The hip radiographic magnication correlates with patient’s BMI

Obese patients have a higher prevalence of total hip arthroplasty (THA) and they are likely to experience a higher rate of pre-operative and post-operative complications. Pre-operative templating is a standard method of THA planning aiming to minimize the risk of complications. The accuracy of pre-operative templating depends on the knowledge of radiographic magnication factor. Whether and to what extent obesity affects radiographic magnication is not well described in literature. The purpose of this study was to determine whether obesity type affects hip radiographic magnication and quantify the relationship between the obesity measured and change in radiographic magnication.


Introduction
Obesity is a rising global epidemic affecting over 10% of the adult population worldwide [1]. Excessive loading due to obesity is known to be associated with a wide range of degenerative musculoskeletal diseases, not least hip and knee osteoarthritis [2,3]. Thus, obese individuals have a substantially higher rate of joint arthroplasties than non-obese ones [4]. Patient obesity presents a complex challenge during total hip arthroplasty (THA) surgery [5] and it has a negative in uence on the THA outcome [6]. The surgery time and subsequent short-term morbidity are higher in obese patients [7]. Obese patients are also at higher risk of THA instability [8], aseptic loosening and infection [4,9].
Acetate or digital templating prior the surgery allows the surgeon to reduce surgery time [10] and choose the correct implant components to restore hip biomechanics [11]. It also reduces the risk of dislocation [12], femoral fractures or loosening [13] and leg length discrepancy [14]. Therefore, pre-operative templating in obese patients requires special consideration [15]. A prerequisite for pre-operative planning is an accurate correction of the radiographic magni cation factor [16]. Radiographic magni cation is a result of the divergence of the X-ray beam and it depends on the mutual position of the X-ray source, patient and detector (Fig. 1). It is reasonable to expect that patients with a higher amount of subcutaneous fat will have more distance between hip joint and detector, thereby higher magni cation of radiograph (Fig. 1). The relation between body habitus and radiographic magni cation has not been quanti ed yet. While some studies report weak dependence between the body mass index (BMI) and measured magni cation [16,17], others conclude that the effect of BMI on pre-operative planning accuracy is negligible [18,19].
The primary aim of this study is to test the hypothesis that obesity type strati ed by BMI affects hip radiographic magni cation. The secondary aim is to quantify the relationship between patient body habitus and the increase in radiographic magni cation. Parameters describing body habitus like mass (m) and body surface area (BSA) were tested as factors in uencing hip magni cation in addition to BMI.

Patients And Methods
The radiographs taken as a part of regular follow-ups were obtained from the records of patients who underwent hip arthroplasty from November 2012 to May 2016 at ve clinics in the Czech Republic. All radiographs were exported from the hospital's picture archiving and communication systems in DICOM format and were anonymized prior to analysis. One randomly chosen hip was chosen in patients with bilateral hip arthroplasty. Patients' mass and height at the time of the follow-up were also obtained from patient records. Patients with incomplete records, e.g. missing mass at follow-up or poor quality of follow-up radiographs (motion artifacts, not clearly visible femoral head) were excluded from the study.
The study includes 303 patients, 125 male and 178 female.
The magni cation of radiographs (M) was determined by comparing the prosthetic femoral head diameter in radiographs (measured by ImageJ [20]) with the implant real size, obtained from the operative notes. The measured size was compared with the real size of the prosthetic femoral head, obtained from the operative notes. The radiographic magni cation M may differ at each clinic because of different radiological equipment and image handling as shown in the recent study [21]. To compare the global effect of obesity on the change in radiographic magni cation, the data from each clinic were normalized relative to zero median. The difference from the median (ΔM) was used in further analysis.
Analysis of original data for each clinic is provided as Supplementary les. Correlation between radiographic magni cation and obesity was evaluated in all patients. In addition to the mass and the body mass index (BMI), it is frequent practice in medicine to estimate the human body surface area (BSA) [22]. BSA of patients was calculated using the Livingston & Lee (2001) formula [23].
The patient's BMI was used to de ne obesity according to the WHO classi cation [1]. The 303 patients were strati ed according to their BMI as follows: underweight <18.4 (2) normal; 18.5 to 24.9 (61); overweight, 25 to 29.9 (119); class-I obese, 30.0 to 34.9 (87); class-II obese, 35.0 to 39.9 (27) and class-III obese, 40.0 or higher (7). Underweight and class-III obese patients were excluded from ANOVA analysis because of the small sample size.
Statistical analysis was performed using R software (R Foundation for Statistical Computing, Vienna, Austria). Groups from different clinics were compared using a Student's t test for normally distributed data. The normality assumption was veri ed graphically. Correlation was calculated with the Pearson correlation coe cient; difference between two dependent correlations was analyzed using the Williams's test. Coe cients of linear regression are expressed as the mean and the standard error of the mean (SE). One-way analysis of variance (ANOVA) was performed to examine the effect of obesity category on radiographic magni cation. The results were then further analysed with Tukey's post-hoc test. A p value 0.05 was regarded as signi cant.

Results
The cohorts from different clinics are comparable in patients' mass, height, BMI, and BSA (t-test, p=0.2). One-way between subject ANOVA was conducted to compare the effect of obesity type on the hip radiographic magni cation for patients classi ed from normal weight to class-II obesity. Obesity has a signi cant effect on radiographic magni cation (F(3,290)=19.24, p<0.001). Post-hoc comparison using Tukey's HSD test indicated that the mean change in radiographic magni cation is not signi cantly different between overweight and class-I obese patients (p=0.117, Fig. 2). The difference between normal weight and overweight patients is on the border of signi cance (p=0.031, Fig. 2).

Discussion
Subcutaneous fat in obese patients increases the distance between the hip joint and the table during pelvis AP projection (Fig. 1). Therefore, the hip radiograph magni cation should be considerably higher in obese patients. The present clinical study of 303 patients after total hip arthroplasty con rmed this assumption.  [25] also showed a weak correlation between BMI and magni cation. The weak correlation observed in previous studies, could be explained by factors others than obesity that in uence magni cation. These effects are shown as data scattering in Fig. 3 and may be attributed to interindividual variation between the patients. The data presented within this study are based on cohort taken from homogenous population in the Czech Republic where 95% of the population are Caucasians. It has been shown that body habitus also depends on race [26] and we may expect a weaker correlation in a non homogenous population. The observed correlation is also affected by the range of studied BMI, e.g. no difference could be found when comparing magni cation in normal and overweight patients (Fig. 2).
The secondary aim of the study was to quantify to what extent obesity affects patient's radiographic magni cation. Patients' mass, BMI and BSA are approximately at the same level of signi cance when predicting hip radiographic magni cation (Tab. 1). Calculation of BSA is not straightforward, and discrepancies between the most of the known BSA formulae can reach 0.5 m 2 for normal patient [27]. Therefore we do not recommend using BSA for magni cation error estimation. . We further observed that for every percent increase in magni cation, there is 5 kg/m 2 increase in BMI (Fig. 3). The value of 5 kg/m 2 corresponds with the range of each obesity type by WHO (2000) [1] and it follows that there is roughly 1 percent difference in radiographic magni cation between obesity types (Fig. 2). More precisely, the radiographic magni cation between obesity types (difference 5 kg/m 2 in BMI) increases for 0.7% and 1.2% for male and female, respectively. The correction for magni cation is higher and more reliable in females than in males (Tab. 1) as females are likely to have higher accumulation of adipose tissue in the gluteo-femoral area [29].
Our study has some limitations. The height and weight were acquired retrospectively from clinical records. Prospective data acquisition might be more reliable. The study cohort contains a limited number of morbidly obese and underweight patients and further study is required to quantify the effect of obesity on radiographic magni cation in these patients. The method based on implanted femoral head as an internal magni cation marker is adopted as a gold standard in measuring hip radiographic magni cation [17]. Within this method, the change in magni cation due to lateral shift is neglected. Analysis of Caucasian patients is presented in the current study. It potentially limits the generalizability of results to patients of African or Asian origin with different adipose tissue distribution [29]. Obesity also affects image quality in radiography due to decreased penetration and attenuation through subcutaneous fat [30]. Increased exposure factors requiring elongated exposure time results in motion artefacts [31]. Therefore, the lower quality of radiographs in obese patients is a potential source of error in radiographic magni cation estimation.
The clinically tolerable margin of magni cation error depends on the steps between implant sizes. Franken et al. (2010) [18] showed, that to template exactly for a speci c implant size, the magni cation error should be less than 2 percent for the ABG-II implant series. Our study shows that the change in radiographic magni cation for class-I obese and class-II obese with respect to normal-weight patients is outside the limit proposed by Franken et al. (2010) [18]. Pre-operative radiographs performed with a magni cation marker at the greater trochanter level could improve accuracy of hip templating in obese patients [32]. However, placing the magni cation marker in obese patients could be di cult as the marker projected image could be outside of the captured eld [33]. Some studies even propose to use xed magni cation factor instead of the magni cation marker method [24,18]. The xed magni cation estimation could be improved by considering the patient's BMI.

Conclusions
Knowledge of the exact dimensions of bone structure plays a key role in pre-operative planning. If the magni cation marker method is not applicable, BMI could be used to estimate the increase in hip radiographic magni cation due to obesity.  Table 1: Pearson correlation coefficient between the patients' parameter and the change in radiographic magnification ΔM. All correlations are statistically significant (p<0.001). Figure 1 Radiographic magni cation in (A) a normal and (B) an obese patient. Owing to the higher position of the hip above the table in the obese patient (B), the image size of the femoral head is larger than in a normal patient (A).

Figure 2
Boxplot interaction between hip radiographic magni cation and obesity. The vertical axis shows the difference from median magni cation determined at each hospital ΔM. The horizontal axis shows the obesity category classi ed by BMI (p reported from Tukey's post-hoc test).

Figure 3
Relationship between patients' BMI and change in hip radiographic magni cation ΔM. The solid black line represents the linear regression; the shaded gray area denotes the 95% con dence intervals.