Nephrocalcinosis in Premature Infants: Variability in Ultrasound Detection

Original Article ⅢⅢⅢⅢⅢⅢⅢⅢⅢⅢⅢⅢⅢⅢ Nephrocalcinosis in Premature Infants: Variability in Ultrasound Detection

Thomas Campfield, MD Francis J. Bednarek, MD CONCLUSION: Mariann Pappagallo, MD There is significant variability among radiologists in the ultrasound identification of nephrocalcinosis in premature infants; a 7.5-MHz ul- Frederick Hampf, MD trasound transducer is associated with less variability in recognizing this John Ziewacz, MD lesion. Jacqueline Wellman, MD Gary Rockwell, MD Gregory Braden, MD Renal calcification is an uncommon problem in children, but recent Patrecia Flynn-Valone, RD reports have called attention to the development of nephrocalcinosis Michael Neylan in premature infants. These calcium oxalate deposits may be identi- Antonio Pangan, MD fied on plain films of the abdomen but are more readily detected by ultrasound.1,2 Estimates of the incidence of nephrocalcinosis in pre- 3 OBJECTIVE: mature infants have varied considerably. Jacinto et al. have reported 4 To measure variability among radiologists in the ultrasound diagnosis of a 65% incidence of nephrocalcinosis, whereas Short and Cooke re- nephrocalcinosis in premature infants. ported a 27% incidence in very low birth weight infants. Differences in image quality obtained using different ultrasound transducers or METHODOLOGY: observer variability in image interpretation may have contributed to In this prospective multicenter study, renal ultrasounds were performed this variation in incidence. No data are available regarding these two on 54 very low birth weight infants using a 5.0- and 7.5-MHz transducer, potential sources of variability. It is also possible that improvements and these ultrasounds were read independently by three radiologists. ␬ in neonatal care, leading to a decrease in the severity of chronic lung coefficients were calculated to assess variability in identification of disease and less exposure to diuretic therapy, may have contributed to nephrocalcinosis among the radiologists. a change in the incidence of this lesion. This study was designed to examine the hypothesis that renal RESULTS: ultrasounds performed using a 7.5-MHz transducer are associated The ␬ coefficient (Ϯ confidence intervals) using a 5.0-MHz transducer with less inter-observer variability compared with a 5.0-MHz trans- was 0.143 (0.108, 0.178); using the 7.5-MHz transducer, the ␬ coefficient ducer in the identification of neonatal nephrocalcinosis. was 0.268 (0.243, 0.293). All three radiologists agreed in their identification of nephrocalcinosis on 3 of 54 ultrasounds using a 5.0-MHz trans- METHODS ducer; a total of 6 of 54 ultrasounds obtained using a 7.5-MHz transducer Patients were read as positive by all three radiologists. This prospective study took place between August 1, 1995 and July 31, 1996 at three neonatal intensive care units. All infants with a birth weight Ͻ1500 gm were considered eligible for the study. Infants were not included in the study for the following reasons: failure to obtain Departments of Pediatrics (T. C., G. R.), Radiology (F. H.), and Medicine (G. B.), Baystate Medical Center, Springfield, MA; Departments of Pediatrics (T. C., G. R.), Radiology (F. H.), informed consent from parents, transfer of the infant to another facil- and Medicine (G. B.), Tufts University School of Medicine, Boston, MA; Division of Neonatol- ity before images could be obtained, or the appearance of hydrone- ogy (F. J. B.), Memorial Health Care, University of Massachusetts Medical Center, Worcester MA; Department of Pediatrics and Obstetrics and Gynecology (F. J. B.), University of Massachu- phrosis or other congenital renal abnormalities on the ultrasound setts School of Medicine (A. P.), Worcester MA; Departments of Pediatrics (M. P.) and Radiology images. In addition, some infants could not be included in the study (J. Z.), University of Connecticut Health Center, Farmington, CT; Diagnostic Radiology (J. W.), Memorial Health Care, Worcester, MA; Department of Nutrition, University of Massachusetts, because one of the ultrasound transducers was broken or unavailable Amherst, MA; and Ross Products Division, Abbott Laboratories, Columbus, OH. when the images were obtained. Renal ultrasounds were performed at

This work was supported by Ross Products Division of Abbott Laboratories. 7 to 8 weeks of age using both a 5.0-MHz sector transducer and a

Address correspondence and reprint requests to Thomas Campfield, MD, Baystate Medical 7.5-MHz linear array transducer. The ultrasound equipment used in Center, Newborn Medicine, W2810, Springfield MA 01199. this study included a Toshiba SSH 140 A (Tokyo, Japan) and an ATL

400 Apogee (Bothell, WA). Three longitudinal and three transverse views of each kidney were obtained. Ultrasound images were obtained at the bedside by radiologists and ultrasound technicians, but their bedside impressions of these images were not included in this study. Patient confidentiality was maintained by covering the patient’s name on the ultrasound hard copy image. Image Interpretation Hard copy images of each renal ultrasound were interpreted independently by three radiologists. Of the three radiologists involved in the study, two had training in ultrasonography; one radiologist had training in pediatric radiology as well as ultrasonography. These radiologists were not aware of the infant’s clinical condition, and they were unaware of other radiologists’ interpretation of these images. The only images shown to these radiologists were those described above; if other views were obtained because of a suspected abnormality, these Figure 1. ␬ coefficients (Ϯ confidence intervals) based on the presence or ab- other images were not included in the study. These criteria were estab- sence of nephrocalcinosis on ultrasounds on 54 premature infants, obtained with 5.0- and 7.5-MHz transducers, read independently by three radiologists. lished as part of the study to minimize bias in interpretation. Images were interpreted as normal or positive for nephrocalcinosis based on the presence of echogenic foci with acoustic shadowing. Positive readings were classified as “focal” or “diffuse” nephrocalcinosis. For the obtained using a 5.0-MHz transducer was 0.143 (0.108, 0.178), ␬ purposes of this study, diffuse nephrocalcinosis was defined as whereas a 7.5-MHz transducer resulted in a value of 0.268 (0.243, nephrocalcinosis involving Ͼ50% of both kidneys. Nephrocalcinosis 0.293). Thus, ultrasounds obtained using a 7.5-MHz transducer re- involving Ͻ50% of one or both kidneys was classified as focal sulted in significantly less variability in interpretation compared with nephrocalcinosis. a 5.0-MHz transducer (Figure 1). The percentage of ultrasounds interpreted as positive by individ- Data Analysis ual radiologists using a 5.0-MHz transducer was 28%, 6%, and 35% Interobserver variability in the diagnosis of nephrocalcinosis was for radiologists 1, 2, and 3, respectively. The percentage of ultrasounds ␬ ␬ measured using the coefficient. The coefficient measures agree- interpreted as positive using a 7.5-MHz transducer was 35%, 11%, and ment among observers above that caused by chance alone. In general, 39% for radiologists 1, 2, and 3, respectively (Figure 2). These propor- ␬ values between 0.00 and 0.20 indicate minimal agreement; at the tions were significantly different for both transducers by ␹-squared ␬ other extreme, values between 0.80 and 1.00 indicate nearly perfect analysis. Because radiologists 1 and 3 were similar with regard to ␬ agreement. values were calculated for both the 5.0-MHz and 7.5- their percentage of positive readings, their image interpretations were MHz transducers, and confidence intervals were used to determine compared using the ␬ statistic. This comparison gave a value of 0.433 whether one transducer gave results significantly less variable than with the 5.0-MHz transducer and a ␬ value of 0.400 with the 7.5-MHz the other. transducer, reflecting only moderate agreement between these two To determine whether the variability observed in ultrasound observers. In other words, although the percentages of positive read- interpretation could be attributed to “over-reading” or “under-readings were similar, different images were interpreted as positive fre- ing” by an individual radiologist, the proportion of positive interpre- quently enough to show only moderate agreement using the ␬ tations for each transducer by each radiologist was determined. These statistic. ␹ proportions were compared using the -squared analysis. The proportion of ultrasounds interpreted as positive by all three To determine whether either transducer was associated with more radiologists using the 5.0-MHz transducer was 3 of 54 ultrasounds frequent identification of nephrocalcinosis, the proportion of ultra- (5.6%), whereas 6 of 54 ultrasounds (11.1%) were interpreted as posi- sounds read as positive for nephrocalcinosis by all three radiologists tive by all three radiologists using a 7.5-MHz transducer. Although was determined for each transducer, and these proportions were then there was a trend toward more uniformly positive image interpreta- compared using Fisher’s exact test. tion with the 7.5-MHz transducer, this difference did not achieve sig- nificance (Fisher’s exact test). The three ultrasounds read as abnor- RESULTS mal with the 5.0-MHz transducer were also uniformly read as A total of 54 patients were studied. The mean birth weight of these 54 abnormal by all three radiologists with the 7.5-MHz transducer. infants was 976 Ϯ 28 gm (mean Ϯ SEM) and the mean gestational There were a total of 38 positive readings among all three radiol- age was 27.1 Ϯ 0.3 weeks; ultrasounds were performed at 49.7 Ϯ 1.8 ogists using the 5.0-MHz transducer; only 2 of these 38 (5.3%) were days of age. interpreted as showing diffuse rather than focal nephrocalcinosis. The ␬ statistic (Ϯ confidence intervals) for renal ultrasounds Similarly, of the 46 positive image interpretations with the 7.5-MHz

Journal of Perinatology (1999) 19(7) 498–500 499 Campfield et al. Ultrasound Detection of Nephrocalcinosis

Figure 2. The percentage of renal ultrasounds on 54 premature infants read as positive for nephrocalcinosis by three radiologists. See text for details. transducer, only 3 (6.5%) were interpreted as diffuse rather than focal transducer. The 7.5-MHz transducer provides better spatial resolution nephrocalcinosis. but less penetration, and this may explain the greater uniformity in the identification of nephrocalcinosis in premature infants with this DISCUSSION transducer.5 The ␬ statistic measures agreement between observers, adjusted for It has been suggested that furosemide therapy should be discon- agreement due to chance alone. Renal ultrasounds are often per- tinued when nephrocalcinosis is identified on a renal ultrasound, or that thiazide be added to blunt the hypercalciuric effect of furo- formed in the neonatal intensive care unit to detect nephrocalcinosis, 6 and patient management may be modified based on the interpreta- semide. Based on the results of this study, caution should be used tion of these images. In this study, we used the ␬ statistic to demon- before diuretic therapy is changed based on a single ultrasound read- strate observer variability in sonographic detection of nephrocalcino- ing by an individual radiologist who may “over-read” subtle ultra- sis. This variability may be of clinical importance. sound findings. In this study, hard copy images of renal ultrasounds were exam- In summary, a 7.5-MHz ultrasound transducer is more sensitive ined by three radiologists, and variability in their interpretations was in identifying nephrocalcinosis in premature infants. There is signifi- measured. A weakness of this study is the use of these hard copy im- cant variability among radiologists in recognizing these predomi- ages rather than real time images. There may have been less variabil- nately focal deposits of calcium oxalate. ity in interpretation if the three radiologists had gathered at each References patient’s bedside to interpret real time images. Conversely, when the 1. Myracle MR, McGahan JP, Goetzman BW, Adelman RD. Ultrasound diagnosis of hard copy images were viewed, the radiologists were unaware of the renal calcification in infants on chronic furosemide therapy. J Clin Ultrasound infant’s clinical condition, and this may contribute to the validity of 1986;14:281–7. the conclusions. Another weakness of this study is failure to have the 2. McCormick FCS, Brady K, Keen CE. Oxalate nephrocalcinosis: a study in autop- radiologists pinpoint the precise location of nephrocalcinosis when sied infants and neonates. Pediatr Pathol Lab Med 1996;16:479–88. ultrasounds were interpreted as positive for focal nephrocalcinosis. 3. Jacinto JS, Modanlou HD, Crade M, Strauss AA, Bosu SK. Renal calcification The identification of nephrocalcinosis could be made with more con- incidence in very low birth weight infants. Pediatrics 1988;81:31–5. fidence if the radiologists had not only agreed on the presence of nephrocalcinosis but also agreed on its precise location. This was not 4. Short A, Cooke RW. The incidence of renal calcification in preterm infants. Arch included as part of the study design because we did not anticipate that Dis Child 1991;66:412–7. nearly all of the ultrasounds interpreted as abnormal would be con- 5. Zagzebski JA. Essentials of Ultrasound Physics. St Louis: Mosby; 1996. sidered to have focal rather than diffuse nephrocalcinosis. 6. Hufnagle KG, Khan SN, Penn D, Cacciarelli A, Williams P. Renal calcifications: a Ultrasounds performed using a 7.5-MHz transducer were associ- complication of long-term furosemide therapy in preterm infants. Pediatrics ated with a trend toward less variability compared with a 5.0-MHz 1982;70:360–3.

500 Journal of Perinatology (1999) 19(7) 498–500