Interobserver Variation In Measurement Research Articles

Effect of patellar luxation on radiographic measurements of tibial plateau angle in small-breed dogs.

OBJECTIVE To determine whether medial patellar luxation would affect radiographic tibial plateau angle (TPA) measurements in small-breed dogs. DESIGN Prospective cross-sectional study. ANIMALS 15 small-breed dogs (25 stifle joints) with grade 2 or 3 medial patellar luxation (5 dogs with unilateral luxation and 10 dogs with bilateral luxation). PROCEDURES Digital mediolateral radiographic images of each affected stifle joint were acquired with the patella in manually reduced (n = 25) and luxated (25) positions. In 2 measurement sessions separated by > 48 hours, 3 observers unaware of patella status (luxated or reduced) measured the TPA in each image twice in random order. Mixed linear modeling was performed to determine the effect of patella status on TPA measurements, and intraobserver and interobserver variation in measurements were calculated. RESULTS TPA measurements by all observers differed significantly between the first and second measurement sessions, but by a mean value of only 0.7°. A few significant differences were identified between 1 pair of observers by patella status and between patella statuses for 1 observer, but all mean differences were ≤ 1.7°. No significant difference in intraobserver variation was identified between patella statuses for any observer. Interobserver variation was not affected by patella status and measurement session. CONCLUSIONS AND CLINICAL RELEVANCE Although some significant differences were identified in radiographic TPA measurements in dogs with medial patellar luxation, depending on whether the patella was luxated or manually reduced, these differences were so small they could be considered clinically unimportant. Consequently, we believe that in small-breed dogs with patellar luxation, patella status would be unlikely to have a clinically meaningful effect on the measured TPA.

A cadaver study validating CT assessment of acetabular component orientation: the Perth CT hip protocol.

The aim was to develop a CT-based protocol to accurately measure post-operative acetabular cup inclination and anteversion establishing which bony reference points facilitate the most accurate estimation of these variables. An all-polyethylene acetabular liner was implanted into a cadaveric acetabulum. A conventional pelvic CT scan was performed and reformatted images created in both functional and anterior pelvic planes. CT images were transferred to a Freedom-Plus Graphics software package enabling an identical, virtual 3D model of the cadaveric pelvis to be created and definitive acetabular cup orientation established. Using coronal and axial slices of the CT scans, acetabular cup inclination and anteversion were measured on five occasions by ten radiographers using differing predetermined bony landmarks as reference points. The intra- and inter-observer variation in measurement of acetabular cup orientation using varying bony reference points was assessed in comparison to the elucidated definitive cup position. Virtually derived definitive acetabular cup orientation was measured showing cup inclination and anteversion as 41.0 and 22.5° respectively. Mean CT-based measurement of cup inclination and anteversion by ten radiographers were 43.1 and 20.8° respectively. No statistically significant difference was found in intra- and inter-observer recorded results. No statistically significant differences were found when using different bony landmarks. CT assessment of acetabular component inclination and anteversion is accurate, reliable and reproducible when measured using differing bony landmarks as reference points. We recommend measuring acetabular inclination and anteversion from the inferior acetabular wall/teardrop and posterior ischium respectively.

Evaluation of antenatal umbilical coiling index at 16-21 weeks of gestation as a predictor of trisomy 21 and other chromosomal defects

To determine whether there is an association between sonographically assessed hyper- or hypocoiling of the umbilical cord and the presence of trisomy 21, to provide reference values for the antenatal umbilical coiling index (aUCI) at a gestational age of 16-21 weeks and to determine whether these measurements are reliable and reproducible. This was a prospective study of 737 pregnancies in which the aUCI was measured between 16 and 21 weeks of gestation by ultrasound at the time of amniocentesis. The aUCI was calculated as the reciprocal value of the mean length of one complete coil in centimeters. We created reference curves and studied the relationship with trisomy 21 and other chromosomal defects. In 30 pregnancies we studied the intra- and interobserver variation in measurements using Bland-Altman plots with associated 95% limits of agreement and intraclass correlation coefficients. aUCI was found to be non-linearly related to gestational age at 16-21 weeks and reference curves were created for the mean aUCI and the 2.3(rd) , 10(th) , 90(th) and 97.7(th) percentiles. There was no significant difference in aUCI values between the reference group (n = 714) and cases with trisomy 21 (n = 16) or other aneuploidies (n = 7) (one-way ANOVA, P = 0.716). There was good intra- and interobserver agreement in aUCI measurements. The aUCI can be measured reliably and varies according to gestational age at 16-21 weeks. The aUCI was not significantly associated with trisomy 21 or other chromosomal defects.

P11.11: Interobserver variation in measurements of Cesarean scar defect and myometrium with 3D ultrasonography

Results: Thirty-eight women were diagnosed of Cesaeran scar defect at TUS. The number of previous Caesarean sections in patients with uterine dehiscence and ectopic pregnancy in Cesarean scar is shown in Table 1. Eleven patients (29%) showed a uterine dehiscence. Seven cases were repaired by laparoscopy and 4 cases required hysterectomy. Five ectopic pregnancies (13%) were diagnosed at TUS. Two patients required hysterectomy, 2 were treated with local methotrexate guided by TUS and 1 was surgically sutured. Two patients (5%) had a complete uterine rupture. One of them diagnosed 6 months after Cesarean section and required hysterectomy. The other which was diagnosed during puerperium period was treated by surgical repair. Conclusions: TUS is useful for detecting Cesarean scar defects providing information for a treatment in case of complications.

Use of a curved-array transducer to reduce interobserver variation in sonographic measurement of thyroid volume in healthy adults.

Sonographic calculation of thyroid volume is used in the diagnosis and follow-up of thyroid diseases. Since the calculated volume of thyroid lobes is highly influenced by the longest (ie, craniocaudal) diameter, we examined whether using a curved-array transducer as opposed to a linear-array transducer to measure the craniocaudal diameter would reduce interobserver variation. Three sonographers with different levels of expertise each used a 5-12-MHz linear-array transducer and a 2-5-MHz curved-array transducer to measure the craniocaudal diameter of both thyroid lobes of 25 healthy volunteers. On the basis of these measurements, thyroid lobe volumes were calculated. Single-factor analysis of variance was used to evaluate the interobserver variations between the measurements made by all 3 observers as well as between measurements taken by pairs of observers. A p value of less than 0.05 was considered significant. Using the linear-array transducer to measure the craniocaudal diameter resulted in significant interobserver variation in thyroid volume calculation (p = 0.02), whereas using the convex-array transducer did not. Using either transducer resulted in a highly significant interobserver variation in measurements of the craniocaudal diameter, although the variation was far more pronounced for measurements made with the linear-array transducer (p = 0.0005) than for those made with the curved-array transducer (p = 0.04). For both transducers, the interobserver variations were most pronounced between the most and the least experienced sonographers. To avoid significant interobserver variation in calculating thyroid lobe volume, we recommend using a curved-array transducer to measure the craniocaudal diameter of the thyroid lobes.

Determination of Prostatic Volume with Transrectal Ultrasound: A Study of Intra-Observer and Interobserver Variation

Purpose We assessed the intra-observer and interobserver variation in measurements of prostatic volume using transrectal ultrasound. Materials and Methods Two volume estimations were performed by 1 observer in 40 patients and by 2 observers in 75. Results Mean prostatic volumes measured ranged from 18.9 to 87.0 ml. (median 39.8) for 1 observer and from 15.5 to 94.5 ml. (median 42.4) for 2 observers. The mean difference plus or minus standard deviation and limits of agreement between the paired volume estimations were 0.1 plus/minus 3.3 ml. and -6.5 to 6.7 ml., respectively, for 1 observer, and -0.3 plus/minus 5.3 ml. and -10.9 to 10.3 ml., respectively, for 2 observers. The mean difference was 9.8 percent for 2 observers compared to 4.6 percent for 1. Conclusions The results show that there is considerable variation in repeated measurements of prostatic volume and the variation is greatest for 2 observers compared to 1 observer.

The hazards of epidemiology, continued.

Professor Holland's May 1995 editorial1 on the hazards of epidemiology has stirred up something of a hornets' nest on EPIDEMIO-L, an electronic mailing list for epidemiologists (subscribe to Listproc@CC.UMontreal.CA). Following an extensive discussion on this list, I have been deputed to respond. Leaving aside the dearth of references to support some very sweeping condemnations, the editorial presents, in some respects, a caricature of epidemiology. Holland's suggestion that nonmedical epidemiologists are unaware of temporal drift in diagnoses and interobserver variation in measurement is simply incorrect. Much basic work on these two issues has been done by nonmedical people, including people who have worked with Holland. Indeed, the problem of measurement error is one of the many reasons why epidemiologists need a wide variety of skills and often come from divergent professional backgrounds. No one would dispute that all epidemiology should include detailed consideration of biological and clinical evidence. It should also include consideration of social, anthropological, chemical, toxicological, and physical evidence. This is usually beyond the competence of any one person, no matter how talented. Serious epidemiologists, regardless of their professional background, will have to become familiar with much of the detail of the biology and statistics of their chosen topic. Clearly, clinicians have a head start on pathophysiology, and statisticians have a head start on mathematics. This does not invalidate the contributions of either. The composition of research teams will vary widely depending on contingency, personal interactions, and funding. There is no magic formula for establishing a winning team. Few if any teams will contain all of the possible contributors, but we believe that successful teams, regardless of composition, will be those that not merely listen to others but actively solicit their views. Modern epidemiology needs clinicians, but it also needs a host of other skills. All of us should perhaps stop thinking ofourselves as physician-epidemiologists, statistician-epidemiologists, or whatever. Of interest is the substantive scientific or policy question and the use of sensible methods to answer interesting questions. The primary degree of the person investigating is surely irrelevant. O Anthony Staines, MRCP, MFPHM, MSc

Reliability of radiographical measurements of spondylolisthesis and extension-flexion radiographs of the lumbar spine

We studied the reliability of extension-flexion radiography by analysing intra-observer and inter-observer variations in measurements of 30 patients with established L5-S1 spondylolisthesis. The highest intra-observer angular variations among three radiologists were found at the L5-S1 level (mean 1.6°, S.D. 1.6°, max. 9°) and the highest sagittal translation also at the L5-S1 level (mean 0.6 mm, S.D. 0.8 mm, max. 4 mm). The highest angular inter-observer variation was found at the L5-S1 level (mean 2.6°, S.D. 2.3°, max. 11°) while the highest variation in sagittal translatory movement was found at the L4–L5 level (mean 1.4 mm, S.D. 1.2 mm, max. 6 mm). The mean intra-observer variation for L5 olisthesis was 1.0 mm (S.D. 0.9 mm, max. 5 mm) and the corresponding inter-observer variation 1.3 mm (S.D. 1.1 mm, max. 6 mm). Thus, in general, the overall consistency and concordance were good. The variations in intra-observer and inter-observer readings were similar. Variability in readings did not depend on the magnitude or abnormality of the readings. on the other hand, variability may occur because of difficulty in defining the exact landmarks used for the measurements. We point out that in certain cases the diagnostic value of one single reading of an extension-flexion radiography examination may be questionable.

A critical evaluation of real-time ultrasonography for the study of gallbladder volume and contraction

Methods for investigating the physiology of the gallbladder in intact man are generally limited by their inconvenience and hazards, which include repeated exposure of the patient to radiation, nasoduodenal intubation, and complicated or time-consuming chemical techniques. We have carefully evaluated real-time ultrasonography for measuring gallbladder volume and contraction. We determined the validity and accuracy of ultrasound in assessing gallbladder volume and emptying in response to a stimulus. Gallbladder volume, calculated from sonography by a method which assumes gallbladder volume is equal to the sum of a series of cylinders, was 0.95 (SD = 0.01) times the true volume of bile instilled into a human gallbladder in vitro. With a rubber balloon model, containing small volumes (3–17 ml), the true volumes were overestimated by ~17% (0.5–3.1 ml). The intra- and interobserver variation in measurement of gallbladder size was usually within 10%. In 16 subjects we compared the fasting gallbladder volumes and response to a standard liquid meal on 2 successive days. The coefficients of variation in fasting volume, maximum percent emptied after meal-stimulated contraction, residual volume, and rate constant of emptying were 15.0, 12.8, 25.3, and 23.8%, respectively. In 15 consecutive subjects gallbladder volume was determined simultaneously by oral cholecystography and by real-time ultrasonography. Using the sum of cylinders technique, ultrasound volumes were 1.03 (SD = 0.11) times the cholecystogram volumes. We conclude that real-time ultrasonography is a simple, accurate, noninvasive and potentially valuable means of studying gallbladder size and emptying in intact man.