RSS Feeds
-
Use standard diagnostic language
Submit responseThe outcomes “correct upstaging” (i.e. avoiding stage-inappropriate surgery) and “incorrect upstaging” (i.e. precluding potential curative surgery) are arguably of equal importance and must be traded of against each other. The authors’ missed to present their findings in standard diagnostic language, i.e. using sensitivity, specificity (these do not depend on prevalence), predictive values (these depend on prevalence) etc. Moreover, their calculation of the “proportion of patients with correctly upstaged disease” (Table 2) does not take into account the difference in prevalence between the randomly allocated groups (28.7% vs. 36.4%) and, thus, hinders valid comparison. As a matter of fact, conventional staging yielded a much higher odds ratio (measuring association) than PET-CT staging (23.4 vs. 12.8, see Table B). While PET-CT staging is better in predicting curative surgery (i.e. the negative predictive value is higher, 81.6% vs. 68.0%), conventional staging is better in predicting stage-inappropriate surgery (i.e. the positive predictive value is higher, 91.7% vs. 74.2%). Albeit, these differences in predicted values become smaller when the difference in prevalence is accounted for (i.e. 81.6% vs. 75.1% and 88.6% vs. 74.2%, based on a prevalence of 28.7%). Would Maziak et al. please comment on this? Thanks.
Table A: Absolute numbers extracted from study flow diagram (Figure 1, p. 224).
Reference
IIIB, IV
< IIIB
Sum
PET-CT
stagingIIIB, IV
23
8
31
< IIIB
25
111
136
Sum
48
119
167
Conventional
stagingIIIB, IV
11
1
12
< IIIB
48
102
150
Sum
59
103
162
Table B: Characteristics of staging procedures (based on Table A).
PET-CT staging
Conventional staging
Estimate
95% CI*
Estimate
95% CI
Prevalence
0.287 (=48/167)
0.224 to 0.360
0.364 (=59/162)
0.294 to 0.441
Sensitivity
0.479 (=23/48)
0.345 to 0.617
0.186 (=11/59)
0.107 to 0.304
Specificity
0.933 (=111/119)
0.873 to 0.966
0.990 (=102/103)
0.947 to 0.998
Positive predictive value
0.742 (=23/31)
0.568 to 0.863
0.917 (=11/12)
0.646 to 0.985
Negative predictive value
0.816 (=111/136)
0.743 to 0.872
0.680 (=102/150)
0.602 to 0.749
Likelihood ratio, positive
7.128 [=(23/48)/(8/119)]
3.430 to 14.811
19.203 [=(11/59)/(1/103)]
2.542 to 145.049
Likelihood ratio, negative
0.558 [=(25/48)/(111/119)]
0.424 to 0.736
0.822 [=(48/59)/(102/103)]
0.726 to 0.930
Odds ratio
12.765 [=(23*111)/(8*25)]
5.117 to 31.841
23.375 [=(11*102)/(1*48)]
2.933 to 186.302
* Confidence intervals were calculated using the software CIA 2.1.1, Trevor Bryant, © 2000, University of Southampton, UK.
Conflict of Interest:
None declared
-
If the objective of the study had been little different
Submit responseInterpretation of PET images is improved by visual correlation with CT.PET/CT have been found useful in improving the accuracy of staging and restaging of lung cancers.But since Morikawa and colleagues' study have demonstrated that 18F-fluoro-2-deoxy-D-glucose(FDG)PET/CT combined with short-tau-inversion-recovery(STIR)MRI improved diagnostic capability for N -staging of lung cancer,wouldn't it have been better if you have used STIR MRI also as a diagnostic tool in your study design?MRI has a better image resolution than CT scan so why not combine PET with MRI than PET with CT scan?
REFERENCES:
1.Vansteenkiste,Johan F.MD,PhD;Stroobants,Sigrid s.MD,PhD.PET scan in lung cancer:current recommendations and innovation.Journal of Thoracic Oncology:January 2006-vol.1,Issue 1:71-73
2.Morikawa M,Demura Y,Ishizaki T,et al.The Effectiveness of 18 F-FDG PET/CT combined with STIR MRI for diagnosing nodal involvement in the Thorax.J Nucl Med.2009;50:81-87
Conflict of Interest:
None declared