Meta-analysis of second-trimester markers for trisomy 21

Authors


  • Corrections added on 22 February 2013 after first online publication: numerators and denominators for some of the analyzed studies have been changed in Tables 1-3, 5 and 9-11. Subsequent adjustments to the results of the meta-analysis have been made. No conclusions were affected.

Correspondence to: Prof. K. H. Nicolaides, Harris Birthright Research Centre for Fetal Medicine, King's College Hospital, Denmark Hill, London SE5 9RS, UK (e-mail: kypros@fetalmedicine.com)

ABSTRACT

Objective

To summarize by meta-analysis the accumulated data on the screening performance of second-trimester sonographic markers for fetal trisomy 21.

Methods

We conducted a literature search to identify studies between 1995 and September 2012 that provided data on the incidence of sonographic markers in trisomy 21 and euploid fetuses at 14–24 weeks' gestation. Weighted independent estimates of detection rate, false-positive rate and positive and negative likelihood ratios (LR) of markers were calculated.

Results

A total of 48 studies was included in the analysis. The pooled estimates of positive and negative LR were, respectively: 5.83 (95% CI, 5.02–6.77) and 0.80 (95% CI, 0.75–0.86) for intracardiac echogenic focus; 27.52 (95% CI, 13.61–55.68) and 0.94 (95% CI, 0.91–0.98) for ventriculomegaly; 23.30 (95% CI, 14.35–37.83) and 0.80 (95% CI, 0.74–0.85) for increased nuchal fold; 11.44 (95% CI, 9.05–14.47) and 0.90 (95% CI, 0.86–0.94) for hyperechogenic bowel; 7.63 (95% CI, 6.11–9.51) and 0.92 (95% CI, 0.89–0.96) for mild hydronephrosis; 3.72 (95% CI, 2.79–4.97) and 0.80 (95% CI, 0.73–0.88) for short femur; 4.81 (95% CI, 3.49–6.62) and 0.74 (95% CI, 0.63–0.88) for short humerus; 21.48 (95% CI, 11.48–40.19) and 0.71 (95% CI, 0.57–0.88) for aberrant right subclavian artery (ARSA); and 23.27 (95% CI, 14.23–38.06) and 0.46 (95% CI, 0.36–0.58) for absent or hypoplastic nasal bone. The combined negative LR, obtained by multiplying the values of individual markers, was 0.13 (95% CI, 0.05–0.29) when short femur but not short humerus was included and 0.12 (95% CI, 0.06–0.29) when short humerus but not short femur was included.

Conclusion

The presence of sonographic markers increases, and absence of such markers decreases, the risk for trisomy 21. In the case of most isolated markers there is only a small effect on modifying the pre-test odds for trisomy 21, but with ventriculomegaly, nuchal fold thickness and ARSA there is a 3–4-fold increase in risk and with hypoplastic nasal bone a 6–7-fold increase. Copyright © 2012 ISUOG. Published by John Wiley & Sons, Ltd.

INTRODUCTION

Several studies have reported that certain features detected during second-trimester ultrasound examination are potential markers for fetal trisomy 21[1-48]. The most widely examined markers are lateral cerebral ventriculomegaly, absent or hypoplastic nasal bone, increased nuchal fold thickness, intracardiac hyperechogenic focus, aberrant right subclavian artery (ARSA), hyperechogenic bowel, mild hydronephrosis and shortening of the femur or humerus. Assessment of the risk for trisomy 21 based on each of these markers necessitates knowledge of their prevalence in trisomic and euploid fetuses.

The aim of this meta-analysis was to examine the screening performance of second-trimester sonographic markers for the detection of trisomy 21.

METHODS

Relevant citations on second-trimester markers for trisomy 21 were extracted from EMBASE and PubMed from 1995 to September 2012 to identify English language articles. Keywords and MeSH terms were combined to generate lists of studies: ‘soft markers’, ‘intracardiac echogenic focus/foci’, ‘ventriculomegaly’, ‘nuchal fold’, ‘nuchal thickness’, ‘echogenic bowel’, ‘hydronephrosis’, ‘pyelectasis’, ‘short humerus’, ‘short femur’, ‘aberrant right subclavian artery’, ‘ARSA’, ‘nasal bone hypoplasia’, ‘absent nasal bone’, ‘nasal bone length’, ‘trisomy 21’ and ‘Down syndrome’. The inclusion criteria were studies reporting on the incidence of one or multiple markers in trisomy 21 and proven or assumed to be euploid fetuses, publication in or after 1995 and minimum and maximum gestational age at examination of 14 and 24 weeks, respectively. The first reviewer (P.C.) sorted all articles by citations and abstract for more detailed evaluation. The second sort was revised by three reviewers (P.C., M.A., K.N.) and all relevant studies were entirely reviewed by the same reviewers. All studies were carefully compared to ensure that we avoided using duplicate reports on the same subjects.

Studies on second-trimester sonographic markers were eligible if first, they included and described both euploid and trisomy 21 fetuses (so that 2 × 2 tables for diagnostic performance of the markers could be constructed), second, the fetal karyotype was unknown at the time of sonographic examination (so as to avoid overt diagnosis bias) and third, chromosomal status of the fetuses was confirmed by either karyotype (the gold standard) or postnatal clinical examination. Prospective and retrospective cohort studies were considered eligible for inclusion if the above criteria were met. In the case of ARSA and absent or hypoplastic nasal bone the number of studies fulfilling these criteria was small and we expanded the selection to include case–control studies. In case of data duplication or overlap, only the largest or most recent study with available data was included.

Information was extracted on study population characteristics, time in pregnancy at which sonography was performed, inclusion and exclusion criteria, study design, outcome assessment and potential verification bias and the main results and conclusions of the study.

Quality and integrity of this review were validated with PRISMA: preferred reporting items for systematic reviews and meta-analyses[49].

Statistical analysis

A meta-analysis was performed to provide a quantitative summary of the test performance of each second-trimester sonographic marker. The Newcastle–Ottawa scale was used to assess the methodological quality of the studies. We assessed the overall diagnostic performance by weighted independent estimation of detection rate (sensitivity), false-positive rate (1 − specificity), positive likelihood ratio (LR; sensitivity / (1 − specificity)) and negative LR ((1 − sensitivity) / specificity). We used both fixed and random effects models to estimate weighted detection rate, false-positive rate and positive and negative LR across studies. The fixed-effects model weighs each study by the inverse of its variance. Random effects incorporate both within-study and between-study variation[50]. Random effects tend to provide wider CIs and are generally preferable, especially in the presence of between-study heterogeneity. Heterogeneity between studies was analyzed using both Higgins' I2 and Q-test and was considered to be high if I2 was over 0.50[51]. To explore the potential effect of different study populations on heterogeneity we performed such analysis for the whole dataset and in the subgroups of studies classified as high risk and screening.

The statistical software package SPSS 20.0 (SPSS Inc., Chicago, IL, USA) and Meta-Analyst (Tufts Medical Center, Boston, MA, USA) were used for data analysis.

RESULTS

The literature search identified 434 potentially eligible studies that were completely reviewed. The inclusion criteria were met by 48 studies (Table S1 online). In the case of ARSA, because there was only one publication, we included an additional study in which the maximum gestational age was 26 rather than 24 weeks[39]. The Newcastle–Ottawa scale assessments for the included studies are presented in Table S1.

Definitions of the markers

In all included studies ventriculomegaly was considered to be present if the diameter of the lateral cerebral ventricle was 10 mm or more, increased nuchal fold thickness was present if the thickness was 6 mm or more and the diagnosis of echogenic bowel required that this was of equal echogenicity to that of bone. The diagnosis of mild hydronephrosis was based on a minimum anteroposterior diameter of the renal pelvis, which varied between studies from 3 mm to 4 or 5 mm. The definitions of short femur, short humerus and hypoplastic nasal bone were based on a cut-off of the respective bone length as a function of gestational age or biparietal diameter, and the cut-offs differed between studies.

Screening performance of sonographic markers for trisomy 21

Screening performances of sonographic markers for trisomy 21 are presented in Tables 1-10 and Figure 1. The pooled estimates of detection rate, false-positive rate and positive and negative LR for trisomy 21 for each marker are summarized in Table 11.

Table 1. Performance of intracardiac echogenic focus in screening for trisomy 21
StudyTypeTrisomy 21EuploidLR+ (95% CI)LR– (95% CI)
n/NDR (95% CI) (%)n/NFPR (95% CI) (%)
  1. Only the first author of each study is given. DR, detection rate; FPR, false-positive rate; HR, high risk; LR+, positive likelihood ratio; LR−, negative likelihood ratio; Sc, screened.

Manning 19989HR3/1717.6 (6.2–41.0)21/8872.4 (1.6–3.6)7.45 (2.46–22.63)0.84 (0.68–1.05)
Sohl 199910HR13/5523.6 (14.4–36.4)150/26395.7 (4.9–6.6)4.16 (2.52–6.86)0.81 (0.70–0.94)
Wax 200013HR2/728.6 (8.2–64.1)25/7723.2 (2.2–4.7)8.82 (2.57–30.28)0.74 (0.46–1.18)
Winter 200014HR16/5330.2 (19.5–43.5)147/31924.6 (3.9–5.4)6.56 (4.23–10.17)0.73 (0.61–0.87)
Sacco 200736HR3/933.3 (12.1–64.6)45/9654.7 (3.5–6.2)7.15 (2.72–18.80)0.70 (0.44–1.11)
Vergani 200841HR1/244.2 (0.7–20.2)18/11291.6 (1.0–2.5)2.61 (0.36–18.79)0.97 (0.90–1.06)
Bottalico 200943HR4/1233.3 (13.8–60.9)30/6284.8 (3.4–6.7)6.98 (2.92–16.71)0.70 (0.47–1.05)
Prefumo 200116Sc2/825.0 (7.2–59.1)239/76883.1 (2.7–3.5)8.04 (2.41–26.88)0.77 (0.52–1.16)
Coco 200425Sc3/1127.3 (9.8–56.6)476/12 6483.8 (3.5–4.1)7.25 (2.75–19.10)0.76 (0.53–1.09)
Schluter 200530Sc27/7337.0 (26.8–48.5)951/16 8915.6 (5.3–6.0)6.60 (4.87–8.97)0.67 (0.56–0.80)
Weisz 200738Sc3/1225.0 (8.9–53.2)104/23204.5 (3.7–5.4)5.58 (2.06–15.13)0.79 (0.57–1.09)
Aagaard-Tillery 200942Sc15/5328.3 (18.0–41.6)345/77254.5 (4.0–5.0)6.34 (4.08–9.85)0.75 (0.63–0.89)
Shanks 200944Sc34/21815.6 (11.4–21.0)2223/62 1113.6 (3.4–3.7)4.36 (3.19–5.95)0.88 (0.83–0.93)
Huang 201046Sc7/2528.0 (14.3–47.6)237/70933.3 (3.0–3.8)8.38 (4.42–15.91)0.75 (0.58–0.95)
Analysis: total       
  Pooled estimate 133/57724.4 (20.9–28.2)5011/126 6883.9 (3.4–4.5)5.83 (5.02–6.77)0.80 (0.75–0.86)
  Heterogeneity       
    Model Fixed effectsRandom effectsFixed effectsRandom effects
    I2 0.4510.946−0.2760.564
    Q 21.874222.6679.40427.515
    P 0.057< 0.00010.7420.011
Analysis: high risk     
  Pooled estimate 42/17725.8 (19.6–33.1)436/10 2123.7 (2.8–4.8)5.82 (4.42–7.66)0.82 (0.73–0.93)
  Heterogeneity     
    Model Fixed effectsRandom effectsFixed effectsRandom effects
    I2 0.1710.878−0.3810.640
    Q 6.03140.8863.62013.893
    P 0.420< 0.00010.7280.031
Analysis: screened     
  Pooled estimate 91/40025.8 (18.1–35.5)4575/116 4764.0 (3.3–4.7)5.83 (4.88–6.97)0.78 (0.70–0.86)
  Heterogeneity     
    Model Random effectsRandom effectsFixed effectsRandom effects
    I2 0.6790.9720.1350.575
    Q 15.599172.6815.78311.771
    P 0.016< 0.00010.4480.067
Table 2. Performance of ventriculomegaly in screening for trisomy 21
StudyTypeDefinitionTrisomy 21EuploidLR+ (95% CI)LR– (95% CI)
n/NDR (95% CI) (%)n/NFPR (95% CI) (%)
  1. Only the first author of each study is given. DR, detection rate; FPR, false-positive rate; HR, high risk; LR+, positive likelihood ratio; LR−, negative likelihood ratio; Sc, screened; V, lateral cerebral ventricle diameter.

Nyberg 19955HRV ≥ 10 mm1/185.6 (1.0–25.8)0/2320.0 (0.0–1.6)0.94 (0.84–1.06)
Deren 19988HRV 10–15 mm2/355.7 (1.6–18.6)4/36740.2 (0.0–0.3)52.49 (9.94–277.26)0.94 (0.87–1.02)
Sohl 199910HRV 10–15 mm3/555.5 (1.9–14.9)13/26390.5 (0.3–0.8)11.07 (3.25–37.76)0.95 (0.89–1.01)
Wax 200013HRV ≥ 10 mm2/728.6 (8.2–64.1)1/7720.1 (0.0–0.1)220.57 (22.51–2161.20)0.72 (0.45–1.14)
Aagaard-Tillery 200942ScV ≥ 10 mm3/545.6 (1.9–15.1)17/77670.2 (0.1–0.4)25.38 (7.66–84.09)0.95 (0.89–1.01)
Analysis: total        
  Pooled estimate  11/1697.5 (4.2–12.9)35/15 0840.2 (0.1–0.4)27.52 (13.61–55.68)0.94 (0.91–0.98)
  Heterogeneity        
    Model  Fixed effectsRandom effectsFixed effectsFixed effects
    I2  0.3220.6790.495−0.993
    Q  4.4239.3365.9391.505
    P  0.3520.0530.2040.826
Analysis: high risk      
  Pooled estimate  8/1158.6 (3.9–17.9)18/73170.2 (0.1–0.6)38.24 (9.97–146.64)0.94 (0.90–0.99)
  Heterogeneity      
    Model  Random effectsRandom effectsRandom effectsFixed effects
    I2  0.5050.7500.662−0.341
    Q  4.0377.9885.9121.492
    P  0.2570.0460.1140.684
Table 3. Performance of increased nuchal fold in screening for trisomy 21
StudyTypeTrisomy 21EuploidLR+ (95% CI)LR– (95% CI)
n/NDR (95% CI) (%)n/NFPR (95% CI) (%)
  1. Only the first author of each study is given. DR, detection rate; FPR, false-positive rate; HR, high risk; LR+, positive likelihood ratio; LR−, negative likelihood ratio; Sc, screened.

Bahado-Singh 19951HR4/850.0 (21.5–78.5)9/6401.4 (0.7–2.7)35.56 (13.76–91.86)0.51 (0.25–1.01)
DeVore 19952HR4/3212.5 (5.0–28.1)13/20000.6 (0.4–1.1)19.23 (6.63–55.78)0.88 (0.77–1.00)
Grandjean 19953HR17/4438.6 (25.7–53.4)273/32058.5 (7.6–9.5)4.54 (3.07–6.70)0.67 (0.53–0.85)
Nyberg 19955HR3/1816.7 (5.8–39.2)1/2320.4 (0.1–2.4)38.67 (4.23–353.10)0.84 (0.68–1.03)
Vintzileos 19966HR9/1464.3 (38.8–83.7)6/4061.5 (0.7–3.2)43.50 (17.95–105.40)0.36 (0.18–0.73)
Deren 19988HR5/2917.2 (7.6–34.6)22/36740.6 (0.4–0.9)28.79 (11.71–70.80)0.83 (0.71–0.98)
Tannirandorn 199911HR2/1910.5 (2.9–31.4)51/21142.4 (1.8–3.2)4.36 (1.14–16.64)0.92 (0.79–1.07)
Vergani 199912HR6/2227.3 (13.2–48.2)16/8981.8 (1.1–2.9)15.31 (6.63–35.37)0.74 (0.57–0.96)
Wax 200013HR0/70.0 (0.0–35.4)1/7720.1 (0.0–0.7)0.01.00 (1.00–1.00)
Viora 200117HR3/339.1 (3.1–23.6)8/20690.4 (0.2–0.8)23.51 (6.53–84.69)0.91 (0.82–1.02)
Bahado-Singh 200218HR28/10825.9 (18.6–34.9)42/56190.7 (0.6–1.0)34.69 (22.37–53.78)0.75 (0.67–0.83)
Sacco 200736HR4/944.4 (18.9–73.3)7/9650.7 (0.4–1.5)61.27 (21.69–173.07)0.56 (0.31–1.00)
Vergani 200841HR8/2334.8 (18.8–55.1)16/11181.4 (0.9–2.3)24.30 (11.58–51.02)0.66 (0.49–0.89)
Bottalico 200943HR2/1216.7 (4.7–44.8)4/6280.6 (0.3–1.6)26.17 (5.29–129.40)0.84 (0.65–1.08)
Schluter 200530Sc24/7332.9 (23.2–44.3)142/16 8910.8 (0.7–1.0)39.11 (27.11–56.41)0.68 (0.58–0.80)
Weisz 200738Sc3/1225.0 (8.9–53.2)46/23202.0 (1.5–2.6)12.61 (4.54–35.00)0.77 (0.55–1.06)
Aagaard-Tillery 200942Sc6/3318.2 (8.6–34.4)24/64730.4 (0.3–0.6)49.04 (21.46–112.08)0.82 (0.70–0.97)
Analysis: total       
  Pooled estimate 128/49626.0 (20.3–32.9)681/50 0241.0 (0.5–1.9)23.30 (14.35–37.83)0.80 (0.75–0.86)
  Heterogeneity       
    Model Random effectsRandom effectsRandom effectsRandom effects
    I2 0.5320.9830.8430.526
    Q 32.072879.95695.45131.625
    P 0.010< 0.0001< 0.00010.011
Analysis: high risk     
  Pooled estimate 95/37825.8 (18.7–34.4)469/24 3401.0 (0.5–2.1)21.87 (12.32–38.81)0.81 (0.75–0.88)
  Heterogeneity     
    Model Random effectsRandom effectsRandom effectsRandom effects
    I2 0.5940.9790.8390.536
    Q 29.557560.51274.32825.854
    P 0.005< 0.0001< 0.00010.018
Analysis: screened     
  Pooled estimate 33/11827.6 (18.9–38.3)212/25 6840.9 (0.4–1.9)32.17 (17.04–60.71)0.75 (0.66–0.86)
  Heterogeneity     
    Model Random effectsRandom effectsRandom effectsRandom effects
    I2 0.5890.9790.7910.643
    Q 2.43248.4904.7892.799
    P 0.296< 0.00010.0910.247
Table 4. Performance of echogenic bowel in screening for trisomy 21
StudyTypeTrisomy 21EuploidLR+ (95% CI)LR– (95% CI)
n/NDR (95% CI) (%)n/NFPR (95% CI) (%)
  1. Only the first author of each study is given. DR, detection rate; FPR, false-positive rate; HR, high risk; LR+, positive likelihood ratio; LR−, negative likelihood ratio; Sc, screened.

DeVore 19952HR6/3218.8 (8.8–35.3)31/20001.6 (1.1–2.2)12.10 (5.43–26.96)0.83 (0.70–0.98)
Nyberg 19955HR1/185.6 (1.0–25.8)5/2322.2 (0.9–4.9)2.58 (0.32–20.90)0.97 (0.86–1.08)
Vintzileos 19966HR0/140.0 (0.0–21.5)4/4061.0 (0.4–2.5)1.01 (1.00–1.02)
Vergani 199912HR0/220.0 (0.0–14.9)7/8980.8 (0.4–1.6)1.01 (1.00–1.01)
Sohl 199910HR9/5516.4 (8.9–28.3)63/26392.4 (1.9–3.0)6.86 (3.60–13.07)0.86 (0.76–0.96)
Wax 200013HR1/714.3 (2.6–51.3)4/7720.5 (0.2–1.3)27.57 (3.51–216.56)0.86 (0.64–1.17)
Bahado-Singh 200218HR23/10821.3 (14.6–29.9)116/56192.1 (1.7–2.5)10.32 (6.88–15.46)0.80 (0.73–0.89)
Sacco 200736HR0/90.0 (0.0–29.9)12/9651.2 (0.7–2.2)1.01 (1.01–1.02)
Vergani 200841HR3/2412.5 (4.3–31.0)9/11290.8 (0.4–1.5)15.68 (4.53–54.32)0.88 (0.76–1.03)
Bottalico 200943HR2/1216.7 (4.7–44.8)10/6281.6 (0.9–2.9)10.47 (2.56–42.73)0.85 (0.66–1.09)
Schluter 200530Sc13/7317.8 (10.7–28.1)252/16 8911.5 (1.3–1.7)11.94 (7.18–19.84)0.83 (0.75–0.93)
Weisz 200738Sc1/128.3 (1.5–35.4)5/23200.2 (0.1–0.5)38.67 (4.88–306.65)0.92 (0.78–1.09)
Aagaard-Tillery 200942Sc8/5514.5 (7.6–26.2)40/77780.5 (0.4–0.7)28.28 (13.89–57.60)0.86 (0.77–0.96)
Analysis: total       
  Pooled estimate 67/44116.7 (13.4–20.7)558/42 2771.1 (0.8–1.5)11.44 (9.05–14.47)0.90 (0.86–0.94)
  Heterogeneity       
    Model Fixed effectsRandom effectsFixed effectsRandom effects
    I2 −0.2200.8940.2970.526
    Q 9.014103.58115.64623.223
    P 0.702< 0.00010.2080.026
Analysis: high risk     
  Pooled estimate 45/30117.1 (13.1–22.1)261/15 2881.5 (1.1–1.9)9.50 (7.13–12.68)0.91 (0.86–0.96)
  Heterogeneity     
    Model Fixed effectsRandom effectsFixed effectsRandom effects
    I2 0.0190.711−0.2360.574
    Q 8.15827.6466.47518.782
    P 0.5180.0010.6920.027
Analysis: screened     
  Pooled estimate 22/14015.9 (10.7–23.0)297/26 9890.6 (0.2–1.6)19.07 (9.21–39.50)0.86 (0.80–0.92)
  Heterogeneity     
    Model Fixed effectsRandom effectsRandom effectsFixed effects
    I2 −0.3070.9820.774−0.138
    Q 0.76555.5044.4170.879
    P 0.682< 0.00010.1100.644
Table 5. Performance of mild hydronephrosis in screening for trisomy 21
StudyTypeRenal pelvisTrisomy 21EuploidLR+ (95% CI)LR– (95% CI)
n/NDR (95% CI) (%)n/NFPR (95% CI) (%)
  1. Only the first author of each study is given. AP, anteroposterior diameter; DR, detection rate; FPR, false-positive rate; HR, high risk; LR+, positive likelihood ratio; LR−, negative likelihood ratio; Sc, screened.

DeVore 19952HRAP ≥ 4 mm6/3218.8 (8.9–35.3)26/20001.3 (0.9–1.9)14.42 (6.38–32.62)0.82 (0.70–0.97)
Nyberg 19955HRAP ≥ 4 mm3/1816.7 (5.8–39.2)5/2322.2 (0.9–4.9)7.73 (2.01–29.79)0.85 (0.69–1.05)
Vintzileos 19966HRAP ≥ 4 mm4/1428.6 (11.7–54.7)16/4063.9 (2.4–6.3)7.25 (2.78–18.89)0.74 (0.53–1.04)
Deren 19988HRAP ≥ 4 mm1/342.9 (0.5–14.9)22/36740.6 (0.4–0.9)4.91 (0.68–35.41)0.98 (0.92–1.04)
Sohl 199910HRAP ≥ 4 mm1/551.8 (0.3–9.6)42/26391.6 (1.2–2.1)1.14 (0.16–8.15)1.00 (0.96–1.04)
Vergani 199912HRAP ≥ 4 mm4/2218.2 (7.3–38.5)18/8982.0 (1.3–3.2)9.07 (3.35–24.59)0.84 (0.69–1.02)
Wax 200013HRAP ≥ 4 mm1/714.3 (2.6–51.3)14/7721.8 (1.1–3.0)7.88 (1.19–52.01)0.87 (0.65–1.18)
Viora 200117HRAP ≥ 4 mm1/333.0 (0.5–15.3)26/20691.3 (0.9–1.8)2.41 (0.34–17.25)0.98 (0.92–1.04)
Sacco 200736HRAP ≥ 4 mm2/922.2 (6.3–54.7)19/9652.0 (1.3–3.1)11.29 (3.07–41.45)0.79 (0.56–1.13)
Vergani 200841HRAP ≥ 4 mm4/2416.7 (6.7–35.9)11/11291.0 (0.5–1.7)17.11 (5.86–49.90)0.84 (0.70–1.01)
Bottalico 200943HRAP ≥ 4 mm3/1225.0 (8.9–53.2)9/6281.4 (0.8–2.7)17.44 (5.39–56.50)0.76 (0.55–1.06)
Coco 200529ScAP ≥ 4 mm2/119.1 (1.6–37.7)364/12 6482.9 (2.6–3.2)6.32 (1.80–22.22)0.84 (0.64–1.11)
Schluter 200530ScAP ≥ 4 mm15/7320.5 (12.9–31.2)355/16 8912.1 (1.9–2.3)9.78 (6.16–15.23)0.81 (0.72–0.91)
Weisz 200738ScAP ≥ 5 mm0/120.0 (0.0–24.3)27/23201.2 (0.8–1.7)1.01 (1.01–1.02)
Aagaard-Tillery 200942ScAP ≥ 3 mm4/557.3 (2.9–17.3)103/77771.3 (1.1–1.6)5.49 (2.10–14.38)0.94 (0.87–1.01)
Carbone 201148ScAP ≥ 4 mm23/21810.6 (7.1–15.3)1213/61 7302.0 (1.9–2.1)5.37 (3.63–7.93)0.91 (0.87–0.96)
Analysis: total        
  Pooled estimate  74/62913.9 (11.2–17.2)2270/116 7781.7 (1.4–2.0)7.63 (6.11–9.51)0.92 (0.89–0.96)
  Heterogeneity        
    Model  Fixed effectsRandom effectsFixed effectsRandom effects
    I2  0.3890.8970.1820.569
    Q  22.910135.85417.11732.492
    P  0.086< 0.00010.3120.006
Analysis: high risk
  Pooled estimate  30/26015.9 (11.3–21.9)208/15 4121.5 (1.2–2.0)9.63 (6.67–13.92)0.93 (0.88–0.98)
  Heterogeneity      
    Model  Fixed effectsRandom effectsFixed effectsRandom effects
    I2  0.3840.7840.1390.519
    Q  14.59941.65110.45318.714
    P  0.147< 0.00010.4020.044
Analysis: screened      
  Pooled estimate  44/36912.6 (7.8–19.7)2062/101 3661.9 (1.5–2.3)6.68 (5.07–8.81)0.91 (0.87–0.96)
  Heterogeneity      
    Model  Random effectsRandom effectsFixed effectsRandom effects
    I2  0.5920.9590.2920.511
    Q  7.35873.4794.2386.132
    P  0.118< 0.0000.3750.190
Table 6. Performance of short humerus in screening for trisomy 21
StudyTypeDefinitionTrisomy 21EuploidLR+ (95% CI)LR– (95% CI)
n/NDR (%) (95% CI)n/NFPR (%) (95% CI)
  1. Only the first author of each study is given. BPD, biparietal diameter; DR, detection rate; FPR, false-positive rate; GA, gestational age; H, humerus; HR, high risk; LR+, positive likelihood ratio; LR−, negative likelihood ratio; MoM, multiples of the median; Sc, screened.

Vintzileos 19967HRH ≤ 0.89 MoM for BPD10/2245.5 (26.9–65.3)49/4939.9 (7.6–12.9)4.57 (2.69–7.76)0.61 (0.41–0.89)
Wax 200013HRH ≤ 0.89 MoM for BPD1/714.3 (2.6–51.3)3/7720.4 (0.1–1.1)36.76 (4.34–311.62)0.86 (0.64–1.16)
Viora 200117HRH ≤ 0.90 MoM for BPD9/3327.3 (15.1–44.2)258/206912.5 (11.1–14.0)2.19 (1.24–3.86)0.83 (0.67–1.03)
Bahado-Singh 200218HRH < 0.90 MoM for BPD30/10827.8 (20.2–36.9)320/56195.7 (5.1–6.3)4.88 (3.53–6.73)0.77 (0.68–0.86)
Bottalico 200943HRH ≤ 0.89 MoM for GA2/1216.7 (4.7–44.8)8/6281.3 (0.7–2.5)13.08 (3.10–55.24)0.84 (0.66–1.09)
Schluter 200530ScH ≤ 0.92 MoM for BPD47/7364.4 (52.9–74.4)1956/16 89111.6 (11.1–12.1)5.56 (4.66–6.63)0.40 (0.30–0.55)
Aagaard-Tillery 200942ScH ≤ 0.90 MoM for BPD3/2611.5 (4.0–29.0)89/38402.3 (1.9–2.8)4.98 (1.68–14.72)0.91 (0.79–1.04)
Analysis: total        
  Pooled estimate  102/28130.3 (17.1–47.9)2683/30 3124.6 (2.8–7.4)4.81 (3.49–6.62)0.74 (0.63–0.88)
  Heterogeneity        
    Model  Random effectsRandom effectsRandom effectsRandom effects
    I2  0.8620.9890.6580.799
    Q  36.273450.25214.60024.926
    P  < 0.0001< 0.00010.024< 0.0001
Analysis: high risk      
  Pooled estimate  52/18229.0 (22.8–36.2)638/95814.2 (2.3–7.7)4.80 (2.80–8.23)0.78 (0.72–0.86)
  Heterogeneity      
    Model  Fixed effectsRandom effectsRandom effectsFixed effects
    I2  0.3240.9810.754−0.039
    Q  4.438154.94412.2052.888
    P  0.350< 0.00010.0160.577
Analysis: screened      
  Pooled estimate  50/9934.0 (3.8–87.1)2045/20 7315.3 (1.0–23.0)5.54 (4.66–6.59)0.61 (0.28–1.35)
  Heterogeneity      
    Model  Random effectsRandom effectsRandom effectsRandom effects
    I2  
    Q  15.832241.5750.03922.029
    P  < 0.0001< 0.00010.844< 0.0001
Table 7. Performance of short femur in screening for trisomy 21
StudyTypeDefinitionTrisomy 21EuploidLR+ (95% CI)LR– (95% CI)
n/NDR (95% CI) (%)n/NFPR (95% CI) (%)
  1. Only the first author of each study is given. BPD, biparietal diameter; DR, detection rate; F, femur; FPR, false-positive rate; GA, gestational age; HR, high risk; LR+, positive likelihood ratio; LR−, negative likelihood ratio; MoM, multiples of the median; Sc, screened.

Grandjean 19954HRF ≤ 0.91 MoM for BPD15/3444.1 (28.9–60.6)495/276317.9 (16.5–19.4)2.46 (1.67–3.63)0.68 (0.51–0.92)
Nyberg 19955HRF ≤ 0.91 MoM for BPD5/1827.8 (12.5–50.9)14/2326.0 (3.6–9.9)4.60 (1.87–11.34)0.77 (0.58–1.03)
Vintzileos 19967HRF ≤ 0.88 MoM for BPD5/2222.7 (10.1–43.4)50/49310.1 (7.8–13.1)2.24 (0.99–5.06)0.86 (0.68–1.08)
Sohl 199910HRF ≤ 0.91 MoM for BPD9/5516.4 (8.9–28.3)42/26391.6 (1.2–2.1)10.28 (5.27–20.06)0.85 (0.76–0.96)
Wax 200013HRF ≤ 0.91 MoM for BPD1/714.3 (2.6–51.3)2/7720.3 (0.1–0.9)55.14 (5.63–540.30)0.86 (0.64–1.16)
Viora 200117HRF ≤ 0.91 MoM for BPD10/3330.3 (17.4–47.3)213/206910.3 (9.1–11.7)2.94 (1.73–5.02)0.78 (0.62–0.97)
Bahado-Singh 200218HRF < 0.90 MoM for BPD30/10827.8 (20.2–36.9)503/56199.0 (8.2–9.7)3.10 (2.26–4.25)0.79 (0.71–0.89)
Vergani 200841HRF ≤ 0.91 MoM for GA4/2416.7 (6.7–35.9)145/111013.1 (11.2–15.2)1.28 (0.52–3.16)0.96 (0.80–1.15)
Bottalico 200943HRF ≤ 0.91 MoM for GA2/1216.7 (4.7–44.8)7/6281.1 (0.5–2.3)14.95 (3.46–64.64)0.84 (0.65–1.09)
Schluter 200530ScF ≤ 0.93 MoM for BPD46/7363.0 (51.6–73.2)2534/16 89115.0 (14.5–15.6)4.20 (3.51–5.03)0.44 (0.32–0.59)
Weisz 200738ScF < 5th percentile for GA1/128.3 (1.5–35.4)111/23204.8 (4.0–5.7)1.74 (0.26–11.48)0.96 (0.81–1.14)
Aagaard-Tillery 200942ScF ≤ 0.91 MoM for BPD16/5628.6 (18.4–41.5)514/77616.6 (6.1–7.2)4.31 (2.83–6.58)0.77 (0.65–0.90)
Analysis: total        
  Pooled estimate  144/45427.7 (19.3–38.1)4630/43 2976.4 (4.7–8.8)3.72 (2.79–4.97)0.80 (0.73–0.88)
  Heterogeneity        
    Model  Random effectsRandom effectsRandom effectsRandom effects
    I2  0.7790.9890.7060.630
    Q  45.344870.63833.97726.993
    P  < 0.0001< 0.0001< 0.00010.005
Analysis: high risk      
  Pooled estimate  81/31326.7 (21.9–32.0)1471/16 3255.7 (3.7–8.7)3.79 (2.47–5.83)0.83 (0.78–0.88)
  Heterogeneity      
    Model  Fixed effectsRandom effectsRandom effectsFixed effects
    I2  0.3470.9830.76−0.245
    Q  10.714407.43929.1325.623
    P  0.218< 0.0001< 0.00010.689
Analysis: screened      
  Pooled estimate  63/14134.1 (11.8–66.8)3159/26 9727.9 (3.8–15.7)4.19 (3.55–4.94)0.70 (0.48–1.02)
  Heterogeneity      
    Model  Random effectsRandom effectsFixed effectsRandom effects
    I2  0.9480.998−0.1740.951
    Q  19.253446.9950.85220.472
    P  < 0.0001< 0.00010.653< 0.0001
Table 8. Performance of aberrant right subclavian artery in screening for trisomy 21
StudyTrisomy 21EuploidLR+ (95% CI)LR– (95% CI)
n/NDR (95% CI) (%) n/NFPR (95% CI) (%)
  1. Only the first author of each study is given. DR, detection rate; FPR, false-positive rate; LR+, positive likelihood ratio; LR−, negative likelihood ratio.

Zalel 2008393/837.5 (13.7–69.4)13/9241.4 (0.8–2.4)26.65 (9.38–75.77)0.63 (0.37–1.08)
Borenstein 2010458/2828.6 (15.3–47.1)14/9321.5 (0.9–2.5)19.02 (8.69–41.62)0.73 (0.57–0.92)
Pooled estimate11/3630.7 (17.8–47.4)27/18561.5 (1.0–2.1)21.48 (11.48–40.19)0.71 (0.57–0.88)
Heterogeneity      
  ModelFixed effectsFixed effectsFixed effectsFixed effects
  I20000
  Q0.2320.0290.2570.203
  P0.6300.8640.6130.653
Table 9. Performance of absent nasal bone in screening for trisomy 21
StudyTypeTrisomy 21EuploidLR+ (95% CI)LR– (95% CI)
n/NDR (95% CI) (%) n/NFPR (95% CI) (%)
  1. Only the first author of each study is given. DR, detection rate; FPR, false-positive rate; HR, high risk; LR+, positive likelihood ratio; LR−, negative likelihood ratio; Sc, screened.

Bromley 200219HR6/1637.5 (18.5–61.4)1/2230.4 (0.1–2.5)83.63 (10.71–653.06)0.63 (0.43–0.92)
Cicero 200323HR11/3432.4 (19.1–49.2)6/9820.6 (0.3–1.3)52.95 (20.81–134.76)0.68 (0.54–0.86)
Vintzileos 200324HR12/2941.4 (25.5–59.3)0/1020.0 (0.0–3.6)0.59 (0.43–0.80)
Cusick 200426HR1/425.0 (4.6–69.9)0/4220.0 (0.0–0.9)0.75 (0.43–1.32)
Tran 200531HR11/3135.5 (21.1–53.1)1/1360.7 (0.1–4.1)48.26 (6.47–360.02)0.65 (0.50–0.84)
Viora 200532HR10/1855.6 (33.7–75.4)2/4170.5 (0.1–1.7)115.83 (27.36–490.37)0.45 (0.27–0.75)
Cusick 200734HR1/119.1 (1.6–37.7)3/3710.8 (0.3–2.4)11.24 (1.27–99.68)0.92 (0.76–1.11)
Gianferrari 200735HR10/2147.6 (28.3–67.6)1/25150.0 (0.0–0.2)1197.62 (160.43–8940.55)0.52 (0.35–0.79)
Persico 200840HR7/2626.9 (13.7–46.1)0/1350.0 (0.0–2.8)0.73 (0.58–0.92)
Odibo 200633Sc5/2222.7 (10.1–43.4)13/24460.5 (0.3–0.9)42.76 (16.67–109.70)0.78 (0.62–0.97)
Analysis: total       
  Pooled estimate 74/21236.1 (29.8–43.0)27/77490.5 (0.3–0.7)66.75 (40.62–109.69)0.71 (0.65–0.78)
  Heterogeneity       
    Model Fixed effectsFixed effectsFixed effectsFixed effects
    I2 0.1890.0780.3870.484
    Q 9.8688.67413.04615.506
    P 0.3610.4680.1610.078
Analysis: high risk     
  Pooled estimate 69/19037.5 (30.8–44.8)14/53030.5 (0.3–0.8)79.23 (44.16–142.15)0.67 (0.59–0.77)
  Heterogeneity     
    Model Fixed effectsFixed effectsFixed effectsRandom effects
    I2 0.1320.1860.4100.530
    Q 8.0658.60411.85814.887
    P 0.4270.3770.1580.061
Table 10. Performance of absent or hypoplastic nasal bone in screening for trisomy 21
StudyTypeDefinition (absent or hypoplastic)Trisomy 21EuploidLR+ (95% CI)LR– (95% CI)
n/NDR (95% CI) (%)n/NFPR (95% CI) (%)
  1. Only the first author of each study is given. BPD, biparietal diameter; DR, detection rate; FPR, false-positive rate; GA, gestational age; HR, high risk; LR+, positive likelihood ratio; LR−, negative likelihood ratio; MoM, multiples of the median; NBL, nasal bone length; Sc, screened.

Bromley 200219HRBPD/NBL ≥ 1111/1668.8 (44.4–85.8)11/2234.9 (2.8–8.6)13.94 (7.17–27.08)0.33 (0.16–0.68)
Cicero 200323HRNBL < 2.5 mm21/3461.8 (45.0–76.1)12/9821.2 (0.7–2.1)50.50 (27.15–94.09)0.39 (0.25–0.59)
Bunduki 200322HRNBL < 5th percentile for GA13/2259.1 (38.7–76.7)82/16005.1 (4.2–6.3)11.53 (7.68–17.32)0.43 (0.26–0.71)
Cusick 200426HRNBL < 3 mm4/4100.0 (51.0–100.0)0/4220.0 (0.0–0.9)0.00
Gámez 200427HRNBL < 2.5 mm5/5100.0 (56.6–100.0)34/18991.8 (1.3–2.5)55.85 (40.03–77.93)0.00
Odibo 200428HRBPD/NBL ≥ 126/1637.5 (18.5–61.4)19/5083.7 (2.4–5.8)10.03 (4.64–21.68)0.65 (0.44–0.95)
Tran 200531HRBPD/NBL ≥ 1815/3148.4 (32.0–65.2)5/1363.7 (1.6–8.3)13.16 (5.17–33.49)0.54 (0.38–0.76)
Viora 200532HRNBL < 2.5 mm14/1877.8 (54.8–91.0)3/4170.7 (0.2–2.1)108.11 (34.09–342.88)0.22 (0.09–0.53)
Cusick 200734HRBPD/NBL ≥ 117/1163.6 (35.4–84.8)16/3714.3 (2.7–6.9)14.76 (7.66–28.41)0.38 (0.17–0.83)
Gianferrari 200735HRNBL < 0.75 MoM18/2185.7 (65.4–95.0)74/25152.9 (2.4–3.7)29.13 (21.92–38.71)0.15 (0.05–0.42)
Sooklim 201047HRBPD/NBL ≥ 123/1030.0 (10.8–60.3)2/3860.5 (0.1–1.9)57.90 (10.84–309.25)0.70 (0.47–1.06)
Odibo 200633ScBPD/NBL ≥ 129/2240.9 (23.3–61.3)161/24236.6 (5.7–7.7)6.16 (3.65–10.40)0.63 (0.45–0.90)
Analysis: total        
  Pooled estimate  126/21059.8 (48.9–69.9)419/11 8822.8 (1.9–4.0)23.27 (14.23–38.06)0.46 (0.36–0.58)
  Heterogeneity        
    Model  Random effectsRandom effectsRandom effectsRandom effects
    I2  0.5470.9160.8780.537
    Q  22.069118.43682.04321.609
    P  0.024< 0.0001< 0.00010.028
Analysis: high risk      
  Pooled estimate  117/18862.0 (50.7–72.2)258/94592.5 (1.7–3.6)26.31 (16.57–41.77)0.43 (0.33–0.56)
  Heterogeneity      
    Model  Random effectsRandom effectsRandom effectsRandom effects
    I2  0.5310.8630.8420.542
    Q  19.20865.48356.86719.655
    P  0.038< 0.0001< 0.00010.033
Figure 1.

Summary receiver–operating characteristics curves with detection rate and false-positive rate (FPR) of sonographic markers of trisomy 21: (a) intracardiac echogenic foci, (b) ventriculomegaly, (c) nuchal fold thickness, (d) echogenic bowel, (e) hydronephrosis, (f) short humerus, (g) short femur, (h) aberrant right subclavian artery, (i) absent nasal bone and (j) absent or hypoplastic nasal bone.

Table 11. Pooled estimates of detection rate (DR), false positive rate (FPR) and positive and negative likelihood ratios (LR+ and LR−) of sonographic markers for trisomy 21 and estimated likelihood ratio (LR) of individual isolated markers
MarkerDR (95% CI) (%)FPR (95% CI) (%)LR+ (95% CI)LR– (95% CI)LR isolated markera
  1. a

    Derived by multiplying the positive LR for the given marker by the negative LR of each of all other markers, except for short humerus. ARSA, aberrant right subclavian artery; NB, nasal bone.

Intracardiac echogenic focus24.4 (20.9–28.2)3.9 (3.4–4.5)5.83 (5.02–6.77)0.80 (0.75–0.86)0.95
Ventriculomegaly7.5 (4.2–12.9)0.2 (0.1–0.4)27.52 (13.61–55.68)0.94 (0.91–0.98)3.81
Increased nuchal fold26.0 (20.3–32.9)1.0 (0.5–1.9)23.30 (14.35–37.83)0.80 (0.74–0.85)3.79
Echogenic bowel16.7 (13.4–20.7)1.1 (0.8–1.5)11.44 (9.05–14.47)0.90 (0.86–0.94)1.65
Mild hydronephrosis13.9 (11.2–17.2)1.7 (1.4–2.0)7.63 (6.11–9.51)0.92 (0.89–0.96)1.08
Short humerus30.3 (17.1–47.9)4.6 (2.8–7.4)4.81 (3.49–6.62)0.74 (0.63–0.88)0.78
Short femur27.7 (19.3–38.1)6.4 (4.7–8.8)3.72 (2.79–4.97)0.80 (0.73–0.88)0.61
ARSA30.7 (17.8–47.4)1.5 (1.0–2.1)21.48 (11.48–40.19)0.71 (0.57–0.88)3.94
Absent or hypoplastic NB59.8 (48.9–69.9)2.8 (1.9–4.0)23.27 (14.23–38.06)0.46 (0.36–0.58)6.58

Estimation of combined likelihood ratio of multiple markers for trisomy 21

The LR for trisomy 21 of individual isolated markers is given in the last column of Table 11. This was derived by multiplying the positive LR for the given marker by the negative LR of each of all other markers, except for short humerus.

The same approach can be used when any combination of two or more markers is detected. For example the positive LRs for mild hydronephrosis and ventriculomegaly are 7.63 and 27.52, respectively. When the ultrasound examination detects these two markers the combined positive LR is 209.98 (7.63 × 27.52) and this must be multiplied by the combined negative LR of all other markers that were not present (0.80 × 0.80 × 0.90 × 0.80 × 0.71 × 0.46 = 0.15) to derive a combined LR of 31.50 (209.98 × 0.15). A spreadsheet to automatically perform these calculations is available online (Appendix S1).

Incidence of trisomy 21 and euploid fetuses in the absence of sonographic markers

The literature search identified 12 studies that examined multiple sonographic markers and reported on the incidence of no markers in trisomy 21 and euploid fetuses (Table 12). In the absence of sonographic markers, the pooled incidences of trisomy 21 and euploid fetuses were 30.9% (95% CI, 23.1–39.9%) and 88.1% (95% CI, 85.3–90.4%), respectively. The LR for trisomy 21 in the absence of sonographic markers was 0.37 (95% CI, 0.29–0.47). Consequently, in the absence of all markers the risk for trisomy 21 was reduced by 2.7-fold.

Table 12. Studies reporting data on the absence of multiple markers in trisomy 21 and euploid fetuses
StudyTypeSonographic markersTrisomy 21 – no markerEuploid – no markerLR+ (95% CI)
n/N% (95% CI)n/N% (95% CI)
  1. Only the first author of each study is given. Cardiac markers are pericardial effusion, tricuspid regurgitation, ventricular disproportion, ventricular septal defect. CPC, choroid plexus cyst; EB, echogenic bowel; EF, intracardiac echogenic focus; HR, high risk; hydro, mild hydronephrosis; LR+, positive likelihood ratio; NBH, nasal bone hypoplasia; NF, increased nuchal fold; Sc, screened; SF, short femur; SH, short humerus; SUA, single umbilical artery; ventric, ventriculomegaly.

Nyberg 19955HRNF, EB, hydro, SF, ventric9/1850.0 (29.0–71.0)350/37493.6 (90.6–95.7)0.53 (0.34–0.85)
Vergani 199912HRNF, EF, EB, hydro, SF, SH10/2245.5 (26.9–65.3)870/89896.9 (95.5–97.8)0.47 (0.30–0.74)
Wax 200013HRNF, EF, EB, hydro, SF, SH, CPC, SUA, clinodactyly2/728.6 (8.2–64.1)688/77289.1 (86.7–91.1)0.32 (0.10–1.04)
Nyberg 200115HRNF, EF, EB, hydro, SF, SH58/18631.2 (25.0–38.2)7541/872886.4 (85.7–87.1)0.36 (0.29–0.45)
Viora 200117HRNF, EB, hydro, SF, SH, CPC10/3330.3 (17.4–47.3)1497/206972.4 (70.4–74.2)0.42 (0.25–0.70)
Bromley 200219HRNF, EF, EB, hydro, SF, SH32/16419.5 (14.2–26.3)575/65687.7 (84.9–90.9)0.22 (0.16–0.30)
Vintzileos 200221HRNF, EF, EB, hydro, SF, SH, CPC, SUA, clinodactyly, sandal gap, short ear7/5313.2 (6.6–24.8)3291/370088.9 (87.9–89.9)0.15 (0.07–0.30)
Cicero 200323HRNF, EF, EB, hydro, SF, SH, CPC, NBH, clinodactyly, sandal gap4/3411.8 (4.7–26.6)694/98270.7 (67.8–73.4)0.17 (0.07–0.42)
Sacco 200736HRNF, EF, EB, hydro, SF, cardiac markers2/922.2 (6.3–54.7)915/96594.8 (93.2–96.1)0.23 (0.07–0.80)
Smith-Bindman 200737ScNF, EF, EB, hydro, SF, SH, CPC115/24546.9 (40.8–53.2)7467/870785.8 (85.0–86.5)0.55 (0.48–0.63)
Weisz 200738ScNF, EF, EB, hydro, SF6/1250.0 (25.4–74.6)2013/232086.8 (85.3–88.1)0.58 (0.33–1.02)
Aagaard-Tillery 200942ScNF, EF, EB, hydro, SF, SH21/5935.6 (24.6–48.3)6775/778387.0 (86.3–87.8)0.41 (0.29–0.58)
Analysis: total       
  Pooled estimate  276/84230.9 (23.1–39.9)32 676/37 95488.1 (85.3–90.4)0.37 (0.29–0.47)
  Heterogeneity       
    Model  Random effectsRandom effectsRandom effects
    I2  0.8150.9850.795
    Q  54.040660.08948.696
    P  < 0.0001< 0.0001< 0.0001
Analysis: high risk     
  Pooled estimate  134/52626.4 (19.2–35.2)16 421/19 14489.0 (84.2–92.4)0.32 (0.24–0.42)
  Heterogeneity     
    Model  Random effectsRandom effectsRandom effects
    I2  0.6960.9890.676
    Q  23.010636.64421.597
    P  0.003< 0.00010.006
Analysis: screened     
  Pooled estimate  142/31644.1 (36.6–51.8)16 255/18 81086.5 (85.5–87.4)0.52 (0.44–0.62)
  Heterogeneity     
    Model  Random effectsRandom effectsRandom effects
    I2  0.6120.8360.600
    Q  2.5756.1002.500
    P  0.2760.0470.287

In nine studies high-risk pregnancies were examined in specialist units[5, 12, 13, 15, 17, 19, 21, 23, 36] and in three the patients had routine second-trimester ultrasound examination[37, 38, 42]. The LR for trisomy 21 in the absence of sonographic markers in specialist units was 0.32 (95% CI, 0.24–0.42) and in the routine examination studies it was 0.52 (95% CI, 0.44–0.62). Consequently, in the absence of all markers the risk for trisomy 21 was reduced by 3.1-fold and 1.9-fold, respectively.

Most studies examining multiple markers did not include absent or hypoplastic nasal bone and ARSA. The combined negative LR obtained by multiplying the values of the individual markers in Table 11, but excluding short humerus, absent or hypoplastic nasal bone and ARSA was 0.40 (95% CI, 0.29–0.58).

Estimation of combined likelihood ratio in the absence of all markers for trisomy 21

The combined negative LR, obtained by multiplying the values of individual markers in Table 11, was 0.13 (95% CI, 0.05–0.29) when short femur but not short humerus was included and 0.12 (95% CI, 0.06–0.29) when short humerus but not short femur was included. Consequently, in the absence of all markers the risk for trisomy 21 would be reduced by 7.7-fold and 8.3-fold, respectively.

DISCUSSION

The findings of this meta-analysis confirm that the incidence of each of the selected second-trimester sonographic markers is higher in trisomy 21 than in euploid fetuses. The pooled estimate of the positive LR was about 5 for intracardiac echogenic focus and short femur or humerus, about 10 for echogenic bowel and mild hydronephrosis, 20 for increased nuchal fold thickness and ARSA and about 25 for ventriculomegaly and absent or hypoplastic nasal bone. Absence of all markers, apart from short humerus, was associated with a combined negative LR of 0.13 and therefore a 7.7-fold reduction in risk. If assessment for ARSA and absent or hypoplastic nasal bone was not included in the ultrasound examination the negative LR was 0.40, with a consequent 2.5-fold reduction in risk.

Our results on intracardiac echogenic focus were similar to those of a previous meta-analysis that included 11 studies, published between 1995 and 2001, on a total of 51 831 pregnancies, of which 333 had trisomy 21, and that reported that the positive LR was 6.2[52]. However, our findings differ from those of Smith-Bindman et al.[53], who examined multiple markers on a combined total of 1930 fetuses with trisomy 21 and 130 365 unaffected fetuses in 56 articles published between 1980 and 1999. After exclusion of cases with major defects the estimated positive LR was 2.8 for intracardiac echogenic focus, 2.7 for short femur, 7.5 for short humerus, 6.1 for echogenic bowel, 1.9 for mild hydronephrosis and 17 for increased nuchal fold. These LRs are considerably lower than in our analysis. The most likely explanation for this is that the majority of studies were in high-risk pregnancies and the sensitivity of the markers in the meta-analysis of Smith-Bindman et al. was lower than in our analysis, presumably because awareness of the potential importance of these markers and therefore the search for their presence was not as widespread in the 1980s and early 1990s as in later years.

There was high heterogeneity in results between the studies, which presumably reflects the large differences in design and focus, including prospective and retrospective cohort studies or case–control studies, with the scans performed in specialist units or routine ultrasound departments, reporting on either one or multiple markers and using different definitions for the presence of a marker. The problem of high heterogeneity in results was not overcome by subanalysis of data derived from screening studies and those involving examination of high-risk pregnancies. Particularly big differences in LRs were observed for nuchal fold thickness and echogenic bowel, presumably reflecting the subjective nature of these markers and greater susceptibility to allocation bias.

We excluded studies published before 1995 because awareness of the possible importance of markers and consequently the specific search for their presence or absence was limited before this time. We focused on studies reporting within the gestational-age range of 14–24 weeks to minimize the potential effect of gestational age on the incidence of the markers. In the majority of studies the median gestational age of the ultrasound examinations was 17–19 weeks. It remains uncertain whether the relative proportion of trisomic and euploid fetuses with a given marker, and therefore the LR, is constant across the gestational-age range of 14–24 weeks. This is true for categorical variables, such as intracardiac echogenic foci and echogenic bowel and even more so for markers based on fixed measurements, such as increased nuchal fold thickness, mild hydronephrosis and ventriculomegaly.

Some of the studies were in women undergoing routine screening during the second trimester but most studies were performed in patients undergoing amniocentesis because their risk was considered to be high owing to either advanced maternal age or abnormal second-trimester serum biochemistry testing. This is reflected in the high incidence of trisomy 21 in most study populations included in the analysis.

A few of the included studies report results on multiple markers, but the majority examined specifically the value of individual markers. There are no studies that systematically examined the possible interrelationship between markers, and it is therefore assumed that they are independent of each other, apart from short femur and short humerus, which have been shown to be highly correlated in both euploid and trisomic fetuses[42]. Similarly, there is no consistent evidence that the incidence of markers is related to maternal age or the results of second-trimester serum biochemical testing or first-trimester combined testing[42]. Surprisingly, three studies reported that there is no significant association between nuchal translucency (NT) thickness at 11–13 weeks' gestation and second-trimester nuchal fold thickness[54-56]. Another study in euploid fetuses reported a weak but significant association between NT and nuchal fold thickness, with a correlation coefficient of 0.1[57]. The same study found that in cases with increased NT, compared to those with normal NT, there was a higher frequency of echogenic bowel (2.4 vs 0.1%), but not intracardiac echogenic focus, pyelectasia or short femur and short humerus[57]. In contrast, another study reported that the incidence of intracardiac echogenic focus during the second trimester was 2.8-fold higher in fetuses with increased NT compared to those with normal NT[16].

Several studies have reported on the use of ultrasonography to modify the risk of aneuploidy in pregnancies with advanced maternal age or abnormal serum biochemistry[22, 58-60]. It was subsequently suggested that, in the estimation of the post-test odds for trisomy 21 based on ultrasound findings during the second trimester of pregnancy, the pre-test odds, derived from maternal age, second-trimester serum biochemical testing or first-trimester combined testing, could be multiplied by the positive LR of each marker found to be present and the negative LR of each marker looked for but not found[61].

In this meta-analysis the combined negative LR of all markers, including short femur but not short humerus, was 0.13, implying that if a systematic ultrasound examination is carried out and all markers are excluded there is a 7.7-fold reduction in risk. This estimated negative LR is similar to the 0.15 reported in a study in which a very detailed scan, including examination for features such as short ears, sandal gap and clinodactyly, was carried out in high-risk pregnancies[21]. However, such reduction in risk requires considerable expertise in scanning and in three studies in women undergoing routine second-trimester ultrasound examination the combined negative LR was 0.52, with a consequent 1.9-fold reduction in risk[37, 38, 42].

The clinical implications of our findings are that firstly, if a systematic second-trimester ultrasound examination demonstrates the absence of all major defects and markers there is a 7.7-fold reduction in risk for trisomy 21; secondly, the detection of any one of the markers during the scan should stimulate the sonographer to look for all other markers or defects; thirdly, the post-test odds for trisomy 21 is derived by multiplying the pre-test odds by the positive LR for each detected marker and the negative LR for each marker demonstrated to be absent; and fourthly, in the case of most isolated markers, including intracardiac echogenic focus, echogenic bowel, mild hydronephrosis and short femur, there is only a small effect on modifying the pre-test odds.

Further studies are needed to establish reference ranges for each biometric marker and to estimate the effect of gestational age on screening performance. In the era of widespread first-trimester screening and selective termination of most affected fetuses the undertaking of high-quality screening studies may ultimately be impossible. In the interim the data arising from this meta-analysis and their interpretation could form the basis for clinical practice. However, as in the case of fetal NT, it is essential that those performing the second-trimester scan receive appropriate training and certification of competence and subject their results to regular audit.

ACKNOWLEDGMENT

This study was supported by a grant from The Fetal Medicine Foundation (UK Charity No: 1037116).

SUPPORTING INFORMATION ON THE INTERNET

The following supporting information may be found in the online version of this article:

Table S1 Characteristics of studies included in the analysis

Appendix S1 Excel spreadsheet allowing automated calculations of the likelihood ratio for any given combination of presence and absence of markers using pooled estimates from the meta-analysis. Please note that this does not provide confidence intervals for estimates of combined likelihood ratios

Ancillary