 |
 |
|
Home |
Current Issue |
Past Issues |
In the Clinic |
ACP Journal Club |
CME |
Collections |
Audio/Video |
Mobile |
Subscribe |
Tools |
Help |
ACP Online
|
Rapid Responses to:
|
|
Electronic letters published:
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif){kind=icon}
Computed Tomography versus Ultrasonography for Suspected Acute Appendicitis
Summary ROC is useful
Summary ROC is useful
Is CT the right answer for Atypical Appendicitis
|
|
|||
|
Teruhiko Terasawa, MD National Hospital Organization, Nagoya Medical Center, C. Craig Blackmore
Send rapid response to journal:
terasawa{at}nnh.hosp.go.jp Teruhiko Terasawa, et al.
|
We thank Drs. Razzak, Shiga, and Goto for their interest in our paper on imaging for adults and adolescents with suspected appendicitis. We agree with Dr. Razzak that cost-effectiveness is critical to selection of appropriate imaging, as is local availability of imaging technologies. Further, different imaging strategies may be appropriate in different subsets of patients. These areas would benefit from further investigation. We agree with Dr. Shiga that summary receiver operating characteristics (ROC) curves (1) and funnel plots may be useful in meta- analysis. We had performed similar analyses but did not include the results because of several concerns. Summary ROC curves may be affected by the assumptions used in curve calculation and are less useful at bedside than likelihood ratios, which enable understanding of disease probabilities. Asymmetrical funnel plots are affected by both publication bias and other factors that cause heterogeneity, including small study effects (2). In response to Dr. Goto's comments, the global sensitivity analysis (GSA) that he used (3) can apply to studies with partial verification bias, where not all participants receive diagnosis confirmation by the single reference standard, but not to differential verification bias, as in the case of appendicitis. Under the GSA approach, the range of disease prevalence in the unverified patients is applied to define the possible range of sensitivity and specificity for the test under consideration. This is appropriate if nothing is known about the subjects who did not receive the reference standard (surgery). However, in the studies included in our review, most of the negative imaging results were verified by a secondary clinical reference standard (4). This differential verification bias will lead to some overestimation of overall accuracy as we discuss in the paper. However, most subjects with appendicitis should have progressed with symptoms and therefore would have been identified through clinical follow-up. It seems certain that Goto's assumptions of a prevalence of undiagnosed appendicitis of 0.10 to 0.50 in subjects without operation but with no appendicitis identified on clinical follow-up is not clinically sensible. The use of clinical follow-up is a valuable if imperfect reference standard, and is clearly superior to the GSA assumption that nothing is known about these patients. In summary, based on the data, although some differential verification bias is present leading to overestimation of diagnostic accuracy for both tests, CT scanning is a more sensitive and specific imaging modality to detect appendicitis in patients of indeterminate clinical suspicion. Teruhiko Terasawa, MD National Hospital Organization, Nagoya Medical Center, 4-1-1 Sannomaru, Naka-ku, Nagoya, Japan 460-0001 C. Craig Blackmore, MD, MPH Harborview Medical Center, 325 Ninth Avenue, Box 359728, Seattle, WA 98104. References 1. Moses LE, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: data-analytic approaches and some additional considerations. Stat Med. 1993;12:1293-316. PMID: 8210827 2. Sterne JA, Egger M, Smith GD. Systematic reviews in health care: Investigating and dealing with publication and other biases in meta- analysis. BMJ. 2001;323:101-5. Review. PMID: 11451790 3. Kosinski AS, Barnhart HX. A global sensitivity analysis of performance of a medical diagnostic test when verification bias is present. Stat Med. 2003;22:2711-21. PMID: 12939781 4. Whiting P, Rutjes AW, Reitsma JB, Bossuyt PM, Kleijnen J. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol. 2003;3:25. PMID: 14606960 Conflict of Interest:None declared |
|||
|
|
|||
|
Toshiya Shiga, MD, PhD Department of Anesthesia, Nippon Medical School Chiba Hokusoh Hospital, Chiba, JAPAN, Zen'ichiro Wajima, Tetsuo Inoue
Send rapid response to journal:
QZX02115{at}nifty.com Toshiya Shiga, et al.
|
Dear Editor: We congratulate Terasawa et al. for demonstrating the advantage of computed tomography over ultrasonography in the diagnosis of acute appendicitis in adults and adolescents (1). We note that although they plotted the false-positive rate against the true- positive rate on their graph, they did not construct a summary receiver- operating characteristic curve (ROC). The summary ROC curve as described by Moses et al (2), is a powerful tool for depicting diagnostic performance and showing how the diagnostic performance is affected by the test threshold. Briefly, the regression model is expressed as D = a + b*S, where D is the diagnostic log odds ratio and S is the measure of how the test characteristics vary with the test threshold. We could not resist plotting summary ROC curves (Figure not shown here) from the 2x2 contingency tables in the article by Terasawa et al. The results is that D = 5.2 - 0.49*S for computed tomography and D = 3.8 - 0.005*S for ultrasonography. Comparing their original figure with ours, it is apparent that the diagnostic performance of computed tomography is superior to that of ultrasonography in cases of appendicitis. The authors state that in most of studies reviewed, appendiceal diameter greater than 6 mm was used as the positive diagnostic criterion for appendicitis. But, judging from the table, most of the studies seem to have the plural criteria. However, both regression coefficients b are between –1.0 and 1.0 (P > 0.05), suggesting that there is not significant variation in diagnostic performance with threshold. Diagnostic log odds ratio (logit transformation of the ratio of positive to negative likelihood obtained by the Mantel-Haenszel method) of computed tomography is estimated to be 5.0 (95% confidence interval (CI), 1.3-8.7), and that of ultrasonography is 3.4 (95% CI, 0.9-5.9), indicating again that computed tomography is superior to ultrasonography. In addition, we created a funnel plot to explore the publication bias. The funnel plot is asymmetric, indicating that publication bias is likely. It is possible that some studies of small sample populations that indicate low diagnostic accuracy for these tests have not been published. Our additional analysis does not mean the conclusion offered by Terasawa et al. needs alteration; rather, it strengthens their conclusion. We believe that they provided an elegant systematic review. (1) Terasawa T, Blackmore CC, Bent S, Kohlwes RJ. Systematic review: computed tomography and ultrasonography to detect acute appendicitis in adults and adolescents. Ann Intern Med. 2004;141:537-46. (2) Moses LE, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: data-analytic approaches and some additional considerations. Stat Med. 1993;12:1293-316. Conflict of Interest:None declared |
|||
|
|
|||
|
Toshiya Shiga, MD, PhD Department of Anesthesia, Nippon Medical School Chiba Hokusoh Hospital, Chiba, JAPAN, Zen'ichiro Wajima, Tetsuo Inoue
Send rapid response to journal:
QZX02115{at}nifty.com Toshiya Shiga, et al.
|
Dear Editor: We congratulate Terasawa et al. for demonstrating the advantage of computed tomography over ultrasonography in the diagnosis of acute appendicitis in adults and adolescents (1). We note that although they plotted the false-positive rate against the true- positive rate on their graph, they did not construct a summary receiver- operating characteristic curve (ROC). The summary ROC curve as described by Moses et al (2), is a powerful tool for depicting diagnostic performance and showing how the diagnostic performance is affected by the test threshold. Briefly, the regression model is expressed as D = a + b*S, where D is the diagnostic log odds ratio and S is the measure of how the test characteristics vary with the test threshold. We could not resist plotting summary ROC curves (Figure not shown here) from the 2x2 contingency tables in the article by Terasawa et al. Comparing their original figure with ours, it is apparent that the diagnostic performance of computed tomography is superior to that of ultrasonography in cases of appendicitis. The authors state that in most of studies reviewed, appendiceal diameter greater than 6 mm was used as the positive diagnostic criterion for appendicitis. But, judging from the table, most of the studies seem to have the plural criteria. However, both regression coefficients b are between –1.0 and 1.0 (P > 0.05), suggesting that there is not significant variation in diagnostic performance with threshold. Diagnostic log odds ratio (logit transformation of the ratio of positive to negative likelihood obtained by the Mantel-Haenszel method) of computed tomography is estimated to be 5.0 (95% confidence interval (CI), 1.3-8.7), and that of ultrasonography is 3.4 (95% CI, 0.9-5.9), indicating again that computed tomography is superior to ultrasonography. In addition, we created a funnel plot to explore the publication bias. The funnel plot is asymmetric, indicating that publication bias is likely. It is possible that some studies of small sample populations that indicate low diagnostic accuracy for these tests have not been published. Our additional analysis does not mean the conclusion offered by Terasawa et al. needs alteration; rather, it strengthens their conclusion. We believe that they provided an elegant systematic review. (1) Terasawa T, Blackmore CC, Bent S, Kohlwes RJ. Systematic review: computed tomography and ultrasonography to detect acute appendicitis in adults and adolescents. Ann Intern Med. 2004;141:537-46. (2) Moses LE, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: data-analytic approaches and some additional considerations. Stat Med. 1993;12:1293-316. Conflict of Interest:None declared |
|||
|
|
|||
|
Junaid A Razzak, MBBS The Aga Khan university, Karachi, Pakistan
Send rapid response to journal:
junaidrazzak{at}yahoo.com Junaid A Razzak
|
I would like to thank Terasawa et. al. for the systematic review comparing CT scan to Ultrasound (US) in the diagnosis of appendicitis in patients with atypical presentations. The article suggested that overall, CT is a better then US.[1] For many of us it clarifies the clinical decision making, for others practicing in low resource setting this may actually complicate the decision making. CT, in many low income countries, often costs many times the monthly salary of the family. It is important to carefully answer the following questions before we recommend CT as a standard of care: 1- What about the cost and cost-effectiveness? A focal appendiceal CT in our practice setting costs about 6 times more than an US. This focal CT would likely tell the clinician if patient does or does not have appendicitis. When confronted with the broader question: What is the cause of this abdominal pain if it is not appendicitis? the cost differences are even greater. A complete abdominal/pelvic CT with PO/IV contrasts costs about 18 times more than the US. It may thus be less expensive to admit/observe a patient than ordering a CT. The only group that will benefit is a relatively small group who would get operated earlier in CT group versus US group. Those with negative tests would still need admission or a close follow-up since none of the tests have 100% specificity. 2- Is CT a better diagnostic test than US for all patients or is CT better than US in some patients while US is better than CT in others? Current literature needs to better define sensitivities and specificities for sub-group of patients. Thus it would help to know if CT is as good for a 30 years old thin male as it is for 30 years old female or a 60 years old diabetic with non-specific abdominal pain. Another issue is being more specific in describing the degree of “a-typicality” or the level of pretest probability. Literature does not answer the question on how to make a decision if a clinician thinks that a patient may have appendicitis but is not sure; versus when the clinician does not think that the patients has appendicitis and is not sure. Thirdly, as most of the literature originated in high income countries, would the conclusions of this study hold true for relatively leaner population in low and middle income countries. [1]Terasawa T, Blackmore CC, Bent S, Kohlwes RJ. Systematic Review: Computed Tomography and Ultrasonography To Detect Acute Appendicitis in Adults and Adolescents Ann Intern Med 2004; 141: 537-546 Conflict of Interest:None declared |
|||
