic registrar, a junior orthopaedic registrar, and a medical student) performed measurements for the inter-observer reliability. All the measurements were conducted on a DICOM viewer InteleViewer™ (Intelerad Medical Systems Incorporated, Montreal, QC, Canada). The authors observed excellent interobserver ICCs for all pre-operative readings of the mechanical axes on the LLRs with an ICC of >0.9 at all experience levels. Sorin et al. [49] determined the inter- and intra-observer reliability of four angular parameters (including the mHKA) for the coronal plane on digital radiographs. Four orthopaedic surgeons specialising in lower-limb arthroplasty blindly measured the angles on 42 radiographic studies twice, with a minimum of 2 weeks between measurements, using an I-sight^TM (Philips, Eindhoven, the Netherlands). The ICC for the mHKA inter-observer reliability was 0.99 (95% CI 0.97–0.99) and greater than 0.9 for the intra-observer reliability for the four raters. This article was subject to methodological criticism by Sabour and Ghassemi [50], who argued that the conclusion reached by Sorin et al. [49] may have been misleading due to the inappropriate use of statistical tests. In a study evaluating the accuracy of high tibial osteotomy planning, Jiang et al. [30] repeated measurements ten times with one of the observers. They obtained an intra-observer ICC of 0.996 (95% CI 0.989–0.999). The authors did not analyse the interobserver variability. Gieroba et al. [28] evaluated the intra- and inter-observer reliability of aHKA measurements on LLRs and compared them with the aHKA measurements on CTs. Three observers with different levels of seniority (a senior surgeon, a surgical fellow, and an orthopaedic trainee) performed the measurements in three separate sessions, each one week apart, using a hospital PACS (Centricity Enterprise Web, version 3.0, GE Medical Systems, FL, USA). An excellent intra-observer correlation was found for the aHKA (Pearson Product Moment (r) = 0.92, range of 0.87–0.95). The interobserver correlation was also excellent for the aHKA (r = 0.91).

Our research used the test-retest method rather than the parallel forms method. Each error distribution was comparable to that obtained with a two-sample t-test, which increased the statistical robustness. We could not compare our results with other published studies in numerical terms because they did not provide basic numerical measures to quantify the agreements between sets of measures (for instance, the mean bias error, standard deviation, mean absolute error, or root mean square error). An intraclass correlation coefficient only determines the average agreement between measures, and this requires contextualisation. The variability in a sample influences the reliability coefficient of a test. A test will tend to have a higher reliability coefficient when its variability is greater (Spearman–Brown prophecy formula).

5. Limitations of the Study

Besides its lack of comparison, our study had several other limitations. Firstly, we did not control for the definition and resolution of the monitors used by the different observers, and these variables may have determined the accuracy of the measurements. Secondly, only one of the eight observers was an expert accustomed in his daily work to the types of measurements used in the study. More experienced examiners tend to perform more precise measurements because their experience allows for a better understanding of anatomy and a better interpretation of the radiographs [51]. For the errors of experience groups to be representative, they should be based on several observers. It is not strictly necessary (as the ANOVA F statistic is still calculated), but it is recommended that each experience group consists of the same number of observers. We thought that in our study, we needed at least three observers in each experience group to have a representative mean of the error in each measurement. Our study did not meet this condition, and so we could not compare the errors according to experience, which would have been an added value. However, the study had strengths because it met the criteria for absolute reliability [36] and because of the robustness of the statistical method used.

6. Conclusions

The high inter- and intra-observer validity and reliability of the adapted TraumaMeter v.873 software for assessing the angular values at the lower extremities led us to conclude that the software is an accurate tool that may be helpful in clinical practice and research for measuring lower-limb angles.