Sonographic Assessment of the Severity and Progression of Autosomal Dominant Polycystic Kidney Disease: The Consortium of Renal Imaging Studies in Polycystic Kidney Disease (CRISP)

Sonographic Assessment of the Severity and Progression of Autosomal Dominant Polycystic Kidney Disease: The Consortium of Renal Imaging Studies in Polycystic Kidney Disease (CRISP) W. Charles O’Neill, MD, Michelle L. Robbin, MD, Kyongtae T. Bae, MD, PhD, Jared J. Grantham, MD, Arlene B. Chapman, MD, Lisa M. Guay-Woodford, MD, Vicente E. Torres, MD, Bernard F. King, MD, Louis H. Wetzel, MD, Paul A. Thompson, PhD, and J. Philip Miller, AB ● Background: The accuracy and precision of ultrasonography (US) in assessing the severity of autosomal dominant polycystic kidney disease (ADPKD) is unknown. Methods: US and magnetic resonance imaging (MRI) were performed at baseline and 1 year on 230 subjects with ADPKD. Ellipsoid volume was calculated from US length, width, and depth, and sequential transverse images were used to measure total and cystic volume directly. These were compared with MRI measurements of kidney volume and cystic volume. Results: Variability between different sonographers ranged from 18% to 42%. Correlations between US and MRI volume were 0.88 and 0.89. The SD of the discrepancy from MRI ranged from 21% to 33% and was unrelated to kidney size or body mass. Kidney length was the most reproducible measurement, and its correlation with MRI volume was 0.84. All patients with an US volume less than 700 cm3 had an MRI volume less than 1,000 cm3, and all patients with an US volume greater than 1,700 cm3 had an MRI volume greater than 1,000 cm3. Increases in volume after 1 year were 12% ⴞ 36% for the ellipsoid method, 6% ⴞ 29% for the direct method, and 4.2% ⴞ 7.2% for MRI. Correlation between US and MRI measurement of fractional cyst volume was 0.80. Conclusion: Sonographic measurement of kidney volume in patients with ADPKD is inaccurate and lacks the precision necessary to measure short-term disease progression. However, sonography can provide an estimate of kidney volume that reflects severity and prognosis in individual patients. Am J Kidney Dis 46:1058-1064. © 2005 by the National Kidney Foundation, Inc. INDEX WORDS: Kidney volume; ultrasonography; magnetic resonance imaging.

A

UTOSOMAL DOMINANT polycystic kidney disease (ADPKD) is a progressive disease characterized by the development of multiple cysts and marked renal enlargement. Be-

From the Department of Medicine, Emory University School of Medicine, Atlanta, GA; Departments of Radiology and Medicine, University of Alabama at Birmingham, Birmingham, AL; Department of Radiology and Division of Biostatistics, Washington University, St Louis, MO; Departments of Medicine and Radiology, University of Kansas Medical Center, Kansas City, KS; and Departments of Medicine and Radiology, The Mayo Foundation, Rochester, MN. Received April 6, 2005; accepted in revised form August 23, 2005. Originally published online as doi:10.1053/j.ajkd.2005.08.026 on October 31, 2005. Supported in part by grant no. U01-DK56956 from the National Institutes of Diabetes and Digestive and Kidney Diseases and by the National Institutes of Health General Clinical Research Center grants M01-RR00039 (Emory University), M01-RR00585 (Mayo Foundation), and M01RR00052 (University of Alabama at Birmingham). Address reprint requests to Charles O’Neill, MD, Emory University, Renal Division WMB 338, 1639 Pierce Dr, Atlanta, GA 30322. E-mail: [email protected] © 2005 by the National Kidney Foundation, Inc. 0272-6386/05/4606-0007$30.00/0 doi:10.1053/j.ajkd.2005.08.026 1058

cause a decline in renal function occurs late in this disorder1 after substantial distortion and destruction of the normal parenchyma has occurred, potential therapies must be instituted before renal function declines. Assessment of the efficacy of these therapies therefore must be based on changes in kidney structure, rather than kidney function. Renal volume and renal cystic volume both correlate with renal outcomes in patients with ADPKD2-5 and may be useful markers of disease progression. However, the best method to measure these parameters is unknown. Ultrasonography (US) was the earliest method used to measure kidney volume in vivo and has the advantage of being widely available, easily performed, and inexpensive. However, despite good correlation with kidney volume ex vivo,6 measurements of normal kidney volume in vivo have shown poor accuracy and reliability.7-9 US has been used to quantitate kidney volume and cyst burden in patients with ADPKD2 and follow up their disease progression,3 but its accuracy and reproducibility in these kidneys are unknown. Sonography also is the test of choice for the diagnosis of ADPKD, which is based on cyst quantity that exceeds the age-dependent incidence in the general population.10,11

American Journal of Kidney Diseases, Vol 46, No 6 (December), 2005: pp 1058-1064

ULTRASOUND MEASUREMENTS IN ADPKD

More reproducible measurements of kidney volume can be obtained by means of computed tomography and have been used to follow up the progression of ADPKD,5,12 but exposes patients to ionizing radiation. This can be avoided with the use of magnetic resonance imaging (MRI), in which renal volume is measured by tracing the kidneys in sequential images and counting voxels, the 3-dimensional equivalent of pixels.4,9,13 Although this measurement has never been compared with true renal volume ex vivo, it shows a high degree of accuracy and reproducibility in other human organs9 and porcine kidneys ex vivo8 and has been used to follow up progression in patients with ADPKD.4 In a previous publication from the Consortium of Renal Imaging Studies in Polycystic Kidney Disease study, MRI yielded accurate measurements of volume in phantom ADPKD kidneys and highly reproducible measurements of kidney volume in patients with ADPKD.13 However, this technique is limited by cost and availability. The Consortium of Renal Imaging Studies in Polycystic Kidney Disease was formed by the National Institutes of Health to identify imaging methods that could detect short-term progression of patients with ADPKD and therefore be useful in evaluating potential therapies. One of the goals of this study is to determine the accuracy and reliability of US measurements of renal volume and cyst burden by comparison with similar measurements performed using MRI. A detailed description of this study, baseline clinical characteristics of the cohort, and baseline MRI measurements have been published.13 METHODS

Patients US and MRI studies were performed at 4 different centers that recruited a total of 241 subjects. Age ranged from 15 to 46 years, and creatinine clearance, either measured or estimated by using the Cockcroft-Gault equation, was greater than 70 mL/min/1.73 m2 (⬎1.17 mL/s/1.73 m2). Glomerular filtration rate, measured as iothalamate clearance, ranged from 41 to 186 mL/min/1.73 m2 (0.68 to 3.10 mL/s/1.73 m2), with a mean of 98 mL/min/1.73 m2 (1.63 mL/s/1.73 m2). Subjects were studied at baseline and then at 1-year intervals for 3 years. MRI was performed at each visit, whereas sonography was performed only at baseline and 1 year.

Sonography Scans were performed with commercial equipment using curvilinear phased-array probes ranging in frequency from 2 to

1059

5 MHz by experienced sonographers (no more than 3 at each site) specifically trained to perform the protocols. Equipment varied at different sites. Two different methods were used to determine renal volume. First, volume was determined from maximum length (L), width (W), and depth (D), using the formula for an ellipsoid (␲/6 ⫻ L ⫻ W ⫻ D). Length and width were obtained from longitudinal images acquired in planes ranging from sagittal to coronal, whereas depth was obtained from transverse images of the midkidney acquired in the plane perpendicular to the longitudinal plane. Volume also was determined directly by measuring the cross-sectional area of the kidney in sequential transverse images and multiplying the sum by the slice interval of 2 cm. All measurements were made by the sonographer during scanning. Fractional cystic volume was determined by measuring the cystic area of each sequential transverse image. This was performed by means of a stereological technique after submission of the images to the Data Coordinating and Image Analysis Center (DCIAC). For all studies, subjects were fasting and were examined in the supine or lateral decubitus position.

Magnetic Resonance Imaging This procedure has been described in detail in a previous publication.13 Briefly, subjects were scanned in the supine position by using 1.5-T scanners. Coronal T2-weighted images were obtained with a fixed slice thickness of 3 mm to cover both kidneys anteroposteriorly during breath hold. Coronal T1-weighted images were obtained with a fixed slice thickness of 3 mm. Kidney volume was determined by summing crosssectional areas of T1-weighted images and multiplying by slice interval. These areas were measured by using a stereological technique in which points of a superimposed grid that overlay the kidney were counted. Cyst volume was determined from T2-weighted images by summing cyst areas for each image and multiplying by slice interval. Cyst area was determined by means of a region-based thresholding method in which the threshold was selected by the analyst. All measurements were performed at the DCIAC. Reliability coefficients for measurements of polycystic kidney phantoms were 0.99 for total volume and 0.89 for cystic volume.13

Data Analysis All studies were transmitted electronically in Digital Imaging and Communications in Medicine format to the DCIAC for analysis. Data in addition to images were submitted to the DCIAC through an Internet-based entry system. All data were entered into a centralized database maintained at the DCIAC and available to the investigators.

RESULTS

Adequate sonographic measurements for determining renal volume by the ellipsoid method were obtained in 230 subjects at baseline and 205 (left kidney) and 206 (right kidney) subjects at 1 year. Adequate sequential transverse images were obtained from 229 (left kidney) and 226 (right kidney) subjects at baseline and 208 subjects at 1 year. MRI measurements of both kidneys were obtained

1060

O’NEILL ET AL

Fig 1. Comparison of baseline measurements of renal volume by means of US and MRI. (Left) Sonographic volume determined by using the ellipsoid formula, n ⴝ 460 kidneys from 230 subjects. Linear regression yielded an r of 0.88. (Right) Sonographic volume determined directly from sequential transverse images, n ⴝ 226 right kidneys and 229 left kidneys. Linear regression yielded an r of 0.89. Solid lines, lines of identity; dashed lines, linear regressions.

for all 241 subjects at baseline and 228 subjects in 1 year. Both US and MRI measurements were not obtained in 11 subjects at 1 year. The reproducibility of US measurements of renal volume was assessed by performing multiple studies by different sonographers using different equipment on 3 subjects whose kidney volumes ranged from 307 to 439 cm3 by means of MRI. Coefficients of variation for renal volume determined by using the ellipsoid method were poor, ranging from 21% to 35%. Of component measurements, length showed the least variability, with coefficients of variation between 4% and 10%. The reproducibility of direct sonographic measurements of renal volume was equally poor, with coefficients of variation ranging from 18% to 42%. This is in contrast to the 1.7% coefficient of variation for MRI measurements of renal volume, determined by scanning 4 patients with ADPKD at each center.13 Comparisons between sonographic and MRI measurements of renal volume are shown in Fig 1. Because results did not differ between right and left kidneys, both kidneys are included. Correlation with MRI volume was similar for the 2 methods (r ⫽ 0.88 for ellipsoid versus r ⫽ 0.89 for direct). Because the distribution of kidney volumes was skewed toward larger volumes, regression was repeated on log-transformed data and resulted in correlation coefficients that were slightly larger (0.91 and 0.92). Better assessment

of sonography is obtained by examining the deviation from MRI volume, described by Bland and Altman,14 in which the mean difference indicates any bias and the SD of differences indicates variability. For the ellipsoid method, US volume was 11% greater than MRI volume, with an SD of 34%. For the direct method, mean difference was 9%, with a SD of 27%. None of the correlations with MRI volume were improved when subsets based on kidney size were analyzed (not shown). There was weak correlation between errors for the ellipsoid and direct methods (r ⫽ 0.46), and for each method, there was correlation between error for the right kidney and left kidney in each patient (r ⫽ 0.48 for ellipsoid; r ⫽ 0.53 for direct). There was weaker correlation between errors in baseline and 1-year measurements (r ⫽ 0.44 for ellipsoid; r ⫽ 0.33 for direct). Errors in either US measurement of kidney volume did not correlate with kidney size or body mass index (all r ⬍ 0.17). The error differed among centers, with 2 centers showing large mean overestimations and 2 centers showing slight underestimations, and this pattern was consistent between the 2 measurements. Differences between baseline and 1-year measurements are listed in Table 1. Both sonographic measurements showed a mean increase in kidney volume during 1 year, but increases were greater than those measured by using MRI. Each measurement used to calculate ellipsoid volume also

ULTRASOUND MEASUREMENTS IN ADPKD

1061

Table 1. Comparison of 1-Year and Baseline Measurements 1 Year – Baseline (% of baseline) Right

Left

Parameter

Mean

SD

Mean

SD

Ellipsoid US volume Length Width Depth Direct US volume MRI volume

11 1.9 1.6 8.4 7.6 4.2

30 11 22 21 31 7.3

14 2.8 0.8 9.0 4.7 4.2

41 13 22 25 27 7.1

showed mean increases after a year. Variability in volume change was much greater for US than MRI, again reflecting the poorer reproducibility of US. However, the difference in length showed much less variability than that in width or depth. The relationship between baseline and 1-year measurements is shown graphically in Fig 2 and includes both kidneys because there was no difference between right and left kidneys. Correlations were similar for the ellipsoid (r ⫽ 0.87) and direct methods (r ⫽ 0.89), but much weaker than that for MRI (r ⫽ 0.99), even after log transformation (r ⫽ 0.91 for either US method). Because sonographic measurement of renal length showed less variability than sonographic measurement of volume, the relationship between length and MRI volume was examined (Fig 3). Linear regression showed r of 0.84, only slightly weaker than the correlation between either sonographic volume measurement and MRI volume.

Theoretically, volume should correlate linearly with the cube of renal length if renal enlargement is symmetric. However, linear regression was not improved (r ⫽ 0.82) when the cube of length was compared with MRI volume (data not shown). To determine whether sonography might be useful in stratifying patients into broad ranges of kidney volume, subjects were divided on the basis of combined kidney volumes measured by means of US (⬍500, 500 to 999, 1,000 to 1,999, and ⱖ2,000 cm3). Combined kidney volumes measured by means of MRI in each group are shown in Fig 4. Although there was considerable overlap between adjacent stratifications, some useful distinctions could be achieved. All patients with a combined sonographic renal volume less than 500 cm3 had a combined MRI volume less than 800 cm3. US volume less than 1,000 cm3 indicated an MRI volume less than 1,250 cm3, whereas US volume greater than 2,000 cm3 indicated an MRI volume greater than 1,250 cm3. Stratification by US length was less useful (data not shown). The other sonographic measurement applied to polycystic kidneys was fractional cystic volume, determined by tracing cysts within each transverse section. To determine the reproducibility of this measurement, 3 investigators independently analyzed 76 different transverse images from 6 subjects. Correlation coefficients of the 3 pairings of investigators were 0.75, 0.77, and 0.84. Comparison of fractional cyst volumes measured by means of MRI and US at baseline is shown in Fig 5. Correlation between the 2 measurements was poor,

Fig 2. Comparison of baseline and 1-year measurements of renal volume by means of US. Sonographic volume determined by (A) the ellipsoid formula, n ⴝ 199 right kidneys and 198 left kidneys; and (B) sequential transverse images, n ⴝ 199 right kidneys and 197 left kidneys. Lines of identity are shown.

1062

O’NEILL ET AL

Fig 3. Correlation between kidney length by means of US and kidney volume by means of MRI. Solid line is the linear regression (r ⴝ 0.84). Baseline measurements; n ⴝ 460.

with r of 0.71, and US yielded lower cystic volumes. Correlation between baseline and 1-year measurements was substantially weaker for US (r ⫽ 0.67) than for MRI (r ⫽ 0.93). DISCUSSION

ADPKD is characterized by progressive cystic enlargement of the kidneys that precedes and predicts the development of renal failure.2-5,15 Thus, renal volume or cystic volume may be clinically useful parameters in assessing risk and progression. The major finding of this study is that sonography lacks the precision necessary to follow up the

Fig 4. Combined MRI kidney volume in different stratifications of combined US kidney volume (ellipsoid method), baseline measurements.

short-term progression of patients with ADPKD. Sonographic measurement of renal volume based on the ellipsoid formula showed poor accuracy and reproducibility in patients with ADPKD. In the aggregate, renal volume was exaggerated, but volume also was undermeasured in many kidneys. The error was not related to kidney size or body habitus, suggesting that it is not caused by patientspecific factors. However, for each examination, the error correlated with the error for the direct measurement of volume by using US, and errors for the right and left kidneys also correlated, indicating examination-specific factors. The large coeffi-

ULTRASOUND MEASUREMENTS IN ADPKD

1063

Fig 5. Comparison of fractional cyst volume measured by means of US and MRI at baseline. Both right and left kidneys are shown; n ⴝ 444. The line of identity is shown.

cient of variation for these measurements when the same patient was studied multiple times by different sonographers and the different errors at different centers suggest that much of the variability is related to the sonographer. Operator dependence is a well-recognized drawback of sonography. The results in ADPKD kidneys are consistent with the use of the ellipsoid formula in normal kidneys. In an early study performed on kidneys ex vivo,6 correlation between US volume and true volume (measured by water displacement) was excellent (r ⫽ 0.97), but there was slight overestimation (6%). However, a more recent ex vivo study showed 24% overestimation.8 The same error occurred with measurements obtained using MRI, indicating that the error lay with the formula. This error also occurred when the formula was used in vivo (US or MRI) and compared with renal volume calculated by means of the MRI voxel method.9 The ellipsoid method also results in large intraobserver and interobserver variability in normal kidneys, ranging from 14% to 31%7,9,16 and similar to the variability observed in ADPKD kidneys. Because 3 separate sonographic measurements are multiplied to obtain the volume, the errors are compounded, potentially leading to large errors. Variability in measurement of length in normal kidneys is only 5%,7,9,16,17 suggesting that most of the error in the ellipsoid volume derives from transverse measurements of width and depth. This also appears to be the case in patients with ADPKD because measurement of length, although

more variable than in normal kidneys, was considerably less than the variability of width or depth measurements. Variability of the latter measurements most likely stems from the difficulty of reproducibly obtaining a transverse image through the midkidney that is precisely perpendicular to the long axis of the kidney and that the kidney crosssection is not a perfect oval, lending some subjectivity to the selection of dimensions to measure. Because the kidney is not a perfect ellipsoid and its shape varies, some investigators advocated an approach essentially identical to the MRI technique: tracing and summing of sequential crosssectional areas.18,19 This should decrease variability because areas are measured directly and then summed, avoiding compounding of errors during multiplication of dimensions for the ellipsoid formula. In ex vivo kidneys, this measure of volume correlated very strongly (r ⫽ 0.93) with kidney weight,19 but neither accuracy nor reproducibility has been determined previously in vivo. In the present study, this method resulted in less overestimation of kidney volume and less variability, although variability was still large. The principal problem encountered with this technique in patients with ADPKD is the difficulty delineating the border of the kidney, particularly the distal border, and the inability to fully image the upper poles in some kidneys. Measurement of kidney length showed the least variability between baseline and 1-year studies of any sonographic parameter, suggesting it is the most reproducible sonographic measurement. How-

1064

O’NEILL ET AL

ever, it did not accurately predict renal volume. Its correlation with renal volume by means of MRI was weaker than the correlation for either US volume measurement, and length was not as useful as volume measurements in stratifying kidney volume. This is almost certainly because kidney growth in patients with ADPKD is not proportional in all dimensions. Thus, renal length alone is not a useful parameter for estimating kidney volume in patients with ADPKD. Measurement of fractional cystic volume by means of sonography also proved difficult and consequently showed much variability compared with MRI. The major problem was differentiation of cysts from parenchyma, indicated by the large variability between analyzers. Measurement of cystic volume by means of MRI showed greater variability from baseline to 1 year than measurement of renal volume by means of MRI. Thus, measurement of cystic volume by any method appears to be inferior to measurement of total renal volume. Although sonography lacks the accuracy and precision necessary to detect short-term progression in patients with ADPKD, it was useful in separating subjects into broad ranges of renal volume. Thus, sonography may be useful in stratifying patients according to risk because renal volume is a risk factor for disease progression.2-5 Sonography is a useful method for diagnosing ADPKD, but the present study clearly shows that current techniques lack the accuracy and reliability to follow up shortterm progression of this disease. Whether newer techniques that decrease operator dependence, such as 3-dimensional sonography, can improve the utility of sonography in patients with ADPKD remains to be determined. Despite the drawbacks of sonography, it remains a useful tool in patients with ADPKD for diagnosis and screening. ACKNOWLEDGMENT The authors wish to acknowledge the excellent work of the following sonographers: Ying Yuan, MD, Kraisith Arya, MD, Duane M. Brakke, RDMS, RVT, Candace S. Spalding, BA, RDMS, RVT, RT(R), Jaina Cox-Mosburg, ASRT(R), RDMS, RVT, Felix A. Hester, RDMS, RVT, RT, and Carl A. Abst, RDMS.

REFERENCES 1. Franz KA, Reubi FC: Rate of functional deterioration in polycystic kidney disease. Kidney Int 23:526-529, 1983 2. Gabow PA, Johnson AM, Kaehny WD, et al: Factors affecting the progression of renal disease in autosomal-

dominant polycystic kidney disease. Kidney Int 41:1311-1319, 1992 3. Fick-Brosnahan GM, Belz MM, McFann KK, Johnson AM, Schrier RW: Relationship between renal volume growth and renal function in autosomal dominant polycystic kidney disease: A longitudinal study. Am J Kidney Dis 39:11271134, 2002 4. King BF, Reed JE, Bergstralh EJ, Sheedy PF, Torres VE: Quantification and longitudinal trends of kidney, renal cyst, and renal parenchymal volumes in autosomal dominant polycystic kidney disease. J Am Soc Nephrol 11:1505-1511, 2000 5. Sise C, Kusaka M, Wetzel LH, et al: Volumetric determination of progression in autosomal dominant polycystic kidney disease by computed tomography. Kidney Int 58:2492-2501, 2000 6. Hricak H, Lieto RP: Sonographic determination of renal volume. Radiology 148:311-312, 1983 7. Sargent MA, Long G, Karmali M, Cheng SM: Interobserver variation in the sonographic estimation of renal volume in children. Pediatr Radiol 27:663-666, 1997 8. Bakker J, Olree M, Kaatee R, De Lange EE, Beek FJA: In vitro measurement of kidney size: Comparison of ultrasonography and MRI. Ultrasound Med Biol 24:683-688, 1997 9. Bakker J, Olree M, Kaatee R, et al: Renal volume measurements: Accuracy and repeatability of US compared with that of MR imaging. Radiology 211:623-628, 1999 10. Ravine D, Gibson RN, Donlan J, Sheffield LJ: An ultrasound renal cyst prevalence survey: Specificity data for inherited renal cystic diseases. Am J Kidney Dis 22:803807, 1993 11. Ravine D, Gibson RN, Walker RG, Sheffield LJ, KincaidSmith P, Danks DM: Evaluation of ultrasonographic diagnostic criteria for autosomal dominant polycystic kidney disease 1. Lancet 343:824-827, 1994 12. Thomsen HS, Madsen JK, Thaysen JH, DamgaardPetersen K: Volume of polycystic kidneys during reduction of renal function. Urol Radiol 3:85-89, 1981 13. Chapman A, Guay-Woodford L, Grantham J, et al: Renal structure in early autosomal dominant polycystic kidney disease (ADPKD): the Consortium for Radiologic Imaging Studies of Polycystic Kidney Disease (CRISP) cohort. Kidney Int 64:1059-1064, 2003 14. Bland JM, Altman DG: Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1:307-310, 1986 15. Gabow PA, Chapman AB, Johnson AM, et al: Renal structure and hypertension in autosomal dominant polycystic kidney disease. Kidney Int 38:1177-1180, 1990 16. Emamian SA, Nielsen MB, Pedersen JF: Intraobserver and interobserver variations in sonographic measurements of kidney size in adult volunteers. Acta Radiol 36:399-401, 1995 17. Ablett MJ, Coulthard A, Lee REJ, et al: How reliable are ultrasound measurements of renal length in adults? Br J Radiol 68:1087-1089, 1995 18. Jones TB, Riddick LR, Harpen MD, Dubuisson RL, Samuels D: Ultrasonographic determination of renal mass and renal volume. J Ultrasound Med 2:151-154, 1983 19. Troell S, Berg U, Johansson B, Wikstad I: Renal parenchymal volume in children. Acta Radiol 29:127-130, 1988