Lumbar spine radiographs — is it time for widespread adoption of posteroanterior projection?

Abstract

Objective:

Phantom studies and a prior patient study have shown up to 53% effective dose reduction when lumbar spine radiographs are acquired posteroanterior (PA) instead of anteroposterior (AP). Since November 2017, Taunton and Somerset NHS Foundation Trust has acquired all standing lumbar spine radiographs PA. The aim of this study was to locally evaluate dose and image quality in both projections and survey current UK practice.

Methods:

80 outpatients having a standing lumbar spine radiograph (40 AP; 40 PA) had their dose–area product recorded at a constant KV and focus film distance. Effective dose was calculated using PCXMC software. Each blinded radiograph was scored against an optimal reference image using European Guidelines criteria. The data were analyzed using Mann–Whitney U tests and linear regression. Eighty radiologists nationally were sent an anonymous survey to establish their current practice.

Results:

A lumbar spine radiograph acquired PA instead of AP reduced effective dose by 41% (p < 0.001) with no difference in image quality (p = 0.9). 21 radiologists completed our survey and only 1 NHS Trust is currently using PA.

Conclusion:

PA lumbar spine radiography reduces patient radiation exposure with no affect on image quality, acquisition time or cost. The majority of NHS Trusts nationally are still using AP and it is time to standardize to PA.

Advances in knowledge:

This patient study provides further good evidence of how reduction in exposure to ionizing radiation can be achieved in lumbar spine radiography and more widespread adoption of PA protocol could improve patient safety.

Background

Lumbar spine radiographs are a commonly ordered investigation with a moderately high radiation dose relative to other plain radiographic investigations. The effective dose (ED) relates to probability of health detriment due to exposure to ionizing radiation; for a lumber spine radiograph ED is typically between 0.1–0.4 mSv, which equates to up to five times the dose from a transatlantic flight.¹ To put that in context, a radiation dose of 1 Sv carries an approximate risk of developing cancer of 5.5% (4.1% in an adult population), meaning a 1 mSv dose carries a risk of approximately 1 in 20,000 and the risk from a lumbar spine is between 1 in 200,000 and 1 in 50,000.^2,3

Given the fundamental principle of ALARP,⁴ it is important radiology departments continually review their current practice with respect to all ionising radiation imaging procedures in light of the evolving evidence.

Over the past 20 years, a number of studies have been published that demonstrate a substantial reduction in individual organ dose and consequently, effective dose when a lumbar spine is acquired posteroanterior (PA) instead of anteroposterior (AP) as is the accepted convention.

Mekis et al⁵ showed the dose received by the testes in the PA projection in sacroiliac joint imaging is 93.1% lower compared to AP (p = 0.020). Davey and England⁶ found mean absorbed organ dose to the stomach and colon (some of the most radiosensitive organs in the abdomen) reduced by 70.4 and 61.1%, respectively with PA compared to AP, whilst the mean effective dose on phantoms fell by 19.8%; there was also no statistically significant change in image quality. Another study by Brennan et al,⁷ showed benefits of reduction in entrance surface dose by 38.6% in female patients and 38.9% in internal phantom dose when using the PA projection.

This reduction in dose is particularly pertinent for the paediatric population given the increased lifetime risk from early radiation exposure. A phantom study⁸ on PA spinal projections predicted a 64% fall in effective dose in children aged 10, and a 65% reduction in adults.

The most recent study by Alukic et al⁹ assessed the dose and image quality when acquiring the radiographs PA in both phantoms and 100 real patients. Their patient cohort had a wider weight range (including measuring abdominal thickness) than previous studies and showed a 53% reduction in effective dose with PA projection, again with no significant effect on radiograph image quality.

Moreover, PA positioning in other radiographic investigations has also demonstrated dose reductions; significant decrease in dose to breast and thyroid^10,11 in scoliosis and clavicle¹² radiography and other organ doses reductions with abdominal^13,14 and pelvic radiography.¹⁵

Given this increasing body of evidence, following discussions with radiology consultants and radiographers at Musgrove Park Hospital in Taunton concerning quality improvement and patient safety, it was agreed that from November 2017 onwards all routine departmental standing lumbar spine radiographs would be acquired PA instead of AP.

The aim of this study was to locally evaluate any change in dose and image quality in patients with the intention of confirming previous study finding prior to adoption of PA protocol (the only other large study involving real patients⁹ had not been published at the time of our protocol change to PA). We also wanted to establish what current national practice is concerning protocols for lumbar spine radiographs and whether other departments have switched to PA in order to improve patient safety by reducing exposure to ionizing radiation.

Methods and Materials

80 outpatients that attended Musgrove Park Hospital in Taunton over a 3 month period in 2017 for a routine standing lumbar spine radiograph which met our inclusion criteria of weighing between 60 and 100 kg and of either sex were included in this study; 40 acquired AP and 40 PA. Patient’s height and weight were recorded and BMI calculated. There were no exclusion criteria for clinical indication or spinal pathology.

The radiographs were acquired on two different Fujifilm, Visionary Suite digital radiography imaging equipment systems (same manufacturer model) at 80KV (since the previous study⁶ demonstrated optimum image quality was acquired at this voltage) and a constant focus film distance (FFD) of 110 cm. Automatic exposure control was enabled and the dose–areaproduct (DAP in µGym²) for each radiograph displayed on the machine was recorded.

The radiographs were centred between the lower costal margin and the iliac crests in accordance with standard methodology^16,17 and for PA projection the patients were turned around with their anterior abdomen against the radiographic plate.

Local departmental ethics approval was granted for this quality improvement project and the planned departmental protocol change from AP to PA agreed prior to data collection. Consequently, no individual patient consent for this study was required during data collection.

Dose calculations

The ED in mSv and entrance surface dose (ESD) in mGy was calculated using the Monte Carlo simulation software PCXMC,¹⁸ a well-established method used in previous studies.^6,7,9,19 The field size, height, weight and sex for each patient as well as KV, DAP and filtration were used for the dose calculations.

Image quality assessment

Each radiograph was assessed by two experienced consultants, a musculoskeletal radiologist and a spinal surgeon against an optimal reference image. The scores were based on the 7 point European Guidelines Criteria²⁰ with weighting factors applied as used in a previous study⁶; e.g. visually sharp reproduction of the cortex and trabecular structures is given a higher weighting than reproduction of the adjacent soft tissue and psoas shadows. Each criterion was scored between 1 and 3; 1, suboptimal and non-diagnostic; 2, lower quality than the reference image; and 3, on a par with the reference image. The maximum score was 48. The windowing of each radiograph was optimised prior to the quality assessment (not by one of the consultant assessors); it was not possible to re-window the radiographs during the scoring process to minimize interrater variability. Each consultant was blinded to the projection and the radiographs were presented in random order.

In addition, the magnification factor of PA vs AP was assessed by measuring the maximal diameter of L3 vertebral body on each radiograph by each reviewer as used previously.⁶

Statistical analysis

The data were not normally distributed so the dose and image quality data were initially analyzed with non-parametric Mann–Whitney U tests. The dose data were then logarithmically transformed (log10) to normalise to facilitate multiple linear regression analysis.

Linear regression analysis allows for potential confounder adjustment in multivariable models. It could be argued that it is difficult to interpret the regression coefficient following log transformation but this was not considered an issue in this context as we were not concerned with the rate of change in dose as a function of the projection. We knew the direction of association from the non-parametric analysis (i.e. radiation dose falls with PA projection) and only interested in any change in significance of the association in both the unadjusted/ simple and adjusted models with additional potential confounders included.

National survey

In November 2018, we disseminated a very short anonymous online survey via social media to 80 musculoskeletal radiologists within the UK with two questions, their hospital name and what projection their radiology department currently uses for acquiring standing lumbar spine radiographs. No personal or patient data were collected.

Results

Even though the patients were not randomized to the groups, the two groups were very similar with respect to their sex and physical parameters (Table 1). Consequently, there is minimal potential to introduce bias between patient characteristics in the AP or PA group.

Table 1. .

Characteristics of AP and PA patient groups (p > 0.05)

	AP	PA
	N = 40	N = 40
Age
Mean (SD)	59.7 (18.2)	59.1 (18.1)
Gender
Number (%) Male	20 (50%)	17 (43%)
Height (cm)
Mean (SD)	170.6 (9.8)	168.8 (9.4)
Mass (kg)
Mean (SD)	76.7 (10.2)	74.6 (11.3)

AP, anteroposterior; PA, posteroanterior; SD, standard deviation.

Considering the non-parametric analysis initially (Table 2), when using PA instead of AP projection, there was a 31% reduction in median ESD from 5.88 to 4.06, but the overall difference was not statistically significant (p = 0.074); the median ED reduced by 41.3% from 0.22 to 0.13, which was highly significant (p < 0.001).

Table 2. .

Calculated doses using Monte Carlo Simulations

	AP	PA	Comparison (Mann–Whitney U test) (% Change)
	Median (IQR)	Median (IQR)
Entrance surface dose (mGy)	5.878 (3.182–9.174)	4.058 (2.019–7.110)	p = 0.074 (- 31%)
Effective dose (mSv)	0.220 (0.147–0.371)	0.129 (0.083–0.206)	p < 0.001 (- 41.3%)

IQR, interquartile range.

With linear regression following log transformation (Table 3) the difference between the ESD in the simple model was insignificant (p = 0.093). Following adjustment for radiography machine, age, sex, height and BMI the association became significant (p = 0.019), suggesting that these parameters have an influence on ESD. In fact, ESD is positivity correlated with BMI (Pearson’s r = 0.529, p < 0.001, n = 80).

Table 3. .

Difference between AP and PA doses tested parametrically with linear regression

	Mean (SD) of logarithmically transformed values		Multiple regression analysis to predict log transformed values:
	AP	PA	Unadjusted group (mean) difference	Group (mean) difference after adjustment for machine, age, sex and BMI
Entrance surface dose	0.721 (0.328)	0.590 (0.357)	−0.130 (95%CI −0.283 to 0.022) p = 0.093	−0.116 (95%CI −0.211 to −0.020) p = 0.019
Effective dose	−0.663 (0.301)	−0.895 (0.306)	−0.232 (95%CI −0.367 to −0.097) p = 0.001	−0.225 (95%CI −0.323 to −0.127) p < 0.001

AP, anteroposterior; PA, posteroanterior; SD, standard deviation.

The reduction in effective dose was highly significant in both the simple (p = 0.001) and adjusted model (p < 0.001) indicating that none of the measured patient parameters were confounding the outcome with a clear association between PA projection and reduction in effective dose.

With reference to Table 4, there were 75% and 69.5% reduction in median absorbed dose in the both the stomach and colon, respectively (two of the most radiosensitive organs in the abdomen) when using a PA projection. Moreover, there were large reductions in absorbed doses to the pelvic organs (e.g. bladder, −64.5%). As expected, there are big increases in absorbed dose to the retroperitoneal organs (kidneys, 798.5%; adrenals, 206%) given these are closest to the radiation source. Besides the pancreas (p = 0.264), these differences in absorbed organ dose between the projections were all statistically significant.

Table 4. .

Individual absorbed organ doses for each projection (mGy)

	AP	PA	Comparison (Mann–Whitney U test) (% change in median)
	Median (IQR)	Median (IQR)
Stomach	0.489 (0.320–0.660)	0.120 (0.075–0.193)	p < 0.001 (-75%)
Colon	0.562 (0.379–0.928)	0.171 (0.118–0.282)	p < 0.001 (-69.5%)
Adrenals	0.061 (0.043–0.111)	0.187 (0.094–0.437)	p < 0.001 (206%)
Kidneys	0.142 (0.114–0.224)	1.276 (0.812–2.026)	p < 0.001 (798.5%)
Pancreas	0.253 (0.161–0.392)	0.201 (0.118–0.372)	p = 0.264 (-20.5%)
Ovaries	0.565 (0.341–0.935)	0.336 (0.209–0.528)	p = 0.004 (-40.5%)
Testicles	0.011 (0.003–0.023)	0.004 (0.002–0.007)	p = 0.002 (-63.6%)
Bladder	0.240 (0.098–0.681)	0.085 (0.039–0.112)	p < 0.001 (-64.5%)
Uterus	0.854 (0.477–1.578)	0.296 (0.217–0.529)	p < 0.001 (-65.3%)

AP, anteroposterior; IQR, interquartile range; PA, posteroanterior.

When comparing the scores of the two assessors for image quality assessment (Table 5), there was no significant difference between individual or combined average scores of the AP and PA radiographs when compared with the reference image (p = 0.939). The magnification factor between PA and AP was 1.05.

Table 5. .

Image quality assessment for the two consultant reviewers

	Spinal surgeon	MSK radiologist	Average of combined scores
	Median (IQR)	Median (IQR)	Median (IQR)
AP	46 (41.5–48)	43.5 (39.5–45.5)	44 (39.8–46.3)
PA	46 (41.5–48)	43 (38–45.5)	43 (41–46.3)
Difference (Mann–Whitney U test)	p = 0.929	p = 0.900	p = 0.939

AP, anteroposterior; IQR, interquartile range; PA, posteroanterior.

Of the 80 radiology consultants nationally that were sent the survey, 21 replied (Table 6). Only 1 hospital out of the 21 (5%) currently performs standing lumbar spine radiographs PA projection.

Table 6. .

Current practice in lumbar spine projection from NHS trusts that completed our survey (n = 21)

West Hertfordshire NHS Trust	AP
University Hospitals of North Midlands	AP
Mid Cheshire Hospital NHS Trust	AP
Addenbrookes NHS Trust	AP
Warrington & Halton Hospitals NHS Foundation Trust	AP
NHS Greater Glasgow and Clyde	AP
Robert Jones and Agnes Hunt Orthopaedic Hospital	PA
University Hospitals of Leicester	AP
London North West NHS Trust	AP
University Hospital of Morecambe Bay NHS Trust	AP
Abertawe Bro Morgannwg University Health Board	AP
Oxford University Hospitals Foundation Trust	AP
Western Sussex NHS Foundation Trust	AP
Queen Victoria Hospital NHS Trust	AP
Imperial Hospital NHS Trust	AP
Leeds Teaching Hospitals NHS Trust	AP
Royal National Orthopaedic Hospital	AP
Bolton NHS Foundation Trust	AP
Pennine Acute Hospitals NHS Trust	AP
Royal Orthopaedic Hospital Birmingham	AP
University Hospital Birmingham NHS Trust	AP

Discussion

The results from this study confirm previous finding that there is a significant reduction in effective dose when acquiring a lumbar spine radiograph PA instead of AP with no associated detrimental impact on image quality. Our reduction in median effective dose of 41.3% is broadly similar to the 53% found in, to our knowledge, the only other large study on real patients.⁹

The reduction in median entrance surface dose of 31% with PA is similar to 39% in the phantom study⁷ and 27% fall in DAP (a component of ESD) in the only other real patient study,⁹ although was not statistically significant with the non-parametric test (p = 0.07). The statistical significance for the reduction in ESD following adjustment in the linear regression model was likely due to the effect of BMI. It is well known that thinner body parts results in less photon attenuation, and therefore less mA required to achieve a satisfactory image.¹⁶ This is particularly true since the imaging equipment was set up with automatic exposure control, under which the mA is varied so that a constant, optimized dose is received by the image detector. In fact, for every 1 cm reduction in body thickness there typically a 25% reduction in mAs.²¹ Although we did not measure waist circumference specifically, this can be implied from BMI. The benefit of the PA projection is, in part, likely due to tissue displacement of the anterior abdominal tissues against the X-ray plate reducing the mA required via automatic exposure control.⁷

Alukic et al only measured DAP and ED; changes to individual organ dose was not investigated by them. We found a decrease in median patient absorbed dose to two of the most radiosensitive organs the colon (69.5%) and stomach (75%), in line with a previous phantom study⁶ (colon 61.1%, stomach, 70.4%). The slightly larger fall in organ dose in our study is likely due to the fact that radiographs were performed on more modern digital radiography equipment and on real patients as opposed to phantoms with associated advantages of tissue displacement.¹³ In addition, we have demonstrated an approximately 800% and 200% increase in the absorbed dose to the kidneys and pancreas, respectively with PA projection. This is to be expected given they are retroperitoneal and closer to the radiation source. However, this has limited impact on the effective dose, and hence long-term radiation risk, given the tissue weighting factor²² for the 14 less radiosensitive remainder tissues (including kidneys and adrenals) combined is 0.12, which equates to the tissue weighting factor for the colon alone.²³

In terms of the scores for image quality, the spinal surgeon scored the images on average three points higher than the radiologist, but there was no significant difference in intrarater image quality scores between the two projections or when the interrater median scores were averaged. Image examples are shown in Figure 1. There was a minimal difference in magnification factor (1.05) between PA and AP, which is lower than the 1.08 stated in the phantom study by Davey et al.⁶ We also considered the reliability of the image scores between the two raters using interclass correlation which was 0.6 indicating moderate reliability. This may improve with more raters, but the difference in combined median scores between the two consultants was minimal; besides, these image quality findings concur with multiple previous studies^6,7,9 establishing no significant difference in quality scores between AP and PA.

Figure 1.

Example radiographs acquired in the two projections for different patients of similar body habitus. AP, anteroposterior; PA, posteroanterior.

We had no specific inclusion criteria regarding spinal pathology as the clinical indications for a lumbar spine radiograph are limited, as are the disease processes that can be comprehensively assessed. We did not perform subgroup analysis for difference in image quality between the two projections in specific pathologies or in those patients attending the emergency department following acute spinal trauma. Consequently, our findings are only applicable to routine outpatient standing radiographs. However, in day-to-day clinical practice, besides assessing degree of degenerative change or sub acute trauma, the gold-standard for more complex pathologies such as nerve root impingement, bone lesions or spinal surgical fixation is CT or MRI.

To our knowledge, no previous study has reviewed current practice for the lumbar spine radiographic imaging within UK radiology departments. Of the 21 radiology consultants at different UK NHS Trusts that completed our survey, only 1 other hospital (Robert Jones and Agnes Hunt Orthopaedic Hospital, Shropshire) is presently using PA projection (Table 6); at the remaining 20 trusts with a reasonably good geographical spread throughout the UK, all acquire images AP. In order to maximize response rate from busy departments, our survey was deliberately kept simple and did not include questions concerning the reasons for their current practice, which likely vary between departments; consequently, we cannot provide further insight in this regard.

This quality improvement study at Musgrove Park Hospital, Taunton is the only one to have calculated individual organ dose in real patients and employed parametric statistics to enable confounder adjustment. It adds to the growing weight of evidence that lumbar spine radiographs acquired PA instead of AP significantly reduces effective dose with no affect on image quality. Our department has been performing all standing lumbar spine radiographs PA since November 2017 reducing patient exposure to ionising radiation. Such a change in protocol has no negative impact on acquisition time or investigation cost. As the majority of UK NHS Trusts are still performing this investigation AP, we recommend that the time has come for widespread standardisation to PA with its associated benefits for patient safety.