Eur Radiol (2011) 21:2604–2617 DOI 10.1007/s00330-011-2221-4

MUSCULOSKELETAL

Diagnosis of bone metastases: a meta-analysis comparing 18FDG PET, CT, MRI and bone scintigraphy Hui-Lin Yang & Tao Liu & Xi-Ming Wang & Yong Xu & Sheng-Ming Deng

Received: 14 December 2010 / Revised: 9 June 2011 / Accepted: 17 June 2011 / Published online: 2 September 2011 # European Society of Radiology 2011

Abstract Objective To perform a meta-analysis to compare 18FDG PET, CT, MRI and bone scintigraphy (BS) for the diagnosis of bone metastases. Methods Databases including MEDLINE and EMBASE were searched for relevant original articles published from January 1995 to January 2010. Software was used to obtain pooled estimates of sensitivity, specificity and summary receiver operating characteristic curves (SROC). Results 67 articles consisting of 145 studies fulfilled all inclusion criteria. On per-patient basis, the pooled sensitivity estimates for PET, CT, MRI and BS were 89.7%, 72.9%, 90.6% and 86.0% respectively. PET=MRI>BS>CT. (“=”indicated no significant difference, P>0.05; “>” indicated significantly higher, P < 0.05). The pooled specificity estimates for PET, CT, MRI and BS were 96.8%, 94.8%, 95.4% and 81.4% respectively. PET=CT=MRI>BS. On H.-L. Yang (*) : T. Liu Department of Orthopaedics, The first affiliated hospital of Soochow University, No188, Shizi Street, Suzhou 215006, People’s Republic of China e-mail: [email protected] X.-M. Wang Department of Radiology, The first affiliated hospital of Soochow University, Suzhou, People’s Republic of China S.-M. Deng Department of Nuclear Medicine, The first affiliated hospital of Soochow University, Suzhou, People’s Republic of China Y. Xu Department of Epidemiology and Biostatistics, Public health school of Soochow University, Suzhou, People’s Republic of China

per-lesion basis, the pooled sensitivity estimates for PET, CT, MRI and BS were 86.9%, 77.1%, 90.4% and 75.1% respectively. PET=MRI>BS>CT. The pooled specificity estimates for PET, CT, MRI and BS were 97.0%, 83.2%, 96.0% and 93.6% respectively. PET>MRI>BS>CT. Conclusion PET and MRI were found to be comparable and both significantly more accurate than CT and BS for the diagnosis of bone metastases. Keywords Bone metastases . PET . MRI . Bone scintigraphy . Meta-analysis

Introduction Bone metastases are the commonest malignant bone lesion. Skeletal involvement occurs in 30%–70% of all cancer patients [1, 2]. Early detection of bone metastases has an important impact on optimal therapy, disease outcome and the quality of life of the patient [3, 4] . Many patients with bone metastases are asymptomatic and metastases are detected incidentally on routine screening or when a cause for rising tumour markers is sought. Symptoms occur mainly when the lesion increases in size, causing extensive bone destruction, which may lead to collapse or fracture, or in the presence of accompanying complications, such as spinal cord compression or nerve root invasion [5, 6]. Imaging is an essential part of the management of bone metastasis [1–4]. There are various morphological and functional imaging techniques for the assessment of malignant bone involvement [1]. The most widely used techniques include 18FDG PET, CT, MRI and bone scintigraphy (BS). They have been proved to be important in the detection of bone metastases. Extensive studies on the diagnostic value of these techniques have been reported,

Eur Radiol (2011) 21:2604–2617

but no consensus has been reached as to the optimal imaging technique for detection of bone metastases. It is because a wide variation in patient populations, imaging techniques, study designs and results exist in these studies, which make it difficult for workers in this field to know the exact diagnostic value of these imaging techniques. A metaanalysis of diagnostic tests represents a powerful tool for summarising findings in the literature by taking into account and enabling analysis of differences between studies [7, 8]. Thus, the purpose of our study is to perform a meta-analysis to compare the value of 18FDG PET, CT, MRI and BS for the diagnosis of bone metastases, and then to find out which is the best diagnostic technique for bone metastases.

Materials and methods

2605

and/or radiographic confirmation by multiple imaging techniques were used as the reference standard. (c) For perpatient or per-lesion statistics, sufficient data were presented to obtain true-positive, true-negative, false-positive and false-negative results of the imaging techniques compared with the reference standard; (d) When data or subsets of data were presented in more than one article, the article with the most details or the most recent article was chosen. The exclusion criteria were as follows: (a) Case reports, letters, editorial, comments, reviews, animal, or in vitro studies and the studies that did not include raw data; (b) The studies were not about 18FDG PET and 99mTc-MDP BS imaging, but about other radio-pharmacy imaging such as 18 F PET, 18FCH PET, 99mTc-MIBI BS and 99mTcO4- BS etc.; (c) The studies in which the results of different imaging techniques were presented in combination and could not be differentiated with regard to performance assessment of tests on an individual technique.

Literature search Data extraction A comprehensive computer literature search [9] of abstracts about studies in human subjects was performed to identify articles about the diagnostic performance of 18FDG PET, CT, MRI or 99mTc-MDP Bone Scintigraphy (BS) imaging for the detection of bone metastases. The MEDLINE, EMBASE, Scopus, Sciencedirect, Springerlink, Web of Knowledge, EBSCO and the Cochrane Database of Systematic Review, from January 1995 to January 2010, were used with the following keywords: (“PET OR positron emission tomography OR FDG OR fluorodeoxyglucose OR CT OR computed tomography OR MRI OR magnetic resonance imaging OR MDP OR bone scan OR bone scintigraphy OR ECT OR emission computed tomography”) AND (“bony metastases OR skeletal metastases OR osseous metastases OR bone metastases”) AND (sensitivity OR specificity OR false negative OR false positive OR diagnosis OR detection OR accuracy). The list of articles was supplemented with extensive cross-checking of the reference lists of all retrieved articles. No language limitation was applied. Selection of studies Two reviewers (L.T, Y.HL) independently assessed potentially eligible studies. After reading all the abstracts, we found the potentially eligible articles and any citations for which a determination could not be made from the abstract. Then we managed to get the full-text of these articles to determine whether they were eligible. The inclusion criteria were as follows: (a) 18FDG PET (including 18FDG PET and PET/CT), CT, MRI or 99mTc-MDP BS imaging was used to identify and characterise bone metastases. (b) Histopathological analysis and/or close clinical and imaging follow-up

The same observers independently extracted relevant data (including study design characteristics and examination results) from each article by using a standardised form. Observers were not blinded with regard to information about the journal name, the authors, the authors’ affiliation or year of publication, as this has been shown to be unnecessary [10]. To resolve disagreement between reviewers, a third reviewer(X, Y)assessed all discrepant items, and the majority opinion was used for analysis. Study design characteristic The QUADAS quality assessment tool was used to extract the relevant study design characteristics of each study. This tool fully described by Penny Whiting is the first systematically developed, evidence-based quality assessment tool to be used in the systematic review of diagnostic accuracy studies [11]. Common characteristics The following common characteristics were recorded: (a) Year of publication. (b) Sample size. (c) Description of study population, which included age and male–female distribution. (d) Tests of reference standard (Histopathological analysis or close clinical and imaging follow-up or radiographic confirmation by multiple imaging techniques). (e) Type of primary tumour. (f) Authors’ country. Imaging features The following imaging features were extracted: For 18FDG PET, these features included system type (PET or PET/CT) and type of analysis (qualitative or quantitative or both). For CT, these features included type of device (section, helical), section thickness and use or not of contrast agent. For MRI, these features included

2606

Eur Radiol (2011) 21:2604–2617

magnetic field strength, use or not of contrast agent, use or not of STIR sequences, and use or not of DWI sequences. For 99mTc-MDP BS, these features included use or not of SPECT (single photon emission computed tomography).

backwards stepwise algorithm, to identify only the most important characteristics. The variate with the highest P value was excluded first. Characteristics were retained in the regression model if P<0.1.

Examination results 2×2 tables including the numbers of true positive, true negative, false positive and false negative results, were extracted both on a per-patient basis and on a per-lesion basis. To avoid selection bias of data sets, all tabulated results for multiple PET, CT, MRI and BS systems and/or techniques were counted as separate data sets.

Subgroup analysis

Statistical analysis Data were separately analysed for PET, CT, MRI and BS. First we calculated pooled sensitivity, specificity, DOR and SROC for each technique. The DerSimonian Laird method (bivariate random effects model) was used for the primary meta-analysis to obtain a summary estimate for sensitivity, specificity and DOR with 95% confidence intervals (CI). The SROC curve is placed over the points to form a smoothed curve which can be achieved using a regression model proposed by Moses et al. [12]. Then we did Z test to find whether the sensitivity, specificity and DOR and of each technique were significantly different from the others. If P<0.05, the result was considered to be statistically significant. All analyses were performed by using Microsoft Excel 2003 (Microsoft, Seattle, WA, USA), SPSS 13.0 for Windows (SPSS, Chicago, IL, USA), and Meta-DiSc [13]. Meta-DiSc, produced by javier.zamora, is freeware software for performing meta-analysis of studies of evaluations of diagnostic and screening tests [13]. Meta-regression analysis To determine whether the results were significantly affected by heterogeneity between individual studies, we performed a meta-regression analysis. We used logit-transformed sensitivity and specificity ln {p/(1 _p)} as the dependant variates, where ln is the natural logarithm. DOR was transformed to In (DOR). Because of the transformation, these values were approximately normally distributed. First, we carried out single factor regression analysis. The following variates were included: year of publication, sample size, type of reference stand, type of primary tumour, authors’ country, and the answers to the 14 questions of the QUADAS quality assessment tool (“yes” vs “no” and “unclear” responses). We considered variates to be explanatory if their regression coefficients were statistically significant (P<0.05). Subsequently, we developed a multivariate regression model with which we used a

We performed a subgroup analysis of the technical differences of each technique. For PET, we compared the following techniques: (a) type of system (PET VS PET/ CT). (b) type of analysis (qualitative analysis VS both qualitative and quantitative analysis). For MRI, we compared the following techniques: (a) axial MRI VS whole body MRI. (b) unenhanced MR imaging VS enhanced MR imaging. (c) using DWI sequences VS not using DWI sequences. For BS, we compared the following techniques: (a) using SPECT VS not using SPECT. No enough datasets were available to perform subgroup analyses for CT.

Results Literature search and selection of studies After the computerised search was performed and reference lists were extensively cross-checked, about 4785 abstracts were identified. We found 148 articles that were potentially eligible after reading all the abstracts. After we read the full texts of these articles, 81 of the 148 relevant articles were excluded because (a) the aim of the articles was not to reveal the diagnostic value of 18FDG PET, CT, MRI or 99mTc-MDP BS for identification and characterisation of bone metastases (n=26); (b) researchers in the articles did not report the reference standard clearly or the reference standard did not meet the inclusion criteria of this meta-analysis (n=6); (c) researchers in the articles did not report data that could be used to construct or calculate true positive, false positive, true negative, and false negative results (n=22); (d) researchers in the articles presented results from a combination of different imaging techniques that could not be differentiated for assessment of single tests (n=13); (e) the studies were not about 18FDG PET and 99mTc-MDP BS, but concerned to other forms of radiopharmacy imaging such as 18 F PET, 18 FCH PET, 99mTc-MIBI BS and 99mTcO4- BS etc. (n=12); (f) the data or subsets of data were also presented in the another article (n=2). At last, 67 articles consisting of 145 studies fulfilled all inclusion criteria and were selected for data extraction and data analysis. Study design characteristics Most studies (Table 1) had optimal design for most of the questions except the examination with the same reference

Eur Radiol (2011) 21:2604–2617

2607

Table 1 Results of the distribution of study design characteristics in 145 studies Question about Study Design Characteristic

Response Yes

No

1 2 3 4 5

Was the spectrum of patients representative of the patients who received the test in practice? Were selection criteria clearly described? Is the reference standard likely to help correctly classify the target condition? Is the time between performance of reference standard and index test short enough? Did the whole sample or a random selection of the sample receive verification by using a reference standard?

126 120 126 141 140

19 25 19 4 5

6 7 8 9 10 11 12 13 14

Did patients undergo examination with the same reference standard regardless of the index test result? Was the reference standard performed independently of the index test? Was the execution of the index test described in sufficient detail to permit replication of the test? Was the execution of the reference standard described in sufficient detail to permit replication of the test? Were the index test results interpreted without knowledge of the results of the reference standard? Was the reference standard interpreted without knowledge of the results of the index test results? Were the same clinical data available when test results were interpreted as would be available in practice? Were uninterpretable/intermediate test results reported? Were withdrawals from the study explained?

33 141 139 69 142 8 143 137 143

112 4 6 76 3 137 2 8 2

Data were the numbers of responses from the QUADAS tool. The numbers indicated how many articles were assigned a score of “Yes” or “No.” The responses of “No” and “Unclear” were summarised as “No”

standard (77.2% for “no” responses to question 6), the description of the execution of the reference standard (52.4% for “no” responses to question 9) and the interpretation of the reference standard results without knowledge of the index test results (94.5% for “no” responses to question 11). However, as the choice for a treatment strategy strongly depends on the outcome of the diagnosis, it is hard to perform an ideal study. The description of the execution of the reference standard remains a problem in the studies of diagnostic performance of various techniques. What is more, as there was often more than one lesion for bone metastases, and maybe the lesion had a deep-seated location, it was difficult to achieve histopathological analysis for all the lesions both for moral reasons and because of technical difficulties. It was ineluctable that ‘clinical and imaging follow-up and/or radiographic confirmation by multiple imaging techniques’ were also used as the reference standard. Common characteristics There were total of 15,221 patients in the studies selected and the ages ranged from 10 to 91 years old. In 106 studies, the sex distribution was described: 5905 patients were male and 4427 patients were female. In 96 studies, imaging data were presented about the identification of patients. In the other 49 studies, imaging data were presented about the identification of lesions. The published year ranged from 1997 to 2010. The reference standard was histopathological analysis in 1 study, clinical and imaging follow-up in 24

studies, radiographic confirmation by multiple imaging techniques in 6 studies, both histopathological analysis and clinical and imaging follow-up in 99 studies, both clinical and imaging follow-up and radiographic confirmation by multiple imaging techniques in 8 studies, and all three kinds in the other 7 studies. The primary tumour was lung cancer in 36 studies, breast cancer in 25 studies, prostate cancer in 9 studies, malignant lymphoma in 4 studies, thyroid cancer in 4 studies, renal cell carcinomas in 4 studies, malignant melanoma in 4 studies, cervical cancer in 3 studies, oesophageal cancer in 3 studies, nasopharyngeal cancer in 2 studies, gastrinomas in 2 studies, malignancies of the upper aerodigestive tract in 2 studies, endocrine gastroenteropancreatic tumours in 1 study, neuroendocrine tumour in 1 study, hepatoma in 1 study and mixed tumours in 44 studies. The authors’ country was the USA in 13 studies, Germany in 30 studies, Japan in 26 studies, China in 22 studies, South Korea in 10 studies, Chinese Taiwan in 10 studies, United Kingdom in 8 studies, Israel in 6 studies, Switzerland in 5 studies, Belgium in 4 studies, France in 2 studies, Netherlands in 2 studies, Sweden in 2 studies, Australia in 1 study, Italy in 1 study, Malaysia in 1 study and Turkey in 1 study. Table 2 presents the data sets included with the corresponding numbers of datasets and patients and reference numbers. A full list of all the articles included with all relevant study characteristics and complete examination results (also including the standardised form for extracting data) is available on request from the authors of this article.

2608

Eur Radiol (2011) 21:2604–2617

Table 2 Study characteristics of the data sets included for each kind of imaging technique

Per-patient

Per-lesion

Technique concerned

Number of data sets

Number of patient (lesions)

Reference Numbers

PET CT MRI BS PET CT MRI BS

26 7 15 48 18 3 7 21

4367 723 1032 4638 1033(6779) 421(983) 358(2874) 1039(6083)

15, 17, 20, 27, 29, 35, 38, 41, 44–46, 48, 52, 55, 56, 65, 67–69, 71, 73, 76–78 29, 50, 54, 66, 76, 77, 79 16, 22, 23, 26, 31, 42, 43, 62, 63, 69, 70, 74, 77, 78 15–22, 24–28, 31, 34–36, 39–42, 44–46, 48–53, 55–57, 63–67, 69–74 17, 30, 32, 33, 37, 44, 47, 48, 58–61, 67, 69, 75, 76 60, 61, 76 14, 22, 59, 61, 63, 69 14, 21, 22, 30, 32, 33, 37, 44, 47, 48, 53, 57, 58, 63, 67, 69, 75, 80

Imaging features For PET (per-patient and per-lesion studies were summarised together), 13 studies used PET/CT, and the other 31 studies used PET. With regard to the analysis method, 22 used qualitative, 13 used both quantitative and qualitative analysis, and in the other 9 it was not clear. For CT, 3 studies used sixteen-section helical, 2 eight-section helical l, 1 four-section helical, 1 dual-section helical and in 3 it was not clear. For most studies, the section thickness was 5 mm (5 studies), except for 1 study (2 mm), 1 study (3 mm), 1 study both 5 mm and 10 mm, and in 2 it was not clear. Most studies used contrast agent, but 3 did not. For MRI, 7 studies used axial skeleton MRI, the other 15 studies used whole body MRI. 15 studies used 1.5 T, 2 used 3.0 T, 1 used 0.5 T, 1 used 0.2, 1.0 and 1.5 T, 1 used both 1 and 1.5 T and the rest 2 did not state clearly what was used. 9 studies used contrast agent, 9 studies did not use contrast agent, 2 studies partly used contrast agent and in 2 studies it was not clear. Most studies used STIR sequences, except for 4 studies. 4 studies used DWI sequences, and the other 18 studies did not. For BS, 15 studies used SPECT and the other 54 studies did not. The amount of trace was <740 MBq in 21 studies (ranging from 450 MBq to 700 MBq), ≥ 740 MBq in 47 studies (ranged from 740 MBq to 925 MBq) and the other 1 did not state the amount clearly. Summary estimates of sensitivity, specificity, DOR and SROC curves on a per-patient basis The pooled sensitivity estimates for PET, CT, MRI and BS were 89.7%, 72.9%, 90.6% and 86.0% respectively. The sensitivity estimates for PET and MRI were significantly higher than those for BS and CT (P<0.05). They were similar for PET and MR (P>0.05), but significantly higher for BS than for CT (P<0.05). In total, PET=MRI>BS>CT. (“=” indicated no significant difference, P>0.05; “>” indicated significantly higher, P<0.05). The pooled specificity esti-

mates for PET, CT, MRI and BS were 96.8%, 94.8%, 95.4% and 81.4% respectively. For specificity, PET=CT=MRI>BS. The pooled DOR estimates for PET, CT, MRI and BS were 232.6, 76.0, 143.9 and 29.4 respectively. For DOR, PET>MRI=CT>BS. The results are shown in Table 3. The SROC curves for FDG-PET, CT, MRI and BS on a per-patient basis are shown in Fig. 1. The SROC curves for MRI and PET showed better diagnostic accuracy than those for CT and BS. Summary estimates of sensitivity, specificity, DOR and SROC curves on a per-lesion basis The pooled sensitivity estimates for PET, CT, MRI and BS were 86.9%, 77.1%, 90.4% and 75.1% respectively. For sensitivity, MRI>PET>BS=CT. The pooled specificity estimates for PET, CT, MRI and BS were 97.0%, 83.2%, 96.0% and 93.6% respectively. For specificity, PET>MRI>BS>CT. The pooled DOR estimates for PET, CT, MRI and BS were 240.3, 6.0, 167.6 and 25.2 respectively. For DOR, PET=MRI>BS>CT. The results are shown in Table 4. The SROC curves for FDG-PET, CT, MRI and BS on a per-lesion basis are shown in Fig. 2. The SROC curves for MRI and PET showed better diagnostic accuracy than those for CT and BS. Meta-regression analysis Following the backwards stepwise regression analysis, several variables were identified as significant predictors of the diagnostic performance of each technique for the assessment of bone metastasis both on a per-patient basis and a per-lesion basis. The detailed results are shown in Table 5. Subgroup analysis The results of the subgroup analysis for the PET, CT, MRI and BS imaging techniques are presented in Table 3 on a per-

Eur Radiol (2011) 21:2604–2617

2609

Table 3 Summary estimates of sensitivity, specificity and Diagnostic Odds Ratio (DOR) for PET, CT, MRI and BS and the results of the subgroup analysis of technique difference on per-patient basis Technique and group PET Overall PET PET-CT Qualitative analysis both qualitative and quantitative analysis CT Overall MRI Overall Axial MRI Whole–body MRI Unenhanced MRI enhanced MRI Using DWI sequences Not using DWI sequences BS Overall using SPECT Not using SPECT a

Number of data sets

Sensitivity (%)

Specificity (%)

DOR

26 19 7 15 7

89.7(87.4–91.6) a 87.1(83.8–89.8) 93.7(90.5–96.0) c 89.0(85.7–91.7) 89.1(85.2–92.4)

96.8(96.2–97.3) a 96.4(95.5–97.1) 97.4(96.4–98.1) 97.5(96.7–98.1) 96.7(95.6–97.6)

232.6(113.5–476.9) a 169.3(76.6–374.3) 486.9(191.2–1240.2) c 224.9 (75.6–753.9) 218.9 (62.0–772.9)

7

72.9(66.6–78.6) b

94.8(92.4–96.6) a

76.0 (23.4–246.4)

15 5 10 6

90.6(86.7–93.7) a 98.3(94.1–99.8) c 85.4(79.3–90.2) 89.9(82.2–95.0)

95.4(93.6–96.8) a 95.4(89.5–98.5) 95.4(93.4–96.9) 98.4(95.4–99.7)

143.9(59.4–348.7) 336.2(78.1–1447.1) c 109.6(38.7–310.0) 172.2(51.3–578.2) c

9 3 12

91.0(86.1–94.6) 95.2(89.1–98.4) 88.1(82.7–92.3)

94.3(92.0–96.1) 89.3(82.5–94.2) 96.6(94.8–97.9)

142.6(41.2–493.3) 158.3(39.2–638.5) 147.9(50.9–429.5)

48 8 40

86.0(84.0–87.8) 82.6(76.2–87.8) 86.5(84.4–88.4)

c

c

81.4(80.0–82.8) b 92.8(89.6–95.2) c 79.9(78.4–81.4)

29.4(19.5–44.2) b 60.8(20.9–177.2) c 25.6(16.4–39.8)

the highest sensitivity, specificity or DOR among PET, CT, MRI and BS

b

the lowest sensitivity, specificity or DOR among PET, CT, MRI and BS

c

significantly higher compared with the opposite subgroup

patient basis and in Table 4 on a per-lesion basis. For PET, the sensitivity and diagnostic OR for PET-CT were significantly higher than those using PET (P<0.05) both on a per-patient basis and a per-lesion basis, while the specificity showed no statistical difference (P>0.05). The sensitivity for using both qualitative and quantitative analysis was significantly higher than for using qualitative analysis with lower specificity (P< 0.05) on a per-lesion basis, while the DOR showed no statistical difference (P>0.05). There were no significant differences found on a per-patient basis. For MRI, the sensitivity and DOR for axial MRI were significantly higher than for whole-body MRI (P<0.05) on a per-patient basis. There were no significant differences found for unenhanced MR imaging and enhanced MR imaging except that the sensitivity for unenhanced MR imaging was significantly higher than that for enhanced MR on a per-lesion basis (P< 0.05).The sensitivity for using DWI sequences was significantly higher than that for not using DWI sequences with lower specificity on a per-patient basis (P<0.05), while the DOR showed no statistical difference (P>0.05). For BS, the specificity and DOR for using SPECT were significantly higher than those for not using SPECT both on a per-patient basis and a per-lesion basis (P<0.05), while the sensitivity showed no statistical difference (P>0.05).

Discussion In this meta-analysis, we obtained summary estimates and summary ROC curves for the diagnostic accuracy of 18 FDG-PET, CT, MRI and BS in the detection of patients with bone metastases. Both on a per-patient basis and a perlesion basis, PET and MRI were found to be comparable and significantly more accurate than CT and BS. Data on subgroup analyses indicated that PET/CT was better than PET, and that the type of analysis for PET imaging would affect the diagnostic result on a per-lesion basis. The sensitivity and DOR for axial MRI were significantly higher than for whole-body MRI on a perpatient basis. There were no significant differences between unenhanced MR imaging and enhanced MR imaging. Using DWI sequences would affect the diagnostic result of MRI on a per-patient basis. For BS, using SPECT was better than not using SPECT. To do a meta-analysis, we needed to search all the qualifying articles published worldwide to avoid selection bias. Thus, in this study, no language limitation was used, and we searched many databases for relevant articles. To minimise bias in the selection of studies and in data extraction, reviewers independently selected articles on the

2610

Eur Radiol (2011) 21:2604–2617

Fig. 1 SROC curves for PET, CT, MRI and BS on a per-patient basis. The middle blue line was the SROC curve and the two others were 95% confidence intervals. The SROC curves for MRI and PET showed better diagnostic accuracy than those for CT and BS

basis of the inclusion criteria, and scores were assigned to study design characteristics and examination results by using a standardised form. It was shown that studies of the diagnostic performance of techniques with methodological shortcomings may cause overestimation of the accuracy of a diagnostic test [81]. Therefore, we evaluated the effect of these characteristics on diagnostic performance by metaregression analysis. The advantage of this regression analysis is that this model accounts not only for the heterogeneity of studies due to different threshold settings but also for the error of estimation of the sensitivity, specificity and DOR values in each study. This random

model also accounts for the residual heterogeneity that may remain even after adjusting for study characteristics and imaging techniques [82]. Publication bias is a potential limitation of any metaanalysis. A meta-analysis conducted by the Cochrane Collaboration showed that a large number of methods have been developed, and when the methods are compared with one another, results can provide different estimates in terms of direction and magnitude of publication bias [83]. In addition, studies about publication bias focus mostly on randomised trials, and these studies are registered; the registration of studies about diagnostic studies is either limited or difficult to achieve. We attempted to examine

Eur Radiol (2011) 21:2604–2617

2611

Table 4 Summary estimates of sensitivity, specificity and Diagnostic Odds Ratio (DOR) for PET, CT, MRI and BS and the results of the subgroup analysis of technique difference on per-lesion basis Technique and group PET Overall PET PET-CT qualitative analysis both qualitative and quantitative analysis CT Overall MRI Overall Unenhanced MRI enhanced MRI BS Overall using SPECT Not using SPECT a

Number of data sets

Sensitivity (%)

Specificity (%)

DOR

18 12 6 7 6

86.9(85.3–88.3) a 80.1(77.7–82.5) 94.2(92.6–95.6) c 66.0(61.6–70.3) 94.0(92.6–95.3) c

97.0(96.5–97.5) a 96.9(96.2–97.5) 97.2(96.4–97.8) 99.6(99.1–99.8) c 95.7(94.7–96.5)

240.3(91.6–630.2) a 175.2(68.7–446.9) 348.1(39.0–3109.4) c 321.8(117.9–877.8) 235.2(37.4–1477.8)

3

77.1(72.9–81.0) b

83.2(79.7–86.2) b

6.0 (0.599–60.7) b

7 3 3

90.4(87.2–93.0) a 93.1(89.2–95.9) c 83.8(77.0–89.2)

96.0(95.2–96.8) 95.9(93.2–97.4) 96.3(95.3–97.1)

167.6(70.8–396.8) a 219.9(37.0–1038.4) c 158.1(41.0–609.8)

21 7 14

75.1(72.9–77.2) 76.8(72.5–80.8) 74.5(71.8–77.0)

93.6(92.8–94.2) 96.3(95.3–97.2) 92.1(91.1–93.0)

c

25.2(10.6–60.3) 65.6(14.7–293.1) 15.6(5.1–47.5)

c

the highest sensitivity, specificity or DOR among PET, CT, MRI and BS

b

the lowest sensitivity, specificity or DOR among PET, CT, MRI and BS

c

significantly higher compared with the opposite subgroup (P<0.05)

publication bias by using an evaluation of whether the size of studies was associated with the results of diagnostic accuracy. In particular, small studies with optimistic results may be published more easily than small studies with unfavourable results. Larger studies with optimistic results may also be published more easily than larger studies with unfavourable results, but this difference is usually smaller. The evaluation of the effect of sample size did not show a better diagnostic performance of smaller studies compared with larger studies in all data sets, except for specificity for PET on a per-lesion basis, while, the coefficient of the size was very small (0.003) with a P value of 0.08. Funnel plot analysis was not performed, as the limited number of data points for some data sets could have decreased the power of detecting publication bias. The reference standard used in this meta-analysis was ‘Histopathological analysis and/or close clinical and imaging follow-up and/or radiographic confirmation by multiple imaging techniques’. Several limitations are present in any meta-analysis study of a diagnostic test. First, there is no accepted gold standard, which is a common barrier to all studies assessing various tumours and different imaging procedures for diagnostic accuracy in the detection of distant metastases. Most of the studies relied on a clinical course and other imaging techniques to confirm the presence of metastatic disease with a subset of the patients undergoing biopsy. Clearly, histopathological correlation could not be obtained for every

lesion in a patient for ethical reasons. We attempted to reveal whether the kind of reference standard influenced the diagnostic accuracy. The result was that no significant differences could be observed except for the sensitivity of PET on a per-patient basis. There were also a number of drawbacks, which perhaps caused incorrect results. As stated in the Standards for Reporting of Diagnostic Accuracy initiative, a reference standard can be either a single method or a combination of methods to establish the presence of the target condition [84]. The major problem, however, was the absence of critical information, such as data on the execution of the reference test, the confidence rating, and these data were insufficiently described or not described in a large subset of articles. Therefore, the Standards for the Reporting of Diagnostic Accuracy initiative was developed to improve the quality of the reporting of diagnostic studies. Another limitation is the consideration of 2×2 tables for multiple CT and MRI techniques as separate data sets. This has been performed to avoid selection bias. We are aware of the dependency in datasets from the same patient population. Analysis of this dependency is not possible with our software, as the random-effects linear regression model with mixed-effects approach is able to adjust for this potential dependency only if the same numbers of data sets in each study are available. We examined this correlation by using the empirical standard error calculated by using the robust variance estimator, also known as the “sand-

2612

Eur Radiol (2011) 21:2604–2617

Fig. 2 SROC curves for PET, CT, MRI and BS on a per-lesion basis. The middle blue line was the SROC curve and the two others were 95% confidence intervals. The SROC curves for MRI and PET showed better diagnostic accuracy than those for CT and BS

wich”estimator, which was described by Fay et al.[85] and also used by Bipat et al.[86]. A further possible limitation of this meta-analysis is that a multiple backward stepwise regression analysis was performed with 19 covariates, and the final model was adjusted for significant variables. In our study, more data sets were available for PET and BS than for CT and MRI. Less significant predictors were found for CT and MRI; this can be explained by the smaller number of data sets for CT and MRI. By adjusting the CT and MRI models with fewer variables, the accuracy of CT and MR imaging could be underestimated. Moreover, because of the few CT imaging studies suitable for subgroup analysis, the sensitivity,

specificity and DOR values may overlap completely or partially. A final limitation of this study was the diversity of tumour types. There is an important bias in detection of bone metastases with 18FDG PET, as the type of primary tumor affects the sensitivity of the tracer. Firstly although 18 FDG PET has been reported as being appropriate for detecting all types of bone metastases—including lytic, sclerotic, and mixed lesions—accumulating data suggest that 18FDG PET is more sensitive in detecting lytic metastases than sclerotic metastases. The latter type of metastases may show uptake of lower intensity compared with lytic lesions or even no increased uptake at all.

Eur Radiol (2011) 21:2604–2617

2613

Table 5 Predictors identified with backward regression analysis for each imaging technique Modality

PET

CT

MRI

BS

Parameter

Per-patient

Per-lesion

Covariates

Coefficienta

P Value

Covariates

Coefficienta

P Value

Sensitivity

Question3

1.55 (0.78)

0.058

Specificity

/

/

/

DOR

−1.32 (0.73) 1.92 (1.08) / / 2.62 (1.04) / 0.17 (0.09)

0.083 0.080 / / 0.053 / 0.086

2.26 (0.60) 1.08 (0.48) −2.80 (0.73) 0.003 (0.001) /

<0.01 0.04 <0.01 0.08 /

Sensitivity Specificity DOR Sensitivity Specificity

Question1 Question5 / / Question9 / Primary tumour

Question 6 Reference stand Question 6 Size / / / / / Question 11

/ / / / −2.93 (1.21)

/ / / / 0.01

DOR Sensitivity Specificity DOR

/ / Authors’ country /

/ / −0.08 (0.05) /

/ / 0.10 /

Question 11 Question 9 Authors’ country Authors’ country

−3.84 −1.97 −0.34 −0.28

0.06 <0.01 <0.01 <0.01

(0.90) (0.46) (0.06) (0.06)

/ no significant predictors were found a

numbers in parentheses are standard error. A positive regression coefficient indicates better discriminatory power of the technique in studies with that characteristic compared with that in studies without the corresponding characteristic, and a negative regression coefficient indicates reduced diagnostic performance in studies with that characteristic

Secondly, if the primary tumor is not 18FDG avid, 18FDG PET is not considered a suitable investigation for staging. In these cases, failure to detect bone metastases may be unrelated to their localization in bone or to their sclerotic nature but to reflect the non–18FDG avidity of the individual tumor [1]. It is ideal to analyse different types of primary tumour separately, however, as bone metastases from different types of primary tumour have lots of common imaging features, and because 44 of the studies enrolled (30.3%) analysed the bone metastases without differentiating among the types of primary tumours, we analysed them together. 18 FDG PET, CT, MRI and 99mTc-MDP BS are widely used for diagnosis of bone metastases. CT and MRI are anatomical imaging techniques that analyse tumour tissues on the basis of their morphological appearance while 18FDG PET and 99m Tc-MDP BS are functional imaging techniques.18FDG PET identifies viable tumours on the basis of higher glycolytic rates in neoplasms than in normal tissue. 18FDG PET is better in lytic than blastic bone metastases, in comparison with BS which is better in blastic bone metaseses. BS identifies bone metastasis by detecting the osteoblastic response to bone destruction by tumour cells and the accompanying increase in blood flow [87, 88]. In practice, we frequently see changes in glucose metabolism before a physical change or symptoms occur. Therefore, PET may have great potential advantage in detecting metastasis. In the present study, 18FDG-PET was shown to be a better

imaging technique than CT and BS. However, PET also has limitations. False-positive results will exist for PET because FDG accumulates in metabolically active tissue, including inflammation and infection, and some normal high FDG uptake tissue such as some muscles will also possibly lead to false-positive results. Another limitation for PET is that 18 FDG might be less sensitive for the detection of osteoblastic metastases, although it is more sensitive for the detection of osteolytic metastases [89]. Furthermore, PET lacks anatomical details, which hinders the localisation and characterisation of increased 18FDG uptake. PET/CT can add important anatomical information to that obtained by PET. The use of hybrid systems with CT can overcome the limitation of PET, as the nature of many benign lesions can be determined by the CT part of the study. Many studies have shown that PET/CT is better than PET alone and CT alone [67–69, 71]. This is also demonstrated in this meta-analysis. The advantage of CT is its good anatomical resolution, soft-tissue contrast and detailed morphology. Both cortical and trabecular bone components are well defined. Because considerable cortical destruction is required for visualisation of a metastasis by CT, the sensitivity of this technique in detecting early malignant bone involvement is relatively low [90]. Moreover, cortical destruction may be especially difficult to determine in the presence of severe osteoporotic or degenerative changes [91]. CT is not sensitive for the assessment of malignant marrow infiltration, although the presence of the latter may occasionally be

2614

suggested because marrow infiltrated by tumour cells is more strongly attenuated compared with normal marrow [92].In this meta-analysis, we found CT had the lowest sensitivity. Magnetic resonance imaging has been found to have an advantage over CT in many indications based on its higher soft tissue contrast. Moreover, MRI has good spatial and contrast resolution. It is the optimal imaging technique for bone marrow assessment. MRI can detect an early intramedullary malignant lesion before there is any cortical destruction or reactive processes. Normal marrow shows a high-intensity signal on T1-weighted imaging, whereas metastases appear as areas of reduced signal, reflecting the replacement of fat in the marrow with tumour. Bone marrow metastases have longer T1- and T2-weighted relaxation times than normal marrow and are usually enhanced after the administration of contrast medium. In the present study, MRI proved to be a comparable imaging technique to PET, and better than CT and BS. However, MRI also has limitations. MRI is less sensitive than CT for detecting cortical bone destruction because cortical bone appears black on T1- and T2-weighted sequences [92]. The specificity of MRI is moderate because of overlap in the appearance of metastases and a variety of benign lesions. In this meta-analysis, both whole body MRI and axial MRI studies were enrolled. The subgroup analysis showed that the sensitivity and DOR for axial MRI were significantly higher than those for wholebody MRI. Although whole body MRI is capable of meeting a variety of needs in a single examination, it has some limitations. Many anatomical structures cannot be assessed: among them are the small bowel and stomach, breast, prostate, joints and coronary arteries. Some of these structures can be assessed by dedicated MRI.Another challenge is the enormous increase in image data in whole body MRI, which not only implies an increase in evaluation time for the radiologist but also increases the potential for false-negative findings. Additionally, the radiologist cannot restrict him- or herself to assessing only the target structures. In particular, the three-dimensional nature and the large fieldof-view of whole body MR necessitate the work-up of all visible structures: bones and soft tissues, parenchymal organs, lymph nodes and the venous system. The radiologist must assume responsibility for chance findings, and the patient must consider the possibility that indistinct findings may be made and that these might have to be assessed with further, potentially invasive tests. Bone scintigraphy is the most commonly used technique for the detection of bone metastases, as it can provide an entire skeletal visualisation within a reasonable amount of time and cost [5, 93] BS has a high false-positive rate decreasing the specificity of BS because of benign processes, such as fractures and degenerative changes which cause increased bone turnover. False-negative find-

Eur Radiol (2011) 21:2604–2617

ings can occasionally result when pure osteolytic metastases are growing rapidly, when bone turnover is slow, or when the site is avascular (photon-deficient lesions; “cold spots”) [94, 95] Furthermore, BS should never be considered diagnostic when it produces equivocal findings (e.g. “suspicious” lesions or a single “hot spot”). In the present study, BS was shown to have the lowest specificity on a per-patient basis. While, the addition of SPECT to the acquisition protocol of BS can improve its diagnostic accuracy for detecting malignant bone involvement [96, 97].SPECT is useful in evaluations of complex areas that are extensively surrounded by soft tissue such as the thoracolumbar spine and pelvis; it can also clarify “hot spots” obtained with other imaging techniques by virtue of its improved contrast resolution. SPECT was useful for distinguishing benign from malignant lesions. In the present study, we demonstrated that BS using SPECT was better than BS without SPECT. In this meta-analysis, analysis was performed both on a per-patient basis and on a per-lesion basis. Although treatment plans are based primarily on the presence of metastatic disease as opposed to the absolute number of lesions, the number, size, location, and surgical margin of the bone metastases are also important. Because of its noninvasive nature, low cost and widespread availability, BS can be used to help distinguish patients with diffuse disease who are not eligible for curative treatment from the group of patients with no bone metastases or the group with a limited number of bone metastases. The patients in the latter group should undergo MR imaging, or 18FDG PET. On the basis of the results of this meta-analysis, PET and MRI were found to be comparable and both significantly more accurate than CT and BS for the diagnosis of bone metastases. PET/CT was better than PET and the type of analysis for PET imaging would affect the diagnostic result on a per-lesion basis. The sensitivity and DOR for axial MRI were significantly higher than those for whole-body MRI. Using DWI sequences would affect the diagnostic result of MRI on a perpatient basis. BS using SPECT was better than BS without SPECT both on a per-patient basis and on a per-lesion basis. We analysed bone metastases of all kinds of tumours together; it would be ideal to analyse different types of primary tumour separately. Thus, in the future, better designed and more specific articles are needed to continue to collect and separately compare the effectiveness of these techniques for different types of primary tumour.

References 1. Even-Sapir E (2005) Imaging of malignant bone involvement by morphologic, scintigraphic, and hybrid modalities. J Nucl Med 46:1356–1367

Eur Radiol (2011) 21:2604–2617 2. Padhani A, Husband J (1998) Bone metastases. In: Husband JES, Reznek RH (eds) Imaging in oncology. Isis Medical Media Ltd, Oxford, pp 765–787 3. Rubens RD (1998) Bone metastases: the clinical problem. Eur J Cancer 34:210–213 4. Rybak LD, Rosenthal DI (2001) Radiological imaging for the diagnosis of bone metastases. Q J Nucl Med 45:53–64 5. Hamaoka T, Madewell JE, Podoloff DA et al (2004) Ueno NT. Bone imaging in metastatic breast cancer. J Clin Oncol 22:2942– 2953 6. Roodman GD (2001) Biology of osteoclast activation in cancer. J Clin Oncol 19:3562–3571 7. Van Houwelingen HC, Zwinderman KH, Stijnen T (1993) A bivariate approach to meta-analysis. Stat Med 12:2273–2284 8. Van Houwelingen HC, Arends LR, Stijnen T (2002) Advanced methods in meta-analysis: multivariate approach and metaregression. Stat Med 21:589–624 9. Deville WL, Bezemer PD, Bouter LM (2000) Publications on diagnostic test evaluation in family medicine journals: an optimal search strategy. J Clin Epidemiol 53:65–69 10. Berlin JA (1997) Does blinding of readers affect the results of meta-analyses? University of Pennsylvania Meta-analysis Blinding Study Group. Lancet 350:185–186 11. Whiting P, Rutjes AW, Reitsma JB et al (2003) The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Meth 3:25 12. Moses LE, Shapiro D, Littenberg B (1993) Combining independent studies of a diagnostic test into a summary ROC curve: dataanalytic approaches and some additional considerations. Stat Med 12:1293–1316 13. Zamora J, Abraira V, Muriel A et al (2006) Meta-DiSc: a software for meta-analysis of test accuracy data. BMC Med Res Meth 6:31 14. Eustace S, Tello R, Carey J et al (1997) A comparison of wholebody turbostir MR imaging and planar 99 m Tc-Methylene Diphosphonate Scintigraphy in the examination of patients with suspected skeletal metastasis. Am J Roentgenol 169:1655–1661 15. Bury T, Barreto A, Daenen F et al (1998) Fluorine-18 deoxyglucose positron emission tomography for the detection of bone metastases in patients with non-small cell lung cancer. Eur J Nucl Med 25:1244–1247 16. Gibril F, Doppman JL, Reynolds JC et al (1998) Bone metastases in patients with gastrinomas: a prospective study of bone scanning, somatostatin receptor scanning, and magnetic resonance image in their detection, frequency, location, and effect of their detection on management. J Clin Oncol 16:1040–1053 17. Schirrmeister H, Guhlmann A, Elsner K et al (1999) Sensitivity in detecting osseous lesions depends on anatomic localization: planar bone scintigraphy versus 18 F-PET. J Nucl Med 40:1623–1629 18. Yip CH, Paramsothy M (1999) Value of routine 99mTc-MDP bone scintigraphy in the detection of occult skeletal metastases in women with primary breast cancer. Breast 8:267–269 19. Lebtahi R, Cadiot G, Delahaye N et al (1999) Detection of bone metastases in patients with endocrine gastroenteropancreatic tumors: bone scintigraphy compared with somatostatin receptor scintigraphy. J Nucl Med 40:1602–1608 20. Moog F, Kotzerke J, Reske SN (1999) FDG PET can replace bone scintigraphy in primary staging of malignant lymphoma. J Nucl Med 40:1407–1413 21. Schirrmeister H, Guhlmann A, Kotzerke J et al (1999) Early detection and accurate description of extent of metastatic bone disease in breast cancer with fluoride ion and positron emission tomography. J Clin Oncol 17:2381–2389 22. Earnest F, Ryu JH, Miller GM et al (1999) Suspected non-small cell lung cancer: incidence of occult brain and skeletal metastases

2615

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

and effectiveness of imaging for detection - pilot study. Radiology 211:137–145 Layer G, Steudel A, Schuller H et al (1999) Magnetic resonance imaging to detect bone marrow metastases in the initial staging of small cell lung carcinoma and breast carcinoma. Cancer 85:1004– 1009 Schirrmeister H, Glatting G, Hetzel J et al (2001) Prospective evaluation of the clinical value of planar bone scans, SPECT, and 18 F-Labeled NaF PET in newly diagnosed lung cancer. J Nucl Med 42:1800–1804 Koga S, Tsuda S, Nishikido M et al (2001) The diagnostic value of bone scan in patents with renal cell carcinoma. J Urology 166:2126–2128 Altehoefer C, Ghanem N, Hogerle S et al (2001) Comparative detectability of bone metastases and impact on therapy of magnetic resonance imaging and bone scintigraphy in patients with breast cancer. Eur J Radiol 40:16–23 Ohta M, Tokuda Y, Suzuki Y et al (2001) Whole body PET for the evaluation of bony metastases in patients with breast cancer: comparison with 99Tcm-MDP bone scintigraphy. Nucl Med Commun 22:875–879 Savelli G, Maffioli L, Maccauro M et al (2001) Bone scintigraphy and the added value of SPECT (single photon emission tomography) in detecting skeletal lesions. Q J Nucl Med 45:27–37 Buchmann I, Reinhardt M, Elsner K et al (2001) 2-(Fluorine-18) Fluoro-2-Deoxy-D-Glucose positron emission tomography in the detection and staging of malignant lymphoma: a bicenter trial. Cancer 91:889–899 Wu HC, Yen RF, Shen YY et al (2002) Comparing whole body 18 F-2-deoxyglucose positron emission tomography and technetium-99 m methylene diphosphate bone scan to detect bone metastases in patients with renal cell carcinomas – a preliminary report. J Cancer Res Clin Oncol 128:503–506 Ghanem N, Altehoefer C, Hogerle S et al (2002) Comparative diagnostic value and therapeutic relevance of magnetic resonance imaging and bone marrow scintigraphy in patients with metastatic solid tumors of the axial skeleton. Eur J Radiol 43:256–261 Yang SN, Liang JA, Lin FJ et al (2002) Comparing whole body 18 F-2-deoxyglucose positron emission tomography and technetium-99 m methylene diphosphonate bone scan to detect bone metastases in patients with breast cancer. J Cancer Res Clin Oncol 128:325–328 Hsia TC, Shen YY, Yen RF et al (2002) Comparing whole body 18 F2-deoxyglucose positron emission tomography and technetium-99 m methylene diphosphate bone scan to detect bone metastases in patients with non-small cell lung cancer. Neoplasma 49:267–271 Prior JO, Barghouth G, Delaloye JF et al (2003) The value of bone marrow scintigraphy using 99mTc monoclonal antigranulocyte antibodies in complement to bone scintigraphy in detecting bone metastases from primary breast cancer. Nucl Med Commun 24:29–36 Zhang C, Zhou Q, Chen YH et al (2003) Comparison of 18 F-FDG PET with 99mTc-MDP bone scan in detection of bone metastases. Chin J Nucl Med 23:201–203 Lee KH, Park JM, Yoon JK et al (2003) Bone scintigraphy of skeletal metastasis in hepatoma patients treated by TAE. HepatoGastroenterology 50:1983–1986 Han LJ, Qu WY, He J et al (2003) Comparison of the clinical values of 18 F-FDG hPET/CT and 99mTc-MDP bone scan in detection of bone metastases. Chin J Nucl Med 23:204–206 Nakamoto YJ, Osman M, Wahl RL (2003) Prevalence and patterns of bone metastases detected with positron emission tomography using F-18 FDG. Clin Nucl Med 28:302–307 Bossche BVD, Dhaeninck E, Winter FD et al (2004) 99mTc depreotide scan compared with 99mTc-MDP bone scintigraphy for the detection of bone metastases and prediction of response to

2616

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.

56.

57.

hormonal treatment in patients with breast cancer. Nucl Med Commun 25:787–792 Schirrmeister H, Arslandemir C, Glatting G et al (2004) Omission of bone scanning according to staging guidelines leads to futile therapy in non-small cell lung cancer. Eur J Nucl Med Mol Imag 31:964–968 Cheran SK, Herndon JE 2nd, Patz EF Jr (2004) Comparison of whole-body FDG-PET to bone scan for detection of bone metastases in patients with a new diagnosis of lung cancer. Lung Cancer 44:317–325 Engelhard K, Hollenbach HP, Wohlfart K et al (2004) Comparison of whole-body MRI with automatic moving table technique and bone scintigraphy for screening for bone metastases in patients with breast cancer. Eur Radiol 14:99–105 Lauenstein TC, Goehde SC, Herborn CU et al (2004) Whole-body MR imaging: evaluation of patients for metastases. Radiology 233:139–148 Abe K, Sasaki M, Kuwabara Y et al (2005) Comparison of 18 FDG-PET with 99mTc-HMDP scintigraphy for the detection of bone metastases in patients with breast cancer. Ann Nucl Med 19:573–579 Nakai T, Okuyama C, Kubota T et al (2005) Pitfalls of FDG-PET for the diagnosis of osteoblastic bone metastases in patients with breast cancer. Eur J Nucl Med Mol Imag 32:1253–1258 Kato H, Miyazaki T, Nakajima M et al (2005) Comparison between whole-body positron emission tomography and bone scintigraphy in evaluating bony metastases of esophageal carcinomas. Anticancer Res 25:4439–4444 Uematsu T, Yuen S, Yukisawa S et al (2004) Comparison of FDG PET and SPECT for detection of bone metastases in breast cancer. AJR Am J Roentgenol 184:1266–1273 Fujimoto R, Higashi T, Nakamoto Y et al (2006) Diagnostic accuracy of bone metastases detection in cancer patients: Comparison between bone scintigraphy and whole-body FDGPET. Ann Nucl Med 20:399–408 Gao Y, Fang J, Liu XY et al (2006) Diagnostic value of nuclide bone imaging for bone metastasis from lung cancer and clinic analysis. Chin J Lung Canc 9:357–361 Groves AM, Beadsmoore CJ, Cheow HK et al (2006) Can 16detector multislice CT exclude skeletal lesions during tumour staging? Implications for the cancer patient. Eur Radiol 16:1066– 1073 Sadik M, Jakobsson D, Olofsson F et al (2006) A new computerbased decision-support system for the interpretation of bone scans. Nucl Med Commun 27:417–423 Liu FY, Chang JT, Wang HM et al (2006) [18 F]Fluorodeoxyglucose positron emission tomography is more sensitive than skeletal scintigraphy for detecting bone metastasis in endemic nasopharyngeal carcinoma at initial staging. J Clin Oncol 24:599–604 Sapir EE, Metser U, Mishani E et al (2006) The detection of bone metastases in patients with high-risk prostate cancer: 99mTc-MDP planar bone scintigraphy, single- and Multi-Field-of-View SPECT, 18 F-Fluoride PET, and 18 F-Fluoride PET/CT. J Nucl Med 47:287– 297 Aslan S, Cetin B, Akinci M et al (2006) Computed tomography in detecting bone metastases of breast carcinoma. Is it better than plain x-ray? Saudi Med J 27:1326–1328 Hur J, Yoon CS, Ryu YH et al (2007) Accuracy of fluorodeoxyglucose positron emission tomography for diagnosis of single bone metastasis: comparison with bone scintigraphy. J Comput Assist Tomogr 31:812–819 Phan HT, Jager PL, Plukker JT et al (2007) Detection of bone metastases in thyroid cancer patients: bone scintigraphy or 18 FFDG PET? Nucl Med Commun 28:597–602 Strobel K, Burger C, Brukhardt S et al (2007) Characterization of focal bone lesions in the axial skeleton: lesions in the axial

Eur Radiol (2011) 21:2604–2617

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

skeleton: scintigraphy compared with SPECT and SPECT fused with CT. AJR Am J Roentgenol 188:467–474 Ito S, Kato K, Ikeda M et al (2007) Comparison of 18 F-FDG PET and bone scintigraphy in detection of bone metastases of thyroid cancer. J Nucl Med 48:889–895 Schmidt GP, Schoenberg SO, Schmid R et al (2007) Screening for bone metastases: whole-body MRI using a 32-channel system versus dual-modality PET-CT. Eur Radiol 17:939–949 Taira A, Herfkens RJ, Gambhir SS et al (2007) Detection of bone metastases: assessment of integrated FDG PET/CT Imaging. Radiology 243:204–211 Pfannenberg C, Aschoff P, Schanz S et al (2007) Prospective comparison of 18 F-fluorodeoxyglucose positron emission tomography/computed tomography and whole-body magnetic resonance imaging in staging of advanced malignant melanoma. Eur J Cancer 43:557–564 Lecouvet FE, Geukens D, Stainier A et al (2007) Magnetic resonance imaging of the axial skeleton for detecting bone metastases in patients with high-risk prostate cancer: diagnostic and cost-effectiveness and comparison with current detection strategies. J Clin Oncol 25:3281–3287 Xu X, Ma L, Zhang JS et al (2008) Feasibility of whole body diffusion weighted imaging in detecting bone metastasis on 3.0 T MR scanner. Chin Med Sci J 23:151–157 Sadik M, Suurkula M, Hoglund P et al (2008) Quality of planar whole-body bone scan interpretations—a nationwide survey. Eur J Nucl Med Mol Imag 35:1464–1472 Kim MR, Roh JL, Kin JS et al (2008) 18 F-Fluorodeoxyglucosepositron emission tomography and bone scintigraphy for detecting bone metastases in patients with malignancies of the upper aerodigestive tract. Oral Oncol 44:148–152 Bistow AR, Agrawal A, Evans AJ et al (2008) Can computerised tomography replace bone scintigraphy in detecting bone metastases from breast cancer? A prospective study. Breast 17:100–105 Wu SQ, Liu JJ, Song SS et al (2008) Comparison of the value of 18 F-FDG PET and 99mTc-MDP bone scan in the detection of bone metastases. Radiol Pract 23:1273–1277 Song JW, Oh YM, Shim TS et al (2009) Efficacy comparison between 18 F-FDG PET/CT and bone scintigraphy in detecting bony metastases of non-small-cell lung cancer. Lung Cancer 65:333–338 Takenaka D, Ohno Y, Matsumoto K et al (2009) Detection of bone metastases in non-small cell lung cancer patients: comparison of whole-body diffusion-weighted imaging (DWI), whole-body MR Imaging without and with DWI, whole-Body FDG-PET/CT, and bone scintigraphy. J Magn Reson Imaging 30:298–308 Venkitaraman R, Cook G, Dearnaley D et al (2009) Whole-body magnetic resonance imaging in the detection of skeletal metastases in patients with prostate cancer. J Med Imaging Radiat Oncol 53:241–247 Min JW, Um SW, Yi JJ et al (2009) The Role of whole-Body FDG PET/CT, Tc 99 m MDP bone scintigraphy, and serum alkaline phosphatase in detecting bone metastasis in patients with newly diagnosed lung cancer. J Kor Med Sci 24:275–280 Costo S, Halley A, Bergot E et al (2009) Usefulness of 99mTcdepreotide scintigraphy for the diagnosis of bone metastases in non-small cell lung cancer stage III–IV. Med Nucl 33:201–210 Kruger S, Buck AK, Mottaghy FM et al (2009) Detection of bone metastases in patients with lung cancer: 99mTc-MDP planar bone scintigraphy, 18 F-fluoride PET or 18 F-FDG PET/CT. Eur J Nucl Med Mol Imag 36:1807–1812 Sohaib SA, Cook G, Allen SD et al (2009) Comparison of wholebody MRI and bone scintigraphy in the detection of bone metastases in renal cancer. Br J Radiol 82:632–639 Xu WN, Xin J, Yu SP et al (2009) Comparison of 18 F-FDG PETCT and 99mTc-MDP bone scintigraphy in detection of bone metastases. J Chin Clin Med Imaging 20:323–327

Eur Radiol (2011) 21:2604–2617 76. Liu NB, Ma L, Zhou W et al (2010) Bone metastasis in patients with non-small cell lung cancer: the diagnostic role of F-18 FDG PET/CT. Eur J Radiol 74:231–235 77. Liu FY, Yen TC, Chen MY et al (2009) Detection of hematogenous bone metastasis in cervical cancer: 18 F-Fluorodeoxyglucose– positron emission tomography versus computed tomography and magnetic resonance imaging. Cancer 115:5470–5480 78. Heusner T, Golitz P, Hamami M et al (2011) “One-stop-shop” staging: Should we prefer FDG-PET/CT or MRI for the detection of bone metastases? Eur J Radiol 78:430–435 79. Putzer D, Gabriel M, Henninger B et al (2009) Bone metastases in patients with neuroendocrine tumor:68 Ga-DOTA-Tyr3-Octreotide PET in comparison to CT and bone scintigraphy. J Nucl Med 50:1214–1221 80. Zhao Z, Li L, Li FL et al (2010) Single photon emission computed tomography/spiral computed tomography fusion imaging for the diagnosis of bone metastasis in patients with known cancer. Skeletal Radiol 39:147–153 81. Lijmer JG, Mol BW, Heisterkamp S et al (1999) Empirical evidence of design-related bias in studies of diagnostic tests. JAMA 282:1061–1066 82. Rutter CM, Gatsonis CA (2001) A hierarchical regression approach to meta-analysis of diagnostic test accuracy evaluations. Stat Med 20:2865–2884 83. Pham B, Platt R, McAuley L et al (2001) Is there a “best” way to detect and minimize publication bias? an empirical evaluation. Eval Health Prof 24:109–125 84. Bossuyt PM, Reitsma JB, Bruns DE et al (2003) Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD initiative. Radiology 226:24–28 85. Fay MP, Graubard BI, Freedman LS et al (1998) Conditional logistic regression with sandwich estimators: application to a meta-analysis. Biometrics 54:195–208

2617 86. Bipat S, van Leeuwen MS, Comans EF et al (2005) Colorectal liver metastases: CT, MR imaging, and PET for diagnosis – metaanalysis. Radiology 237:123–131 87. Tryciecky EW, Gottschalk A, Ludema K (1997) Oncologic imaging: interactions of nuclear medicine with CT and MRI using the bone scan as a model. Semin Nucl Med 27:142–151 88. Deeks J (2001) Systematic reviews of evaluations of diagnostic and screening tests. BMJ 323:157–162 89. Cook GJ, Houston S, Rubens R et al (1998) Detection of bone metastases in breast cancer by 18FDG PET: differing metabolic activity in osteoblastic and osteolytic lesions. J Clin Oncol 16:3375–3379 90. Muindi J, Coombes RC, Golding S et al (1983) The role of computed tomography in the detection of bone metastases in breast cancer patients. Br J Radiol 56:233–236 91. Karnholz R, Sze G (1991) Current imaging in spinal metastatic disease. Semin Oncol 18:158–169 92. Vogler JB, Murphy WA (1988) Bone marrow imaging. Radiology 168:679–693 93. Blake GM, Park-Holohan SJ, Cook GJ et al (2001) Quantitative studies of bone with the use of 18 F-fluoride and 99mTc-methylene diphosphonate. Semin Nucl Med 31:28–49 94. Cook GJ, Fogelman I (2000) The role of positron emission tomography in the management of bone metastases. Cancer 88:2927–2933 95. Loeffler RK, DiSimone RN, Howland WJ (1975) Limitations of bone scanning in clinical oncology. JAMA 234:1228–1232 96. Roland J, van den Weygaert D, Krug B et al (1995) Metastases seen on SPECT imaging despite a normal planar bone scan. Clin Nucl Med 20:1052–1054 97. Sedonja I, Budihna NV (1999) The benefit of SPECT when added to planar scintigraphy in patients with bone metastases in the spine. Clin Nucl Med 24:407–413