Several biomarkers for predicting intrauterine growth restriction (IUGR) have been proposed in recent years. However, the predictive performance of these biomarkers has not been systematically evaluated.
Several biomarkers for predicting intrauterine growth restriction (IUGR) have been proposed in recent years. However, the predictive performance of these biomarkers has not been systematically evaluated.
To determine the predictive accuracy of novel biomarkers for IUGR in women with singleton gestations.
Electronic databases, reference list checking and conference proceedings.
Observational studies that evaluated the accuracy of novel biomarkers proposed for predicting IUGR.
Data were extracted on characteristics, quality and predictive accuracy from each study to construct 2 × 2 tables. Summary receiver operating characteristic curves, sensitivities, specificities and likelihood ratios (LRs) were generated.
A total of 53 studies, including 39 974 women and evaluating 37 novel biomarkers, fulfilled the inclusion criteria. Overall, the predictive accuracy of angiogenic factors for IUGR was minimal (median pooled positive and negative LRs of 1.7, range 1.0–19.8; and 0.8, range 0.0–1.0, respectively). Two small case–control studies reported high predictive values for placental growth factor and angiopoietin-2 only when IUGR was defined as birthweight centile with clinical or pathological evidence of fetal growth restriction. Biomarkers related to endothelial function/oxidative stress, placental protein/hormone, and others such as serum levels of vitamin D, urinary albumin : creatinine ratio, thyroid function tests and metabolomic profile had low predictive accuracy.
None of the novel biomarkers evaluated in this review are sufficiently accurate to recommend their use as predictors of IUGR in routine clinical practice. However, the use of biomarkers in combination with biophysical parameters and maternal characteristics could be more useful and merits further research.
Intrauterine growth restriction (IUGR) is defined as a failure of the fetus to achieve its optimal growth potential and constitutes a major clinical and public health problem, mainly in the developing world. It is considered a heterogeneous syndrome associated with hypertensive disorders of pregnancy, smoking, infection, undernutrition and unexplained factors. IUGR fetuses are at greater risk of perinatal death, birth hypoxia, neonatal complications, impaired neurodevelopment and manifestations of the metabolic syndrome in adult life such as type 2 diabetes, coronary heart disease and hypertension.[4-8]
Although no preventive interventions are available at present, being able to predict IUGR reliably would be valuable because: (i) it would identify fetuses that require early referral to secondary care and closer surveillance (ii) identifying at-risk fetuses would allow specific preventive interventions to be tested (iii) studies of predictors of IUGR could improve our understanding of the biological and pathological mechanisms that cause fetal growth restriction, leading potentially to better interventions (iv) identifying low-risk fetuses would avoid the use of unnecessary interventions, and (v) accurate prediction, and prevention, of IUGR could be an early stage in a public health strategy that aims to avoid the adult consequences of fetal growth restriction.
In the last decade, several biomarkers have been proposed as predictors of IUGR.[9-11] Some have been assessed systematically, including five serum metabolites for prenatal screening of aneuploidy and open neural tube defects, as individual or combined biomarkers. However, to our knowledge, the predictive performance of other biomarkers for IUGR has not been adequately evaluated in a systematic manner.
Therefore, the objective of this systematic review and meta-analysis was to identify, and determine the accuracy of, biomarkers proposed from the year 2000 onwards for the prediction of IUGR in women with singleton gestations.
The systematic review was conducted following a prospectively prepared protocol and reported using the checklist recommended by the STARD initiative for reporting systematic reviews of diagnostic test accuracy.
An initial search was performed in PubMed, Embase, Cinahl, Lilacs and Medion (all from inception to 31 July 2012), and Google Scholar using a combination of keywords and text words related to biomarkers (‘biomarker’, ‘marker’), prediction (‘prediction’, ‘screening’, ‘diagnosis’, ‘accuracy’, ‘sensitivity’, ‘specificity’, ‘likelihood ratio’), and IUGR (‘intrauterine growth restriction’, ‘intrauterine growth retardation’, ‘fetal growth restriction’, ‘fetal growth retardation’, ‘impaired fetal growth’, ‘small for gestational age’, ‘small for date’, ‘small for gestation’). In the initial search, we chose those biomarkers proposed from the year 2000 onwards for the prediction of IUGR. A further computerised search was conducted using keywords and text words for each of the biomarkers identified in the initial search and keywords and text words for IUGR described previously. Congress proceedings of international society meetings of maternal and fetal, and reproductive medicine, as well as international meetings on fetal growth, reference lists of identified studies, textbooks, previously published systematic reviews, and review articles were also searched. In addition, we contacted investigators involved in the field to locate unpublished studies. Language restrictions were not applied.
Studies were included if: (i) they were cohort, cross-sectional or case–control studies that evaluated the accuracy of biomarkers for predicting IUGR or small-for-gestational-age (SGA) infants in women with singleton gestations at any level of risk (ii) the biological samples were collected before the clinical onset of IUGR and, if possible, before 30 weeks of gestation; and (iii) they allowed construction of 2 × 2 tables of accuracy. Studies were excluded if: (i) they were case series or reports, editorials, comments or reviews without original data (ii) biomarker data were reported only as mean or median values (iii) they did not report accuracy test estimates or reported insufficient data to construct a 2 × 2 table; or (iv) biomarkers were evaluated in women with suspected IUGR/SGA fetuses, pre-eclampsia, diabetes or multiple gestation. If a study based its results on mixed pregnancies (singleton and multiple), it was not considered for inclusion in the review unless singleton pregnancy data were extractable separately,
We excluded five serum biomarkers used in screening for aneuploidy and open neural tube defects (human chorionic gonadotrophin, unconjugated estriol, inhibin A, pregnancy-associated plasma protein A and alphafetoprotein) because a comprehensive systematic review published in 2008 found that those biomarkers had low predictive accuracy for SGA. These findings have been confirmed in large studies published after the meta-analysis.[15-20] Uterine artery Doppler ultrasonography was also excluded because it is not considered a biomarker. Genetic biomarkers were not included in this review because they require a different methodological approach and meta-analytic techniques.
One reviewer (AC-A) screened titles and abstracts of all identified citations and selected potentially eligible studies. Then, all full-text articles were assessed by the same reviewer for inclusion and data extraction, and a 10% sample of the papers was examined by a second independent reviewer (JV). We resolved any disagreements by discussion and consensus. For multiple or duplicate publication of the same data set, we included only the most recent or complete study.
Acceptable reference standards for IUGR are traditionally based on birthweight below the tenth, fifth or third centile for gestational age or birthweight at least two standard deviation (SD) units less than the mean for gestational age regardless of birthweight reference or standard used. SGA is commonly defined as a birthweight below the tenth centile for gestational age. In the majority of situations the term refers to size rather than growth. Most clinicians and researchers define IUGR as being the same as SGA, which combines both constitutionally small and pathological fetuses.
Based on these principles, we anticipated that a key methodological issue of our review would be the definition of IUGR used in the included studies. The choice of an appropriate reference standard is very important because estimates of predictive performance of biomarkers are based on the assumption that the test is being compared to a reference standard that is 100% sensitive and specific. However, no reference standard for the diagnosis of IUGR is 100% sensitive or 100% specific.
Since fetal size alone cannot distinguish pathological fetuses from constitutional SGA fetuses, the use of additional, integrated indicators of fetal and placental health has been proposed to enhance the accuracy of defining true IUGR. The indicators proposed include amniotic fluid volume, biophysical profile, uterine and umbilical artery Doppler ultrasound, and placental pathology among others. Although an optimal scheme of combining birthweight with indicators of fetal and placental health remains to be determined, we have taken this integrated approach into account to define IUGR in assessing the methodological quality of the studies. Our aim is to identify IUGR subgroups that can be related to specific biomarkers.
The methodological quality of the included studies was assessed by at least one reviewer using a modified version of the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) tool. We evaluated five items believed to be important for the quality of studies evaluating the predictive accuracy of biomarkers for IUGR. Each item was scored as ‘yes’, ‘no’ or ‘unclear’. The items evaluated and their interpretation, were as follows:
Data were extracted from each article by means of a standardised and pilot-tested data collection form. The following information was extracted from each article: study characteristics (design, prospective or retrospective data collection, recruitment of women, blinding of test results, completeness of follow up and reporting of withdrawals); participants (inclusion and exclusion criteria, sample size, demographic characteristics, and country and date of publication); description of the biomarker test (gestational age at sampling, frequency of test, sampling site, analytical method used and cut-off level); reference standard used (IUGR definition); and use of preventive/therapeutic interventions for IUGR during pregnancy.
For each study, for all cut-off values defining abnormality, and for several IUGR categories (birthweight below tenth centile, birthweight below fifth centile, birthweight below tenth or fifth centile with additional clinical or pathological evidence of fetal growth restriction, and IUGR requiring delivery before 34 weeks of gestation), we extracted numbers of true-positive, false-positive, true-negative, and false-negative results. When predictive accuracy data were not available, we recalculated them from the reported results including scatter-plot graphs. In studies where serial biomarker samples were collected, we extracted data separately for each gestational period. Ten authors were contacted in an attempt to obtain additional data.
Data extracted from each study were arranged in 2 × 2 contingency tables. When these tables contained cells for which the value was 0, we added 0.5 to each cell to allow calculations to be performed. Sensitivity and specificity were calculated for each biomarker and for all reported cut-off values and outcomes. For each biomarker we planned to plot sensitivities and specificities in receiver operating characteristic (ROC) plots according to gestational age at testing (<20 and ≥20 weeks of gestation) and IUGR definition (below tenth centile and below fifth centile). We then constructed summary ROC curves regardless of cut-offs used by means of a bivariate random-effects approach and calculated area under the summary ROC curves with their corresponding 95% confidence intervals (CIs). This measure allows for comparison of the predictive accuracy of the test for different outcomes. Two-sided P < 0.05 was considered to be statistically significant.
Meta-analyses were performed using subgroups of studies with similar characteristics such as gestational age at testing and outcome measures to minimise clinical heterogeneity. Pooled estimates of sensitivity and specificity with 95% CIs were calculated using a bivariate, random-effects, meta-regression model. Thereafter, we derived likelihood ratios (LRs) with 95% CIs from the pooled sensitivities and specificities for each outcome reported. LRs indicate by how much a given test result raises or lowers the probability of having the condition and so allow interpretation of the results for use in clinical practice. It has been suggested that LRs >10 for a positive test result and LRs <0.1 for a negative test result provide convincing predictive evidence. Moderate prediction can be achieved with LRs values of 5–10 and 0.1–0.2 whereas those <5 and >0.2 provide only minimal prediction.
For each biomarker, we planned to calculate the post-test probability of IUGR by using LRs generated from meta-analyses or individual studies for positive and negative biomarker test results at a range of different pretest probabilities of IUGR (5%, 10% and 15%).
Heterogeneity of the results among studies was investigated through visual examination of forest plots of sensitivities and specificities, and ROC plots. In addition, the quantity I2 was used to assess statistical heterogeneity. I2 values ≥50% indicated a substantial level of heterogeneity. We explored potential sources of heterogeneity by performing meta-regression analysis of subgroups defined a priori (study design, sample size, study quality, IUGR definition used and gestational age at testing). Publication and related biases were assessed visually by examining the symmetry of funnel plots and statistically by using the Egger's regression test. P < 0.1 indicated significant asymmetry.
The bivariate models were fitted using the NLMIXED procedure (SAS 9.1 for Windows; SAS Institute, Inc., Cary, NC, USA). Summary ROC curves were constructed using the RevMan (Review Manager) 5.1.7. The remaining analyses were performed using SPSS version 15.0 (SPSS Inc, Chicago, IL, USA).
Figure 1 summarises the processes used to identify and select studies. The searches produced 5383 citations, of which 215 were considered relevant. In all, 162 studies were excluded, the main reasons being the lack of data to construct 2 × 2 tables (43%) and not being a test accuracy study (25%). Fifty-three studies including 39 974 women, evaluating a total of 37 novel biomarkers, met the inclusion criteria.[32-84] Box 1 lists novel biomarkers for predicting IUGR that were identified in this review.
Table S1 details the individual characteristics of the included studies (see Supplementary material). Thirty studies (57%) were performed in European countries and ten (19%) in North America. Only seven studies (13%) were conducted in developing countries. There were 31 case–control and 22 cohort studies. The sample size in the cohort studies ranged from 63 to 6016 (median, 485) women. The number of case participants enrolled in case–control studies ranged from 8 to 296 and the corresponding number of controls ranged from 8 to 3592. Seventeen case–control studies had ≤40 IUGR cases. In cohort studies, the incidence rates for IUGR below the tenth centile ranged from 2.7 to 17.5% (median, 8.7%); the range for IUGR below the fifth centile was 4.9 to 11.2% (median, 5.1%). Two cohort studies evaluated biomarkers in populations at high risk for IUGR.[38, 40] Biomarkers in the following biological samples were included in the
review: serum or plasma (46 studies), amniotic fluid (four studies) and urine (three studies). Forty-five studies evaluated biomarkers at ≤22 weeks and the remaining eight at 23–35 weeks of gestation. Four studies evaluated biomarkers serially in two gestational periods. The definition of IUGR used birthweight below tenth centile in 27 studies; below fifth centile in 18 studies; below third centile in three studies; below tenth, fifth and third centiles in one study; below tenth and third centiles in one study; ≤2 SD of the mean in two studies; and was unreported in one study.
The quality assessment of the included studies is shown in Figure 2. Only three studies (6%) fulfilled all five criteria, 12 (23%) fulfilled four criteria and the remaining 37 (71%) fulfilled three or fewer criteria. The most common shortcomings were in study design and the reference standard used. Only six studies used an appropriate reference standard that included birthweight centile with additional clinical or pathological evidence of fetal growth restriction.[38, 39, 45, 48, 53, 60] Moreover, information on blinding and selection criteria was unclear in 30 and 25 studies, respectively.
Thirteen studies reported data on placental growth factor (PlGF),[32-35, 37, 38, 40-46] three on soluble fms-like tyrosine kinase-1,[38, 40, 43] two on soluble endoglin,[41, 43] and one each on vascular endothelial growth factor and angiopoietin-2. Figure 3 shows the summary ROC curves of PlGF for predicting IUGR. Overall, PlGF had a low predictive accuracy for IUGR (area under the summary ROC curve 0.66, 95% CI 0.44–0.87; Figure 3A). Summary ROC curves of PlGF for predicting IUGR according to gestational age at testing and birthweight centile are depicted in Figure 3B. The greatest area under the summary ROC curve was for PlGF measured at ≥20 weeks of gestation for predicting IUGR (0.68, 95% CI 0.29–1.00) followed by PlGF for predicting IUGR below the tenth centile (0.67, 95% CI 0.58–0.76) and fifth centile (0.64, 95% CI 0.03–1.00), and PlGF measured at <20 weeks of gestation (0.58, 95% CI 0.38–0.79), although the differences were not statistically significant. Overall, the predictive accuracy of angiogenic factors for IUGR was minimal (median pooled positive and negative LRs of 1.7, range 1.0–19.8, and 0.8, range 0.0–1.0, respectively; Table 1). One small case–control study (nine cases/79 controls), which used a definition of IUGR based on birthweight centile and additional clinical (abnormal Doppler ultrasound or oligohydramnios) and pathological (abnormal placental pathology) evidence of fetal growth restriction, reported a sensitivity of 100% and a specificity of 95% for PlGF (positive and negative LRs of 19.8 and 0.0, respectively). Another, similar, small case–control study (21 cases/23 controls) using a definition of IUGR based on birthweight below the tenth centile with abnormal umbilical artery Doppler ultrasound reported a sensitivity of 92% and a specificity of 78% for angiopoietin-2 (positive and negative LRs of 4.3 and 0.1, respectively).
|Biomarker||Outcome||No of studies||No of women||Pooled sensitivity,% (95% CI)||Pooled specificity,% (95% CI)||Positive likelihood ratio (95% CI)||Negative likelihood ratio (95% CI)||I2 (%)|
|PlGF||IUGRa||10[32-35, 37, 42-46]||5709||38 (35–42)||71 (70–72)||1.3 (1.2–1.5)||0.9 (0.8–0.9)||83|
|IUGR <10th centile||5[32, 34, 35, 37, 46]||3917||46 (41–51)||68 (66–69)||1.4 (1.2–1.6)||0.8 (0.7–0.9)||74|
|IUGR <5th centile||5[33, 42-45]||1792||33 (29–37)||80 (78–82)||1.7 (1.4–2.0)||0.8 (0.8–0.9)||88|
|IUGRa (<20 weeks at testing)||7[32, 33, 35, 42-44, 46]||2451||30 (26–34)||82 (80–83)||1.6 (1.4–1.9)||0.9 (0.8–0.9)||80|
|IUGRa (≥20 weeks at testing)||5[33, 34, 37, 43, 45]||3739||49 (44–53)||64 (63–66)||1.4 (1.2–1.5)||0.8 (0.7–0.9)||85|
|IUGR with additional evidence of fetal growth restriction**||1||88||100 (70–100)||95 (88–98)||19.8 (7.6–51.3)||0.0 (0.0–0.3)||NA|
|IUGR <5th centile in women with abnormal uterine artery Doppler||2[38, 40]||124||65 (43–82)||67 (58–76)||2.0 (1.3–3.0)||0.5 (0.3–1.0)||0|
|PlGF slope||IUGR <10th centile||1||346||59 (51–67)||50 (43–57)||1.2 (1.0–1.5)||0.8 (0.6–1.0)||NA|
|sFlt-1||IUGR <5th centile||1||447||48 (40–56)||67 (61–72)||1.4 (1.1–1.8)||0.8 (0.7–0.9)||NA|
|IUGR <5th centile in women with abnormal uterine artery Doppler||2[38, 40]||124||75 (53–89)||60 (50–69)||1.9 (1.3–2.6)||0.4 (0.2–0.9)||57|
|Δ sFlt-1||IUGR <10th centile||1||346||48 (40–56)||54 (47–60)||1.0 (0.8–1.3)||1.0 (0.8–1.2)||NA|
|sFlt-1/PlGF||IUGR <5th centile||1||411||48 (40–56)||67 (61–72)||1.4 (1.1–1.8)||0.8 (0.7–0.9)||NA|
|IUGR <5th centile in women with abnormal uterine artery Doppler||2[38, 40]||124||75 (53–89)||57 (47–66)||1.7 (1.2–2.4)||0.4 (0.2–1.0)||63|
|VEGF||IUGR ≤2 SD mean birthweight||1||49||56 (27–81)||88 (74–95)||4.4 (1.6–12.2)||0.5 (0.2–1.1)||NA|
|Endoglin||IUGR <5th centile||1||355||61 (52–69)||67 (60–73)||1.8 (1.4–2.3)||0.6 (0.5–0.7)||NA|
|Δ endoglin||IUGR <10th centile||1||346||19 (14–27)||92 (88–95)||2.4 (1.4–4.3)||0.9 (0.8–1.0)||NA|
|Angiopoietin||IUGR <10th centile||1||44||71 (50–86)||78 (58–90)||3.3 (1.5–7.5)||0.4 (0.2–0.7)||NA|
|IUGR with additional evidence of fetal growth restriction**||1||36||92 (67–99)||78 (58–90)||4.3 (1.9–9.4)||0.1 (0.0–0.7)||NA|
Five studies evaluated homocysteine,[47-51] two each 8-oxo-7,8 dihydro-2-deoxyguanosine (8-OHdG)[56, 57] and isoprostanes,[55, 56] and nine several other inflammatory biomarkers, including leptin, asymmetric dimethylarginine, endothelial cell adhesion molecules, fibronectin, lactate dehydrogenase, pentraxin 3, cytokines, C-reactive protein and folate. Homocysteine, 8-OHdG and isoprostanes were tested at <20 weeks of gestation in all included studies. Overall, the predictive accuracy of endothelial function/oxidative stress-related biomarkers for IUGR was minimal (median positive and negative LRs of 2.0, range 0.8–19.2, and 0.8, range 0.0–1.1, respectively; Table 2). Two studies showed high positive LRs, one for endothelial cell adhesion molecules (9.0–19.2) and the other for fibronectin (13.3), but at the expense of poor negative LRs (0.5–0.9).
|Biomarker||Biological sample||Outcome||No of studies||No of women||Pooled sensitivity,% (95% CI)||Pooled specificity,% (95% CI)||Positive likelihood ratio (95% CI)||Negative likelihood ratio (95% CI)||I2 (%)|
|Endothelial function/oxidative stress-related biomarkers|
|Homocysteine||Blood||IUGRa||5[47-51]||9007||27 (23–31)||84 (84–85)||1.7 (1.5–2.0)||0.9 (0.8–0.9)||89|
|IUGR <10th centile||2[48, 50]||2088||26 (19–33)||89 (87–90)||2.3 (1.7–3.1)||0.8 (0.8–0.9)||0|
|IUGR <5th centile||2[49, 51]||6170||29 (25–34)||81 (80–82)||1.6 (1.3–1.9)||0.9 (0.8–0.9)||97|
|Leptin||Blood||IUGR <10th centile||1||139||63 (41–81)||72 (63–79)||2.2 (1.4–3.5)||0.5 (0.3–0.9)||NA|
|Asymmetric dimethylarginine||Blood||IUGR <10th centile||1||43||75 (47–91)||55 (38–71)||1.7 (1.0–2.8)||0.5 (0.2–1.3)||NA|
|sVCAM-1||Blood||IUGR <10th centile||1||1404||16 (7–30)||98 (97–99)||9.0 (3.9–20.7)||0.9 (0.7–1.0)||NA|
|sICAM-1||Blood||IUGR <10th centile||1||1404||42 (28–58)||98 (97–99)||19.2 (11.5–32.1)||0.6 (0.5–0.8)||NA|
|F2-Isoprostane||Amniotic fluid||IUGR <3rd centile||1||114||100 (84–100)||72 (63–80)||3.6 (2.6–5.0)||0.0 (0.0–0.3)||NA|
|8-OHdG||Urine||IUGR <10th centile||2[56, 57]||597||38 (28–48)||78 (74–81)||1.7 (1.3–2.4)||0.8 (0.7–0.9)||3|
|Isoprostane||Urine||IUGR <10th centile||1||508||16 (7–32)||80 (76–83)||0.8 (0.3–1.8)||1.1 (0.9–1.2)||NA|
|Fibronectin||Blood||IUGR <10th centile||1||130||57 (33–79)||96 (90–98)||13.3 (5.0–35.0)||0.5 (0.2–0.8)||NA|
|LDH||Amniotic fluid||IUGR <10th centile||1||93||88 (53–98)||82 (73–89)||5.0 (2.9–8.4)||0.2 (0.0–1.0)||NA|
|Pentraxin 3||Blood||IUGR with additional evidence of fetal growth restriction**||1||72||50 (25–75)||57 (44–68)||1.2 (0.6–2.2)||0.9 (0.5–1.6)||NA|
|Cytokines||Blood||IUGR <10th centile||1||128|
|Interferon γ||35 (24–48)||87 (78–93)||2.8 (1.4–5.6)||0.7 (0.6–0.9)||NA|
|Interleukin-1ra||54 (42–67)||82 (71–89)||3.0 (1.7–5.1)||0.6 (0.4–0.8)||NA|
|Interleukin-12||65 (52–76)||66 (55–76)||1.9 (1.3–2.8)||0.5 (0.4–0.8)||NA|
|Eotaxin||30 (20–43)||85 (74–91)||1.9 (1.0–3.8)||0.8 (0.7–1.0)||NA|
|RANTES||68 (56–79)||70 (59–80)||2.3 (1.6–3.5)||0.5 (0.3–0.7)||NA|
|C-reactive protein||Blood||IUGR <5th centile||1||5969||5 (3–8)||97 (97–98)||1.8 (1.1–3.1)||1.0 (0.9–1.0)||NA|
|Folate||Blood||IUGR <5th centile||1||5774||30 (25–35)||80 (79–81)||1.5 (1.2–1.8)||0.9 (0.8–1.0)||NA|
|Placental proteins/hormone-related biomarkers|
|IGFBP-1||Blood||IUGR <10th centile||1||172||24 (12–43)||91 (85–95)||2.7 (1.1–6.5)||0.8 (0.7–1.0)||NA|
|IGFBP-3||Amniotic fluid||IUGR <10th centile||1||162||69 (54–81)||61 (52–69)||1.8 (1.3–2.4)||0.5 (0.3–0.8)||NA|
|ADAM-12||Blood||IUGR <5th centile||3[65-67]||1947||12 (10–14)||95 (93–96)||2.2 (1.6–3.1)||0.9 (0.9–1.0)||43|
|PP-13||Blood||IUGRa||4[68-71]||4456||22 (17–26)||91 (90–91)||2.3 (1.8–2.8)||0.9 (0.8–0.9)||0|
|IUGR <5th centile||3[68, 69, 71]||3854||34 (27–42)||91 (90–92)||3.6 (2.8–4.7)||0.7 (0.6–0.8)||0|
|IUGR <10th centile||2[70, 71]||3066||12 (8–17)||94 (93–95)||2.0 (1.4–3.0)||0.9 (0.9–1.0)||0|
|Activin A||Blood||IUGR <5th centile||1||635||50 (42–59)||47 (43–51)||1.0 (0.8–1.1)||1.1 (0.9–1.3)||NA|
|PGH||Blood||IUGR <5th centile||1||180||47 (35–59)||58 (49–66)||1.1 (0.8–1.6)||0.9 (0.7–1.2)||NA|
|SP1||Blood||IUGRa||3[33, 74, 75]||944||32 (26–38)||87 (85–90)||2.5 (1.9–3.3)||0.8 (0.7–0.9)||0|
|Annexin A5||Amniotic fluid||IUGR <10th centile||1||145||67 (42–85)||67 (59–74)||2.0 (1.3–3.1)||0.5 (0.2–1.0)||NA|
|HGF||Blood||IUGRa||2[77, 78]||46||77 (53–90)||72 (54–85)||2.8 (1.5–5.3)||0.3 (0.1–0.8)||0|
|Albumin:creatinin ratio||Urine||IUGR <10th centile||1||257||78 (45–94)||59 (53–65)||1.9 (1.3–2.8)||0.4 (0.1–1.3)||NA|
|Vitamin D||Blood||IUGRa||2[80, 81]||1562||51 (45–57)||54 (51–57)||1.1 (1.0–1.3)||0.9 (0.8–1.0)||0|
|Thyroid function||Blood||IUGR <5th centile||1||3804|
|THS||3 (1–6)||98 (97–98)||1.1 (0.5–2.6)||1.0 (1.0–1.0)||NA|
|Free T4||5 (3–8)||98 (97–98)||1.9 (1.0–3.6)||1.0 (0.9–1.0)||NA|
|Free T3||2 (1–5)||98 (97–98)||0.8 (0.3–2.1)||1.0 (1.0–1.0)||NA|
|Metabolomic profile||Blood||IUGR <10th centile||2[83, 84]||156||74 (63–82)||81 (71–88)||3.9 (2.4–6.3)||0.3 (0.2–0.5)||0|
Three studies provided data on a disintegrin and metalloprotease (ADAM)-12,[65-67] four on placental protein 13 (PP-13),[68-71] three on pregnancy-specific β-1-glycoprotein,[33, 74, 75] two on hepatocyte growth factor,[77, 78] and one each on insulin-like growth factor binding proteins 1 and 3, activin A, placental growth hormone and annexin A5. ADAM-12 and PP-13 were assessed at <20 weeks of gestation in all included studies. In general, these biomarkers had a low predictive accuracy for IUGR (pooled sensitivities between 12 and 77%, median 34%; specificities 47–95%, median 87%; positive LRs 1.0–3.6, median 2.2; and negative LRs 0.3–1.1, median 0.8) (Table 2).
Two studies evaluated serum levels of vitamin D,[80, 81] two evaluated metabolomic profile,[83, 84] and one each urinary albumin : creatinine ratio and thyroid function tests. Overall, these biomarkers had low positive LRs (0.8–3.9) and high negative LRs (0.3–1.0; Table 2).
As a consequence of the low predictive accuracy of all novel biomarkers identified and evaluated in this study, no further calculation of post-test probabilities of IUGR was performed. There was a substantial level of heterogeneity (I2 ≥ 50%) among studies in 10 of 19 meta-analyses performed. Meta-regression analysis showed that heterogeneity was mainly explained by study design, IUGR definition used, and gestational age at testing. In general, studies with a case–control design, that used birthweight below the tenth centile to define IUGR and that tested biomarkers at ≥20 weeks of gestation tended to overestimate the accuracy of the predictive tests. All funnel plots showed no asymmetry, either visually or statistically (P > 0.10 for all, by Egger test).
This systematic review shows that, at the present time, there is no clinically useful biomarker for predicting IUGR in women with a singleton gestation. Overall, none of the 37 novel biomarkers evaluated in our review showed a high predictive accuracy for IUGR. Subgroup analyses according to birthweight centile used to define IUGR and gestational age at testing did not improve predictive accuracy. Nevertheless, two small case–control studies reported better predictive values for two angiogenic factors when a definition of IUGR based on both birthweight centile and additional clinical or pathological evidence of fetal growth restriction was used.[39, 45] It was noteworthy that only 13% of the studies included in the review were conducted in developing countries despite IUGR being a major public health problem in such countries.
The poor performance of the biomarkers evaluated in this review could be explained by the following. (i) The multifactorial nature of the IUGR syndrome. There are many causes of IUGR including maternal, fetal and placental factors. Therefore, given the pathophysiological heterogeneity of the condition, the limited clinical utility of any individual biomarker for predicting IUGR is understandable. (ii) The use of an inadequate definition of IUGR in the included studies. Only six of the 53 studies included in the review used a definition for IUGR that went beyond birthweight for gestational age. It is well known that not all small fetuses are growth restricted because some are just constitutionally small. Thereby, it is possible that many fetuses defined as growth-restricted in these studies were actually healthy, constitutionally small but not growth-restricted. (iii) Differences in predictive accuracy of biomarkers according to the gestational age at which the sample is collected. For example, we found that PlGF had a lower sensitivity and a higher specificity when measured at <20 weeks than at ≥20 weeks. (iv) Other factors, such as differences in storage time of the samples, frequency of sampling, methods of analysis (sensitivity and specificity of the assays, techniques used, thawed and refrozen of the samples), cut-off values used for defining abnormality, ethnicity, and statistical analysis used.
This systematic review used a similar methodology to that used in our previous review on novel biomarkers for the prediction of spontaneous preterm birth. The strengths of our review lie in its compliance with stringent criteria for performing a rigorous systematic review of predictive test accuracy. These included the use of a prospective protocol designed to address a research question; extensive and continually updated literature searches without language restrictions; strict assessment of the quality of the studies; the use of contemporary statistical methods, recently recommended for meta-analyses of diagnostic and predictive tests; the performance of subgroup analyses according to gestational age at testing and IUGR definition; the exploration of potential sources of heterogeneity; the quantitative way of summarising the evidence, and the assessment of a wide range of biomarkers instead of only a few.
Some potential limitations of our review must be considered. First, the reliability of the results of a meta-analysis is limited by the methodological quality of the studies included in the review. In our review, less than one-third of included studies met at least four predefined quality criteria. Thirty-one of the 53 studies included in the review had a case–control design. Studies evaluating tests in a diseased population and a separate control group overestimate the diagnostic performance compared with studies that use a clinical population. In addition, case–control studies do not permit accurate estimation of the screening effectiveness in the general population. The issue of an appropriate reference standard has been discussed above. Second, there was substantial heterogeneity among individual studies and results in about half the meta-analyses performed. We explored the sources of heterogeneity as thoroughly as possible and only partial explanations were provided by the study design, definition of IUGR used and gestational age at testing. Third, information on selection criteria and blinding was omitted or could not be determined in about half the studies included. Poor reporting of selection criteria is relevant because neonates with IUGR associated with congenital or chromosomal anomalies or conditions such as pre-eclampsia, diabetes or infection could have been included in the study. Unmasking of clinicians to test results most likely resulted in women with abnormal test results being followed up more carefully or receiving therapy, with both events biasing evaluation of any biomarker's predictive accuracy. Fourth, 69 studies were excluded because they did not report sufficient information to construct a 2 × 2 table resulting in a potential loss of relevant data. The great majority of these studies found there were no statistically significant differences in mean or median concentrations of novel biomarkers between women with an IUGR neonate and those with a non-IUGR neonate. Fifth, the number of studies and IUGR cases available for analysis of most novel biomarkers is still too small for us to draw conclusions. Finally, several studies using the same biomarker to predict IUGR differed substantially in the threshold selected to distinguish normal from abnormal results. Taking into account all of these methodological issues, results must be interpreted with caution.
There is growing interest in the use of combinations of tests because no single biomarker has yet been shown to meet all the requirements of a clinically useful predictive test for IUGR in women with a singleton gestation. Some recent studies have reported on the accuracy of the combination of tests for predicting SGA.[13, 87, 88] A systematic review by Hui et al. reported low predictive values for combinations of serum biomarkers used in prenatal screening for aneuploidy and open neural tube defects. Karagiannis et al. developed a model for predicting SGA below the fifth centile based on combinations of maternal characteristics and biophysical and biochemical markers. At a fixed specificity of 90%, this model yielded a sensitivity of 47% for all SGA, 73% for preterm SGA, and 46% for term SGA. Papastefanou et al. reported that a combination of maternal characteristics and first-trimester ultrasound parameters and biochemical markers had a sensitivity of 55% for predicting SGA up to the fifth centile at a 20% screen positive rate.
In conclusion, none of the novel biomarkers evaluated in this review are sufficiently accurate to recommend their use as a predictor of IUGR in routine clinical practice. However, the use of biomarkers in combination with biophysical parameters and maternal characteristics could be more useful and merits further research. Before further, large-scale, prospective, longitudinal studies are conducted to establish the real potential of novel biomarkers as individual predictive tests for IUGR, stronger evidence is needed linking potential biomarkers with specific subgroups of the IUGR syndrome. In addition, future studies should involve populations from developing countries where the condition is more prevalent.
All authors conceived and designed the study. AC-A extracted and analysed the data, interpreted the results, and drafted the manuscript. The other authors (ATP, SHK and JV) critically revised the draft. The final version of the paper was approved by all authors.
No ethics approval was required.
Departmental funds were used to support the study.
ATP and SHK are supported by the Oxford Partnership Comprehensive Biomedical Research Centre with funding from the Department of Health NIHR Biomedical Research Centres funding scheme.