Meta analysis of low dose spiral CT screening for lung cancer in high risk population

Navigation:Home > Respiratory Medicine > Lung Cancer > Meta analysis of low dose spiral CT screening for lung cancer in high risk population

1 Shang Wenli 2 Zhang Yang Shuanying 1 Zhang Wei 1 Huo Shufen 1 Liu Yanfeng 1 Zheng Huadong 3 Lin Jie Du 1 Chapter 31 Department of respirat


1 Shang Wenli 2 Zhang Yang Shuanying 1 Zhang Wei 1 Huo Shufen 1 Liu Yanfeng 1 Zheng Huadong 3 Lin Jie Du 1 Chapter 3

1 Department of respiration, the Second Affiliated Hospital of Xi'an Jiao Tong University Medical College, Xi'an 710004

Department of respiratory medicine, Baoji Central Hospital, Shaanxi 2, China

3 elderly Department of respiration, Second Affiliated Hospital of Xi'an Jiao Tong University

Correspondent: Yang Shuanying

Abstract: Objective: To evaluate the value of low-dose spiral CT (low-dose spiral computed tomography) in the screening of lung cancer by Meta analysis. Methods: to search the related literature during 1996 -2010 in PubMed and Cochrane database; according to the inclusion criteria to screen eligible literature; scientific evaluation and quality classification of statistical information on the literature; by using a random effects model for data analysis; the likelihood ratio (likelihood ratio LR), the weighted sensitivity and specificity (Pooled Sensitivity and Specificity) and summary receiver operating characteristic curve (summary receiver operator characteristic curve, SROC curve) was comprehensively evaluated the value of LDCT in screening for lung cancer. Results: a total of 12 articles were included in the study, which were composed of 9 independent studies, with a total of 28024 subjects. The quality evaluation results suggest that the literature quality is better; LDCT value in screening for lung cancer with +LR was 3.35, -LR was 0.29, the weighted sensitivity was 0.735 (0.730 ~ 0.740), the specificity was 0.791 (0.747 ~ 0.830), the area under the SROC curve (Area under, curve, AUC) 0.8473. Conclusions: the value of LDCT reported in most independent research units is consistent with the value of screening for early lung cancer. The results of this study indicate that LDCT is of high value in early lung cancer screening, but it still needs high quality randomized controlled trials to evaluate its clinical value more accurately.

Key words: low dose spiral CT, early lung cancer, meta analysis, screening

Role of low-dose CT in lung cancer screening: a spiral scan Meta-analysis

[Abstract] Objective: Using meta analysis to evaluate the value of LDCT in screening for lung cancer. Methods: Searching the relevant documentation from PubMed ISI Web, of Knowledge, Cochrane and other databases from 1996 to 2010; screening eligible literature according to the inclusive criteria literatures were gathered as statistical; information assessed, the quality of scientific and classified; using random effect model to analyze data to evaluate the value of; LDCT in screening for lung cancer comprehensively by the likelihood ratio (LR), the pooled sensitivity, specificity and summary receiver operating characteristic curve (SROC curve) statistical indicators. Results: 12 English literatures were collected with 9 independent R Esearches and a total of were brought into the 28024 cases study. Heterogeneity test showed that the homogeneity of the study was good, which showed the quality of literature collected in this study was preferable the value of; the + LR was 3.35, -LR was0.29, the weighted sensitivity of 0.735 (0.730 ~ 0.740), the specificity (was0.791 0.747 ~ 0.830) and SROC area under the curve (Area under curve AUC to0.8473 using LDCT screening) by lung cancer. Conclusion: The reports of the most independent research on the LDCT screening for lung cancer were simply consistent. Using LDCT to screen for lung cancer had a high value through the summary of relevant research results, but it still needed high-quality random control trail to assess its v Alue in clinic more accurately.

Table 1 details of the literature


research center

Sample size (person)

Age (age)

Smoking history



High risk population

Quality score (grade)

Ugo Pastorino et al. [7]


One thousand and thirty-five

More than 50

More than 20



Takeshi Nawa et al. [8]


Seven thousand nine hundred and fifty-six


Not require



Shusuke Sone et al. [9]


Five thousand four hundred and eighty-three


Not require



Claudia I.Hensch et al. [10-11]


One thousand

More than 60

More than 10



Stefan Diederich et al. [12-13]


Eight hundred and seventeen

> 40

More than 20



Giulia Veronesia et al. [14]


Five thousand two hundred and one

More than 50

More than 20



S. J. Swensen et al. [15]


One thousand five hundred and twenty

More than 50

More than 20



John K. Gohagan et al. [16]


One thousand six hundred and sixty


More than 20



Ravi J. Menezes et al. [17]


Three thousand three hundred and fifty-two

More than 50

More than 10



Results: 2 literature evaluation evaluation shows that this research into the scientific literature, quality evaluation for a or B, sensitivity analysis showed that the literature has good stability, and a large number of samples are representative, can be used for Meta analysis.

3 data analysis: a total of 12 articles were selected and included in the study, with a total of 28024 subjects aged over the age of 40.

(1) heterogeneity test: the results of this study show that the logarithmic correlation coefficient =0.417 (Spearman), P=0.265, P, >, and of the logarithm of sensitivity and (1- specificity) shows that there is no threshold effect, as shown in table 2. Calculated DOR values of Cochran-Q=30.9, df=8, P=0.0001, P < 0.05, indicating the presence of non threshold effects. As figure 1. The results showed that there was heterogeneity in the literature, which was related to the population, age, and so on. Therefore, the stochastic model was used to analyze meta.

Table 2:Analysis of Diagnostic Threshold

Spearman correlation coefficient: 0.417 0.265 p-value=

(Logit (TPR) vs Logit (FPR)

------------------------------------------------------------------------------- Moses'model (D = a + bS)

Weighted regression (Inverse Variance)

Var Coeff. Error T Std. p-value


A 2.532 0.291 8.703 0.0001

B (1) -0.076 0.179 0.425 0.6836


Tau-squared = is (Convergence achieved after 8 iterations 0.5357 estimate)

Restricted Maximum Likelihood estimation (REML)

No. studies = 9

Figure 1 shows the diagnostic odds ratio of forest map, the red dot represents a specific value, the line represents the 95%CI arrow data not display area.

(2) weighted sensitivity and specificity: because of the heterogeneity of the data provided in the literature, the random effects model. DerSimonian-Laird statistical method was used to calculate the combined DOR value, sensitivity and specificity, and the 95%CI was calculated. DORDL=12.37 (7.17-21.35) (see Figure 1). The pooled weighted sensitivity and specificity of the samples after the indirect merger, the combined sensitivity of =0.735 (0.730 ~ 0.740), specificity =0.791 (0.747 ~ 0.830) (see Table 2). Figure 2, 3, for its forest map, you can visually see the size of the weighted sensitivity and specificity for each document. Aggregated weighted positive likelihood ratio (+LR) =3.35 and negative likelihood ratio (-LR) =0.29, as shown in Figure 4, 5.

Table 3:Summary Specificity and Sensitivity

Study Spe [95% Conf. Iterval.] Sen [95% | Conf. Iterval.] --------------------------------------------------------------------------------------

Ugo Pastorino | 0.820 0.795 - 0.844 0.773 0.546-0.922

Takeshi Nawa | 0.739 0.730 - 0.749 0.975 0.868-0.999

Shusuke Sone | 0.953 0.947 - 0.958 0.458 0.314-0.608

Claudia I. Henschke | 0.787 0.760 - 0.812 0.794 0.621-0.913

Stefan Diederich | 0.580 0.545 - 0.615 0.692 0.482-0.857

Giulia Veronesia | 0.476 0.463 - 0.490 0.859 0.770-0.923

S J. SWENSEN | 0.492 0.466 - 0.517 0.880 0.688-0.975

John K. Gohagan | 0.812 0.792 - 0.830 0.750 0.588-0.873

Ravi J. Menezes | 0.834 0.821 - 0.847 0.862 0.753-0.935 --------------------------------------------------------------------------------------

Pooled | 0.735 0.730 - 0.740 0.791 0.747-0.830


Heterogeneity chi-squared = 45.72 (d.f.= =) P = 0

Inconsistency (I-square) = 82.5

Figure 2, figure, a forest map of the sensitivity and specificity of LDCT in early lung cancer screening, figure 4, 5 is a forest map with positive likelihood ratio and negative likelihood ratio. The red dot map (-) said the numerical study on the index, through the line to the corresponding 95% CI. The last diamond red dot is the combined value. If the dot is small, the linear distribution is wider, the research on the accuracy of small weights given in weighted small; conversely, the study of precision, weight is also large.

(3) the production of SROC curve using metadisc software: Table 2 shows linear correlation b=-0.076, no statistically significant difference of 0, so that the SROC curve is symmetrical, so Symmetric and Weighted least Squars curve, as shown in Figure 6 is included in the SROC curve fitting of literature and 95%CI, AUC=0.8473.

Figure 6 SROC curve

Figure 6 is a summary of the ROC curve, the SROC curve, which is included in the study in 9. The coordinate X axis was 1 - specific; the Y axis was sensitive. The red dot in the figure indicate the actual value of the included studies, if the dot is small, weight when given the small; conversely, the right is significant. A blue line in the middle of the graph indicates the SROC curve obtained by this group, with two blue lines on either side of the 95%CI.

Three, discussion:

Lung cancer is one of the most malignant tumors with the highest morbidity and mortality, the most serious threat to human health and life. Therefore, the effective screening method for early lung cancer is an urgent problem in the world. At present, the main methods used in the study of lung cancer screening include imaging, sputum cytology and the recent development of molecular pathology, fiberoptic bronchoscopy and fluorescent fiber optic bronchoscopy. In early twentieth Century 70, 80s, the United States, Japan's Medical Center has useful chest X-ray film and sputum cytology as a screening method for lung cancer, multi center, large sample, randomized study, but the research institutions the results showed: Although chest X-ray can detect some lung cancer, but it is not sensitive to the small lesions and subtle lesions the sensitivity of sputum cytology, only originated in the large airways of central type lung cancer have higher, clinically active treatment, the survival rate of patients with lung cancer for 5 years did not improve, and ultimately failed to reduce overall mortality, so they did not recommend a chest X-ray and sputum cytology as a means of [2,18-22] screening of early lung cancer. Chest CT scan is currently recognized as the display image of chest lesions most sensitive examination method, but if the routine dose of complete lung acquisition, especially the scanning body will absorb considerable X-ray radiation dose, the equipment load is too large. Therefore, to reduce the exposure dose and not affect the image quality of LDCT has become the focus of current research. Since the lung itself is an air bearing tissue, the contrast between the air in the alveoli and the surrounding soft tissue is high, and many studies have shown that LDCT does not reduce the ability to detect pulmonary nodules [23-25].

Meta analysis is a powerful scientific research method, it is actually a higher, more macro vision of all relevant research to explore the comprehensive analysis, to seek its inherent laws. In 80s, the Research Report of meta analysis began to increase and gradually applied to various fields of clinical medicine. This study used meta analysis method to comprehensive analysis and research of the research center on LDCT screening for lung cancer, according to the quality evaluation of literature is a or B, and the sample size is large, high credibility, but a meta-analysis of DOR value of Cochran-Q=30.9, df=8, P=0.0001, the existence of non threshold effects showed that all the results between considering the heterogeneity, the reasons are: 1 included literatures were English literature, language bias and publication bias; 2 research institutions in the selection of equipment (such as ordinary CT, monolayer and multilayer spiral CT) and scanning parameters (scanning duration, slice thickness, tube voltage, tube current) is not uniform; 3 study of the age, sex, race, smoking index is not unified; 3 different readers, the imaging results determine the different.

This paper using a random effects model in MetaDisc1.4 software, analyzed the results showed that LDCT in the early stage of the pooled sensitivity in lung cancer screening was 0.735, fluctuations in the 0.46~0.98, the specificity was 0.791, fluctuations in the 0.48~0.95, the fluctuations are large, because this not only with the heterogeneity, and the focus of the location, at the same time the lesion density, lesion size, lesion length (early lesions not detected) on the properties and the research object itself, and this study only analyzes LDCT in the first screening in sensitivity and specificity.

The positive likelihood ratio (+LR) =3.35 and negative likelihood ratio (-LR) of aggregated weighted =0.29. LR is a composite index that reflects both sensitivity and specificity. It is not affected by the prevalence, and fully reflects the value of diagnostic tests. +LR shows that the probability of correct judgment of the diagnostic test is positive, and the probability of error is positive. The greater the ratio, the greater the probability that the positive is true. The smaller the -LR value, the greater the possibility that the test result is true negative. The bigger the +LR, the smaller the -LR, the higher the diagnostic value of the diagnostic test. Studies have shown that when +LR is greater than 10 and less than 0.1 -LR, with a diagnosis of convincing effect, when +LR is greater than 5 and less than 0.2 -LR, [26] has stronger diagnostic efficacy, evaluation of other indicators of diagnostic test value is SROC curve and AUC. Effect of SROC curve is not affected by the changes of threshold, through graphics and area directly to the diagnostic test, the more accurately reflect the linear relationship between sensitivity and specificity, accuracy of AUC on behalf of the diagnostic tests, the results of this study area under the SROC curve was 0.8473, indicating that LDCT in the early diagnosis of lung cancer screening in high the.

At present, LDCT as a major issue: screening for lung cancer (1) false positives: studies have reported that [27] is less than 5mm, pulmonary nodules, malignant ratio is about 3%, and more than 10mm in malignant nodules of up to 50% per cent. Although LDCT can detect more lung cancer, the detection rate of benign nodules also increases. False positive may lead to patients with invasive iatrogenic anxiety, and increase the additional cost; (2) can ultimately improve the 5 year survival rate of patients with lung cancer: the lead time bias, the bias refers to the screening time of diagnosis and clinical diagnosis of the time difference, is interpreted as a result of screening for prolonged survival this time, on the surface of prolonged survival time, is actually ahead of time by screening leads to error diagnosis, early diagnosis is not dead time delay, just ahead of time makes the diagnosis from the time of diagnosis is 5 years survival rate increased, and there was no real change rate.

In conclusion, the results of the present study suggest that LDCT can screen more lung cancer, especially early lung cancer, so LDCT is currently the most effective means of screening for lung cancer. But still need to carry out more large sample, multicenter, randomized, controlled trial to grasp more comprehensive information, including screening criteria (smoking history, age group) and screening interval, and a clear LDCT screening can reduce mortality in patients with lung cancer.


[1] Gazdar AF, Minna Molecular detection early cancer. Nati Cancer Inst, 1999, 91:299-301, J, JD., of, lung

[2] Melamed MR, Flehinger BJ, Zaman MB, Heelan RT, Parchick WA, Martini Screening lung results of the Sloan-Kettering study New York. Chest, 1984; 86: 44 - - 53 - cancer: - Memorial, in - for

[3] Tockman MS. and from lung in screened the Johns Hopkins Study. Chest, 1986; 89: 324 - - 325, cancer, population:, a (mortality) Survival

[4] Fontana RS, Sanderson DR, Woolner LB, et al. cancer the program. J Occup Med, 1986; 28: 746 - - 50, Lung, Mayo

[5] Zhang Boheng, screening and diagnosis of Meta analysis. See Wang Jiyao, ed. evidence based medicine and clinical practice. 2 ed. Science Press, 2006; 134-139.

[6] Zhang Tiansong, bell. Fumiaki application of Meta-Disc software in the analysis of diagnostic tests in Meta. Evidence based medicine, 2008; 8: 97-108.

[7] Ugo Pastorino, Massimo Bellomi, Claudio Landoni, et al. Early lung-cancer detection with spiral CT and positron emission tomography in heavy smokers: 2-year results. Lancet, 2003; 362: 593 - 597

[8] Takeshi Nawa, Tohru Nakagawa, Suzushi Kusano, et al. Cancer Using Spiral CT: of Baseline 1-Year Follow-up Studies. Chest, 2002; 122; 15-20 (Results); Lung (and) Screening

[9] Shusuke Sone, Shodayu Takashima, Feng Li, et al. screening lung with mobile computed tomography scanner. Lancet, 1998; 351: 1242 - - 1245, for - spiral, cancer

[10] Henschke CI. Lung Action overall design and fi ndings from screening. Cancer Suppl, 2000; 89:2474 - - 2482; baseline Cancer Early Project:

[11] Henschke CI, Naidich DP, Yankelevitz DF, et al. Lung Action Project:initial fi ndings on screening. Cancer, 2001; 92:153 - - 159 (Early - repeat)

[12] Diederich S, Wormanns D, Semik M, et al. for lung with low-dose CT: prevalence in 817 asymptomatic Radiology, 2002; 222:773 - - 781 - early, cancer, spiral, smokers.

[13] Diederich S, Thomas M, Semik M, et al. Screening for early lung cancer with low-dose spiral computed tomography: results of annual follow-up examinations in asymptomatic smokers. Eur Radiol, 2004; 14:691, 702

[14] Giulia Veronesia, Massimo Bellomi, Mulshine JL, et al. Lung cancer screening with low-dose computed tomography: A non-invasive diagnostic protocol for baseline lung nodules. Lung Cancer, 2008; 61: 340 - 349

[15] SWENSEN SJ, JETT JR, SLOAN JA., et al. for Cancer Low-Dose Spiral Tomography. Am J Respir 2002; 165:508 - 513 - Screening Lung with

[16] Gohagan JK, Marcus PM, Fagerstrom RM, et al. Final results of the Lung Screening Study, a randomized feasibility study of spiral CT versus chest X-ray screening for lung cancer. Lung Cancer, 2005; 47:9, 15

[17] Menezes RJ, Roberts HC, Paul NS, et al. cancer using computed tomography at-risk individuals: Toronto experience. Lung Cancer, 2010; 67:177 - - 183, screening, in, The, low-dose

[18] Hayabuchi N, Russell WJ, Murakami J, et al. for cancer a fixed by chest radiography. Radiology, 1983; (148): 369 - 373 (population) Biennial in lung

[19] MuhmJR, Miller WE, Fontana RS, et detected cancer during screening program four2month chest radiographs. Radiology, 1983; 148; 609 - - 615 - Al (Lung) a

[20] Heelan RT, Flehinger BJ, Melamed MR, et lung Non-small-cell cancer: results the New screening program. Radiology, 1984; 151; 289 - - 293 (of) York

[21] Strauss GM, Gleason RE, Sugarbaker lung Screening cancer: another look; a different Chest, 1997; 111: 754-768 view. (DJ) for

[22] Eddy Screening for cancer. Intern Med 1989, 111: 232-237 DM. (lung) Ann

[23] Gartenschlager M, Schweden F, Gast K, et al. Detection low-dose conventional-dose spiral Eur Radiol, 1996; 8:609 - - 614, with - CT., Pulmonarynodules:

[24] Rusinek H, Naidich DP, McGuiness G, et al. nodule low-dose conventional CT. Radiology, 1998; 209:243 - - 249 (detection: - versus)

[25] Diederich S, Lenzen H, Windmann R, et al. CT pulmonary experimental and studies. Radiology, 1999; 213:289 - - 298, of - clinical, Low-dose

[26] Jacschke R, Guyart GH, Sackett DL. Users'guides to the medical literature III. How to use an artical about a diagnostic test B. What are the results and will they help me in Earing for my patients The Evidence-Based Medicine Working? Group.JAMA, 1994; 271:703-707

[27] Kim YH, Lee KS, Primack SL, et al. pulmonary on accompanying surgically lung cancer:likelihood malignancy. J Thorac Imaging, 2002; 17:40 - - 46, nodules, resectable, of, CT

Cerebral Vascular Disease,Acne,Heart Disease,Deaf,Headache,Std,Condyloma Acuminatum,Fibroid,Pneumonia,Brain Trauma,。 Rehabilitation Blog 

Rehabilitation Blog @ 2018