The Intraoperative Pelvic Radiograph During Total Hip Arthroplasty: Is It Reliable to Estimate Leg Length?

Background: Symptomatic leg length discrepancies (LLDs) are a significant complication after total hip arthroplasty. Many surgeons incorporate an intraoperative anteroposterior pelvis radiograph, to help prevent LLD; however, obtaining a high-quality radiograph is often difficult. The purpose of this study is to evaluate the accuracy and reliability of estimating LLD using different radiographic reference landmarks on suboptimal anteroposterior pelvis radiographs. Material and methods: We obtained 2 pelvis Sawbones models with attached femurs and created a true shortening of the left femur of the experimental model by 7 mm. We then obtained a series of radiographs manipulating each model in standardized increments for a total of 66 different permutations of suboptimal radiographs. Each radiograph was evaluated for LLD by 2 separate orthopedic surgeons using reference lines bisecting the following anatomic landmarks: ischial tuberosities, acetabular teardrops, obturator foramina, sacroiliac joints, and the femoral heads, to the lesser trochanters. The accuracy and reliability of each line were then analyzed.
Results: The obturator foramina line yielded the most reliable LLD estimates with an intraclass correlation coefficient of 0.939. This reference line was also the most accurate, with an average difference of 1.5 mm from the true LLD (P < .001), with 95% confidence to be within 1.8 mm.
Conclusion: The obturator foramen reference line on an intraoperative radiograph is an accurate and reliable tool that should be utilized by joint replacement surgeons to estimate LLD even if the radiograph is suboptimal. This estimate is reliably reproduced among multiple observers and puts the estimate within 1.8 mm of a true LLD.

Introduction

Symptomatic leg length discrepancies (LLDs) are a significant complication after total hip arthroplasty among the leading causes of patient dissatisfaction [1]. These LLDs are usually preventable. There are many documented techniques to estimate leg lengths intraoperatively including manual palpation of bilateral lower extremity landmarks. Preoperative scanograms can be used to anticipate anatomic irregularities affecting leg lengths. Intraoperative radiographs can be compared to preoperative films and templating. Computer navigation and robotics also offer solutions, but are not consistently available. No method is perfect, and many surgeons adopt multiple techniques and use them as checkpoints throughout the surgery.

Using the intraoperative radiograph to estimate limb length has been shown to be reproducible [2]. A technique that connects lines from static landmarks on the pelvis and measuring their distances to the lesser trochanters has been demonstrated to be reproducibly accurate and precise in prior studies [2,3]. Unfortunately, these reference lines may be disturbed when the anteroposterior (AP) pelvis radiograph is suboptimal. This is especially true for the acetabular teardrop, a commonly used landmark. When the surgical approach requires a lateral decubitus position, getting an optimal AP pelvis radiograph presents challenges, which sometimes leads the surgeon to settle for a suboptimal radiograph or risk added case time by repeating radiographs. Maloney and Keeney defined a perfect AP pelvis radiograph as one that lacks pelvic tilt and obliquity, has symmetric obturator foramina, as well as a distance of less than 3 cm from the tip of the coccyx to the pubic symphysis [3]. Any deviations due to suboptimal radiographs can affect LLD estimations. The purpose of this study is to find the reference line on a suboptimal AP pelvis radiograph that can most reliably estimate LLD.

Material and methods

We obtained 2 adult pelvis Sawbones models with attached proximal femurs (1 control, 1 experimental). These models had 125-degree neck/shaft angles, and the attached femurs measured 20 cm in length (Fig. 1). We took a series of AP pelvis radiographs of each model starting with the pelvis in neutral position (Fig. 4). The models were held in the desired positions by supporting them with foam wedges. The distal ends of the femurs were taped to each other with a wooden dowel rod to maintain fixed positions. Orientation of the pelvis was maintained via the use of a centering crosshair on a platform. A level was placed over the pubic symphyses to obtain accurate rotational permutations (Fig. 5). The pelvis was then rotated around a vertical axis, 5, 10, 15, 20, and 30 degrees to the right and left, measured with a level. We took radiographs in each position and repeated the entire series of rotations with the pelvis then placed obliquely 15 degrees to the left and then 15 degrees to the right in the coronal plane. A goniometer and centering crosshairs on the underlying platform were used to obtain accurate oblique permutations of the models (Fig. 6). This resulted in 33 different positional iterations, representing suboptimal radiographs.

Figure 1

Adult Sawbones pelvis with attached proximal femurs.

Figure 4

Birdseye view of pelvis in neutral rotation and obliquity. The underlying radiographic platform with crosshairs was used to calibrate position. Foam wedges were used to hold the pelvis in desired positions.

Figure 5

A level is used on top of the pubic symphyses to assess rotation of the pelvis model.

Figure 6

The centering crosshairs on the underlying platform and a goniometer are used to add obliquity to the pelvis model.

We then shortened the left proximal femur of the experimental model at the femoral neck by removing a 1-cm cylindrical section, translating to a 7-mm true femur length shortening (Fig. 2). We then repeated all 33 radiographs. Our final analysis included a total of 66 different radiographs, 33 for the control and 33 for the experimental model.

Figure 2

(a) Anterioposterior (AP) pelvis radiograph of Sawbones model. (b) One centimeter cylinder removed from left hip femoral neck (7 mm true femur shortening).

All radiographs were then viewed in the Picture Archiving and Communications System by a fifth-year orthopedic surgery resident and a fellowship-trained joint replacement surgeon. They estimated LLDs by using reference lines bisecting the following anatomic landmarks—the inferior aspects of the ischial tuberosities, the inferior aspects of the acetabular teardrops, the inferior aspects of the obturator foramina, the inferior aspects of the sacroiliac (SI) joints, and the center of the femoral heads—and measuring the distance to the medial-most prominences of the right and left lesser trochanters (Fig. 3).

Figure 3

(a) Our measured reference lines demonstrated: (1) the inferior aspects of the sacroiliac joints (red), (2) the center of the femoral heads (yellow), (3) the inferior aspects of the acetabular teardrops (green), (4) the inferior aspects of the obturator foramina (orange), (5) the inferior aspects of the ischial tuberosities (blue). (b) The distance from these lines to the medial-most prominences of the lesser trochanters was recorded.

Each observer then estimated the LLD between right and left lower extremities based off the 5 reference lines. The LLD between the control and experimental groups was analyzed with the goal of finding which of the reference lines provides the most reliable estimate. We accomplished this with the intraclass correlation coefficient (ICC). ICCs give a measure of the reliability between 2 raters. Values below 0.4 indicate poor reliability, between 0.4 and 0.59 fair reliability, between 0.60 and 0.74 good reliability, and above 0.75 indicates excellent reliability [4]. Reliability was assessed via two-way random ICC with absolute agreement. The 95% confidence intervals (CIs) of the ICCs were also calculated.

We further subdivided the data into 2 groups, a minor and a major malposition group. The minor group included those radiographs taken with 0, 5, 10, and 15 degrees of rotatory deviation. The major group included those radiographs taken with 20 and 30 degrees of rotatory deviation (Tables 1 and 2).

Table 1

Difference from true in minor malpositioned (5-, 10-, and 15-degree rotation) radiographs.

	Minor rotation			One-sample t-test difference from -7 mm
	Mean (mm)	Std. deviation	Difference from 7 mm	P value
SI joint avg.	−4.7775	3.2047	2.2225	.006
Teardrop avg.	−3.365	3.53937	3.635	<.001
Obturator avg.	−5.34	2.85886	1.66	.018
Fem head avg.	0.1325	5.69994	7.1325	<.001
Ischium avg.	−9.0325	2.78096	−2.0325	.004

Table 2

Difference from true in major malpositioned (20- and 30-degree rotation) radiographs.

	Major rotation			One sample t-test difference from -7 mm
	Mean (mm)	Std. deviation	Difference from 7 mm	P value
SI joint avg.	−5.55	4.872	1.45	.325
Teardrop avg.	−4.175	9.62295	2.825	.331
Obturator avg.	−5.5833	1.80332	1.41667	.02
Fem head avg.	−0.9333	5.15343	6.06667	.002
Ischium avg.	−9.4042	2.4902	−2.40417	.007

We additionally evaluated measurements from the experimental group to identify those with the lowest average difference from the true measurement. The average difference from the true shortened value (7 mm) was calculated for each reference line with a 1-sample t-test using −7 as the reference value. A positive value indicated an underestimated value of LLD compared with the true 7-mm difference. A negative value indicated an overestimate of LLD compared with the true 7-mm difference. This was also done separately for the major and minor malposition subgroups to evaluate their influences on the error. The standard error of measurement was calculated for each measurement using the following formula: . Where σ represents the standard deviation. The r represents the reliability value obtained from the ICC. The 95% CI was also calculated, and the upper bound was selected as the cutoff at which values below that number likely represent measurement error and above are likely to represent a true difference. The measure with the lowest value for measurement error was also evaluated further to determine if the average difference varied based on minor or major malposition utilizing analysis of variance (ANOVA). Analyses were performed utilizing SPSS version 26, and alpha was set at P < .05 to declare significance (IBM Corp. Released 2019. IBM SPSS Statistics for Windows, Version 26.0, Armonk, NY: IBM Corp.)

Results

Our 2 observers estimated the LLD using each of the 5 reference lines on all 66 radiographs (33 from the control group, and 33 from the experimental group) for a total of 330 measurements.

The reference line connecting the inferior aspects of the obturator foramina had the highest ICC in all groups, 0.939, 0.944, and 0.935 in the combined, minor, and major malposition groups, respectively (Tables 3 and 4).

Table 3

Intraclass correlation, all cases.

All cases	ICC	95% CI
Reference line	Two-way random, absolute agreement	Lower bound	Upper bound
SI joint	0.744	0.613	0.835
Teardrop	0.634	0.464	0.759
Obturator	0.939	0.902	0.962
Fem head	0.902	0.841	0.939
Ischium	0.903	0.843	0.941

Table 4

Intraclass correlation, minor and major malposition groups.

Reference line	ICC
	Minor	Major
SI joint	0.855	0.635
Teardrop	0.745	0.612
Obturator	0.944	0.935
Fem head	0.929	0.868
Ischium	0.918	0.866

In the experimental group, the obturator foramina reference line produced the smallest difference from the true discrepancy, 1.5 mm (P = .001). The inferior SI joint reference line averaged a 1.9-mm difference. The acetabular teardrop reference line averaged a 3.3-mm difference. The center of the femoral head reference line averaged a 6.7-mm difference. The inferior ischial tuberosity reference line averaged a negative 2.1-mm difference. All differences were statistically significant. The purpose of statistical significance here is to demonstrate that enough radiographs were included in our experiment and that the differences detected were unlikely to be due to chance (Table 5). Our true LLD was 7 mm.

Table 5

Estimated LLD difference from true LLD (7 mm) using the average between the 2 raters for each reference line.

Experimental group
						One-sample t-test difference from -7 mm
	N	Mean (mm)	Std. deviation	Std. error of mean (SEM)	Difference from 7 mm (mm)	P value
SI joint avg.	33	−5.0758	3.79467	0.66057	1.92424	.006
Teardrop avg.	33	−3.7333	6.28976	1.09491	3.26667	.005
Obturator avg.	33	−5.4864	2.46672	0.4294	1.51364	.001
Fem head avg.	33	−0.3455	5.37477	0.93563	6.65455	<.001
Ischium avg	33	−9.1409	2.6053	0.45352	−2.14091	<.001

The upper bound of the 95% CI for the standard error of the mean (SEM) for the obturator foramen reference line difference from true is 1.8 mm (Table 6). This was the lowest SEM of all measures. Furthermore, there is no significant difference in measurement error for the obturator foramen reference line between the minor and major malposition groups (Table 7).

Table 6

Standard error of the mean (SEM) and lower and upper bounds of the confidence intervals (CI) for each reference line.

	SEMa	Lower 95% CI (mm)	Upper 95% CI (mm)
SI joint	1.9	1.8	5.7
Teardrop	3.8	3.7	11.3
Obturator	0.6	0.6	1.8
Fem head	1.7	1.6	5.0
Ischium	0.8	0.8	2.4

The standard error of the mean for each reference line in the experimental group calculated with the following formula: .

Table 7

Measurement error of the obturator foramen reference line compared between minor and major malposition groups.

Reference line
Obturator foramen	N	Mean (mm)	Standard deviation	Standard error	95% Confidence interval for mean (mm)
					Lower bound	Upper bound	Min. (mm)	Max. (mm)	P value
Minor	20	−1.66	2.85886	0.63926	−2.998	−0.322	−7.25	2	.794
Major	12	−1.42	1.80332	0.52057	−2.5624	−0.2709	−4.2	1.7
Total	32	−1.57	2.48547	0.43937	−2.4649	−0.6726	−7.25	2

Discussion

Many primary total hip replacements are still performed from a posterior approach in the lateral decubitus position. Intraoperative radiographs can be a helpful guide; however, attempts to obtain an optimal AP pelvis radiograph are often limited by patient positioning and sterility and can cause undue delay. Inadvertent LLD after total hip arthroplasty is an unfortunate complication that presents lifelong challenges to the patient. In 2007, LLD was the second highest cause for malpractice lawsuits against orthopedic surgeons according to the American Association of Hip and Knee Surgeons [1,5]. Heaver et al. in 2013 analyzed the reliability of LLD estimations using reference lines on a suboptimal AP pelvis radiograph of a Sawbones model without LLD [6]. They concluded that the ischial tuberosity reference line was least distorted by hip positioning [6]. We felt we could improve on this study by using a model with a created and measurable LLD as well as increasing the number of positional permutations. We also focused on analyzing the accuracy and reliability of the different reference lines and divided our samples into minor or major malposition categories to see if perhaps a new choice reference line would emerge with increasing distortion.

We acknowledge that there are many methods to estimate LLD, and even though the surgeon performs their due diligence in preoperative preparation, discrepancies still occur. LLD of less than 5 mm is usually asymptomatic, whereas LLDs greater than 2 cm are always symptomatic [7]. Thus, in our study, we appreciate that all our analyzed reference lines yielded average differences from true of less than 5 mm, except for the femoral head reference line, which averaged a 6.7-mm difference. Given that cases of osteoarthritis typically are not going to produce a LLD between the center of the femoral head and the lesser trochanter, this reference line is unlikely to be used clinically. This demonstrates that many reference lines may be useful in estimating LLD if one’s goal was simply to be within 5 mm of the contralateral limb. However, we demonstrate that the obturator foramen reference line is the most accurate. It yielded the smallest average difference from the true LLD at 1.5 mm. The upper bound of the 95% CI for the SEM measure was 1.8 mm; therefore, an estimated LLD of greater than 1.8 mm is likely to represent a true difference. Additionally, this value is also likely to be within 1.8 mm from the true difference. For most patients, 1.8 mm represents a clinically insignificant LLD [[7], [8]]. We assessed the measurement error of this reference line in both the minor and major malposition groups and found no significant difference.

Of the 5 previously described reference lines we analyzed, the obturator foramen reference line was also the most reliable between observers (ICC 0.94) in estimating LLD. The SI joint and acetabular tear drop references lines were the least reliable (ICC 0.74 and 0.63, respectively), likely because these landmarks become the most distorted and difficult to identify on a suboptimal radiograph. This provides strong support that the obturator foramen reference line can be used reliably and accurately in an intraoperative setting, when compared with other commonly used reference lines, even if the radiograph is suboptimal. Although estimating LLD on the intraoperative radiograph is only one of the many safety checkpoints surgeons use, the obturator foramen is the reference line of choice and will be incorporated in our practice.

Our study is limited by our Sawbones models, which do not fully simulate the live pelvis in an intraoperative environment. Our method of LLD estimation is confined to the hip joint and does not account for any LLD that could be attributed to intrapelvic or suprapelvic pathology. Additionally, we realize that proximal femurs are difficult to manipulate in a live intraoperative environment. Therefore, it is important that this technique be considered along with other methods of assessing LLD during total hip arthroplasty.

Conclusions

For those surgeons who obtain intraoperative radiographs to assess leg lengths, the quality of the radiograph is often suboptimal due to patient positioning (commonly lateral decubitus), sterility, and time. This study supports the use of a reference line bisecting the inferior aspects of the obturator foramina, referenced against the medial prominences of the lesser trochanters, even when the radiograph is suboptimal. This estimate is accurate and reliably reproduced among multiple observers and puts the estimate within 1.8 mm of a true LLD.

Conflicts of interest

The authors declare that there are no conflicts of interest.

For full disclosure statements refer to https://doi.org/10.1016/j.artd.2022.03.009