MR Imaging of Large Multinodular Goiters: Observer Agreement on Volume Versus Observer Disagreement on Dimensions of the Involved Trachea

Transcription

1 MR Imaging of Large Multinodular Goiters: Observer Agreement on Volume Versus Observer Disagreement on Dimensions of the Involved Trachea Steen J. Bonnema 1 Peter B. Andersen Dorthe U. Knudsen Laszlo Hegedüs 1 OBJECTIVE. MR imaging and sonography are considered to be among the most reliable methods available for estimating goiter volume. Our aim was to assess the observer variability of MR imaging for estimating the volume of a large multinodular goiter and the dimensions of the related trachea. Additionally, we compared the goiter volume estimates from MR imaging with those from sonography. SUBJECTS AND METHODS. The effect of high-dose 131 I therapy on the thyroid gland and the impact on the trachea in 3 patients with a large multinodular goiter (range in volume, 1 73 ml) were monitored by observers unaware of duplicated measurements on MR imaging (n = 68) before, 1 week after, and 1 year after 131 I therapy. In goiters exclusively cervically located (n = 1), cross-sectional planimetric sonographic measurements (n = 4) were performed simultaneously with MR imaging. RESULTS. The mean intraobserver difference for the MR imaging measurements of goiter volume was.1 ml (1.4%, p =.3), and the coefficient of variation (CV) ± SD was 3.6% ±.6%. The mean interobserver difference was.4 ml (.3%, p =.86), and the CV ± SD was 4.1% ± 3.5%. Compared with MR imaging, sonography underestimated goiter volume; the mean percentage difference between the volume estimates on MR imaging and those on sonography (volume estimated on MR imaging volume estimated on sonography) was 19.5% (95% limits of agreement:.% to 83.7%), and the CV ± SD was 15.% ± 1.4%. The mean interobserver difference in the MR imaging measurement of tracheal volume along the goiter extension was 7.4% (95% confidence interval: %) and that of the smallest cross-sectional area of the trachea was 7.9% (95% confidence interval:.9 13.%). The corresponding CV ± SD were 8.1% ± 6.6% and 1.3% ± 1.3%, respectively. CONCLUSION. For the estimation of goiter volume, MR imaging has low intra- and interobserver variations. In contrast, the determination of tracheal dimensions using MR imaging has a high variability and, thus, is imprecise. Sonography significantly underestimates thyroid volume compared with MR imaging in patients with a large goiter. Received August, 1; accepted after revision January 3,. Supported by grants from The Agnes and Knut Mørks Foundation and The A. P. Møller Relief Foundation. Presented in part at the annual meeting of the European Thyroid Association, Warsaw, August 1. 1 Department of Endocrinology, Odense University Hospital, DK-5 Odense, Denmark. Address correspondence to S. J. Bonnema. Department of Radiology, Odense University Hospital, DK-5 Odense, Denmark. AJR ;179: X// American Roentgen Ray Society P atients with a large goiter that gives rise to compressive symptoms is a common clinical problem, especially in elderly persons. Although surgical thyroidectomy is the standard therapeutic recommendation [1], radioiodine ( 131 I) therapy may be an alternative. Thus, we [] and others [3] have previously shown that high-dose 131 I therapy can reduce the size of large goiters by 3 4% within the first year. In patients undergoing 131 I therapy, a valid estimate of the goiter volume is essential because this variable must be included in the algorithm for 131 I dose calculation. Furthermore, large goiters commonly compromise the trachea and may occasionally provoke severe respiratory distress. This problem is not easy to anticipate because there is a clear discrepancy between symptoms and respiratory function in patients with a goiter [4]. The smallest cross-sectional tracheal area as measured on MR imaging correlates well with the level of respiratory function []. A matter of concern is that during 131 I therapy the tracheal lumen may be further reduced in susceptible individuals []. Therefore, attention should also be focused on the tracheal anatomy in patients with a large goiter, particularly if 131 I therapy is being considered. Because valid estimates of the volume of the goiter and dimensions of the involved trachea are essential in some patients, the aim of our study was to assess the inter- and intraobserver variabilities of MR measurements of these structures. Also, because sonography is AJR:179, July 59

2 Bonnema et al. the standard method by which to estimate the volume of small goiters [5], a comparison of sonographic measurements with MR imaging measurements was considered relevant. Subjects and Methods Study Population and Design In a previously published article [], we documented the efficacy of 131I therapy in treating patients with large goiters. In summary, during a period of years, 3 patients (1 women, two men; 15 euthyroid, eight hyperthyroid) with a median age of 67 years (range, 4 86 years) who had a large, benign, multinodular goiter and who exhibited major compressive symptoms were consecutively treated with high-dose 131I therapy during hospitalization. Within 1 year, the mean goiter volume had been reduced from 311 ml (range, 1 73 ml) to 15 ml (range, ml). The effect of the 131I therapy was monitored by MR imaging and sonographic measurements at regular intervals. The study was approved by the ethics committee of the county of Funen, Denmark. All patients provided signed informed consent. MR Imaging MR imaging of the neck and thorax was performed on a superconducting system (Gyroscan T5 II; Philips, Eindhoven, The Netherlands) operating at.5 T. T1-weighted images (TR/TE, 7/15) were obtained in the axial, coronal, and sagittal planes using a standard neck coil (Fig. 1A). The slice thick- Fig year-old woman with large multinodular goiter. A, Coronal T1-weighted MR image shows compression of trachea (white arrow) and left-sided substernal extension (black arrows). B and C, Axial T1-weighted MR images show marked contours of goiter (B) and tracheal lumen (C). A B 6 C AJR:179, July

3 ness was 8 mm with an interslice gap of.8 mm covering the entire thyroid gland. On each axial slice, the cross-sectional area of the thyroid and that of the trachea were measured manually by drawing a line along the contours of the thyroid and the tracheal lumen (Figs. 1B and 1C), and the volume was calculated by multiplying the measured areas with the slice thickness and interslice gap. Two experienced observers (observers A and B) calculated, in a blinded design, three variables: the total goiter volume, the smallest cross-sectional area of the tracheal lumen, and the total tracheal volume along the initial thyroid extension. One observer (observer A) made blinded double calculations (first and second measurements) of the goiter volume. With the patients in a standardized supine position, MR imaging was performed before, 1 week after, and 1 year after 131 I therapy. Thus, after excluding one unsuccessful measurement in one patient, MR imaging data were obtained from the 3 patients on 68 occasions. Sonography We used a specially equipped compound scanner (type 1846; Brüel & Kjœr, Copenhagen, Denmark) mounted with a 5-MHz transducer on a static scanner arm. The principle by which the thyroid volume was estimated is comparable to the MR imaging method just described. With the patient in the supine position and the neck hyperextended, we obtained cross-sectional scans covering the entire gland successively with an interslice gap of.5 cm. On each image, the area outlining the thyroid gland was drawn manually on the screen with a cursor. The volume of the thyroid was estimated using a computerized calculation that incorporated all areas and gap distances [6]. Because of the overlying thoracic skeleton, a valid sonographic measurement is not feasible in goiters with a substernal extension. In our study, 1 of the 3 patients had an exclusively cervical goiter, thus allowing reliable sonographic measurements. Goiter volume was assessed in these 1 patients before and 1 year after 131 I therapy (4 measurements) by one experienced observer (observer C), and these measurements were compared with the corresponding MR imaging data. Statistical Analysis Data are presented as means (±SD) or medians (range). The paired Student s t test and Wilcoxon s signed rank test were used to test for differences in the mean and median values, respectively. Simple linear regression analysis and Spearman s rank correlation coefficient (r) were used to test for correlation between two series of measurements. The intraobserver, interobserver, and intermethod variabilities were determined according to the principles of Bland and Altman [7], and the mean difference and 95% limits of agreement between two measurements were calculated. Because the differences between two estimates were found at each side of zero, the calculation of the mean percentage difference was based on log-transformed data, thereby making a decline of a variable equivalent to an increase. In addition, the coefficient of variation (CV) ± SD is given for each variable. A p value of less than.5 was considered statistically significant. MR Imaging Observers of Multinodular Estimates Goiters of Goiter and Tracheal Volumes and of Tracheal Area TABLE 1 Using Sonography and MR Imaging Variable Estimated and Modality Used Note. Dash ( ) indicates that data were not obtained. First Estimate Observer A Second Estimate Observer B Observer C Sonography Cervical goiter volume (n = 4 measurements) Mean ± SD (ml) 167 ± 86 Median (ml) 175 Range (ml) MR imaging Cervical goiter volume (n = 4 measurements) Mean ± SD (ml) 19 ± 87 Median (ml) 18 Range (ml) Total goiter volume (n = 68 measurements) Mean ± SD (ml) 79 ± ± ± 13 Median (ml) Range (ml) Tracheal volume (n = 68 measurements) Mean ± SD (cm 3 ) 19 ± 6 1 ± 7 Median (cm 3 ) 19 Range (cm 3 ) Smallest cross-sectional area of trachea (n = 68 measurements) Mean ± SD (mm ) 83 ± 7 89 ± 8 Median (mm ) 8 87 Range (mm ) TABLE Data for Variability and Correlations from Thyroid and Tracheal Imaging of Patients with Large Multinodular Goiters Variability Mean Value Difference Between Two Estimates 95% Limits of Agreement Mean % Note. For the intra- and interobserver variations, 68 duplicated measurements were obtained. For the intermethod variation, 4 duplicated measurements were obtained. Individual data are shown in Figures 11. Spearman s rank correlation coefficient = r. For all r values, p <.1. a Calculated as observer A s second estimate minus observer A s first estimate. b The actual result was.1 ml; however, for intraobserver measurements, mean difference is defined as equal to zero. c The actual result was 1.4%; however, for intraobserver measurements, mean difference is defined as equal to zero. d Calculated as observer B s estimate minus observer A s first estimate. 95% Limits of Agreement e Calculated as estimate based on MR imaging minus estimate based on sonography. Coefficient of Variation ± SD (%) Intraobserver variation a Goiter volume. ml b c ± Interobserver variation d Goiter volume.4 ml ± Tracheal volume 1.4 cm ± Smallest cross-sectional 5.8 mm ± area of trachea Intermethod variation e Goiter volume.9 ml ± r AJR:179, July 61

4 Bonnema et al. Results The estimates of each observer are shown in Table 1. According to the MR imaging measurements, the goiters (n = 3) ranged in volume from 61 to 733 ml with a mean of 8 ml. In the subgroup with an exclusively cervical goiter (n = 1), the largest volume was 388 ml (normal range, 1 8 ml [6]). The mean volume determined using MR imaging is significantly higher than the mean determined using sonography (19 vs 167 ml, respectively; p <.5). Comparison of the mean (or median) values of tracheal volume or of the smallest cross-sectional area of the trachea reveals highly significant differences between the estimates of observers A and B (p <.1). By contrast, neither intraobserver (p =.3) nor interobserver (p =.86) variation was significant in regard to the goiter volume estimates. Table shows the variability of the duplicated measurements. The difference and correlation plots of the same variables are given in Figures 5 and Figures 6 9, respectively. The mean intraobserver (i.e., repeatability) and interobserver differences of the MR imaging measurements of goiter volume were.1 ml (1.4%) and.4 ml (.3%), respectively. Because there is neither reason to believe that the two blinded measurements affected each other nor that the process of measurement changed with time, the mean intraobserver difference by definition is given as zero in the table. The limits of agreement were slightly narrower for the intraobserver observations, for which 95% of the measurements would not range below 11.3% or above 1.8% of a second measurement. However, a marked difference between the two measurements in absolute terms was occasionally apparent. Thus, the greatest intra- and interobserver differences were 44 and 57 ml, respectively. Intraobserver Difference of Two Measurements (ml) Mean Goiter Volume (ml) Fig.. Graph shows intraobserver difference of two measurements (difference plot) of goiter volume. Solid line indicates mean value of all measurements (n = 68), and dotted lines denote 95% limits of agreement (mean ± 1.96 SD). Interobserver Difference of Two Measurements (ml) Mean Goiter Volume (ml) Fig. 3. Graph shows interobserver difference of two measurements (difference plot) of goiter volume. Solid line indicates mean value of all measurements (n = 68), and dotted lines denote 95% limits of agreement (mean ± 1.96 SD). 6 AJR:179, July

5 MR Imaging of Multinodular Goiters Although Figures 8 and 9 show a high correlation of the tracheal measurements, the difference plots (Figs. 4 and 5) reveal the high variability of these same data. The 95% confidence interval for the mean difference of the tracheal volume was % and that of the tracheal area was.9 13.%, reflecting a clear bias between the two observers in their measurements of the trachea. A difference between the two observers estimates that is less than 38.8% for tracheal volume and 58.6% for tracheal area cannot be regarded as statistically significant, as evidenced by the 95% limits of agreement (Table ). With a few exceptions, observers using sonography underestimated the goiter volume by a relatively large margin compared with MR imaging (Table and Figs. 1 and 11). The relationship between the two methods found by simple linear regression is shown in Figure 11. The mean difference clearly expanded with an increase in mean volume (Fig. 1). The mean percentage difference between the two imaging methods (MR imaging sonography) was 19.5%. The corresponding 95% limits of agreement ranged from.% to 83.7%, meaning that a volume estimate based on sonography cannot be considered significantly dissimilar to one based on MR imaging unless the sonographic measurement ranges beyond these percentages of the MR imaging estimate. Discussion How is the true volume of the thyroid assessed? This issue has been an important area of research for many years [5], mostly because the epidemiologic surveillance of endemic goiter requires reliable methods. Moreover, in daily clinical practice, a valid thyroid evaluation is crucial [5], particularly if 131 I therapy is being considered. An exact volume estimate in Fig. 4. Graph shows interobserver difference of two measurements (difference plot) of smallest crosssectional area of trachea. Solid line indicates mean value of all measurements (n = 68), and dotted lines denote 95% limits of agreement (mean ± 1.96 SD). Interobserver Difference of Two Measurements (mm ) Smallest Cross-Sectional Area of the Trachea (mm ) 1 Interobserver Difference of 3 Two Measurements (cm ) Fig. 5. Graph shows interobserver difference of two measurements (difference plot) of tracheal volume along initial goiter extension. Solid line indicates mean value of all measurements (n = 68), and dotted lines denote 95% limits of agreement (mean ± 1.96 SD) Tracheal Volume Along the Initial Goiter Extension (cm ) AJR:179, July 63

6 Bonnema et al. vivo is essentially impossible to determine directly, and given the high vascularity of the gland, neither surgical nor postmortem specimens can be considered the gold standard. Although simple and without cost, neck palpation is a notoriously imprecise method with a very high interobserver variation both with regard to the assessment of the type of disease (diffuse or nodular) [8] and to the determination of goiter size [9]. Therefore, the thyroid must be evaluated using other techniques, each of which has limitations. The advantages and limitations of MR imaging and CT for thyroid imaging have recently been reviewed [1]. MR imaging provides high-resolution three-dimensional images allowing exact identification of the boundaries of any anatomic structure. This information is useful for evaluation of a large multinodular goiter that is irregular in shape. In contrast, thyroid scintigraphy is often disabled in this setting because of the inconsistent functioning of a multinodular goiter, and the competence of sonography is reduced in cases of substernal goiter extension, which are frequently encountered. Furthermore, MR imaging has been shown to be superior to CT in assessing cases of substernal goiters when imaging findings are compared with operative findings [11]. The variability of MR imaging for volume esti- Observer A's First Volume Estimates (ml) Observer A's Volume Estimates (ml) Observer B's Second Volume Estimates (ml) Observer B's Volume Estimates (ml) Fig. 6. Graph shows intraobserver correlation of measurements (n = 68) of goiter volume. Regression line is shown. Fig. 7. Graph shows interobserver correlation of measurements (n = 68) of goiter volume. Regression line is shown Observer A's Estimates of Tracheal Area (mm ) Observer A's Estimates of 3 Tracheal Volume (cm ) Observer Estimates B's of Tracheal Area (mm ) Observer B's Estimates of Tracheal Volume (cm ) Fig. 8. Graph shows interobserver correlation of measurements (n = 68) of smallest cross-sectional area of trachea. Regression line is shown. Fig. 9. Graph shows interobserver correlation of measurements (n = 68) of tracheal volume along initial goiter extension. Regression line is shown. 64 AJR:179, July

7 MR Imaging of Multinodular Goiters Difference of Measurements Based on MR Imaging vs Sonography (ml) mates of large goiters has been evaluated previously by Huysmans et al. [1]. In duplicated measurements, they found an intraobserver CV ± SD of.% ±.% and an interobserver CV ± SD of 4.1% ±.%. In that study, a high correlation of volume estimates was obtained from coronal, sagittal, and axial measurements. The high reproducibility was confirmed in our study comprising more than three times the number of measurements with a similar interslice thickness and slice gap as in the former study. The relationship between two observations is traditionally given as rank correlation, as in the study of Huysmans et al. However, as argued by Bland and Altman [7], a correlation coefficient of two measurements of the same object is not very meaningful. A more relevant parameter of agreement is the mean difference between the two measurements. By this approach, the limits of agreement can be calculated, either in absolute terms or as a percentage. Therefore, whether a difference between two values is statistically significant can be more directly assessed. We found a mean difference of the goiter volume estimates of approximately 1% with only a marginal aberration between intra- and interobserver agreements. Thus, our data, in addition to those of Huysmans et al., lead us to conclude that MR imaging is precise in this setting. Whether MR imaging also has a high accuracy remains to be confirmed, but for now, this high-precision method should be preferred to other techniques when determining the size of a large goiter. Cervical compression and dyspnea are common symptoms among patients with a goiter. However, the subjective feeling of respiratory distress correlates poorly with the impact of the goiter on the upper airways as verified by a lung function test [4]. On CT scans, the tracheal cross-sectional areas also show a poor relationship with lung function [13]. However in an earlier study [], we found that MR imaging estimates of the tracheal area correlate significantly with inspiratory capacity in patients with a large goiter. In particular, if 131 I therapy is being considered, one should pay attention to the involved section of trachea because this type of therapy occasionally causes the thyroid to enlarge by 1 5% during the early postradiation period [, 14]. Although other authors [3] have used MR imaging for tracheal imaging in patients with a large goiter, this application has not previously, to our knowledge, been the object of a closer evaluation. Our data show that the interobserver agreement was poor with regard to tracheal area as well as tracheal volume. Furthermore, a systematic measurement bias between the two observers was noticed. Thus, despite the high resolution of the MR images, the poor precision of the tracheal measurements probably reflects the small dimensions of the structure being examined. In consequence, if tracheal dimensions are monitored on MR imaging, these dimensions should be based on repeated measurements (as was the case in our former study []) Mean Goiter Volume (ml) Fig. 1. Graph shows difference between MR imaging and sonographic goiter volume measurements (difference plot). Solid line indicates mean value of all measurements (n = 4), and dotted lines denote 95% limits of agreement (mean ± 1.96 SD). Estimates of Goiter Volume on Sonography (ml) to increase the accuracy of the measurement; however, even with repeated measurements, the reproducibility is still low. We suggest that in evaluating a goiter s impact on the trachea imaging findings cannot be used alone and that a lung function test should be included. Although not evaluated in this study, volume calculations made from thyroid scintigrams, which have been in use for many Estimates of Goiter Volume on MR Imaging (ml) Fig. 11. Graph shows correlation between MR imaging and sonographic goiter volume measurements (n = 4). Regression line is shown. [Volume on sonography = 1.3 ml + (.93 volume on MR imaging)] r =.93, p <.1. AJR:179, July 65

8 Bonnema et al. years because of the lack of other techniques, will be discussed. The applied principles and validation are based on older studies [15 17], most of which included patients with exclusively diffuse goiters. Compared with goiter volumes calculated using surgical or postmortem specimens, scintigraphic volume estimates varied considerably [15 17]. In the more recent study by Huysmans et al. [1] of large multinodular goiters, thyroid volume estimates from 131 I scintigraphy were compared with those from MR imaging. The intra- and interobserver CVs of the scintigraphic measurements ranged from 5% to 1%. Correlation with MR imaging was poor (CV = 17%, r =.67), and scintigraphy underestimated the volume by up to 8%. In patients with a moderately enlarged goiter, Wesche et al. [18] estimated thyroid volume using 131 I scintigraphy based on two different formulas and compared these estimates with sonographic calculations, using the same technique as we used in our study. Correlation of each of the formulas with sonography relied on whether a diffuse or nodular goiter was being measured. Furthermore, considerable disagreement differences of up to nearly % was seen between the two methods. These results indicate that thyroid volume measurements on scintigraphy are unreliable and should not be used. Sonography has been used to estimate thyroid volume for several years [1, 5]. As is the case with scintigraphy, different methods of thyroid measurement on sonography have been adopted. The ellipsoid method evaluated by Brunn et al. [19] and Knudsen et al. [] over- and underestimated, respectively, thyroid volume when compared with autopsy specimens as the reference. The CV of this method was 16% []. The method used in our study requires special equipment and is based on computerized calculations from cross-sectional areas covering the entire gland [6]. The problem with estimating the volume of an irregularly shaped goiter, for which the ellipsoid method is invalid, can thereby be overcome. Evaluation in several studies of this method has shown a high accuracy when compared with findings from surgical or postmortem specimens [6, 1, ]. In our hands, the precision of this method is high: The CV was 5% for intraobserver duplicated measurements [6]. However, validation has until now been based on studies that included only small or moderately enlarged goiters. Our present data on more voluminous goiters further clarify the precision of the method. Compared with MR imaging, the correlation was poor. If we consider MR imaging to be the most definite method available, then sonography clearly leads observers to underestimate goiter volume in most cases. This result cannot be explained by inadequate sonographic measurement due to a substernal extension because these goiters were excluded from this part of the study. Large multinodular goiters, irregular in shape and structure, often profoundly expand widely and laterally in the neck, so the divergence most likely is caused by difficulties in defining the goiter contours by the sonographic method. Increasing the transducer frequency will not solve this problem because imaging of some parts of the thyroid will be more detailed at the expense of poorer penetration into the deep part of the neck. We did not perform duplicated sonographic measurements, which would have enabled us to calculate observer variation. However, because of the marked inaccuracy of sonography compared with MR imaging and because the differences between the two methods extended on both sides of zero, the variability of the sonographic method is probably high. In goiters less than 15 ml in volume, the divergence between MR imaging and sonography was less pronounced, and in these cases we believe that MR imaging can be omitted to reduce cost. Although sonography is still useful for imaging-guided fine-needle biopsy, MR imaging should be preferred in patients with large goiters if a precise volume estimate is essential for example, in patients for whom 131 I therapy is being considered. References 1. Bonnema SJ, Bennedbaek FN, Wiersinga WM, Hegedüs L. Management of the nontoxic multinodular goitre: a European questionnaire study. Clin Endocrinol (Oxf) ;53:5 1. Bonnema SJ, Bertelsen H, Mortensen J, et al. The feasibility of high dose iodine 131 treatment as an alternative to surgery in patients with a very large goiter: effect on thyroid function and size and pulmonary function. J Clin Endocrinol Metab 1999;84: Huysmans DAKC, Hermus ARMM, Corstens FHM, Barentsz JO, Kloppenborg PWC. Large, compressive goiters treated with radioiodine. Ann Intern Med 1994;11: Gittoes NJL, Miller MR, Daykin J, Sheppard MC, Franklyn JA. Upper airway obstruction in 153 consecutive patients presenting with thyroid enlargement. Br Med J 1996;31: Hegedus L. Thyroid ultrasound. Endocrinol Metab Clin North Am 1;3: Hegedüs L, Perrild H, Poulsen LR, et al. The determination of thyroid volume by ultrasound and its relationship to body weight, age, and sex in normal subjects. J Clin Endocrinol Metab 1983; 56: Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurements. Lancet 1986;1: Jarløv AE, Hegedüs L, Gjørup T, Hansen JM. Observer variation in the clinical assessment of the thyroid gland. J Intern Med 1991;9: Jarløv AE, Hegedüs L, Gjørup T, Hansen JM. Accuracy of the clinical assessment of thyroid size. Dan Med Bull 1991;38: Jennings A. Evaluation of substernal goiters using computed tomography and MR imaging. Endocrinol Metab Clin North Am 1;3: Belardinelli L, Gualdi G, Ceroni L, Guadalaxara A, Polettini E, Pappalardo G. Comparison between computed tomography and magnetic resonance data and pathologic findings in substernal goiters. Int Surg 1995;8: Huysmans DA, de Haas MM, van den Broek WJ, et al. Magnetic resonance imaging for volume estimation of large multinodular goitres: a comparison with scintigraphy. Br J Radiol 1994;67: Melissant CF, Smith SJ, Perlberger R, Verschakelen J, Lammers JW, Demedts M. Lung function, CTscan and X-ray in upper airway obstruction due to thyroid goitre. Eur Respir J 1994;7: Nygaard B, Faber J, Hegedüs L. Acute changes in thyroid volume and function following 131 I therapy of multinodular goitre. Clin Endocrinol (Oxf) 1994;41: Allen HC, Goodwin WE. The scintillation counter as an instrument for in vivo determination of thyroid weight. Radiology 195;58: Kelly FJ. Observations on the calculation of thyroid weight, using empirical formulae. J Clin Endocrinol Metab 1954;14: Himanka E, Larsson LG. Estimation of thyroid volume. Acta Radiol 1955;43: Wesche MF, Tiel-van Buul MM, Smits NJ, Wiersinga WM. Ultrasonographic versus scintigraphic measurement of thyroid volume in patients referred for 131 I therapy. Nucl Med Commun 1998;19: Brunn J, Block U, Ruf G, Bos I, Kunze WP, Scriba PC. Volumetric analysis of thyroid lobes by real-time ultrasound [in German]. Dtsch Med Wochenschr 1981;16: Knudsen N, Bols B, Bülow I, et al. Validation of ultrasonography of the thyroid gland for epidemiological purposes. Thyroid 1999;9: Berghout A, Wiersinga WM, Smits NJ, Touber JL. Determinants of thyroid volume as measured by ultrasonography in healthy adults in a non-iodine deficient area. Clin Endocrinol (Oxf) 1987;6: Yokoyama N, Nagayama Y, Kakezono F, et al. Determination of the volume of the thyroid gland by a high resolutional ultrasonic scanner. J Nucl Med 1986;7: AJR:179, July