Glandularity and mean glandular dose determined for ... - Oncoline

INSTITUTE OF PHYSICS PUBLISHING Phys. Med. Biol. 51 (2006) 1807–1817

PHYSICS IN MEDICINE AND BIOLOGY

doi:10.1088/0031-9155/51/7/012

Glandularity and mean glandular dose determined for individual women at four regional breast cancer screening units in The Netherlands J Zoetelief1, W J H Veldkamp2, M A O Thijssen3 and J T M Jansen1 1 Department of Radiation, Radionuclides and Reactors, Faculty of Applied Sciences, Delft University of Technology, Mekelweg 15, 2629 JB Delft, The Netherlands 2 Department of Radiology, Leiden University Medical Center, Albinusdreef 2, 2300 RC Leiden, The Netherlands 3 National Expert and Training Centre for Breast Cancer Screening, Radboud University Nijmegen Medical Centre, Geert Grooteplein-Zuid 10, 6525 GA Nijmegen, The Netherlands

E-mail: [email protected]

Received 28 November 2005, in final form 6 February 2006 Published 16 March 2006 Online at stacks.iop.org/PMB/51/1807 Abstract The nationwide breast cancer screening programme using mammography has been in full operation in The Netherlands since 1997. There is concern that the mean glandular doses due to mammography might be differing between different regions of the country due to differences in glandularity and compressed breast thickness. To investigate regional differences, glandularity, compressed breast thickness and mean glandular dose were determined for individual breasts during screening at mammography units at four locations in The Netherlands. Differences in glandularity were observed, which could be related qualitatively to differences in age of the participants at the different locations. Mean glandular dose depends on compressed breast thickness, glandularity and technical conditions of screening. The lowest average value of the mean glandular dose was found for the unit in Amsterdam. This is most likely due to the use of the Mo/Rh anode/filter combination at this unit, in addition to the Mo/Mo combination. At the other three units, almost exclusively the Mo/Mo anode/filter combination was used. Differences in mean glandular dose averaged per unit could be related mainly to differences in tube-current exposure-time product values. Consequently, it is concluded that differences in mean glandular dose at different units are marginal.

1. Introduction The nationwide breast cancer screening programme using mammography has been in full operation in The Netherlands since 1997. Screening is performed in nine regions of the country employing 63 screening units and 28 central reading units. Annually, about one 0031-9155/06/071807+11$30.00 © 2006 IOP Publishing Ltd Printed in the UK

1807

1808

J Zoetelief et al Table 1. Characteristics of the screening units at each location.

Focus-to-table distance, dFt (cm) Air kerma at dFt per PIta (mGy (mAs)−1) Mo/Mo at 26 kV Mo/Mo at 27 kV Mo/Mo at 28 kV Mo/Rh at 28 kV Film dimensions 18 cm × 24 cm Film dimensions 24 cm × 30 cm Tilt of compression plate 18 cm × 24 cm (cm) 24 cm × 30 cm (cm) Thickness of Mo filter (µm) Thickness of Rh filter (µm) Tube voltage range for Mo/Mo (kV) Tube voltage range for Mo/Rh (kV) Automatic filter selection Mo or Rh? a

Amsterdam

Rotterdam

Heerenveen

Tholen

63

58.5

58.5

58.5

0.087 0.099 – 0.097 Yes No

0.094 0.107 0.122 – Yes Yes

0.094 0.107 0.120 – Yes No

0.086 0.098 0.111 – Yes Yes

0.1 – 30 25 26–27 27–30

0.7 1.1 30 25 24–31 –

0.8 – 30 25 24–31 30

0.8 0.8 30 25 24–31 –

Yes

No

No

No

Tube-current exposure-time product, PIt.

million women in the age group of 50–75 years are invited for screening and approximately 80% of the invited women actually participate. For the implementation of a population breast cancer screening programme using mammography, it is a prerequisite that the benefits of screening are considerably greater than the risks induced due to the use of ionizing radiation. There is concern that the mean glandular doses might be appreciably different between different regions of the country, e.g. cities versus countryside, due to differences in the size and composition of the breast. If these concerns were true, there might be regions of the country where the benefit-to-risk ratio might be less favourable than the national average. Therefore, a study was performed into the composition of the female breast, in terms of glandularity, i.e. the average fraction of glandular tissue, the compressed breast thickness and the mean glandular dose per mammogram in four regional units in The Netherlands. Among the regional units, two were in large cities and two in the countryside. The units have a wide spread across the country. Possible differences in glandularity, compressed breast thickness and mean glandular dose for different regions will be investigated. 2. Materials and methods Four regional units were approached by the National Expert and Training Centre for Breast Cancer Screening in The Netherlands. The units were selected as participating in the quality control programme of breast screening units, since this facilitated the calibration measurements, and for a spread among city and countryside and over the geography of The Netherlands. The participating units are located at Amsterdam, Rotterdam, Tholen and Heerenveen. Amsterdam and Rotterdam are the two largest cities of The Netherlands, located in the west of the country. Tholen and Heerenveen are located in the south-west and north-east of The Netherlands, respectively, in much less densely populated areas. Some characteristics of the mammography units are given in table 1. For all units, the maximum compression force is 200 ± 10 N (20 ± 1 kgf). However, for each mammogram the compression force was applied on an individual basis but never exceeding the maximum.

Glandularity and mean glandular dose at four Dutch breast screening units

1809

A method was used to derive glandularity of the individual compressed breasts using mammography by comparing measured attenuation in the actual compressed female breast during mammography and calculated attenuation in a breast model as a function of compressed breast thickness, glandularity and radiation quality (Jansen et al 2005). Compressed breast thickness was derived using a method similar to that of Burch and Law (1995). They proposed a thickness measurement based on the imaging of lead markers placed on the compression plate and by calibration of the breast thickness on the imaged positions of the lead markers. Measured attenuation was obtained as exit air kerma divided by incident air kerma. Incident air kerma was obtained by calibration of x-ray tube output in the presence of the compression plate at the relevant anode/filter combinations and tube voltages at each unit (table 1). Exit air kerma was obtained from the optical density measured using a calibrated film digitizer for a relevant region of the breast imaged on the mammogram, i.e. excluding the region of the pectoral muscle and outer region only containing skin and underlying adipose tissue. A correction to derive the exit air kerma has to be made as a function of compressed breast thickness since the anti-scatter grid absorbs an increasing amount of scattered radiation with increasing breast thickness. Attenuation was determined for each pixel in the region of interest of the mammogram and plotted as a histogram of attenuations for each mammogram. To calculate the attenuation in the female breast, the Monte Carlo N-Particle radiation transport code (MCNP) (Briesmeister 2000) was used. The compressed female breast was modelled as half a cylinder with a radius of 8 cm. The cylinder was cut in half along the central cylinder axis. The central part simulates the breast tissue and the outer part of the compressed breast model simulates the skin and underlying adipose tissue. For the central region three compositions have been used, i.e. 100% adipose tissue, a mixture of 50% adipose tissue and 50% glandular tissue by mass and 100% glandular tissue. The outer part of the compressed breast model is modelled using a 5 mm thick layer of adipose tissue. The overall breast thickness is varied and values of 11–110 mm have been used (Jansen et al 2005). The glandularity of the breast is determined by comparing the measured attenuation for a compressed breast thickness with the computed attenuation for the same breast thickness, tube voltage and anode/filter combination for the three compositions (0, 50 and 100% glandular tissue). The methods for interpolation are presented by Jansen et al (2005). The average compositions of adipose and glandular tissues are used in the Monte Carlo calculations and it should be noted that individual tissue compositions can deviate. If the individual adipose composition is attenuating less than average and the woman has a low glandularity, then it is possible that the computed glandularity becomes negative. The mean glandular dose per mammogram can be derived from incident air kerma, Ka,i, using a conversion factor g(t, p, a/f, VT) determined as a function of the compressed breast thickness (t), the glandularity of the breast (p), the actual anode/filter combinations (a/f) and the tube voltage (VT): ¯ G = Ka,i g(t, p, a/f, VT ). D

(1)

The conversion coefficients g(t, p, a/f, VT) were calculated previously as published by Klein et al (1997). The breast model used consists of central region of a mixture of glandular and adipose tissues and a superficial layer of 5 mm of adipose tissue representing skin and underlying adipose tissue. Tissue compositions were according to Hammerstein et al (1979). To illustrate the influence of the relevant parameters on g, one parameter is varied while the other parameters remain fixed. The initial values of the parameters are compressed breast thickness t = 5 cm, glandularity of the breast p = 0.5, anode/filter combination a/f = Mo anode material with 30 µm Mo filtration and tube voltage VT = 28 kV. Varying the compressed

1810

J Zoetelief et al

Figure 1. Glandularity as a function of compressed breast thickness for four screening units in Amsterdam, Rotterdam, Heerenveen and Tholen. Results are given for each individual breast and, if applicable, for Mo/Mo and Mo/Rh anode/filter combinations.

breast thickness from 3 cm to 7 cm results in a change of the conversion factor g(t, p, a/f, VT) from 0.313 mGy mGy−1 to 0.130 mGy mGy−1. Changing a 100% adipose breast, p = 0, to a 100% glandular breast, p = 1, modifies the conversion factor g from 0.236 mGy mGy−1 to 0.153 mGy mGy−1. Changing the filter from 30 µm Mo to 25 µm Rh results in a conversion factor change from 0.187 mGy mGy−1 to 0.216 mGy mGy−1. A tube voltage variation from 25 kV to 30 kV results in a conversion factor change from 0.168 mGy mGy−1 to 0.198 mGy mGy−1. The calibration measurements of x-ray tube output, optical density as a function of dose and corrected for breast thickness and of the breast thickness measurements were made at each unit before and after the collection of mammograms. The x-ray tube output remained constant within 1% over the period the mammograms were obtained. At the units in Amsterdam, Rotterdam, Heerenveen and Tholen, 927, 912, 1234 and 1126 mammograms were made and collected, respectively, at the end of 2003 and the beginning of 2004 and analysed. About 3% of the films had to be excluded from the analysis since data were lacking. The films were digitized and the histograms of attenuation were made. Measured attenuation and calculated attenuation were compared using information on compressed breast thickness, anode/filter combination and tube voltage used. The glandularity was determined for each breast. Incident air kerma, combined with the conversion factor g(t, p, a/f, VT), resulted in mean glandular dose for each breast according to equation (1). 3. Results and discussion The glandularity as a function of the compressed breast thickness is given for each of the four screening units in figure 1. Although all units had the option of selecting a molybdenum (Mo) or a rhodium (Rh) filter, in practice this option was used for the unit in Amsterdam, and


1811

only for one pair of breasts in Heerenveen. In Rotterdam and Tholen, the Mo/Rh anode/filter combination is not used at all although the National Expert and Training Centre for Breast Cancer Screening recommends to use the Mo/Rh anode/filter combination in the case of compressed breasts with a thickness in excess of 6 cm. This is due to practical reasons, i.e. the selected filter cannot be easily identified. Consequently, this might result in using the Rh filter for thin breasts and related poor image quality. The unit in Amsterdam is capable of automatically selecting the filter and the tube voltage on the basis of a small pre-exposure. In about half of the cases in Amsterdam the Mo filter is selected and in the other cases the Rh filter. The selection is as expected, i.e. Mo filter in the case of thinner breasts and for smaller values of glandularity in the case of approximately average breast thickness. For larger glandularities at approximately average breast thickness and for thicker compressed breasts, the Rh filter is selected. The use of the Mo/Rh anode/filter combination is expected to result in lower doses as more penetrating radiation is used for breasts with larger attenuation, e.g. Klein et al (1997). In table 2, average, standard deviation and standard deviation of the mean glandularity for each of the mammography units in Amsterdam, Rotterdam, Heerenveen and Tholen are shown. To test if two regions have significantly (p = 0.05) different average values, take the absolute difference of the average value and divide it by the square root of the summed squared standard errors of the mean. If this ratio is greater than 1.96 then the difference is significant, otherwise it is not. The highest values for glandularity are observed for relatively thin breasts (figure 1) and amount to 2.15, 1.82, 1.32 and 2.52 for the units in Amsterdam, Rotterdam, Heerenveen and Tholen, respectively. It should be noted that for Amsterdam and Tholen the minimum compressed breast thickness of about 2.2 cm is smaller than that in Rotterdam and Heerenveen where the value is about 3.0 cm (figure 1). It should be borne in mind that for a 2 cm compressed breast thickness and the breast model used for the calculations, the thickness consists half of adipose tissue representing skin and underlying adipose tissue and half of the central breast region. Maybe for thinner breasts the layer of skin and adipose tissue is smaller than that for thicker breasts. Values of glandularity exceeding 1 are also observed in other studies at small breast thicknesses (Klein et al 1997, Dance et al 2000). For all breasts, 2.5% of the glandularities are in excess of 1. In the analysis, the calculated glandularities are used even if they are outside the range from 0 to 1. For the computation of the conversion factor, g, the interpolation from 100% adipose, mixture of 50% adipose and 50% glandular and 100% glandular changes in an extrapolation but still excess glandularities can be handled. The minimum glandularities in Amsterdam and Rotterdam are slightly larger than zero, whereas for Tholen the minimum value is −0.14 and three values below zero are observable. For Heerenveen the minimum value is −0.37 and for 25 out of 1234 breasts the derived glandularity is below zero, a cluster of three breasts of approximately average thickness showing relatively low values. For 28 breasts out of a total of 4199, i.e. only about 0.7% of all breasts, a too low value is found, when the total number of mammograms studied is taken into account. Apparently, for 0.7% of all breasts the conditions that the individual adipose tissue composition is causing less attenuation than the average adipose tissue composition and that these breasts have a low glandularity hold true. Under these circumstances, negative computed glandularities can be expected. The glandularity averaged over all four centres is 32%, which is reasonably close to the values of 35% and 43% found by Klein et al (1997), but considerably smaller than the average glandularity assumed in most dosimetry protocols, i.e. 50% glandularity, e.g. EC (1996). The average glandularity in Amsterdam is higher than that in the other centres. This may be partly due to the fact that in Amsterdam a considerable number of women in the age group of

1812 Table 2. For each of the mammography units in Amsterdam, Rotterdam, Heerenveen and Tholen, the average value ± standard deviation, standard deviation of the mean are given for the glandularity, compressed breast thickness (cm), mean glandular dose (mGy), dose conversion factor g (mGy mGy−1) (equation (1)) and tube-current exposure-time product (mAs). In addition, the tests according to Press et al (1987) of the skewness of the distributions of the compressed breast thickness and of the mean glandular dose as shown in figures 3 and 4, respectively, are supplied.

Glandularity Compressed breast thickness (cm) Mean glandular dose (mGy) Conversion factor, g (mGy mGy−1) Tube-current exposure-time product (mAs) Number of mammograms Standard deviation of skewness Skewness of thickness Ratioa of thickness Skewness of mean glandular dose Ratioa of mean glandular dose a

Amsterdam

Rotterdam

Heerenveen

Tholen

0.50 ± 0.30, 0.010 5.42 ± 1.36, 0.04 1.04 ± 0.43, 0.01 0.193 ± 0.025, 0.001 50.1 ± 24.8, 0.8 927 0.080 0.24 3.0 2.17 27.0

0.31 ± 0.18, 0.006 5.92 ± 1.16, 0.04 1.16 ± 0.41, 0.01 0.174 ± 0.023, 0.001 48.5 ± 15.4, 0.5 912 0.081 0.22 2.8 2.22 27.3

0.19 ± 0.16, 0.004 6.19 ± 1.19, 0.03 1.63 ± 0.65, 0.02 0.175 ± 0.024, 0.001 71.9 ± 27.1, 0.8 1234 0.070 0.09 1.3 0.87 12.5

0.31 ± 0.23, 0.007 5.82 ± 1.26, 0.04 1.32 ± 0.49, 0.01 0.179 ± 0.028, 0.001 58.6 ± 19.3, 0.6 1126 0.073 0.23 3.1 0.77 10.5

Ratio is skewness divided by the standard deviation of skewness.

J Zoetelief et al


1813

Figure 2. Fraction of participants as a function of age. The results are given separately for each screening unit.

Figure 3. Fraction of breasts as a function of compressed breast thickness. The results are given separately for each screening unit.

45–49 years participated compared to that in the other centres (figure 2) and a relatively large number of women with thinner compressed breasts were observed (figure 3). An increasing glandularity with decreasing age has usually been observed (Klein et al 1997, Dance et al 2000), as well as an increasing glandularity with decreasing breast thickness (Zoetelief et al 1989). The relatively low value for the average glandularity in Heerenveen might be attributed to the relatively larger number of women with larger compressed breast thickness (figure 3) and the relatively higher ages (figure 2) of the participating women compared to the other centres. A decrease in glandularity with increasing breast thickness and increasing age is commonly observed (Dance et al 2000, Klein et al 1997). The differences in glandularity between cities (Amsterdam versus Rotterdam) and between different locations at the countryside (Tholen versus Heerenveen) are as large as the differences between the average of cities and the average of countryside. Qualitatively, the differences can be explained as shown in the previous paragraph but most likely not explained quantitatively. Differences in glandularity between north (Amsterdam and Heerenveen) and south (Rotterdam and Tholen) are absent (table 2). Differences in averaged glandularity between east (Heerenveen) and west (Tholen) are significant (p < 10−9), but smaller than those between cities and countryside.

1814

J Zoetelief et al

The distribution of the ages of the participants is given in figure 2. For Amsterdam, the young age group of 45–49 years is more than four times as large as for the other screening units. For the highest age groups, Heerenveen shows the largest relative number of participants. The compressed breast thickness is given in table 2 and figure 3 for women at each of the four screening units. The results in figure 3 suggest that the distribution is close to Gaussian. The largest compressed breast thickness in each of the four centres is 10.3, 10.2, 10.8 and 10.7 cm for Amsterdam, Rotterdam, Heerenveen and Tholen, respectively. The maximum values are quite similar. The compressed breast thickness averaged over all participating women is approximately 5.82 cm. This is slightly higher than the values of 5.59 and 5.08 cm found for two groups of patients by Klein et al (1997), but in the range of values for compressed breast thickness found by Eklund et al (1993). The standard deviations of compressed breast thickness found by Klein et al (1997) of 1.25 and 1.26 cm are quite similar to the values in table 2. Mean glandular dose values are given for each unit in table 2. The largest average glandularity is found for the participants in Amsterdam, whereas the compressed breast thickness is smallest at this unit (table 2). The net result of these two counteracting characteristics will partly compensate. The lowest average value for the mean glandular dose is found at this unit. This is also due to the use of the Mo/Rh anode/filter combination for thicker breasts and breasts of higher glandularity at approximately average thickness. To support the latter statement, the compressed breast thickness region was selected for Amsterdam. This thickness region includes 156 mammograms made with the Mo filter and 110 mammograms with the Rh filter. The average and standard deviation of the mean of the breast thickness are 4.92 ± 0.02 cm and 5.14 ± 0.03 cm for the mammograms with Mo and Rh filters, respectively. The average and standard deviation of the mean of the glandularity are 0.447 ± 0.012 and 0.533 ± 0.015 for the mammograms with Mo and Rh filters, respectively. As already expected, breast thickness and mean glandularity are larger when using the Rh filter. This would result in a higher mean glandular dose when using the Rh filter if the x-ray spectrum would remain the same as with the Mo filter. The x-ray spectrum, however, does change with the filter selection, and average value and standard deviation of the mean of mean glandular dose are 0.988 ± 0.015 mGy and 0.831 ± 0.019 mGy when using Mo and Rh filters, respectively. This demonstrates dose reduction by the use of the Rh filter instead of the Mo filter. For the other units than Amsterdam, the Mo/Rh anode/filter combination is not used except for two breasts at the unit in Heerenveen. The use of the Mo/Rh anode/filter combination at the unit in Amsterdam is most likely due to the automatic selection of the filter, which was only possible at this more modern unit. Also the tilt of the compression plate is smallest for this unit (table 1). In table 2, average value, standard deviation and standard deviation of the mean of dose conversion coefficient g(t, p, a/f, VT) are shown for each of the screening units at the various locations. The average dose conversion coefficient for Amsterdam is about 10% higher than the average value of the other three units. This is due to the use of the Rh filter in Amsterdam. If the results for the units in Rotterdam, Heerenveen and Tholen, i.e. those units where almost exclusively the Mo/Mo anode/filter combination was used, are intercompared there are still appreciable differences in the average values of the mean glandular dose. Differences in glandularity cannot explain the differences in mean glandular dose, because on the basis of glandularity (table 2) the smallest average value of the mean glandular dose would be expected in Heerenveen, whereas the highest dose value was found. The small differences in compressed breast thickness cannot compensate for the difference in glandularity between Rotterdam and Tholen compared to Heerenveen. On the basis of compressed breast thickness (table 2), a slightly smaller or almost equal average value for mean glandular dose would be


1815

expected for Tholen compared to Rotterdam, but the mean glandular dose in Rotterdam is significantly smaller. A more likely cause for the average mean glandular dose differences might be derived from table 2, where considerable differences in average values for the tube-current exposuretime product are found for the units in Rotterdam, Heerenveen and Tholen. The average tube-current exposure-time product values for Rotterdam, Heerenveen and Tholen are 48.5, 71.9 and 58.6 mAs, respectively. Suppose that it is assumed that the differences in average values of the mean glandular dose are due to differences in average values for tube-current exposure-time product. If it is further assumed that output corrections have to be made according to table 1, employing average tube voltage values of 27.3, 26.9 and 27.4 kV for Rotterdam, Heerenveen and Tholen, respectively, then starting from the Rotterdam values, on the basis of the values for the tube-current exposure-time products and outputs, an average value of the mean glandular dose for Heerenveen a value of 1.62 mGy (compared to 1.63 in table 2) and for Tholen of 1.29 mGy would be derived (compared to the value of 1.33 mGy for Tholen in table 2). The differences in tube-current exposure-time product, corrected for output, seem to almost completely explain the differences in results between the units in Rotterdam, Heerenveen and Tholen. Differences in tube-current exposure-time product at the various units may be due to differences in the selected level of optical density, differences in speed of the film–screen combinations or differences in film processing. Although differences in glandularity (table 2) and differences in compressed breast thickness exist between the units in Rotterdam, Heerenveen and Tholen, average values of the mean glandular dose are similar if the results are corrected for average tube-current exposuretime product values. Differences due to city versus countryside location in the country are absent between Rotterdam, Heerenveen and Tholen. The values for the unit in Amsterdam cannot be directly compared because of the use of the Rh filter and the relatively large age group from 45 to 49 years. For the air kerma per unit of tube-current exposure-time product using the Mo filter, glandularity and compressed breast thickness, the screening unit Amsterdam is similar compared with Tholen (tables 1 and 2). In table 1, it is shown that the air kerma per unit tube-current exposure-time product is about 13% lower for a Rh filter than that for a Mo filter. As Amsterdam uses roughly in half of the cases the Rh filter, this would lower the mean glandular dose in Amsterdam by about 7%. The average tube-current exposure-time product is about 15% lower in Amsterdam compared to Tholen (table 2) and the dose conversion factor is about 8% higher in Amsterdam compared to Tholen (table 2). As a result, one can expect roughly the mean glandular dose in Amsterdam to be about 14% lower than that in Tholen. In table 2, it is shown that the mean glandular dose in Amsterdam is 21% lower than that in Tholen. About two thirds of this difference would be roughly explained by the difference in the use of the Rh filter. For risk assessment for the national breast screening programme, generally a value of 2 mGy for the average value of the mean glandular dose was used. At all units, lower average values of the mean glandular dose were found (table 2). The results of Amsterdam suggest that improvements in the technique, i.e. the automatic use of a Mo or Rh filter, may result in dose reduction. The highest mean glandular dose values in the present study are less than 5 mGy (table 2). This implies that there is only an increase in the mean glandular dose of a factor of at maximum about 2.5 compared to the value of 2 mGy used for risk assessments until now for the national breast screening programme. It has to be borne in mind that the uncertainty in risk factors for tumour induction is in the order of a factor of 2. The highest doses occur for a relatively small number of participants in the screening programme as can be concluded from figure 4. The peak in the distributions of the mean

1816

J Zoetelief et al

Figure 4. Distribution of mean glandular dose. The results are given separately for each screening unit.

glandular dose is for Amsterdam in the range of 0.5–1 mGy, whereas for the other units the peaks are in the range of 1.0–1.5 mGy. It can also be observed that the distribution of the mean glandular dose (figure 4) is log-normal whereas it was concluded that the distribution of compressed breast thickness (figure 3) is close to Gaussian. These hypotheses were subjected to a skewness test according to Press et al (1987). Skewness was calculated as N 1 xj − x¯ 3 . (2) skew(x1 , . . . , xN ) = N j =1 σ The standard deviation of skewness was calculated as 6 standard deviation of skewness = , (3) N where N is the number of mammograms and σ the standard deviation (Press et al 1987). Press et al state that in real life it is good practice to believe in skewness only when they are several or many times as large as the standard deviation of skewness. In table 2, the ratio of skewness and standard deviation of skewness is shown for compressed breast thickness and for mean glandular dose for each of the screening units at the different locations. For compressed breast thickness, the ratio in table 2 varies from 1.3 to 3.1, i.e. at maximum a few times. For mean glandular dose, the ratio in table 2 varies from 10.5 to 27.3, i.e. from several times to many times. The test confirms the qualitative conclusions drawn from figures 3 and 4. 4. Conclusions The mammography unit in Amsterdam automatically selects the filter and the tube voltage. In over 50% of the mammograms the Mo/Rh anode/filter combination is used, i.e. for thicker compressed breasts and for approximately average breast thicknesses in the case of higher glandularity. The use of the Mo/Rh anode/filter combination leads to dose reduction. Although it is recommended to select the Rh filter for compressed breasts in excess of 6 cm thickness, only for two breasts in Heerenveen is this filter selected. For the other units only the Mo/Mo anode/filter combination was used, for practical reasons, i.e. the selected filter cannot be easily identified.


1817

The glandularity averaged over the four screening units is 0.32. This value is similar to that reported by Klein et al (1997), but different from the value of 50% by weight recommended in most national and international (European) dosimetry protocols. Values of glandularity below 0 and above 1 are rare. The influence of regional differences, e.g. city versus countryside, east versus west and north versus south, is not large in terms of glandularity and breast thickness. In terms of mean glandular dose, differences seem completely absent for Rotterdam, Tholen and Heerenveen. For Amsterdam, a direct comparison is not possible due to the difference in technique, i.e. the use of the Mo/Rh anode/filter combination. The participation as a function of age indicates that participation at ages 50–59 is highest, followed by a decrease with age. Only in Amsterdam was there a significant group of younger participants, i.e. in the age range of 45–49 years. The distribution of compressed breast thickness seems close to normal, but the distribution of mean glandular dose seems log-normal. The differences in technical parameters are appreciable, e.g. differences in tube-current exposure-product values at the different units for similar other conditions. The mean glandular dose averaged for each unit is smaller than 2 mGy, the value used for risk assessment for the national breast screening programme. The largest values are not high compared to 2 mGy, i.e. from 3.33 to 4.84 mGy. Acknowledgments This work was supported in part by the Dutch ‘College voor Zorgverzekeringen (Health Care Insurance Board)’, previously ‘Ziekenfondsraad’. The authors would like to thank BSc L J Oostveen and MSc P A M van de Looi of the LRCB for their contribution to this project. References Briesmeister J F 2000 MCNP—A General Monte Carlo N-Particle Transport Code (Los Alamos, NM: Los Alamos National Laboratory) Manual LA-13709-M, version 4C Burch A and Law J 1995 A method for estimating compressed breast thickness during mammography Br. J. Radiol. 68 394–9 Dance D R, Skinner C L, Young K C, Beckett J R and Kotre C J 2000 Additional factors for the estimation of mean glandular breast dose using the UK mammography dosimetry protocol Phys. Med. Biol. 45 3225–40 EC (European Commission) 1996 European Protocol for Dosimetry in Mammography (Luxembourg: Office for the Official Publications of the Commission of the European Communities) Eklund S, Thilander A, Leitz W and Matsson S 1993 The impact of anatomic variations on absorbed radiation doses in mammography Radiat. Prot. Dosim. 49 167–70 Hammerstein G R, Miller D W, White D R, Masterson M E, Woodard H Q and Laughlin J S 1979 Absorbed radiation dose in mammography Radiology 130 485–91 Jansen J T M, Veldkamp W J H, van Woudenberg S, Thijssen M A O and Zoetelief J 2005 Method for determination of the mean fraction of glandular tissue in individual female breasts using mammography Phys. Med. Biol. 50 5953–67 Klein R, Aichinger H, Dierker J, Jansen J Th M, Joite-Barfuß S, S¨abel M, Schulz-Wendtland R and Zoetelief J 1997 Determination of average glandular dose with modern mammography units for two large groups of patients Phys. Med. Biol. 42 651–71 Press W H, Flannery B P and Vetterling W T 1987 Numerical Recipes. The Art of Scientific Computing (Cambridge: Cambridge University Press) Zoetelief J, de Wit N J P and Broerse J J 1989 Technical and dosimetric aspects of quality control in mammography Technical and Physical Parameters for Quality Assurance in Medical Diagnostic Radiology: Tolerances, Limiting Values and Appropriate Measuring Methods ed B M Moores, F E Stieve, H Eriskat and H Schibilla BIR-Report 18 pp 143–6