Strong Bone Asia 2013 Osteoporosis in ASEAN Radiation Physics Principles of DXA Basic Statistics for DXA Essential Anatomy for DXA Chris Schultz Scientist-in-Charge (Bone Densitometry) Royal Adelaide Hospital, Adelaide, Australia Copyright 2005-2013 This material is Copyright by the Council of the Australian & New Zealand Bone & Mineral Society (ANZBMS) and may not be copied or transmitted by any means without written permission from the Council of the ANZBMS.
Strong Bone Asia 2013 Osteoporosis in ASEAN Radiation Physics Ionising Radiation in Bone Densitometry Copyright 2005-2013 This material is Copyright by the Council of the Australian & New Zealand Bone & Mineral Society (ANZBMS) and may not be copied or transmitted by any means without written permission from the Council of the ANZBMS.
Electromagnetic Spectrum Wavelength Frequency (Energy) Ionising 10-9 10-6 10-3 1 10 3 10 6 10 9 ev Radio waves Powers of ten represent orders of magnitude difference, with respect to the visible spectrum, of the photon energy (or frequency) Also represent the approx. photon energy in electron volts (ev; 1eV = 1.6 x 10-19 J) Infra-red Ultra violet Visible DXA X-rays Therapy X & rays ANZBMS 2005-2013 - Slide 3
Key properties of Diagnostic X-rays Diagnostic X-rays have photon energies in the range 20-150keV Absorption & scattering of X-rays occurs in matter - including human tissue. X-rays interact with atoms (ie electron clouds) not molecules or cells X-rays produce biological effects that are seen as direct or indirect actions on living cells Johns & Cunningham (1983) ANZBMS 2005-2013 - Slide 4
Penetration of Matter by Diagnostic X-rays aluminium lead photon energy X-ray beam Basis of X-ray attenuation, imaging, & shielding Aluminium selectively attenuates low energy photons (beam hardening) removing rays that don t contribute to the image ( filtration ) Body tissue allows some of the beam through without interaction, creating the image. This is the primary beam. The rest of the beam causes the radiation dose Lead blocks the beam and is suitable for radiation shielding ANZBMS 2005-2013 - Slide 5
Interactions between X-rays & Matter Absorption with production of secondary electrons &/or new X-rays Scattering resultant photon has lower energy and new direction Heggie et al (2001) Attenuation is the sum of these - the reduction in the forward intensity of the X-ray beam as it traverses matter - depending on: The physical density & atomic composition of the material The energy spectrum of the X-ray beam ANZBMS 2005-2013 - Slide 6
Inverse Square Law of radiation intensity or dose When the distance from a point radiation source is doubled, the intensity is reduced by one quarter. This is the basis of radiation protection by distance I 2 I 2 I 1 d d 2 1 2 2 I 1 source d 2 d 1 ANZBMS 2005-2013 - Slide 7
Biological Effects of Ionising Radiation CELL DAMAGE ALTERED METABOLISM & FUNCTION TRANSFORMATION REPAIR CELL DEATH Radiosensitivity of organelles and cells varies DNA region > 100 times more sensitive than cytoplasm Ion pairs created in or near DNA by radiation (and by chemical environmental insults) Therefore rapidly dividing cells more susceptible to damage (eg ovaries, testes, bone marrow, intestinal epithelium) The foetus and young children are therefore more sensitive to the effects of radiation Hallenbeck (1994) ANZBMS 2005-2013 - Slide 8
Types of Radiobiological Effects (A): Deterministic Directly caused by the radiation (eg destruction of bone marrow cells following radiotherapy level dose) Cellular repair mechanisms overwhelmed by the severity of the insult and most cells die Occur only above a specific threshold - typically as high as 200 times yearly background dose Above this threshold severity of effect dosedependent eg; ICRP60 (1990) ANZBMS 2005-2013 - Slide 9
Types of Radiobiological Effects (B): Stochastic Stochastic effects are assumed to arise from damage to a single cell, from low radiation doses The majority of damaged cells die, but a few abnormal cells survive The effect cannot be causally related to a particular radiation incident, because the consequences usually arise years later The probability of occurrence, but not the severity of a stochastic effect is proportional to the dose received It is possible that there is no safe dose, though at low doses this is an assumption based on mathematical modelling and not evidence based Most radiation protection environments and calculations (including bone densitometry) are concerned with assessing and predicting stochastic effects eg ICRP60 (1990) ANZBMS 2005-2013 - Slide 10
Strong Bone Asia 2013 Osteoporosis in ASEAN Radiation Physics Quantitative Radiation Protection Copyright 2005-2013 This material is Copyright by the Council of the Australian & New Zealand Bone & Mineral Society (ANZBMS) and may not be copied or transmitted by any means without written permission from the Council of the ANZBMS.
Long-term Biological Effect of Exposure to Ionising Radiation The biological consequences of ionising radiation are among the most predictable of all environmental hazards This is because: the magnitude of radiation exposure can often be recorded (or estimated) reasonably accurately deep follow-up studies have been carried out on thousands of victims for up to 65 years The key to predicting long term detriment is to devise a quantitative measure of the radiobiological effect ANZBMS 2005-2013 - Slide 12
Calculating Effective Dose (E = H T. T ) Definition: Effective Dose (E) sum of the equivalent doses (H T ) for each organ or discrete exposed tissue multiplied by tissue weighting factor ( T ) for that organ sum for all tissues and organs within the exposed region of the body SI Unit : sievert (Sv) E in bone densitometry often quoted in microsievert ( Sv) Effective Dose is the most biologically direct, quantitative measure of the long term risk of severe detriment (cancer, inherited effects) from the stochastic effects of radiation Tissue or Organ Tissue Weighting Factor ( T ) * Gonads 0.08 Bone marrow (red), Colon, Lung, Stomach, Breast (each) Bladder, Liver, Oesophagus, Thyroid (each) Skin, Bone surface, Brain (each) Other organs/tissues (entire) 0.12 0.04 0.01 0.13 Total 1.0 relative radio-sensitivities of organs/tissues ICRP60 (1990), ICRP103 (2007) ANZBMS 2005-2013 - Slide 13
Calculation of stochastic risks for Adults Risk of long term effects Multiply the effective dose in Sv by the coefficient below for the relevant population & detriment Exposed population Detriment (Sv -1 ) All Cancer Severe hereditary effects Total Adult Workers 0.041 0.001 0.042 Whole Population 0.055 0.002 0.057 Note that workers are more resistant to radiation ICRP103 (2007) ANZBMS 2005-2013 - Slide 14
Calculation of Stochastic Risks for Adults Equivalent examples Ionising Radiation Person receiving a man-made dose equal to background radiation per year (~2.5mSv) for 10 years has (2.5 x10-3 ) x 10 x 0.05 = ~0.1% extra subsequent lifetime risk of cancer that could be fatal (ie above normal risk ~25%) Roads Person has about the same (0.1%) risk of dying on Australian roads over a 10-year period ICRP60 (1990) Australian Bureau of Stats (2002) ANZBMS 2005-2013 - Slide 15
Standard DXA Effective Doses ( Sv) Scan Mode PA Spine (L1-L4) Prox. Femur (inc. ovaries) Total Body (ex. ovaries) Total body (inc. ovaries) QDR 4500/Delphi Fan Beam a Expert Fan Beam b Norland (Pencil Beam) Systems c 7 4 1 5 4 1 3 <1 <0.5 3 <1 d <1 Notes: * Lateral spine can deliver up to 2-6 x higher dose than PA, depending on low vs high definition * Quoted values are effective doses, not the more frequently-quoted entrance surface doses (ESD), which are easier to determine but are 2 steps removed from biological significance. * In postmenopausal women, the ovaries no longer carry a significant organ weighting (w T ). Refs: a: Patel et al. (1996). b, c; Established from manufacturer quoted entrance exposures or doses, & Bezakova et al. (1997), Huda & Moran (1996). ANZBMS 2005-2013 - Slide 16
Comparative doses in medical diagnosis & daily life Comparison Dose ( Sv) BERT * Natural Background (all sources) 5-8/day a 1 day Trans-Pacific Return Flight (cosmogenic burden) ~50 b ~4-6 days Standard chest X-ray 25-60 c 3-7 days Mammography ~450 d ~56 days Thoraco-lumbar Lateral & AP X-ray set 500-700 e 71-100 days Thoracic CT or Whole-body PET >7000 f 1000 days Two PA DXA scans (excluding LVA) <12 g < 1 day * Background Equivalent Radiation Time (days), based on average background dose Refs: a: UNSCEAR (2000). b:adapted from Hallenbeck (1994). c,d,e: Standard values from variety of sources, inc. Johns & Cunningham (1983) & NRPB Vol. 10 #1 (1999); f: Price et al calc.mtp, SCGH, Perth, based on 16-slice CT & GSO crystal PET camera (2005). g: See previous slide ANZBMS 2005-2013 - Slide 17
Strong Bone Asia 2013 Osteoporosis in ASEAN Radiation Physics Minimising Radiation Doses to Patient & Technologist Copyright 2005-2013 This material is Copyright by the Council of the Australian & New Zealand Bone & Mineral Society (ANZBMS) and may not be copied or transmitted by any means without written permission from the Council of the ANZBMS.
Explaining the Risk-Benefit of Radiation to Patients Yearly natural background ~ >100 times most DXA Radiation dose in intercontinental flights >5-10 times DXA Other commonly accepted risk factors (eg car accidents) outweigh radiation risks (be prudent though!) Benefits of the DXA investigation - likely impact on quality & length of life (~30% untreated lifetime fracture risk, reduced by >half by proper management, in the spine) The deep knowledge of risks of radiation (eg ICRP, IAEA), compared with other hazards ANZBMS 2005-2013 - Slide 19
Patients Scanning children Extra duty of care must be applied to children Reference ranges acquired using children should be subjected to the strictest of justifications Care should be taken by Technologist to ensure that no scan is wasted Radiosensitive organs (eg gonads) should be shielded, wherever possible Children would not be scanned using QCT, except as pqct (eminently suited to children) Dose limits for children recognise their developmental status eg; ICRP85 (2000); ARPANSA, #8 (2004); Hall, Paediatric Radiology (2002) ANZBMS 2005-2013 - Slide 20
Patients Pregnant or Possibly pregnant Exposures to the axial skeleton incur the highest burden Possibly pregnant patients require particular duty-of-care Pregnancy status of reproductive women should always be established ask! The relative risk depends on the stage of pregnancy. The most sensitive period is 6-10 weeks Bone densitometry is never so urgent that it cannot be postponed However, in all bone densitometry, in the event of any inadvertent dose to the foetus, stochastic risk is small, compared to the natural occurrence of 4-6% congenital risks Summary In all cases of the inadvertent use of DXA, the pregnant patient may be reassured of negligible risks eg; Given-Wilson (1993); Lloyd et al (1998); NHMRC (1986) ANZBMS 2005-2013 - Slide 21
Strategies for Minimising Technologist Doses (I) General ALARA Strategies Time Minimise contact with X-ray beam. Dose is proportional to time spent Distance Maximise distance from source. Protection afforded by inverse square law Shielding Screen (if necessary) between technologist and X-ray source, particularly if work station in same room as DXA. A lead glass screen of 0.25 mm lead-equiv. thickness represents >4HVL for DXA scattered radiation >93% of radiation blocked ANZBMS 2005-2013 - Slide 22
Personal Radiation Monitoring: Judging the Need Some Radiation Regulators require personal radiation monitoring of DXA Technologists, at least initially Application for exemption can be made on basis of dose records that show no significant readings (eg one year) If staff feel safer with personal radiation monitoring, then it might be prudent to offer it, until they acknowledge the lack of need Exceptions, suggestive of continuous monitoring, are: When work practices habitually require more physical interaction between technologist and patient during scanning (eg paediatrics) In the case of QCT using a whole body scanner, the technologist will be a licensed radiographer ANZBMS 2005-2013 - Slide 23
Strategies for Minimising Technologist Doses Pregnant staff ICRP 60 recommends that the dose to the surface of mothers abdomen during pregnancy should be <2 msv For bone densitometry operators this is easily avoided without taking extra precautions Pregnancy in the workplace should be handled with sensitivity. In DXA-only workplaces, the exposure to the foetus will always be negligible, but perception of risk creates anxiety Concern can be dealt with by education and negotiation. A leaded apron costs <$1000 and absorbs the vast majority of dose (0.25 mm lead equiv. Gown ~ >4 HVL of scattered DXA radiation) ICRP60 (1990) ANZBMS 2005-2013 - Slide 24
Strategies for Minimising Doses to Technologists Specific Strategies Strictly minimise exposure to direct beam (relevant for when restricting patient movement) Check that Safety signs (for both ionising radiation & lasers) & lights (if relevant) are logical and obeyed by personnel & patients ionising laser Conduct systematic, regular Best Practice Quality Assurance (reduces need for repeat scanning) Seek licensed maintenance & repair personnel ANZBMS 2005-2013 - Slide 25
Strong Bone Asia 2013 Osteoporosis in ASEAN Principles of Dual Energy X-ray Absorptiometry Copyright 2005-2013 This material is Copyright by the Council of the Australian & New Zealand Bone & Mineral Society (ANZBMS) and may not be copied or transmitted by any means without written permission from the Council of the ANZBMS.
Photon Absorption Mass attenuation coefficient Photon attenuation: Depth = attenuation Atomic number = attenuation photon energy = attenuation Mass attenuation coefficient ( ) for a given material: decreases with photon energy rises with atomic number Sharp changes associated with energy levels (shells) of atom ANZBMS 2005-2013 - Slide 27
Mass Attenuation coefficient (cm 2 /g) Mass attenuation coefficient & ideal energies 2 distinct energies can differentiate between tissue and bone Photon absorption water, tissue, muscle are similar in fat in bone 10000.00 1000.00 100.00 10.00 Soft Tissue 1.00 Adipose Tissue Muscle, Skeletal Cortical Bone 0.10 Water Calcium 0.01 1.00 10.00 100.00 1000.0 Photon Energy (KeV) ANZBMS 2005-2013 - Slide 28
Theory of mass calculation (1) Solve for two masses bone and soft tissue Beam attenuation soft tissue low energy : high energy Equivalent to ratio of mass attenuation coefficients at each energy M M bone tissue I ln( I μ 0high high bonelow I ln( I μ R R 0high high tissuelow bone tissue ).R μ μ μ ).R μ μ μ tissue bonehigh bone I ln( I.R tissuehigh bonelow bonehigh tissuelow tissuehigh 0low low tissue I ln( I.R 0low low bone ) ) ANZBMS 2005-2013 - Slide 29
Mass Attenuation coefficient (cm 2 /g) Theory of mass calculation (2) Scan spine, femur, forearm and total body (any site) Fat correction possible Ideal energies calculated to be between ~20Kev and ~100KeV 10000.00 44KeV 100KeV 1000.00 100.00 10.00 Soft Tissue 1.00 Adipose Tissue Muscle, Skeletal Cortical Bone 0.10 Water Calcium 0.01 1.00 10.00 100.00 1000.00 Photon Energy (KeV) ANZBMS 2005-2013 - Slide 30
Strong Bone Asia 2013 Osteoporosis in ASEAN Principles of Dual Energy X-ray Absorptiometry DXA Scanner structure, X-ray generation Beam collimation Copyright 2005-2013 This material is Copyright by the Council of the Australian & New Zealand Bone & Mineral Society (ANZBMS) and may not be copied or transmitted by any means without written permission from the Council of the ANZBMS.
Scanner structure 10 00 10 01 00 01 11 11 01 01 10 01 00 10 11 Diagram adapted from Wahner HW et al J Nucl Med. (1984) ANZBMS 2005-2013 - Slide 32
X-ray generation Hologic Switch tube kv between 100 & 140kV (was 70 & 140kV) Pulse tube at power line frequency Calibrate & correct for beam hardening using filter wheel R tissue constant (corrected calibration empirically) CdWO 4 detector initially single, followed by array ANZBMS 2005-2013 - Slide 33
X-ray generation LUNAR Factory calibrated to ashed bone Rare earth filter (Ce or Sm) uses K shell absorption of photon energy as filter Tube ma & speed changes alter dynamic range R tissue corrected pixel by pixel using tissue adjacent to bone CdZnTe in narrow fan beam configuration, without need for dead time correction Phantom block for Quality Assurance checks engineering specification only ANZBMS 2005-2013 - Slide 34
Pencil beam vs Fan Beam Pencil beam 1:1 relationship of source & detector Relatively slow scan time Fan Beam Faster scan Inverse square law applies at edges (reduced photon flux) Magnification effect (source location proximity to patient bone) Narrow fan beam Rectilinear type scans Some beam overlap Potentially less artefact Slower than full fan beam ANZBMS 2005-2013 - Slide 35
Strong Bone Asia 2013 Osteoporosis in ASEAN Principles of Dual Energy X-ray Absorptiometry 3 tissue component approximation Copyright 2005-2013 This material is Copyright by the Council of the Australian & New Zealand Bone & Mineral Society (ANZBMS) and may not be copied or transmitted by any means without written permission from the Council of the ANZBMS.
3 compartments from 2 DXA is 2 compartment model bone and soft tissue Possible to derive fat and lean mass Relies on documented fat and lean R values Limited by masking due to bone areas estimate ANZBMS 2005-2013 - Slide 37
3 compartment derivation Given R material μ μ materiallow materialhigh and Fraction fat + Fraction lean = 1 (ie Tissue Total ) then so F F fat lean M M fat lean (R (R fat (R (R M R R fat M ) ) R) R fat tissue tissue lean lean.f lean fat.f ) lean ANZBMS 2005-2013 - Slide 38
Strong Bone Asia 2013 Osteoporosis in ASEAN Principles of Dual Energy X-ray Absorptiometry Limitations of DXA Copyright 2005-2013 This material is Copyright by the Council of the Australian & New Zealand Bone & Mineral Society (ANZBMS) and may not be copied or transmitted by any means without written permission from the Council of the ANZBMS.
Density DXA limitations Measurement is 2 dimensional DXA density is area density Normal Osteoporotic ANZBMS 2005-2013 - Slide 41
DXA limitations Measurement integrates all bone Changes in projection change results (up to 0.5SD for >15 ) Normal Rotated 22 ANZBMS 2005-2013 - Slide 42
DXA limitations Photon flux must be adequate to distinguish between tissue types Low enough to not affect R tissue or R bone ANZBMS 2005-2013 - Slide 43 ANZBMS 2005-2013 - Slide 43
DXA limitations Artefacts can affect calculations Oral contrast Intravenous contrast ANZBMS 2005-2013 - Slide 44
Strong Bone Asia 2013 Osteoporosis in ASEAN Basic anatomy for bone densitometry Copyright 2005-2013 This material is Copyright by the Council of the Australian & New Zealand Bone & Mineral Society (ANZBMS) and may not be copied or transmitted by any means without written permission from the Council of the ANZBMS.
Cancellous Bone Bone Microstructure NORMAL BONE OSTEOPOROTIC BONE ANZBMS 2005-2013 - Slide 47
Cortical Bone Bone Microstructure NORMAL BONE OSTEOPOROTIC BONE ANZBMS 2005-2013 - Slide 48
Bone Microstructure 2 1 3 1. Periosteum membrane covering the outside of the bone. 2. Bone marrow cavity 1 3 3 2 1 3. Endosteum membrane lining the inside of the cortical bone and the trabeculae ANZBMS 2005-2013 - Slide 49
Cortical bone Shafts of long bones Cortical rim of vertebrae ANZBMS 2005-2013 - Slide 50
Distribution of cortical and trabecular bone ANZBMS 2005-2013 - Slide 51
Lumbar Spine Essential Anatomy Proximal Femur Forearm ANZBMS 2005-2013 - Slide 52
Lumbar spine anatomy with relevance to bone densitometry ANZBMS 2005-2013 - Slide 53
Vertebral Anatomy Cortical rim Body Transverse Process Cancellous bone Pedicle Superior articular process Lamina Vertebral foramen Spinous process ANZBMS 2005-2013 - Slide 54
Projectional anatomy of the lumbar spine Lateral view Posterior view Anterior view ANZBMS 2005-2013 - Slide 55
Vertebral Appearance in Densitometry ANZBMS 2005-2013 - Slide 56
Anatomical variations of lumbar Spine Number of vertebrae Approximately 7% of people have missing ribs on T12 Only 2% had 6 lumbar vertebrae ANZBMS 2005-2013 - Slide 57
Anatomical variations of lumbar vertebrae 7.5% of subjects have 4 lumbar Vertebrae. Ribs on L1 are very rare. ANZBMS 2005-2013 - Slide 58
Sacralization L5 sacralization is a common anatomical anomaly, where the L5 vertebral body becomes incorporated into the S1 vertebral body. ANZBMS 2005-2013 - Slide 59
Basic anatomy of the proximal femur with relevance to bone densitometry ANZBMS 2005-2013 - Slide 60
Anatomy of the Hip Femoral neck Femoral head Greater trochanter Lesser trochanter ANZBMS 2005-2013 - Slide 61
Anatomy of Lesser Trochanter ANZBMS 2005-2013 - Slide 62
X ANZBMS 2005-2013 - Slide 63
Effect of rotation of the Lesser Trochanter ANZBMS 2005-2013 - Slide 64
Anatomy of the Forearm Ulna Ulna styloid Radius 33% Shaft 5 mm Ultra -distal Trabecular 5% 35% 75% ANZBMS 2005-2013 - Slide 65
Strong Bone Asia 2013 Osteoporosis in ASEAN Introduction to Statistics for Bone Densitometry Copyright 2005-2013 This material is Copyright by the Council of the Australian & New Zealand Bone & Mineral Society (ANZBMS) and may not be copied or transmitted by any means without written permission from the Council of the ANZBMS.
Why Make Measurements? Places the observation in universally recognised context To compares one observation with another over time, in a single individual or group, or between individuals or groups To compare an observation with a reference range made up of many similar measurements of normal individuals, in order to determine how unusual it is To examine cause and effect relationships between different measured variables ANZBMS 2005-2013 - Slide 67
Frequency Description of a Measured Variable Magnitude of distribution Where does a measurement usually lie? Three common measures of the predominant value of a measurement Mode the most frequently occurring value Median the value where half the measurements lie above it and half lie below Mean an average that gives greater weighting to larger values ( values / N) 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 10 = (1+ 2+ 3+ 3+3+4+4+4+ 4+ 4+ 4+ 4+5+ 5+ 5+5+5+ 5+6+ 6+ 6+ 6+ 6+ 7+ 7+ 8+ 8+9+ 9+ 10)/30 = 5.27 ANZBMS 2005-2013 - Slide 68
Frequency Description of a Measured Variable Measures of the spread of the distribution Within what limits does a measurement usually lie? Range The difference between the lowest and highest values Inter-quartile Range Difference between the top of the first (lowest) quartile and the bottom of the fourth (highest) quartile. 50% of values lie between these two values* Standard Deviation (SD) Estimate of the spread that gives greater weighting to values distant from the Mean * Deciles could also be used in replacement of quartiles n 1 SD = SQRT[(5.27-1) 2 + (5.27-2) 2 + (5.27-3) 2 + + (5.27-9) 2 + (5.27-10) 2 /(30-1)] = 2.12 x i x 2 8 7 6 5 4 3 2 1 0 Lower quartile Range Inter-quartile range Upper quartile 1 2 3 4 5 6 7 8 9 10-2 -1 0 +1 +2 ANZBMS 2005-2013 - Slide 69
Symmetric Distributions the Bell-shaped Curve Most statistical analysis assume measured variable is distributed symmetrically (a bell-shaped curve) Three reasons: Multiple observations of many biological variables follow a bellshape or a mathematical function of it (eg log-normal) Variables not following a bell-shape often have the distribution of means of subgroups of its distribution tending to bell shaped. This powerful result places this type of curve at the centre of statistical analysis! The curve has a mathematical formula so mathematics of the statistical analysis based on the curve is greatly simplified Formally known as the Normal (Gaussian) Distribution ANZBMS 2005-2013 - Slide 70
mean Characteristics of Normally Distributed Data Variables following a Normal Distribution: Describe the predominant value as the Mean = median = mode Describe the spread of data under curve as Standard Deviation 68% of data is within ± 1 SD of the mean 95% of data is within ± 2 SD of the mean 99.89% of data is within ± 3 SD of the mean 50% data 50% data 16.7% data 16.7% data 2.5% data 2.5% data -3-2.5-2 -1.5-1 -0.5 0 0.5 1 1.5 2 2.5 3 ANZBMS 2005-2013 - Slide 71
Mean Does a Measurement belong to a Particular Data Set? The Z-statistic Does belong to the distribution? ie What is the probability that it is so different from the other points that it is not a member of the group? To calculate the Z-statistic Choose data distribution to compare single data point with Need distribution Mean & SD (must be approximately normal) Z = (X-Mean)/SD X-Mean 1 SD -2.5-2 -1.5-1 -0.5 0 0.5 1 1.5 2 2.5 Standard Deviations X ANZBMS 2005-2013 - Slide 72
Measurement Accuracy Systematic Errors Occur when repeated measurements made on the same subject generally agree, but in comparison with a more physically valid technique, are found to differ Difficult to estimate, unless a Gold Standard technique able give correct answer for comparison eg ashing & weighing cadaveric bones following DXA scanning Therefore repeat measurements may agree but may all be wrong Used to calculate the accuracy of the technique (eg DXA), usually expressed as a % variation from the real value, determined using the gold standard ANZBMS 2005-2013 - Slide 73
Measurement Precision Random Errors Occur when repeated measurements made on the same subject do not agree, even though subject s physiological state is unchanged Caused by: fluctuations in the operating characteristics of the machine random operator variations in patient positioning random variations in scan acquisition and scan analysis Easiest errors to estimate, as distribution usually Normal The distribution of random errors is used to calculate the precision of the measurement ANZBMS 2005-2013 - Slide 74
Accuracy vs Precision in clinical densitometry poor precision good accuracy (larger random error) Dual-energy QCT good precision poor accuracy (larger systematic error) DXA QUS Photogrammetry good precision good accuracy (small random & systematic errors) The ideal ANZBMS 2005-2013 - Slide 75
Determining Precision of Bone Densitometry Precision measurements should be made on the population they will be applied to (ie patients), rather than healthy volunteers A good precision estimate requires >28 subjects measured twice Each subject should be removed from scanning couch following first scan, and relocated for the second Such data can be used for short term precision calculation Ethics a source of duplicate scans Certain subjects likely to be reviewed regularly (and therapy varied) may benefit from duplicate measurements at baseline because a bad baseline corrupts future comparisons Duplicate measurements in a clinical practice may also be justified if transferring patients to a new DXA machine Bonnick & Lewis (2002) ANZBMS 2005-2013 - Slide 76
x Calculating Precision of Bone Densitometry SD (s) of the distribution of precision errors (ie random errors of duplicate measurements) is calculated: Where: d i = x i1 - x i2 is the difference in the i th duplicate data set n = the number of subjects Precision error is often expressed as coefficient of variation (CV%) s CV where is the Mean of the 2n measurements The closer to 0% the CV% the greater the reproducibility di 2n 2 s % 100 x Bonnick & Lewis (2002) ANZBMS 2005-2013 - Slide 77
Precision & Accuracy in Bone Densitometry Precision Accuracy Central DXA PA spine 1-2% 4-10% Lateral spine 2-4% 5-15% Femur 1.5-3% 6% Forearm 1% 5% Total body 1% 3% pdxa Forearm 1-2% 4-6% Calcaneus 1-2% 4-6% QCT Spine QCT 1.5-4% 5-15% Peripheral QCT 1-2% 2-8% Genant HK, et al. J Bone Miner Res 1996; 11(6):707 ANZBMS 2005-2013 - Slide 78
The Importance of Good Precision Precision is probably the most important factor to consider in routine bone densitometry High precisional error usually indicates poor work practices or machine compliance Better precision allows smaller changes in BMD to be detected at any selected confidence level and permits any change in bone status to be detected sooner Precision errors are minimised by: regular quality assurance scans to detect machine drift attention to technique - patient positioning, scan acquisition & analysis Significance of small changes in BMD can be assessed by a knowledge of the Least Significant Change (LSC) ANZBMS 2005-2013 - Slide 79
Least Significant Change (1) Serial measurements of BMD are important to manage the patient (eg input into decision as to whether to vary Rx) Important to know the smallest magnitude of change in BMD which is significant ie due to true change rather than random error The Least Significant Change (LSC) for two BMD measurements is the minimum change necessary between serial DXA scans to be certain of a real change in BMD LSC depends on: precision of BMD technique degree of statistical confidence demanded (ie p<0.05, p<0.01) ANZBMS 2005-2013 - Slide 80
Least Significant Change (2) LSC Z * 2 * s where: Z a is the Z-statistic for chosen significance level (eg Z = 1.96 if p<0.05) s is the precision error expressed as the SD of the distribution of differences in duplicate measurements Usually, we are interested in a significance of p<0.05 so: LSC =1.96 x 1.4 x s ~ 2.8 x s If s = 0.01 g/cm 2 then LSC = 0.028 g/cm 2 If s = 0.15 g/cm 2 then LSC = 0.42 g/cm 2 ANZBMS 2005-2013 - Slide 81
LSC (as %) Importance of good precision in follow-up studies Difference in serial measurements to guarantee a real change 8 4 good poor LSC ~2.8 CV 0 0 1 2 3 DXA precision (CV%) ANZBMS 2005-2013 - Slide 82