Conference PaperPDF Available

DEXA Over Estimates Bone Mineral Density in Patients with Total Joint Replacements: The Critical Importance of Correcting Automated Error for Improved Screening Accuracy

Authors:

Abstract and Figures

INTRODUCTION: Osteopenia and osteoporosis are modifiable health factors, particularly for the population group who are most likely to receive a total joint replacement. The World Health Organization has established dual-energy x-ray absorptiometry (DEXA) as the standard assessing bone mineral density (BMD). The purpose of this study is to determine if there are irregularities noted in the diagnosis of or osteopenia (T-Score<-1) or osteoporosis (T-Score<-2.5) in patients with in situ implants. METHODS: Thirty-eight older patients (m=15, 61±7yr, BMI 32.3±5.7 kg·m2 | f=23, 67±6yr, BMI 31.0±6.2 kg·m2) who previously underwent total knee arthroplasty (TKA) for a single limb within the past year consented to have their total body and regional BMD assessed by DEXA (iDXA, GE®). To determine the regional effect of TKA implants on BMD analysis of the legs by DEXA, BMD was recorded as an average of both legs and individually for the CONTROL (no implant) and TKA (with implant) limbs. To determine the effect of the implants on total body measures, total-body BMD was analyzed cumulatively as well as for the non-TKA side only. The impact of automated T-scores were then calculated as a common metric for diagnosis of osteopenia and osteoporosis using enCORE.v16 software (GE®) software and standardized national norms for bone densitometry. Within the analysis software, implant materials were automatically detected and removed as artifact (typical analysis). To examine the effectiveness of the software of artifact removal, additional measures were taken and scored when implant artifact was manually removed (manual artifact correction –MAC). Finally, a correction factor was developed for application in scenarios where artifact removal was not performed either via software or manual removal. A mixed-model ANOVA was used to compare BMD and t-scores between the control limb, the TKA limb, and the TKA limb following MAC (TKA-MAC) as well as for comparing total-body BMD, total-body BMD-MAC, and BMD assessed from the non-TKA side only. Significant interactions were followed by a Tukey’s post hoc test for pairwise comparisons. To validate the correction factor, 45 additional TKA patients were recruited who also underwent DEXA (m=30, 64±6yr, BMI 32.6±6.1 kg·m2 | f=15, 67±8yr, BMI 31.5±5.2 kg·m2). A Bland-Altmon analysis was used to detect proportional measurement bias between the BMD values corrected using the correction factor compared to MAC. Type I error was set as α = 0.05 for all analyses. For pairwise comparisons of control and TKA limbs as well as total-body BMD and BMD assessed from the non-TKA side, effect size was also calculated using a Cohen’s D statistic. RESULTS: When using standard automatic artifact removal software, limbs with TKA implants had significantly higher BMD (~15% difference, p<0.01) compared to control limbs. This resulted in a 2.7% (range 0.8-6%) overestimation of total body BMD (per implant) (Fig 2). These results were consistent across different model types (GE, HEXA). As a result, patients who would have been observed to have low leg (18% of patients) or total BMD (5.3% of patients) would have gone un-diagnosed without manual correction to the image or the BMD value (p<0.05). Manual correction correlates closely to nonimplant BMD. In the event of automated / non-manually corrected DEXA scans, the following correction factors were developed to correct for discrepancies: Limb BMD: BMD x (1-0.127[per implant]), Total body BMD: BMD x (1-0.030[per implant]). When validating the use of the correction factor in our secondary population, Bland-Altman analysis revealed no bias between BMD values adjusted using the calculated correction factors (Limb BMD =1.316 ± 0.028 | Total body BMD =1.315 ± 0.024) compared to MAC (Limb BMD =1.321 ± 0.017 | Total body BMD =1.324 ± 0.025). Similarly, the linear relationship between observed between predicted and MAC measures as well as no significant difference detected between mean also indicated instrument agreement for limb (R2=0.96), and total body (R2=0.99) BMD (p<0.05). Therefore, the correction factors represent a validated proxy to correct for implant error in the event that a scan has not undergone proper MAC for implant artifact. DISCUSSION: Conventional, gold standard means for evaluating bone health via DEXA over-estimates limb and overall BMD in patients with in situ total knee replacements, and therefore under-estimates incidence of low BMD, osteopenia, and osteoporosis in at-risk populations. Patient’s who have joint implants should be flagged as such and closer scrutiny of their DEXA scans should be performed. Normalization of values can occur through thorough manual artifact removal, or by applying correction factors described in this study. Future research should focus on overall national prevalence of this phenomenon and which patients’ are most at-risk for having incorrect DEXA study results. SIGNIFICANCE: The results presented in this investigation may aid and alert practitioners to more closely scrutinize DEXA measures taken on patients with joint replacements for improved diagnostic quality and identification of patients who may benefit from interventions designed to mitigate fragility based fracture risk.
Content may be subject to copyright.
Based on the mean prediction error for single leg, legs, and total body BMD, the following
correction factors were developed to correct for artifact error in the event of automated /
non-manually corrected DEXA scans:
Single Leg BMD BMD x (1-0.127)
Legs BMD BMD x (1-0.066 [per implant])
Total body BMD BMD x (1-0.030 [per implant])
To validate the correction factor for in situ knee implants, forty-five additional patients were
recruited from the same population and underwent the same analysis using DEXA (m=30,
64±6yr, 176.8±7.4cm, 102.0±19.2kg, BMI 32.6±6.1 kg·m2 | f=15, 67±8yr, 163.1±6.0cm,
84.1±16.2kg, BMI 31.5±5.2 kg·m2). A Bland-Altmon analysis for instrument agreement was
used to detect proportional measurement bias between the BMD values corrected using the
correction factor compared to MAC. Type I error was set as α = 0.05 for all analyses.
FIGURE: Data plots representing the results of a Bland-Altman analysis for instrument agreement between adjusted BMD
corrected by either by either manual artifact correction (MAC) or application of correction factors. Significance for the
initial detection of disagreement and post hoc regression to detect proportional bias was set at α = 0.05.
KEY FINDING: Bland-Altman analysis revealed no proportional bias between BMD values
adjusted using the calculated correction factors (Limb BMD = 1.316 ± 0.028 | Total body
BMD = 1.315 ± 0.024) compared to MAC (Limb BMD = 1.321 ± 0.017 | Total body BMD =
1.324 ± 0.025). Similarly, the linear relationship between observed between predicted and
MAC also indicated instrument agreement for limb (R2=0.96), and total body (R2=0.99) BMD
(p<0.05).
BACKGROUND
METHODS
Total Body & Limb Specific
Bone Densitometry
Results & Follow-up Analysis
Correction Factor For Knee Implants
DEXA Model Specific?
Standard software based artifact removal typically used in DEXA analysis results in an
overestimation of BMD in patients with joint implants that may result in misclassification of bone
health and risk for low BMD, and osteoporosis.
Therefore, those with total-joint replacements who may stand to benefit from intervention may
remain at an elevated risk for fragility fractures.
Importantly, manual artifact correction / removal can successfully overcome software based BMD
overestimation.
Therefore, it is critical that technicians screen for patients who have joint replacements and take
appropriate image correction measures. It is also critical that physicians take note of regional limb
measures (example: right leg, left leg, etc) as well as total body measures to ensure that BMD
overestimation and misdiagnosis does not occur in this population.
On the occasion that a patient presents with a DEXA report whereby manual artifact correction was
not properly performed for one or more implants, the correction factors generated here may be of
use for approximating a corrected BMD value.
However, we caution the reader that these findings are limited to primary knee implants. Further
research will be required to determine the degree to which other types of joint replacements result in
overestimations of BMD by standard automated software analysis.
DEXA Overestimates Bone Mineral Density in Patients with Total Joint Replacements:
The Critical Importance of Correcting Automated Error for Improved Screening Accuracy
Bradley S. Lambert, Ph.D; Katharine D. Harper, M.D.; Terry A. Clyburn, M.D.; Stephen J. Incavo, M.D.
1
Orthopedic Biomechanics Research Laboratory, Department of Orthopedics and Sports Medicine, Houston Methodist Hospital;
RESULTS
REFERENCES: [1] Wright NC, Looker AC, Saag KG, Curtis JR, Delzell ES, Randall S, Dawson-Hughes B. The
Recent Prevalence of Osteoporosis and Low Bone Mass in the United States Based on Bone Mineral Density at
the Femoral Neck or Lumbar Spine. J Bone Miner Res. 2014 Nov; 29(11): 2520–2526. [2] El Maghraoui A, Roux C.
DXA scanning in clinical practice. QJM. 2008 Aug;101(8):605-17. [3] Kanis JA. Assessment of fracture risk and its
application to screening for postmenopausal osteoporosis: synopsis of a WHO report. WHO Study
Group, Osteoporos Int, 1994, vol. 4 (pg. 368-81). [4] United States Census Bureau. 2017. https://www.census.gov/.
Accessed 2018 Apr 11. [5] Sloan M, Premkumar A, Sheth NP. Projected Volume of Primary Total Joint
Arthroplasty in the U.S., 2014 to 2030. J Bone Joint Surg Am. 2018;100:1455-60
Osteopenia and osteoporosis are important modifiable health factors.
In 2010, it was estimated that 10.2 million adults were diagnosed with
osteoporosis and that 43.4 million adults had low bone mass (1).
Dual energy x-ray absorptiometry (DEXA) allows for accurate
screening for osteoporosis, estimation of fracture risk, and monitoring
of patients undergoing treatment (2).
The World Health Organization has established DEXA as the
standard densitometric technique for assessing bone mineral density
(BMD) in postmenopausal women and based the definitions of
osteopenia and osteoporosis on its results (3).
In the same population group that osteoporosis and osteopenia
present, patients are also most likely to receive a total joint
replacement. Most patients who receive a total hip arthroplasty (THA)
or total knee arthroplasty (TKA) are between 65 and 85 years of age.
Current total THA and TKA volume in the United States is
approximately one million / year, and is expected to grow to ~1.5
million/year by the year 2030 (4,5).
Typically, automated software is relied upon to remove image artifact
of non-tissue origin (such as implant materials).
Therefore the purpose of this study was to evaluate the analysis of
DEXA scans in patients with total knee arthroplasty to determine if
there are irregularities noted in the measurement of BMD and the
diagnosis of low BMD, osteoporosis, or osteopenia in patient’s with in
situ implants.
Thirty-eight older patients (m=15, 61±7yr,
178.1±7.7cm, 103.5±24.1kg, BMI 32.3±5.7 kg·m2 |
f=23, 67±6yr, 163.0±6.2cm, 82.7±19.4kg, BMI
31.0±6.2 kg·m2) who previously underwent total knee
arthroplasty (TKA) for a single limb within the past year
consented to have their total body and regional bone
mineral density assessed by DEXA scan (iDXA, GE®;
commonly used in radiology clinics for assessment of
bone densitometry).
To determine the regional effect of TKA implants on
BMD analysis of the legs by DEXA, BMD was
recorded as an average of both legs and individually
for the CONTROL (no implant) and TKA (with implant)
limbs.
The impact of implants on automated T-scores were
then calculated as a common metric for diagnosis of
osteopenia and osteoporosis using enCORE software
and standardized national norms for bone
densitometry (6).
Within the analysis software, implant materials were
automatically detected and removed as artifact by the
software (typical analysis). To examine the
effectiveness of the software of detecting and
removing implant artifact from the analysis, additional
measures were taken and scored when implant artifact
was manually removed from the images.
Finally, a correction factor was calculated for
application in scenarios where adequate artifact
removal did not occur via automated software or
manual techniques.
Statistical Analysis: A mixed-model ANOVA was used
to compare BMD and t-scores between the control
limb, the TKA limb, and the TKA limb following manual
artifact correction (TKA- MAC) as well as for
comparing total-body BMD, total-body BMD-MAC, and
BMD assessed from the non-TKA side only. All
detected significant interactions were followed by a
Tukey’s post hoc test for pairwise comparisons.
Houston Methodist OBRL
5505 W. Loop South Freeway, Houston, TX 77081
Phone: (___) ___-____
Statistically 68% of repeat scans fall within 1SD (± 0.010 g/cm² for Total Body Total); USA (NHANES 1999-2004) Total Body, Male Reference Population (v100); Matched for Age, Sex, Ethnic
Date created: 09/25/2018 3:27:02 PM 16 [SP 2]; Filename: 00hkfp4ilb.meb; Total Body; 100:0.19:153.85:15.6 0.00:-1.00 2.40x3.04 13.2:%Fat=34.0%; 0.00:0.00 0.00:0.00; Scan Mode: Standard; 3.0
µGy
GE Healthcare Page: 1 of 1 Lunar iDXA
ME+210719
Patient: INC1-058, INC1-058 Referring Physician: (not specified)
Birth Date: 11/30/1953 Age: 64.8 years Patient ID: INC1-058
Height: 72.0 in. Weight: 187.0 lbs. Measured: 09/24/2018 10:49:17 AM (16 [SP 2])
Sex: Male Ethnicity: White Analyzed: 09/25/2018 3:26:37 PM (16 [SP 2])
Total Body Bone Density
COMMENTS:
Densitometry: USA (NHANES 1999-2004) (Enhanced Analysis)
Region
BMD
(g/cm²)
AM
Centile
AM
Z-score
Head 2.200 - -
Arms 0.841 - -
Legs 1.227 - -
Trunk 0.936 - -
Ribs 0.841 - -
Spine 1.104 - -
Pelvis 0.926 - -
Total 1.146 29 -0.6
FIGURE: Data are presented as mean±SEM for BMD (g·cm2) of the legs and total body measured using DEXA (A & B) with corresponding T-scores
calculated for each measure (C & D). Regional limb comparison (A&C): CONTROL, limb with no implant present; TKA, limb with TKA implant
assessed using standard iDXA software analysis; TKA-MAC, limb with TKA implant following manual artifact removal. Total-Body Comparison (B&D):
Total Body, total-body BMD with implant included assessed using standard iDXA software analysis; Non-TKA Side Only, BMD assessed from the
lateral side of the body with no implant; Total Body – MAC, total body BMD analyzed following manual artifact removal. Individual patient data are
also shown for those classified as normal, osteopenic (T < -1), or osteoporotic (T < -2.5) regionally (E) and for the total-body across each analysis
comparison (F).
KEY FINDINGS:
Those with TKA implants had significantly higher BMD (~15% difference, p<0.01) compared to control limbs when
standard automated artifact removal was used.
This resulted, on average, in a 2.7% (range 0.8-6%) overestimation of total body BMD (for a single implant).
When analyzing the individual limbs, 7 of the 38 patients (18.4%) were found to have low BMD falling into the
osteopenic (T-score < -1) or osteoporotic (T-score < -2.5) categories. This was masked in the TKA limb whereby
0/38 (0%) patients were observed to have a T-score below 1 indicating that the implant resulted in an overestimation
of BMD.
With regards to the impact of a single knee implant on the total body analysis (Figure 2F), 2 of 38 patients (5.3%)
were observed to be osteopenic when analyzing the non-TKA side only compared to the actual total body
measurement (including the TKA implant) where 0/38 (0%) patients were observed to be osteopenic.
Manual artifact correction (MAC) was found to successfully ameliorate the overestimation of BMD for both regional
and total-body measures (Figure 2A – 2F).
All patients in this study
underwent DEXA analysis
from the same model of
DEXA.
However, we did observe this
error to be present in a single
case scan performed using a
separate DEXA model
(Horizon, Hologic, San Diego,
CA) where the overestimation
of BMD was present similar to
the scans in this investigation.
Therefore, more research is
needed to determine the
extent that other software
designs have issue with
artifact removal.
Orthopedic Biomechanics
Research Laboratory
Discussion & Conclusions
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.