Calcif Tissue Int (2006) 79:199 206 DOI: 10.1007/s00223-005-0302-6
Clinical Investigations Application of a Triage Approach to Peripheral Bone Densitometry Reduces the Requirement for Central DXA but is not Cost Effective Elizabeth J. Harrison, Judith E. Adams Clinical Radiology, Imaging Science and Biomedical Engineering, The University of Manchester, Stopford Building, Oxford Road, Manchester, M13 9PT, UK
Received: 19 December 2005 / Accepted: 25 May 2006 / Online publication: 11 September 2006
Abstract. A method proffered for the interpretation of measurements from peripheral dual energy X-ray absorptiometry (pDXA) is a triage approach to stratify patients into one of three risk categories; (i) high-treat, (ii) medium-refer for central DXA and (iii) low-reassure. The aim of this study was to apply the triage approach to measures from peripheral scanners and risk indices and stratify patients into one of three risk categories (i), (ii) or (iii). 207 post-menopausal women had central DXA from which they were categorised as non-osteoporotic or osteoporotic. Additional peripheral scans of the left calcaneus were performed on three scanners (GE Lunar Achilles and PIXI, McCue CubaClinical). From demographic details four risk indices were calculated and algorithms combining measures from peripheral scanners and one risk index were obtained. All peripheral measures, risk indices and combination algorithms were good at identifying women at risk of osteoporosis (ROC areas: 0.67 0.82). Each tool stratified varying numbers of osteoporotic and non-osteoporotic women into each risk category using the triage approach. One combination algorithm (PIXI & osteoporosis indices of risk (OSIRIS)) performed best by minimising misclassification (10% non-osteoporotic, 10% osteoporotic) and reducing requirement for central DXA to 36%. However the cost of implementing the triage approach for PIXI & OSIRIS was greater (263%) than central DXA (100%) scanning all women. Although the triage approach was an effective tool at identifying women at risk of osteoporosis the unnecessary treatment of non-osteoporotic women in the high risk category make it impractical. Therefore an alternative more cost-effective method has been suggested. Key words: Peripheral dual energy X-ray absorptiometry (pDXA) — Quantitative ultrasound (QUS) — Risk indices — Triage application
Currently the diagnosis of osteoporosis by bone densitometry as defined by the World Health Organisation Correspondence to: J. E. Adams; E-mail:
[email protected]
(WHO) is ideally applied to central dual energy X-ray absorptiometry (DXA) measurements of the hip and spine. Unfortunately, the provision of central DXA in some countries, including the UK, is limited [1]. Consequently there is growing interest in smaller, peripheral bone densitometry scanners that are more portable, relatively inexpensive and could be used in a wide range of environments (e.g. general practice) making bone assessment more accessible to a greater number of patients. Despite their potential there is uncertainty as to how best to utilise and interpret results obtained from peripheral scanners. In December 2004 the United Kingdom National Osteoporosis Society (NOS) released guidelines for the appropriate use of peripheral X-ray absorptiometry scanners for the management of osteoporosis [2]. The guidelines are for use in post-menopausal women and suggest how peripheral scanners might be used as an adjunct to central DXA. The guidelines suggest a triage approach to interpret measurements obtained from peripheral scanners by determining two thresholds to classify patients into one of three categories i) treatment recommended, particularly if accompanied with other relevant risk factors, ii) refer for central DXA and iii) no further action required if no low trauma fractures are present. Alternative methods of determining which postmenopausal women are at increased risk of osteoporosis include indices based on factors which may affect bone health. A number of simple algorithms which allow rapid detection of patients at increased risk have been developed. These include the simple calculated osteoporosis risk estimation (SCORE), osteoporosis self assessment tool (OST), osteoporosis risk assessment instrument (ORAI) and osteoporosis indices of risk (OSIRIS) screening tools which calculate risk from age, weight, oestrogen use, ethnicity and fracture history [3 8]. Once the risk has been determined the patients are stratified into a low, medium or high risk category.
200
E. J. Harrison, J. E. Adams: Triage Approach for Diagnosis of Osteoporosis
Table 1. Calculation of risk indices for osteoporosis by different published methods: Osteoporosis risk assessment instrument (ORAI) [4, 5], Osteoporosis indices of risk (OSIRIS) [7], Simple calculated osteoporosis risk estimation (SCORE) [8] ORAI Age
Non-traumatic fracture
>75 yrs 65 74 yrs 55 64 yrs 17. For OST patients were classified as having a low risk if OST > )1, a medium risk if OST was between )3 to )1 and high risk if OST < )3. For OSIRIS, patients were classified as having a low risk if OSIRIS > 1, a medium risk if OSIRIS was between )3 and 1 and a high risk if OSIRIS < )3. For SCORE patients were classified as having a low risk if SCORE < 7, a medium risk if SCORE was between 7 and 15 and a high risk if SCORE > 15. Combination Algorithm - Peripheral Scanner & Risk Index. The 90% sensitivity value and the 90% specificity value were calculated and were used as thresholds to stratify patients into one of the three categories i, ii or iii. Cost Effectiveness Scan cost data were drawn from two sources. Firstly, central DXA costs were obtained from a dedicated bone densitometry
The study population was composed of 70 osteoporotic women (age: mean 61, SD 4 yrs) and 137 non-osteoporotic women (age: mean 62, SD 4yrs). Characteristics of the study group and a summary of the peripheral scanner results are shown in Table 2. The coefficient of variation (CV) and standardized coefficient of variation (SCV = within-subject standard deviation/ 90 percentile range [17]) was calculated for each peripheral scanner. Achilles CV 3%, SCV 4.6%; CubaClinical CV 3.3%, SCV 4.6% and PIXI CV 1.8%, SCV 2.5%. Table 3 shows the regression coefficients and the integers used to calculate the combination algorithms, PIXI & OSIRIS, CubaClinical & OSIRIS and Achilles & OSIRIS. The areas under the ROC (AUCÕS) curves were 0.77 Achilles, 0.75 CubaClinical, 0.80 PIXI, 0.67 SCORE, 0.67 ORAI, 0.70 OSIRIS, 0.69 OST, 0.78 CubaClinical & OSIRIS, 0.82 PIXI & OSIRIS and 0.81 Achilles & OSIRIS. There was no significant difference (P > 0.05) between the AUCÕs with the exception of PIXI & OSIRIS and SCORE (P < 0.05); PIXI & OSIRIS and ORAI (P < 0.05); Achilles & OSIRIS and ORAI (P < 0.05). The triage threshold values for the peripheral scanners and combination algorithms are shown in Table 4.
202
E. J. Harrison, J. E. Adams: Triage Approach for Diagnosis of Osteoporosis
Table 2. Demographic details of the study population. Achilles (A), CubaClinical (C), PIXI (P) (*P < 0.001) Non-Osteoporotic patients
Age (yrs) Height (m) Weight (kg) A Stf ATscore C BUA (dB/MHz) CTscore P BMD (g/cm2) PTscore
Osteoporotic patients
Mean
SD
Min
Max
Mean
SD
Min
Max
Mean difference (P)
61 1.60 70.4 77 )1.8 66.61 )1.39 0.463 )0.46
4 0.07 12.6 13.5 1.0 12.07 0.73 0.084 1.05
55 1.45 46.0 35 )5 19.90 )4.21 0.131 )4.6
70 1.77 106.0 108 0.6 94.30 0.28 0.669 2.1
62 1.58 62.1 64 )2.8 55.34 )2.1 0.369 )1.64
4 0.07 10.8 11.7 0.9 13.73 0.83 0.081 1.01
55 1.42 43.0 38 )4.8 22.00 )4.11 0.237 )3.3
70 1.71 91.0 92 )0.6 97.07 0.45 0.647 1.8
)0.7 (P = 0.29) 0.02 (P = 0.10) 8.3* 13* 1* 11.27* 0.71* 0.094 * 1.18*
Table 3. Regression coefficients (b) used to produce three combination algorithms Achilles & OSIRIS (A & OSIRIS), PIXI & OSIRIS (P & OSIRIS) and CubaClinical & OSIRIS (C & OSIRIS) Variables b A & OSIRIS ATscore OSIRIS C & OSIRIS CTscore OSIRIS P & OSIRIS PTscore OSIRIS
)0.996 )0.237 )0.996 )0.214 )0.981 )0.149
Combination Significance algorithm