2006 4
UDC 10497 Research and application of synthetical evaluation in the Down s syndrome screening system 430070 2006 5 2006 5 2006 4
I
5% Meta II
Abstract Bioinformatics is the interdiscipline of computational molecular biology and information processing science. With the rapid development of computing techniques biological technologies are reshaping the human society. It brings new development for the screening of Down s syndrome by combining the computing techniques with the bioinformatics. Down s syndrome is an usual chromosome disease because of the abnormality of fetus chromosome. It will take big tragedy to the family. So it is a very meaningful job to study the method of screening the DS in the first and second trimester especially using the computing technique. The workload of the collection statistics and analyzing for the information can be depressed through using the DS screening software. The accuracy of the screening result is also improved. Using the software the data can be stored integrally. This is useful for the further study and diagnoses. The information for screening DS in the pregnancy period and the revise information are concluded here the main point and innovation are listed below: 1. The measures used for screening DS pregnancies in the first and second trimester has been concluded here; 2. The information related to the DS in the first and second trimester has been concluded here such as ultrasonic screening methods maternal serum biochemistry ; 3. The efficacy of the respective information are analyzed by combining theory of information and synthetical evaluation; 4. The screen information and its revise information for our system has been chosen though the statistical methods using single diathesis methods and multiple diathesis methods; 5. Three screen fields were evaluated by meta-analysis method. After the evaluation the best field has been chosen for the further study; 6. The function of the screening system is realized at last though the software. This screening system is used for determining the risk of DS pregnancies in the second trimester the risk of the NTD pregnancies is determined at the same time. It is proved that this system is according to the requirement of the DS screening system. Key words: Down s Syndrome Synthetical evaluation information theory meta-analysis III
1...1 1.1...1 1.1.1...1 1.1.2...1 1.2...2 1.2.1...2 1.2.2...2 1.3...3 1.3.1...3 1.3.2...4 1.4...6 2...7 2.1...7 2.1.1...7 2.1.2...7 2.1.3...9 2.2...11 2.2.1...11 2.2.2...11 2.2.3...12 2.2.4 SROC...16 2.3...20 3...21 3.1...21 3.2...24 3.3...25 3.4 β-hcg...28 i
3.5...28 3.6...29 4...31 4.1...31 4.1.1...31 4.1.2...32 4.2 Meta...33 4.2.1...33 4.2.2...33 4.2.3 SROC...35 4.2.4...38 4.3...39 5...40 5.1...40 5.2...40 5.3...40 5.4...41 5.4.1...41 5.4.2...42 5.4.3...43 5.4.4...43 5.4.5...45 5.5...46 6...48 6.1...48 6.2...48...50...54...55...56 ii
1 1.1 1.1.1 1866 J.Langdon.Down 21 (trisomy 21) J.Langdon.Down Down's Syndrome 1959 Lejune G 21 21 21 2 3 1.1.2 5 1/2 1 3 25 0 1
16.2 50% 5 5--6 8% 40 2.6% 50 15 1.2 1.2.1 2004 2 26 600 1.2.2 90 90 ( ) 21 2
1.3 1.3.1 1. 9 12 16~20 200--400ml 99% 0.3%(0.1%--0.4%) [1] 2 3 2. 1983 Simoni 1 DS [2] 1 9-11 0.6% [1] 3. 48-72 0.5%-1.5% 0.3%-1.5% 3
99% 1.3.2 5% 1. 10-13 (Crown Rump Length CRL) (Nichol Translucency NT) NT 2.5mm (NT (Cut-off) 2.5 [3][4] ) NT CRL 10-13 2. 4
/ (gold immunochromatographic assay GICA) 90 [5] 3. 69% 15-20 70% 15 20 86% [6] 4. Duchenne' s 5
[7] 1/10 5 --1/10 9 21 [10] 1.4 5 6
2 2.1 2.1.1 2.1.2 7
X N a 1, a 2, L an x a1 a2 K a n = px ( ) p1 p2 K a1 (2-1) Shannon p 1, p 2, pn 3 [11] f ( p p p ) 1 2 L f ( p p L p ),, n 1 2,, L n pi ( i = 1,2, L, N) 1 1 1 f (,,, ) gn ( ) N N L N = (2-2) ' ' ' ( ) = ( + + ) + + + f pp,, L, p f( p p L p), p, L, p ( p p L p) fpp (,, L, p) (2-3) 1 2 n 1 2 k k+ 1 N 1 2 k 1 2 k ' pk p = k k 1, 2,, K p + p + + p = L (2-4) L 1 2 Shannon f ( p p L p ) 1 2 k,,, n N n i i i= 1 3 f ( p1, p2, L, p ) = C p log p (2-5) 8
C 2 10 e bit Hartley nat Shannon H( p p L p ) N i i i= 1,,, N 1 2 N H( p1, p2, L, p ) = p log p (2-6) X X H( p ) = 0 Shannon 2.1.3 9
U={u 1 u 2 u m } V={v 1 v 2 v n } m n v11 v12 v13 R = L L L (2-7) vn 1 vn2 v n3 A={ a 1 a 2 a m } A R = B (2-8) B={ b 1 b 2 b n } 1 M (, ) b = max[min( a, r ),min( a, r ), L,min( a, r )], j = 1,2 L, n (2-9) j 1 1j 2 2 j m mj min max ( ) 2 M( ) b = max[ ar, a r, L, a r ], j = 1, 2 L, n (2-10) j 11j 2 2j m mj 1 3 M( ) b = min[min( a, r ) + min( a, r ) + L + min( a, r )], j = 1, 2 L, n (2-11) j 1 1j 2 2 j m mj a b = min(1, a+ b) m k = 1 ( a r ) 1 k kj m k = 1 a k = 1 4 M( +) bj = ar 11j + a2r2 j + L + amrmj, j = 1,2 L, n (2-12) 10
5 M( ) b = min( ar + a r + L + a r ), j = 1, 2 L, n (2-13) j 11j 2 2j m mj M (, ) M( ) M( ) M (, ) M( +) M( ) 4 1 2 3 2.2 2.2.1 Logistic OR OR 95% P t Logistic 2.2.2 P<0.10 ( ) Logistic (stepwise ) α=0.10 Logistic 11
2.2.3 (Meta- analysis) Meta Meta [10] Cohen [11] (NHST Null Hypothesis Significance Testing) Schmidt [12] Meta Abrahantes [17] Meta Meta-analysis Meta above beyond behind after Glass 1976 (American Education Research Association) Glass 1. 2. 3. Glass 12
[53] Glass 1. Meta 2. Meta 3. Meta [14] Meta [15] Meta Meta Meta ( ) (effect size) d (effect size) Meta ( t z F ) d d = ( Me Mc)/ Sc (2-14) t r d [15][16] d d Hedge 1982 10 1.5 d = wd / w (2-15) d w Meta w 13
2 = 2 /(8 + ) (2-16) w N d N Cohen 0. 20 0. 80 0. 50 [15] d 70 20 Meta 50 [14] ( ) ( homogeneity of effect size) Meta Meta [17] Rosenthal Rubin (1982) 2 2 x = ( w( d d) ) (2-17) d d w x 2 k 1 k [22] ( ) Hox Meta ( 2 ) [18] 14
d j = δ j + ej (2-18) d j j δ j e j j e j δ 2 j d j 2 r Fisher Z δ j 2 δ j = 1( n j 3) (2-19) n j [23] δ j δ j = β0 + β1z1j + β2z2 j + L + βpzpj + μj (2-20) Z ( ) p 2 μ j 2 [19] μ j δ 2 jp d j = β0 + β1z1j + β2z2 j + L + βpzpj + μj + ej (2-21) 2 d j = β0 + μj + ej (2-22) β 0 δ 2 μ (homogeneous) μ δ 0 δ 0 j 2 μ (heterogeneous) 2 ( ) 2 [19] 2 ID 1 ID CONS( 1) ( Fisher Z) CONS 2 ( Fisher Z) 1 1 1 2 ( ) δ x 2 x 2 2 μ 2 μ 15
d j se..( dj ) x 2 2 d j d j = (2-23) se..( dj ) [19] Meta Meta Meta Meta ROC Meta (true positive rate TPR) (false positive rate FPR) logit ROC SROC (Summary Receiver Operating Characteristic SROC) SROC 2.2.4 SROC SROC X ( ) X ( ) S (True positive rate TPR ) (False positive rate FPR ) (1 ) (TPR ) ( FPR ) S TPR FPR logit 16
U V TPR = ln( ) 1 TPR FPR = ln( ) 1 FPR (2-24) (2-25) U V U V Kardau D=V-U S=V+U (2-24 ) ( TPR) ( FPR) TPR /1 D = ln FPR /1 = ln (2-26) TPR FPR S= ln 1 TPR 1 FPR = ln D S [20] D ( ) ( ) D S (2-27) S=0 S>0 S<0 U V D S D S D=A+B S (2-28) A S=0 B (D) (S) 0 0 A TPR FPR TPR FPR 17
SROC (2-28) 1 ( ) A B 2 D a,b,c,d W 1 1 1 1 1 W= var ( D) = + + + a b c d 1 (2-29) a,b,c,d 0 0.5 ( a + 0.5 ),( b + 0.5 ),( c + 0.5 ),( d + 0.5) Moses Shapiro Littenberg 1993 SROC 3 S D (S D) (S D) 1/3 9 3 10 4 3 11 3 4 3 S D B ( S,D 1 1) ( S,D 2 2) ( S D ) A B 18
A B DS -DS S-S 1 2 2 1 = (2-30) 2 1 D D S-S 2 1 = (2-31) 2 1 SROC 1. TPR FPR (2-28) D S TPR FPR TPR FPR -A/( 1-B) 1 FPR TPR= 1+e FPR (1+ B) /(1 B) 1 (2-32) SROC 2. AUC (area under the curve) [22] a x a x AUC = + 1 b 1 x 1 b 1 x 1 (1 + b)(1 b) (1 + b)(1 b) exp( )( ) 1 exp( )( ) dx 0 (2-33) 2 2 AUC var( ) var( $ AUC AUC AUC AUC = a) + var( b$ ) + 2 cov( a$, b$ ) a b a b (2-34) a b 1 (1 + b) /(1 b) (1 + b) /(1 b) dx 0 (2-35) AUC 1 a x a x = ( )exp( ) ( ) 1+ exp( )( ) a 1 b 1 b 1 x 1 b 1 x SROC b 0 AUC OR AUChom = 2 [( OR 1) ln( OR) ] ( OR 1) (2-36) OR=exp(a) AUChom OR SE( AUC $ hom) = [( OR + 1) ln( OR) 2( OR 1)] SE( a) 3 ( OR 1) SROC b 0 AUC (2-37) 19
OR AUChom = ( OR 1) ln( OR) 2 ( OR 1) [ ] (2-36) OR=exp(a) AUChom OR SE( AUC $ hom) = [( OR + 1) ln( OR) 2( OR 1)] SE( a) (2-37) 3 ( OR 1) 3. Q* Q* ROC SROC TPR =1- FPR ROC exp(a/2) Q*= = OR [1+ OR ] (2-38) 1+exp(a/2) 2 SE(Q*)= OR 2( OR +1) SE(a) $ (2-39) 4. Z Q * [20] ( ) z= Q Q SE (Q ) + SE (Q ) (2-40) * * 2 * 2 * 1 2 1 2 2.3 Meta SROC 20
3 3.1 1. NT (Nuchal Translucency) 10 14 14 21 21 18 13 45XO 10-13 1985 Benacerraf [23] (NT) 2. (CRL) [24] 62.5% CRL 5 CRL 10-13 (Crown Rump Length CRL) (Nuchal Translucency NT) NT 2.5mm (NTCut-off 2.5 [25][26] ) 21
CRL CRL NT ( [27] ) 3-1 CRL(mm) 31 32 33 34 35 36 37 38 39 NT(mm) -- -- 1.01 1.03 1.04 1.06 1.08 1.1 1.12 CRL(mm) 40 41 42 43 44 45 46 47 49 NT(mm) 1.14 1.15 1.16 1.18 1.21 1.23 1.24 1.25 1.26 CRL(mm) 55 56 57 58 59 60 61 62 63 NT(mm) 1.28 1.3 1.32 1.34 1.36 1.38 1.41 1.43 1.47 CRL(mm) 64 65 NT(mm) 1.49 1.52 3. 4. 5 Awwad [28] 1994 20-28 418 Aloka SSD 650 418 10 (4 21 6 18 ) 408 408 5. Nybery [34] 18 4 (22.2%) 22
/ 0.9 [35] =-0.966+0.866 (BPD) [35] 6. (umbilical artery pulsatility index UAPI) Martinez [36] 534 10-13 26 UAPI 95 19 11 UAPI 95 57.8% 95.1% UAPI UAPI 7. Jauniaux [37] 13 21 4 ( 10-14 ) 5 (30.8%) Jauniaux [38] 1996 21 8. / Grandjean [39] 14-24 / 0.88 35% 4.6% 0.85 15% 2.3% BPD( )/ (FL) PBD/ 1.80 40% 97.8% 2.2% [35] 3-2 BPD/FL [36] BPD/FL( 1.5SD) BPD/FL( 1.5SD) 15 1.93 19 1.68 16 1.93 20 1.58 17 1.76 21 1.54 18 1.74 22 1.47 23
( + )/ Johnson [37] 814 36 ( + )/ 1.75 OR 15.3 53% 93% 0.32 0.97 6.9% 3.2 88% 21 8% 2% 21 20% 3-3 [38] 30 <1/1000 40 1/105 30 1/900 42 1/60 35 1/400 44 1/35 36 1/300 46 1/20 37 1/230 48 1/16 38 1/180 49 1/12 39 1/135 24