| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1
Department of Radiology, University Hospital Groningen, Hanzeplein 1, 9713 GZ
Groningen, The Netherlands.
2
Department of Internal Medicine, University Hospital Groningen, 9713 GZ
Groningen, The Netherlands.
3
Department of Family Practice, University of Groningen, Antonius Deusinglaan
4, 9713 AW Groningen, The Netherlands.
4
Assessment of Radiological Technology program, Department of Epidemiology
& Biostatistics and Department of Radiology, Erasmus University Medical
Center Rotterdam, P. O. Box 1738, 3000 DR Rotterdam, The Netherlands.
5
Department of Health Policy and Management, Harvard School of Public Health,
718 Huntington Ave., Boston, MA 02115.
Received January 6, 2000;
accepted after revision January 29, 2001.
Supported by the Northern Centre for Health Care Research, University of
Groningen; by the University Hospital Groningen; and with a PIONIER award from
The Netherlands Organization for Scientific Research.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. The sonographic report and follow-up data of 494 patients who had undergone a primary sonographic examination were retrospectively reviewed. Sensitivity and specificity of sonography for the diagnosis of an abdominal malignancy—that is, a primary tumor or metastasis—were determined. Multivariate logistic regression analysis was performed to determine the incremental value of sonography, and a prediction rule was derived.
RESULTS. An abnormality on sonography—that is, a mass, ascites, pleural effusion, hydronephrosis, or focal intraparenchymal heterogeneity suggestive of a mass—had a sensitivity for abdominal malignancy of 86% and a specificity of 94%. In the multivariate analysis, the sonographic findings were found to have significant incremental value (odds ratio = 74) after adjustment for the clinical determinants. In patients younger than 38 years with no previous malignancy, no palpable mass, normal liver function test results, and negative findings on sonographic examination, the risk of an abdominal malignancy was less than 1 in 500.
CONCLUSION. Our results suggest that sonography may be useful in excluding an abdominal malignancy when used in a primary care setting in patients with abdominal complaints who are at low risk for a malignancy.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Excluding an abdominal malignancy requires an extensive workup with CT scanning and MR imaging [4,5,6]. Although such a workup is clearly indicated in patients with a moderate to high suspicion of malignancy, in a setting such as primary care with a low prevalence, a less expensive workup with sonography may be more realistic. A limitation of a sonographic workup is that it will detect tumors of the gastrointestinal tract only if they are large. Although sonography is commonly used in the workup of patients in primary care with abdominal complaints, its diagnostic value in excluding abdominal malignancies is largely unknown.
Our aims were to determine the sensitivity and specificity of upper abdominal sonography in detecting or excluding abdominal malignancy and to evaluate the additional value of abdominal sonography to previously obtained diagnostic information.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The selection procedure resulted in data for 392 consecutive patients referred for sonography by internists and 506 consecutive patients referred by general practitioners. For patients referred by their general practitioner, only limited information was available in the hospital information system. To obtain complete data sets of these patients, we conducted a postal survey of 150 patients randomly selected from the general practitioners' patient records.
A questionnaire concerning further investigations, treatment, hospital referrals, and the final diagnosis was mailed with a letter to the general practitioners of all selected patients (n = 150). The general practitioners filled out the questionnaires using data from their patient records and returned completed questionnaires to the Department of Radiology at the University Hospital Groningen. Nonrespondents received a telephone call reminder after 4 weeks. Of the 150 mailed questionnaires, 142 (95%) were returned. Four of the returned questionnaires contained no data because patients had moved, resulting in a completion rate of 92%.
Complete data sets were available for 392 patients referred by internists and for 138 patients referred by general practitioners. Patients with a known diagnosis at the time of the sonography and those referred for follow-up examinations (n = 36) were excluded. The remaining 494 patients consisted of a subgroup of 130 patients originally referred by general practitioners and a subgroup of 364 patients referred through the internal medicine outpatient department.
Final Diagnosis
The final diagnoses of patients referred by internists were extracted from
discharge letters, and the final diagnoses of patients referred by general
practitioners were taken from their questionnaires. In 112 patients, the final
diagnosis was made at surgery or by using findings of CT, endoscopic
retrograde cholangiopancreaticography, repeated abdominal sonography,
histologic examination, or a combination of these procedures. In the remaining
patients, 1-year clinical follow-up was used as the reference standard. We
assumed that any diagnosis made beyond the 1-year follow-up period concerned
newly developed disease.
The final diagnoses were divided according to organ system [7]. Patients with an abdominal malignancy were distinguished as a separate group.
Categorizing Sonographic Findings
Sonographic examinations were performed and reported by radiologists and by
residents-in-training. Only residents in the second part of their training
period performed sonographic examinations on their own, with help always being
available when needed; otherwise, residents were closely supervised during
every examination.
The presence of abnormalities on the sonographic examination was determined from the report. (Sonographic images were not reviewed retrospectively.) Sonographic abnormalities were divided into four hierarchic nonoverlapping categories. All findings were categorized without knowledge of the follow-up and final diagnoses.
A category 1 abnormality meant that a mass was seen on sonography. Category 2 comprised abnormalities suggestive of malignancy that always required further workup, such as ascites, pleural effusion, hydronephrosis, or focal intraparenchymal heterogeneity suggestive of a mass. Category 3 consisted of abnormalities that required interpretation in relation to the patient's symptoms and that could require therapy, watchful waiting, or no further workup. This category included renal pelvis dilatation or ureter dilatation; urinary retention; liver enlargement (liver length > 15 cm) or spleen enlargement (spleen length > 12 cm); gallbladder abnormalities; common bile duct dilatation with no history of cholecystectomy; gall-, renal, and ureteral stones; fatty liver infiltration; small liver; pain in the pancreatic head induced by pressure from the transducer during scanning; abnormal configuration of the stomach (echogenic area in the stomach wall); aneurysm; and a large inferior vena cava or an inferior vena cava with no response to respiration. Category 4 included sonographic examinations with normal findings and those with abnormalities that generally required no further workup or therapy, such as liver or renal cysts, hemangioma, gallbladder polyps (all protrusions into the gallbladder lumen), common bile duct dilatation in patients with a history of cholecystectomy, calcifications in the liver or spleen, fatty changes of the pancreas, fat in the region of the falciform ligament, thinning of the renal cortex, duplication of the renal collecting system, dromedary humps, horseshoe kidney, ectopic kidney, and tortuosity of the aorta.
A few patients were known to have an abdominal abnormality (e.g., aortic aneurysm, hydronephrosis) but the sonographic examination was performed for other or new complaints. In these cases the known abnormalities were not taken into consideration.
Data Analysis
An abdominal malignancy was defined as a primary tumor or metastasis in any
abdominal organ (liver, spleen, pancreas, kidneys, stomach, intestine, or
lymph nodes). Sensitivity and specificity of sonography in revealing an
abdominal malignancy were determined. Sensitivity was calculated as the
proportion of patients with an abdominal malignancy who had positive test
results and specificity as the proportion of patients without a malignancy who
had negative test results. We established three criteria for considering test
results positive: findings in category 1 only; findings in both categories 1
and 2; and findings in categories 1, 2, and 3.
Stepwise logistic regression analysis was performed to identify clinical determinants and sonographic findings that might be predictive of abdominal malignancy. Determinants that showed univariate associations (p < 0.25) with abdominal malignancy were entered into the multivariate analysis. In accordance with the sequence of the diagnostic workup in clinical practice, we first included all potentially relevant diagnostic determinants obtained from the patient history and physical examination. Stepwise analysis was performed as a backward-stepping procedure based on a likelihood ratio test, with a p value greater than 0.10 for exclusion from the model. Subsequently, the same approach was applied to findings from blood analysis after they were added to the most efficient patient history and physical examination model. Similarly, findings from the sonographic examination were added. This method enabled us to evaluate whether data from the physical examination and subsequent data from the blood analysis and sonography had an incremental value in the diagnosis of an abdominal malignancy compared with the information obtained from the previous step in the workup. Because the number of events in our study was small (n = 22), we recoded the clinical risk factors as one clinical variable to limit the total number of independent variables to a maximum of three. The clinical variable indicated the presence of a previous malignancy, a palpable mass, or abnormal findings on liver function tests. Age was included as a continuous variable.
The goodness of fit of the estimated models was evaluated using the Hosmer-Lemeshow goodness-of-fit test. The area under the receiver operating characteristic curve was determined for each model.
To enable the use of the regression model in clinical practice, a prediction rule was constructed for predicting abdominal malignancy. For the presence of each characteristic in the regression model, a score was calculated on the basis of the regression coefficients. These scores were added into a sum that, through the logistic formula, corresponded with a predicted probability of abdominal malignancy. (Predicted probability of abdominal malignancy = 1 / (1 + e-score), where "score" depends on the model [8]).
The data analysis was performed using the statistical software package SPSS for Windows, version 8.0 (Statistical Package for the Social Sciences, Chicago, IL) and Stata statistical software, version 4.0 (Stata, College Station, TX).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A final diagnosis of physical illness was made in 351 patients (71%). The final diagnoses according to organ system were vascular, gastroesophageal, intestinal, liver or biliary tract, pancreatic, renal or urinary tract, and other diseases (mainly blood disorders and infectious, rheumatic, systemic, and abdominal complaints caused by the side effects of drugs). Twenty-seven patients had malignancies, of which 22 were abdominal.
Table 1 shows the univariate distribution of the clinical characteristics for patients with and without an abdominal malignancy. Most clinical characteristics were indicative of abdominal malignancy, except categories 3 or 4 sonographic findings. In 37 patients a mass (category 1 abnormality) was seen on sonography (Table 1). Seventeen were malignant tumors or metastases, four masses were benign but required workup, nine were benign without any consequences, and seven were false-positive findings. The univariate odds ratio for detecting a malignancy on the basis of visualizing a mass on sonography was 77 (95% confidence interval, 26-229) (Table 1).
|
Nineteen abdominal malignancies detected on sonography were identified as a mass (category 1 abnormality) and/or as ascites, pleural effusion, hydronephrosis, or focal intraparenchymal heterogeneity suggestive of a mass (category 2 abnormalities). In one patient splenomegaly was found as a symptom of non-Hodgkin's disease. In another patient a dilated renal pelvis and three renal cysts were diagnosed that on follow-up were found to be renal cell carcinoma. In one patient who proved to have an abdominal malignancy, no abnormalities were detected on sonography.
In 17 of 22 patients with an abdominal malignancy, a mass was seen on sonography. Defining a mass (category 1 abnormality) as a positive test result yielded a sensitivity for sonography of 77.3% and a specificity of 95.8% (Table 2). Sensitivity increased to 86.4% when a more lenient positivity criterion was used; namely, if category 1 or 2 abnormalities were considered positive test results, with a slight decrease in specificity to 93.6%. The addition of category 3 abnormalities, which were findings nonspecific for malignancy but possible indirect consequences of a malignancy, increased the sensitivity to 95.5% but at the expense of specificity, which decreased to 61.4%.
|
Logistic regression was used to evaluate the association between the variables and the presence of an abdominal malignancy. Table 3 shows the multivariate odds ratios of the relevant variables obtained with stepwise logistic regression analysis.
|
The overall diagnostic model (model II), including all relevant variables, had an area under the receiver operating characteristic curve of 0.91. This model showed that age and the presence of at least one clinical risk factor were important predictors of malignancy. The sonographic examination result (included as a dichotomous variable indicating a category 1 or 2 abnormality) was an independent determinant of the presence of a malignancy (multivariate odds ratio = 74). A reduced model (model I) that included age and the presence of at least one positive clinical risk factor had a decreased area under the receiver operating characteristic curve.
We found no interaction terms with a statistically significant contribution to the models. The Hosmer-Lemeshow goodness-of-fit test was not significant for each model, which suggests that the models were adequately fitted.
Table 4 presents the probability of an abdominal malignancy by age for patients with at least one clinical risk factor versus those with no clinical risk factors depending on whether sonographic information is available and whether sonographic results are negative or positive. As expected, positive sonographic results are associated with a high probability of malignancy, even in the absence of clinical risk factors. For example, a 30-year-old person without clinical risk factors (i.e., no malignancy in the history, no palpable mass, and normal liver function test findings) has a probability (before sonography) of an abdominal malignancy of 0.7%, which will increase to 11% if the sonographic examination has positive findings. In patients without a clinical risk factor, the probability of malignancy will be substantially reduced by negative sonographic findings. For example, a 40-year-old person with no clinical risk factors has a probability of an abdominal malignancy of 1%, which will be reduced to 0.2% after negative sonographic findings. Even in the presence of a clinical predictor, the probability of malignancy will be reduced. For example, a 30-year-old person with at least one clinical risk factor has a probability of abdominal malignancy of 3.5%, which will decrease to less than 1% if the sonographic examination is negative. In patients younger than 38 years with no previous malignancy, no palpable mass, normal liver function tests, and negative sonographic findings, the risk of an abdominal malignancy is less than 1 in 500.
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
As expected, positive sonographic findings are associated with a high probability of malignancy, irrespective of the clinical presentation, and generally will require further workup. Futhermore, our study shows that in the presence of at least one clinical risk factor, the probability of a malignancy is high; and although negative sonographic findings reduce the probability, the probability is still greater than 1% in patients at 50 years and older. More important, the results suggest that in patients with no previous malignancy, no palpable mass, normal findings on liver function tests, and negative sonographic findings, the risk of abdominal malignancy is very low.
This study was a first step in determining the role of sonography in the diagnosis of abdominal malignancy. One could argue that positive sonographic results will always be followed by a more extensive workup to confirm the findings and to plan treatment, that negative sonographic results cannot be trusted and so would be followed by CT, and that sonography thus has no role in the workup. Whereas this argument is true for patients with a moderate to high suspicion of a malignancy, in a setting such as primary care with a low prevalence, a less expensive workup with sonography only may be reasonable. The usefulness of such an approach depends on the probability of a malignancy, the diagnostic value of the test, the loss in effectiveness (and change in cost) associated with missing the diagnosis, and the loss in effectiveness (and change in cost) associated with further workup in someone without a malignancy. Although making the trade-offs involved would require a formal decision and cost-effectiveness analysis, some insight can be obtained by considering the change in probability of abdominal malignancy from before a sonographic examination to after finding negative sonographic results. If we accept that the probability of an abdominal malignancy with negative sonographic findings is low enough to justify performing no further workup, then sonography has been useful in patient management.
Based on a negative history, normal physical examination, and normal laboratory tests, the probability of an abdominal malignancy in our data set was 2.0%. If, in addition, the sonographic findings are negative (i.e., in the absence of a mass, ascites, pleural effusion, hydronephrosis, or focal intraparenchymal heterogeneity suggestive of a mass), the probability of an abdominal malignancy would further decrease to 0.6%. Thus, if decreasing the probability of an abdominal malignancy from 2.0% to 0.6% is considered useful and would avoid further workup, then sonography has a role in excluding this diagnosis in the initial workup of patients with abdominal complaints. If a 0.6% probability is still considered unacceptably high, one can impose an age restriction. For example, in patients younger than 38 years in the absence of a clinical finding, the probability of a malignancy when sonographic results are negative was less than 0.2% (1 in 500).
That the Hosmer-Lemeshow goodness-of-fit test was not significant for each regression model indicates that we developed adequately fitting models. In spite of this favorable result, we must be cautious in interpreting our findings. The main limitation of our study is that it is based on retrospective data of a heterogeneous patient group. This heterogeneity is shown by the widely differing final diagnoses, which vary from psychiatric disorders to malignancies. On the other hand, we evaluated only primary diagnostic sonographic findings and were able to collect data from a large patient group. The group consisted of all consecutive patients referred by internists and a randomly selected group of patients referred by general practitioners. Given the high response rate of the questionnaire survey among general practitioners, we assume that the sample is representative. A second limitation of this study is that the power of our models was limited by the small number of abdominal malignancies. The incidence of abdominal malignancy was low (5%) in our population. Statistically, the probability of detecting variables that are theoretically associated with the presence of an abdominal malignancy becomes higher as prevalence increases. Given the limitations of the data set, the predictive instrument that we developed should be considered exploratory at best. Before the prediction rule is applied in routine practice, it would need to be validated in a prospective trial that ideally would also consider patient reassurance and financial consequences.
In our hospital we evaluate the liver, gallbladder, spleen, pancreas, kidneys, large vessels, and bladder when examining the abdomen with sonography. The stomach and colon are not evaluated in a standard examination. However, large tumors in the gastrointestinal tract may be found incidentally. Clearly, sonography is not the examination of first choice for the gastrointestinal tract, which is a limitation of using sonography as a screening test in patients with nonspecific abdominal complaints to exclude abdominal malignancies. However, tumors of the stomach and colon will generally manifest themselves through specific signs and symptoms, prompting an examination with endoscopy or barium studies.
How does sonography compare with other modalities for detecting abdominal malignancy? We did not find any studies addressing the detection of abdominal malignancies in general, although we did find studies on diagnostic imaging tests for specific organs or tumors. The performances of the various tests differ depending on the organ or tumor involved. For example, renal cancer can be detected with sonography, CT, or MR imaging. Jamis-Dow et al. [9] found that sonography and CT performed equally in detecting small renal masses. The results of Semelka et al. [10] suggest that MR imaging is moderately better than CT for the detection of renal masses. Tanaka et al. [11] examined the accuracy of abdominal sonography for the detection of malignancy. The sensitivity and specificity of sonography were 98.0% and 95.9%, respectively. Brambs and Claussen [12] also concluded that sonography is useful in the detection of pancreatic tumors. Furthermore, Hagspiel et al. [13] found that superparamagnetic iron oxide—enhanced high-field-strength MR imaging is superior to dynamic CT and percutaneous sonography for detecting liver metastases. All these studies except the one by Tanaka et al. were performed in patients with known malignancies.
In conclusion, the results of this study suggest that abdominal sonography may be useful in excluding abdominal malignancy when used in a primary care setting in patients with abdominal complaints who are at a low risk for malignancy. We recommend further validation of our findings in a prospective study.
Acknowledgments
We thank T. Wierstra of the Department of Data Processing and
Computerization (Dienst informatieverwerking en automatisering), University
Hospital Groningen, for making the data available. We are grateful to the
general practitioners for their kind cooperation, and we thank Mereke Gorsira
for help in preparing the manuscript.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
3-cm) renal
masses: detection with CT versus US and pathologic correlation.
Radiology
1996;198:785
-788
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
