11Screening for clinical problems

11.1. Gestational diabetes

Clinical question

What is the diagnostic value and effectiveness of screening tests to identify women at risk of diabetes in pregnancy?

Previous NICE guidance (for the updated recommendations see below)

The evidence does not support routine screening for gestational diabetes and therefore it should not be offered. [B]

11.1.1. Introduction and background

Gestational diabetes is defined as carbohydrate intolerance resulting in hyperglycaemia of variable severity with onset or first recognition during pregnancy and with a return to normal after birth.823 It includes women who have both diabetes and impaired glucose tolerance (IGT). Definitions and diagnosis in pregnancy are blurred by the fact that blood glucose levels are higher in pregnancy and there is an overlap between women who are clearly diabetic (and at increased risk) and women who are technically diabetic but are actually not at increased risk. Women who develop gestational diabetes are at increased risk of developing type 2 diabetes in later life823 and the escalating rise in the incidence of this in the population at large creates a compelling argument for screening healthy women in pregnancy, whose subsequent health may benefit from education about diet and lifestyle. However, a decision to implement screening of healthy women in pregnancy has to be made on a judgement of the contribution of each of the following:

  • the potential reduction in perinatal morbidity and mortality
  • the possible reduction in maternal morbidity remembering that increased obstetric intervention may bring about an iatrogenic increase in maternal morbidity
  • the increase in health service expenditure
  • the potential long-term health benefits for the woman.

There has been uncertainty about the value of screening for gestational diabetes for many years and indeed this uncertainty was reflected in the previous antenatal care guideline. However, the recent Australian Carbohydrate Intolerance Study in Pregnant Women (ACHOIS) trial824 group showed that women treated for gestational diabetes had a significantly lower rate of serious perinatal complications compared with women with routine care. These women had a higher rate of induction of labour than the women in the routine care group.

Not only has there been uncertainty about the value of screening but there is little agreement about a suitable screening method. A UK survey of obstetric units in 1999825 indicated that, of the blood tests, 43% used the random blood glucose (RBG) test, 11% used random plasma glucose (RPG), and 10% used a 50 g glucose challenge test (GCT). Sixty-seven percent used a risk factor assessment. An earlier survey in 1994826 involving one district health authority in England found a variety of screening practices for gestational diabetes and in fact only eight out of 18 hospitals operated a screening policy. Six undertook RBG, one undertook fasting blood glucose and one a GCT. They noted that GCT was the most thoroughly evaluated method of screening for gestational diabetes. A survey of gynaecologists in Italy827 reported that 53% (151/283) carried out screening with a glucose load. Of these, 36% gave a 50 g GCT to all women, 17% a 100 g GCT to all women and 40% restricted the test to women with risk factors. In an American survey,828 98.5% of clinicians used the 50 g GCT.

A well-conducted RCT (the ACHOIS trial) has provided evidence for the effectiveness of treating mild gestational diabetes.988 [EL = 1++] The trial allocated 490 women with IGT (2 hour 75 g oral glucose tolerance test (OGTT) fasting level< 7.8 mmol/litre and 2 hour level of 7.8–11.0 mmol/litre) to treatment and 510 women with IGT to routine care. The rate of serious perinatal outcomes (prospectively defined for the purpose of the study as a compound outcome including shoulder dystocia and perinatal mortality) among babies was significantly lower in the intervention group (1% versus 4%; P = 0.01). The number needed to treat to prevent a serious outcome in a baby was 34. There was no significant difference between groups in maternal quality of life.

11.1.2. Risk factors

The use of risk factors such as obesity, country of family origin and the birth of a previous macrosomic baby have been used by healthcare practitioners for many years and indeed often appear as alerts on antenatal care notes.

Description of included studies and findings

A Health Technology Assessment (HTA) in 2002483 [EL = 2+] conducted a systematic review on screening for gestational diabetes. The results showed that the risk factors for gestational diabetes included obesity, advanced maternal age, family history of diabetes, minority ethnic background, increased weight gain in early adulthood and current smoker.

The HTA review included a retrospective analysis in the UK (1992)829 [EL = 2−] aimed at determining the frequency of gestational diabetes according age, BMI, parity and ethnic origin in women without known pre-existing diabetes and to analyse the influence of risk factors separately for each ethnic group. 170/11 205 (1.5%) women were diagnosed with gestational diabetes. Women with gestational diabetes were significantly older (32.3 versus 28.3 years; P< 0.001), had higher BMI (27.7 versus 23.8 kg/m2; P< 0.001) and more likely to be from an ethnic minority (55.4% versus 15.3%; P< 0.0001). Rates of gestational diabetes by ethnicity were white 0.4% (26/6135), black 1.5% (29/1977), South East Asian 3.5% (20/572) and Indian 4.4% (54/1218). After adjusting for age, BMI and parity, the RR (with white as the reference category) was as follows: black 3.1 (95% CI 1.8 to 5.5), South East Asian 7.6 (95% CI 4.1 to 14.1), and Indian 11.3 (95% CI 6.8 to 18.8).

An observational study in Australia (1995)830 [EL = 3] sought to determine the proportion of women with gestational diabetes missed if testing was confined to risk factors. The results showed that women without gestational diabetes were significantly younger (26.4 versus 28.1 years; P< 0.02) and had a lower BMI (24.2 versus 25.9 kg/m2; P< 0.05) than women with gestational diabetes. Thirty-one women (39.2%) with gestational diabetes had no historical risk factors and would have been missed if only selective testing had been undertaken.

A case–control study in Australia (2001)831 [EL = 2+] assessed risk factor screening as a practical alternative to universal screening. The results were as follows: for age ≥ 25 years OR 1.9 (95% CI 1.3 to 2.7), for BMI ≥ 27 kg/m2 OR 2.3 (95% CI 1.6 to 3.3), for high-risk racial heritage OR 2.5 (95% CI 2.0 to 3.2), and for family history of diabetes OR 7.1 (95% CI 5.6 to 8.9). It was found that, by using these four criteria for screening, 313 cases (0.6%) would have been missed and could have saved screening up to 1025 women without gestational diabetes (17%).

A US RCT (2000)832 [EL = 2+] compared a risk factor-based screening programme with a universally based one. The risk factor group were given a 3 hour 100 g OGTT at 32 weeks of gestation if any risk factor was present. The universal screening group was given a 50 g GCT and then a 3 hour 100 g OGTT if the plasma glucose at 1 hour was ≥ 7.8 mmol/litre. The results showed that the PPVs of risk factors were as follows: first-degree relative with type 2 diabetes 6.7%, first-degree relative with type 1 diabetes 15%, previous baby >4.5 kg 12.2%, glycosuria in current pregnancy 50%, macrosomia in current pregnancy 40%, and polyhydramnios in current pregnancy 40%. The detection rate using the universal screening was significantly higher than for risk factor screening, at 2.7% versus 1.45%.

A study in Denmark (2004)833 [EL = 2−] retrospectively investigated the power of pre-screening to identify gestational diabetes and screening to predict adverse clinical outcomes. Risk factors for developing gestational diabetes were used for pre-screening. Pregnant women with at least one risk factor were offered capillary fasting blood glucose (cFBG) in weeks 20 and 32. If the cFBG measurements were ≥ 4.1 mmol/litre and< 6.7 mmol/litre, then a 3 hour 75 g OGTT was offered. If cFBG values were ≥ 6.7 mmol/litre, the woman was diagnosed as having gestational diabetes. The most frequent pre-screening risk factors were BMI ≥ 27 kg/m2 (present in 65% of cases) and age ≥ 35 years (present in 16% of cases). No single factor seemed the best indicator for gestational diabetes. The highest OR for developing gestational diabetes was 9.04 (95% CI 2.6 to 63.7) for glycosuria.

A cross-sectional 5 year investigation in the Netherlands (2006)834 [EL = 2−] examined the clinical usefulness of antepartum clinical characteristics, along with measures of glucose tolerance, in Dutch multi-ethnic women with gestational diabetes for their ability to predict type 2 diabetes within 6 months of delivery (early postpartum diabetes). The following risk factors were assessed for all women: age and gestational age at entry into the study; pre-pregnancy BMI; ethnicity; obstetric and clinical history, including the onset of early postpartum diabetes; and pregnancy outcome. The results showed that apart from family history of diabetes no other risk factor showed an association with the development of early postpartum diabetes.

A prospective population-based study in Sweden [EL = 2+] offered all non-diabetic pregnant women a 75 g OGTT at 28–32 weeks of gestation.835 Traditional risk factors used were family history of diabetes (first-degree relative), obesity (≥ 90 kg), prior LGA baby (≥ 4500 g) or prior gestational diabetes. The results showed that women who did not take the OGTT were more likely to be multiparous and of non-Nordic origin but were less likely to have a family history of diabetes, prior macrosomic baby or prior gestational diabetes. Of the women who were given OGTT, 1.7% were diagnosed with gestational diabetes. The risk factors with the strongest association were prior gestational diabetes (12/61, OR 23.6, 95% CI 11.6 to 48.0) and prior macrosomic baby (9/61, OR 5.59, 95% CI 2.68 to 11.7). Other risk factors were family history of diabetes (13/61, OR 2.74, 95% CI 1.47 to 5.11), non-Nordic origin (13/61, OR 2.19, 95% CI 1.18 to 4.08), weight ≥ 90 kg (8/61, OR 3.33, 95% CI 1.56 to 7.13), BMI ≥ 30 kg/m2 (11/61, OR 2.65, 95% CI 1.36 to 5.14) and age ≥ 25 years (55/61, OR 3.37, 95% CI 1.45 to 7.85).

A systematic review published in the USA in 2007836 [EL = 2++] examined the rates and factors associated with recurrence of gestational diabetes among women with a history of gestational diabetes. A total of 13 studies were included. The results showed the recurrence rate of glucose intolerance during subsequent pregnancies varied markedly across studies. The most consistent predictor of future recurrence appeared to be nonwhite race/ethnicity, although the racial breakdowns within a study were not always clearly described. The recurrence rates varied between 30% and 84% after the index pregnancy. The recurrence rates were higher in the minority populations (52–69%) as compared with lower rates found in non-Hispanic white populations (30–37%). No other risk factors were consistently associated with recurrence of gestational diabetes across studies. Other risk factors, such as maternal age, parity, BMI, OGTT levels and insulin use, inconsistently predicted development of recurrent gestational diabetes across studies.

Evidence summary

Evidence shows that risk factors for developing gestational diabetes are pre-pregnancy obesity, advanced maternal age, prior gestational diabetes, family history of diabetes, minority ethnic background, prior macrosomic baby ≥ 4.5 kg, increased maternal weight gain in early adulthood and being a current smoker. The recurrence rates for gestational diabetes varied between 30% and 84% after the index pregnancy.

The alternative to the use of risk factors is the use of some form of biochemical test, either of urine or blood.

11.1.3. Accuracy of biochemical screening tests

Urine test for glucose

Two studies have been identified in this section.

Description of included studies

A US-based retrospective observational study (3217 women) (1995)494 [EL = II] assessed the ability of urine testing for glucose to predict gestational diabetes or pregnancy outcomes. For this review, only the prediction of gestational diabetes has been taken into consideration. Study participants had complete urinalysis at the first prenatal visit and dipstick at each subsequent visit together with a screening 50 g GCT at 24–28 weeks. Women with at least two urinalysis tests during the first two trimesters were included. 2965 women were categorised into two groups, negative or positive for glycosuria. Those with positive GCT screens (cut-off 7.78 mmol/litre (140 mg/100 ml)) started a 3 day carbohydrate load, and had a 100 g glucose tolerance test (GTT).

A German study (1990)493 [EL = II] compared urine and blood screening tests to detect gestational diabetes. Random urine glucose screening values from each antenatal visit of 500 consecutive pregnant women were compared with a serum glucose test done at 28 weeks of gestation after ingestion of a 50 g glucose-containing beverage. A positive test of a serum glucose level of 7.78 mmol/litre (140 mg/100 ml) or more was followed by a 100 g 3 hour OGTT. Glycosuria was considered present if a trace or greater values were found on at least two prenatal visits. Severe glycosuria was defined as a 2+ (13.9 mmol/litre (250 mg/100 ml)) level or greater on urine screening on at least two prenatal visits.

Findings

The US study found a higher incidence of gestational diabetes in women with positive glycosuria in the first two trimesters (12.8% versus 2.9% for negative screens). The sensitivity of glycosuria in the first trimester as a predictor of gestational diabetes was 7.1%, specificity was 98.5%, PPV was 12.8% and NPV was 97.1%.

In the German study any degree of glycosuria had a sensitivity of 27.3%, specificity of 83.5%, efficiency of 81% and PPV of 7.1%. Severe glycosuria had sensitivity of 18.2%, specificity of 96.9%, and PPV of 21.1%. The incidence of glycosuria was not increased in gestational diabetics when compared with pregnant women with normal glucose tolerance. Severe glycosuria occurred in only 18% of these patients.

Random blood glucose (RBG) test

Two studies have been identified in this section (Table 11.1)

Table 11.1. Random blood glucose test.

Table 11.1

Random blood glucose test.

Description of included studies

A prospective population-based study conducted in Sweden (2004)837 [EL = II] aimed to find out whether repeated RBG, with different cut-off levels, with or without anamnestic factors, could be an effective universal screening test method for identifying high-risk women for the OGTT as the second step. All pregnant women without diabetes (n = 4918) visiting the maternal healthcare clinics over a 2 year period were offered a 75 g OGTT between 28–32 weeks of gestation. RBG was proposed every 4–6 weeks.

A study in Kuwait (1988)838 [EL = II] tested the predictability of an RPG test in women who had their last meal within 2 hours and those who had their last meal more than 2 hours previously. Two hundred and seventy-six unselected pregnant women had RPG followed by 75 g OGTT at 28–32 weeks of gestation.

Findings

In the Swedish study traditional risk factors and values of repeated RBG measurements were registered as well as results of the OGTT in terms of fasting blood glucose and 2 hour blood glucose. A total of 3616 women had an OGTT. Results showed that an RBG cut-off level ≥ 8.0 mmol/litre as the only indicator for an OGTT was optimal for detecting gestational diabetes with regard to sensitivity (47.5%) and specificity (97.0%). It had the same sensitivity for detecting gestational diabetes as using traditional risk factors, but reduced the need to carry out the OGTT from 15.8% to 3.8% of the population.

The Kuwait study used the Lind and Anderson threshold,839 7.0 mmol/litre if eaten within 2 hours, 6.4 if eaten after 2 hours. This gave a sensitivity of 16%, specificity of 96% and PPV of 47%. Using the 90th percentile of study group sensitivity of 29%, specificity of 89% and PPV of 38% were reported.

50 g glucose challenge test (GCT)

Description of included studies

A total of four studies tested the diagnostic value of 50 g GCT (Table 11.2). All studies had an evidence level of II.

Table 11.2. Glucose challenge test.

Table 11.2

Glucose challenge test.

Findings

Four studies499,840–842 in which a diagnostic test was performed on all participants showed sensitivities of 79.8%, 59%, 59% and 78.9% and specificities of 42.7%, 91%, 92%, and 87.2%, respectively. The PPVs were 24.5%, not reported, 32% and 13.8%, respectively.

Comparison studies

Three studies were identified in this section (Table 11.3).

Table 11.3. Comparison studies.

Table 11.3

Comparison studies.

Description of included studies

A prospective study in Germany (2003)843 [EL = II] tested the usefulness of glucose meters in screening pregnant women for gestational diabetes. One hundred and ninety-three pregnant women were administered the 50 g GCT and their blood glucose levels were simultaneously measured with five portable meters and a HemoCue. The results were compared with a standard hexokinase method. A cut-off value of 7.8 mmol/litre was used. The six portable meters used were Accu-Chek, EuroFlash, GlucoTouch, HemoCue, OneTouch and Precision Plus.

A US-based randomised trial with no control (1992)844 [EL = II] compared three carbohydrate sources: 50 g glucose polymer, 50 g standard glucose solution and 50 g milk chocolate bar. A New Zealand-based RCT (1985)845 [EL = II] compared the 100 g glucose screening test with 100 g glucose polymer test.

Findings

All meters showed an excellent correlation (r >0.9; P< 0.01). The various sensitivities were as follows: Accu-Chek 84%, EuroFlash 100%, GlucoTouch 98%, HemoCue 57%, OneTouch 92%, Precision Plus 90%. The specificities were Accu-Chek 98%, EuroFlash 79%, GlucoTouch 86%, HemoCue 100%, OneTouch 92%, Precision Plus 91%.

The overall sensitivity in the American study was 60%, for standard glucose 33.3% and for polymer 100%. The specificities for overall, standard glucose and polymer were 84%, 73.6% and 92.8%, respectively, and PPVs were 16%, 9% and 49%, respectively.

In the New Zealand-based study the sensitivity of the glucose polymer test was 89%, the specificity was 81% and the PPV was 29%.

Fasting plasma glucose (FPG)

Two studies were identified that tested the diagnostic value of FPG (Table 11.4).

Table 11.4. Fasting plasma glucose test.

Table 11.4

Fasting plasma glucose test.

Description of included studies

A Brazilian study (1998)498 [EL = II] used baseline data from a cohort study of consecutive pregnant women to evaluate the performance of FPG as a screening test for gestational diabetes as defined by WHO in an unselected group of pregnant Brazilian women. The study included 5579 women aged ≥ 20 years with gestational ages of 24–28 weeks at the time of testing and no previous diagnosis of diabetes. A standardised 2 hour 75 g OGTT was performed in 5010 women.

A cross-sectional, population-based study in Sweden (2006)846 [EL = II] evaluated the diagnostic properties of cFBG as a screening test in an unselected low-risk Swedish population (n = 3616). They compared cFBG (measured at 28–32 weeks of gestation) with traditional risk factors (registered) and repeated (4–6 times during pregnancy) random capillary glucose measurements as screening models for gestational diabetes. A 75 g OGTT was used to diagnose gestational diabetes.

Findings

The Brazilian study showed that for the detection of gestational diabetes an FPG of 4.94 mmol/litre (89 mg/100 ml) jointly maximises sensitivity (88%) and specificity (78%), identifying 22% of the women as test-positive. Lowering the cut-off point to 4.5 mmol/litre (81 mg/100 ml) increases sensitivity (94%), decreases specificity (51%) and identifies 49% of women as test-positive. For detection of IGT, a value of 4.72 mmol/litre (85 mg/100 ml) jointly maximises sensitivity and specificity (68%), identifying 35% of women as test-positive. A cut-off point of 4.72 mmol/litre (85 mg/100 ml) for the detection of gestational diabetes gives sensitivity of 94% and specificity of 66%.

The Swedish study found that 1.52% (55/3616) of women were diagnosed before 34 weeks of gestation. For cFBG cut-off values between 4.0 and 5.0 mmol/litre, the sensitivity ranged between 87% and 47% and specificity between 51% and 96%. The LR+ and LR− were best at ≥ 5.0 mmol/litre. The combination of traditional risk factors with cFBG only slightly increased the sensitivity as compared with the use of cFBG alone.

Jelly beans

Two studies were identified in this section (Table 11.5).

Table 11.5. Jelly bean studies.

Table 11.5

Jelly bean studies.

Description of included studies

A US study (1999)847 [EL = II] tested the hypothesis that a standardised dose of jelly beans could be used as an alternative sugar source to the 50 g glucose beverage to screen for gestational diabetes. This prospective study recruited 160 pregnant women at 24 to 28 weeks of gestation to compare two sugar sources for serum glucose response, side effects, preference, and ability to detect gestational diabetes. Patients were randomly given 50 g glucose beverage or 28 jelly beans (50 g simple carbohydrate) and serum glucose values were determined 1 hour later. A 100 g 3 hour OGTT was performed finally.

Another American study (1995)848 [EL = II] tested the diagnostic value and patient tolerance of jelly beans as an alternative to a 50 g glucose solution. Pregnant women between 26 and 30 weeks of gestation were recruited to participate in the study. Each participant was given cola beverage containing 50 g of glucose and blood glucose was tested 1 hour later. Within 2 weeks of this test, each patient ate 18 jelly beans and had glucose level tested within 1 hour. Within 2 weeks of the jelly bean test, all participants were given a 3 hour 100 g OGTT.

Findings

In the US study 136 participants completed the study and a comparison of efficacies of jelly beans and 50 g glucose beverage as sugar sources in detection of gestational diabetes was made. There was not much difference between serum glucose values after ingestion of jelly beans (6.49 ± 1.31 mmol/litre (116.9 ± 23.6 mg/100 ml)) and of 50 g glucose beverage 6.47 ± 1.5 mmol/litre (116.5 ± 27.0 mg/100 ml). There was a significantly lower incidence of side effects after consumption of the jelly beans (20%) as compared with 50 g glucose beverage (38%). Seventy-six percent of the participants preferred jelly beans and 24% preferred the 50 g glucose beverage.

In the second study the sensitivity, specificity and PPV of the cola beverage using 7.78 mmol/litre (140 mg/100 ml) as the threshold were 46%, 81% and 18%, respectively. The sensitivity, specificity and PPV of jelly beans using a threshold of 6.67 mmol/litre (120 mg/100 ml) were 54%, 81%, and 20%, respectively. Participants tolerated jelly beans better than the cola beverage.

In order to compare the various blood tests for screening gestational diabetes, likelihood ratios were calculated (Table 11.6).

Table 11.6. Likelihood ratios for three blood tests.

Table 11.6

Likelihood ratios for three blood tests.

Effectiveness of screening tests

Description of included studies

A US-based RCT832 [EL = 2+] compared a risk factor-based screening programme with a universally based one. The risk factor group had a 3 hour 100 g OGTT at 32 weeks if any risk factor for gestational diabetes was present. The universal group had a 50 g GCT and if the plasma glucose at 1 hour was ≥ 7.8 mmol/litre, a formal 3 hour 100 g OGTT was then performed.

A study in Denmark (2004)833 [EL = 2−] retrospectively investigated in 1 year the clinical outcome of pregnant women in relation to separate components of the pre-screening procedure, presence of gestational diabetes and the capillary blood glucose 120 minutes after glucose load (cBG120 min) concentration after a 75 g glucose load. The aim was to investigate the power of the pre-screening to identify gestational diabetes and for the screening to predict adverse clinical outcomes.

A cross-sectional 5 year investigation in the Netherlands (2006)834 [EL = 2−] examined the clinical usefulness of antepartum clinical characteristics, along with measures of glucose tolerance, in Dutch multi-ethnic women with gestational diabetes for their ability to predict type 2 diabetes within 6 months of delivery (early postpartum diabetes). The following data were collected for all women: age and gestational age at entry into the study; pre-pregnancy BMI; ethnicity; obstetric and clinical history, including the onset of early postpartum diabetes; pregnancy outcome; level of fasting C-peptide; and glycaemic parameters of 50 g 1 hour GCT and 100 g 3 hour OGTT (diagnostic OGTT). Eleven of 168 or 6.6% of women developed early postpartum diabetes.

A prospective cohort study (1998)849 [EL = 2+] in the United Arab Emirates compared the outcome of pregnancy in women with GCT screening levels ≥ 7.7 mmol/litre and ≥ 8.3 mmol/litre. Pregnancy outcomes were compared for the following groups:

  1. GCT ≥ 7.7 and< 8.3 mmol/litre (194 women)
  2. GCT ≥ 8.3 mmol/litre (194 women)
  3. GCT< 7.7 mmol/litre (194 women matched for age, parity and weight with Group B)

The screening test used was blood glucose 1 hour after a 50 g glucose load (GCT) given in fasting state between 28 and 32 weeks. If the blood glucose was ≥ 7.7 mmol/litre then 3 hour GTT was given.

A prospective cohort study of 6854 participants (2005)850 [EL = 2+] in the USA evaluated the association between obesity, GCT and pregnancy outcome. A 50 g GCT was performed at 24–28 weeks of gestation and a screening value of ≥ 7.22 mmol/litre (130 mg/100 ml) was followed by a 100 g OGTT. For the purpose of analysis, women were categorised by pre-pregnancy BMI and by different GCT thresholds. Maternal outcome was defined by the rate of pre-eclampsia, gestational age at delivery, caesarean section rate and the need for labour induction. Neonatal outcome was defined by fetal size (macrosomia/LGA), arterial cord pH, respiratory complications and neonatal intensive care unit (NICU) admission.

A prospective study (1987)851 [EL = 2+] in a midwestern US population compared the value of routine versus selective diabetes screening in a group of predominantly middle-class, healthy, Caucasian pregnant women. 2000 women were divided into two groups (they were otherwise similar):

  1. those to undergo routine screening between 24 and 28 weeks of gestation
  2. those to be tested selectively in the presence of standard risk factors.

The screening test involved a 50 g GCT followed by a 3 hour OGTT if necessary.

A prospective randomised study (1995)852 [EL = 2+] in China was conducted to determine the relationship between the 50 g GCT and pregnancy outcomes. In this study, 622 pregnant women underwent a 50 g GCT and a 75 g OGTT was performed if the screening test value was ≥ 7.8 mmol/litre.

Findings

The American study showed that universal screening detected a gestational diabetes prevalence of 2.7%, 1.45% more than in the risk factor-screened group. Universal screening for gestational diabetes was found to be superior to risk factor-based screening as it detected more cases, facilitated early diagnosis and is associated with improved pregnancy outcomes.

The results of the Danish study showed that screening using a cFBG of 4.1 mmol/litre was unable to predict gestational diabetes and adverse outcome. The best predictor of complicated delivery was a high BMI. The best predictor of fetal adverse outcome was cBG120 min ≥ 9.0 mmol/litre after a 75 g glucose load. Identical pregnancy complications were present in gestational diabetes and non-gestational diabetes.

The Netherlands study showed that only a family history of diabetes showed an association with early postpartum diabetes. ROC curve analysis identified all three GCT parameters, including fasting glucose concentration, as poor diagnostic tests, with a PPV of 22%, whereas PPV associated with the area under the diagnostic OGTT curve increased progressively over the duration of the test from 20.6% to 100%. Using a 3 hour OGTT glucose area threshold of 35.7 mmol·h/litre resulted in 100% sensitivity and 100% specificity, identifying the 11 women who developed early postpartum diabetes.

In the UAE study, 197/3400 or 5.8% of women were considered to have abnormal GTT plus 199/3400 or 5.8% had IGT. There was no significant difference in pregnancy-induced hypertension between groups. Preterm delivery was significantly higher in Group B. Birthweight above 4.5 kg was 4% in group C, 6% in group A and 9% in group B. The Apgar score >6 at 1 minute found no significant differences between groups.

In the US-based study, a positive GCT result (GCT ≥ 7.22 mmol/litre (130 mg/100 ml)) was identified in 2541/6854 or 37% of women. In this study, 464/6854 or 6.8% of women were diagnosed with gestational diabetes. In both groups of screening results (>7.22 mmol/litre (130 mg/100 ml) and< 7.22 mmol/litre (130 mg/100 ml)), the women who were obese were significantly older, gained more weight during pregnancy and had a lower rate of nulliparity in comparison with the women who were not obese. The women who were obese had higher rates of macrosomia, LGA and induction of labour. No difference was found in mean birthweight, the total rate of ceasarean section, preterm delivery, 5 minute Apgar score ≤ 7, mean arterial cord pH, NICU admission or a need for respiratory support in comparison with non-obese women in both groups of screening results. A gradual increase in the rate of macrosomia, LGA and ceasarean section was identified in both obese and non-obese women in relation to increasing GCT severity categories.

The midwestern American study showed that the incidence of gestational diabetes in the selectively screened group was twice (19/453, 4.2%) that in the routinely screened group (21/1000, 2.1%). Glucose intolerance without a risk factor was found in only one case (1/1000, 0.1%) in the routinely screened group.

In the Chinese study, 103/622 or 16.6% of women underwent the diagnostic test, among whom 32 were identified as having gestational impaired glucose tolerance (GIGT) and 12 as having gestational diabetes. The sensitivity of 50 g GCT was 42.7% (44/103). The incidences of oedema–proteinuria–hypertension syndrome (EPH syndrome), PROM, fetal macrosomia, operative deliveries and perinatal morbidity were higher in women with GIGT/gestational diabetes than in women without GIGT/gestational diabetes.

11.1.4. Women’s views on screening for gestational diabetes

Description of included studies

A prospective survey (2002)853 [EL = 2−] in Australia surveyed women on their experiences of being screened for gestational diabetes in a hospital that screens all women in pregnancy. They tested the hypothesis that women with a positive result on the screen test will experience a reduction in quality of life, their health and that of their baby when compared with women with a normal screening result. The study took place at a level III teaching hospital with a high-risk pregnancy service and NICU. A Spielberger State-Trait Anxiety Inventory, Edinburgh Postnatal Depression Scale and Short Form 36 Item Health Survey were used to study the main outcome measures: anxiety, depression, health status, concerns about the health of the baby and perceived health. Prior to being screened, a total of 158 women participated in the study whereas 51 women participated after being screened.

A prospective cohort study (1997)854 [EL = 2+] in Canada investigated whether false positive results of 50 g GCT for gestational diabetes were associated with adverse psychological effects. Women between 12 and 14 weeks of gestation with no previous history of diabetes or gestational diabetes were included. In this study, 897 women had complete data both at enrolment and at 32 weeks, including 88 who had false positive GCT results. A total of 809 women completed questionnaires at baseline, 32 weeks and 36 weeks of gestation.

Findings

The Australian study found no differences in the levels of anxiety, depression or the women’s concerns about the health of their babies. When women positively screened for gestational diabetes were compared with negatively screened women, the positively screened group had significantly lower health perceptions, were significantly less likely to rate their health as ‘much better than one year ago’ and were significantly more likely to rate their health as ‘fair’ rather than ‘very good’ or ‘excellent’.

The Canadian study showed that, at 32 weeks, 20% of women with false positive GCT results significantly perceived their health as excellent as compared with 38% of women with negative results or not tested. These results were sustained at 36 weeks. The study showed no significant association between false positive test result and anxiety levels, depression or woman’s concern for health of baby. These results were neither significant between baseline and 32 weeks nor at 36 weeks.

11.1.5. Clinical characteristics and screening

Description of included studies

A Canada-based prospective study (1997)855 [EL = 2+] tested the hypothesis that using clinical characteristics for assessing women’s risks of gestational diabetes could enhance the efficiency of screening. In this study, 3131 women were randomly divided into two groups – a derivation group and a validation group. The screening strategies were derived from the derivation group data which were then tested in the validation group by comparing the effectiveness and efficiency with those with usual care. The strategies used were: no screening for low-risk women, usual care for intermediate-risk women, and universal screening with lower thresholds (plasma glucose values of 130 mg/100 ml (7.2 mmol/litre) or 128 mg/100 ml (7.1 mmol/litre) for high-risk women.

Findings

In the Canadian study there was a 34.6% reduction (95% CI 32.3% to 37.0%) in the number of screening tests performed after using the new strategies. The detection rate of gestational diabetes with new strategies was 81.2–82.6 % compared with the 78.3% detected through usual care. There was a significant reduction in the percentage of false positive screening tests, from 17.9% with usual care to 16.0% or 15.4% (P< 0.001) with the new strategies, depending on the threshold values for high-risk women.

Evidence summary for Section 11.1

Owing to the heterogeneity among studies for different screening tests there is no obvious best test available to screen for gestational diabetes. The evidence for the accuracy of these tests is further undermined by the fact that the thresholds for sensitivity and specificity are determined by the individual researchers rather than with reference to an agreed standard.

There is low-grade evidence from the effectiveness studies that IGT in pregnancy or frank gestational diabetes is associated with macrosomia, and possible increases in the incidence of pre-eclampsia and preterm delivery. On the other hand obesity was the factor most likely to be associated with complicated delivery and family history seemed to relate to post-delivery diabetic risk.

The ACHOIS study seems to suggest that treating women who have mild gestational diabetes in pregnancy is likely to be effective in reducing the risk of complications.

There is some evidence suggesting that receiving a positive screen result reduces women’s health perceptions and makes them more likely to rate their health as ‘fair’ rather than ‘very good’ or ‘excellent’.

Health economics – screening, diagnosis and treatment of gestational diabetes

The effectiveness of screening, diagnosis and treatment (including monitoring) for gestational diabetes was identified by the GDG as a priority for health economic analysis. The analysis of cost-effectiveness was addressed through joint work with the GDG for the NICE Diabetes in Pregnancy guideline.636 The methods and results from the health economic modelling are summarised here; further details are provided in Appendix F.

A systematic search of the literature identified 337 studies potentially related to the clinical question. After reviewing the abstracts, 33 articles were retrieved for further appraisal and eight have been included in this section of the review. Two papers were identified in the literature that examined the cost-effectiveness of screening for and treating gestational diabetes, six papers were identified that examined the cost-effectiveness of screening only for gestational diabetes and one paper was found that examined the cost-effectiveness of treating gestational diabetes. None of these papers was suitable for answering the question addressed in the guideline. The results of the systematic review are reported in Appendix F.

The recently published ACHOIS trial972 demonstrated potential benefit of treatment for mild gestational diabetes. Evidence of clinical effectiveness is not always sufficient for a treatment to be considered cost-effective – often the people that would benefit from treatment must be identified from a group of people who do not require treatment. This is the case with gestational diabetes; the cost-effectiveness of screening and treatment for gestational diabetes are highly interdependent. As a result, a single decision-analytic health economic model was developed to help the combined GDGs make recommendations in relation to screening and treatment for gestational diabetes (see Appendix F for a full description of the model structure, data inputs and results).

The health economic model considered screening based on risk factors (age, ethnicity, BMI and family history of gestational diabetes) and/or blood tests (RBG, FPG and a 1 hour 50 g GCT) followed by a diagnostic test (2 hour 75 g OGTT). The possibility of universal diagnostic testing was also considered. The treatment alternatives considered in the model were diet, oral hypoglycaemic agents (glibenclamide and metformin) and insulin. All women diagnosed with gestational diabetes were assumed to undertake self-monitoring of blood glucose, regardless of which form of treatment they were using.

Under the baseline assumptions of the health economic model, two screening strategies were not dominated by other strategies. A strategy of offering an OGTT to women from high-risk ethnic backgrounds had an ICER of £3,678 compared with no screening or treatment (strategy 2). A strategy of offering an OGTT to all women defined by the American Diabetes Association (ADA) as being at high risk of gestational diabetes (i.e. women older than 25 years or BMI above 27 kg/m2 or family history of diabetes or high-risk ethnic background; strategy 6) had an ICER of £21,739 compared with screening based on ethnicity alone. However, the combined GDGs expressed concern over the number of women that would undergo an OGTT if the ADA screening strategy were recommended because a large proportion of women would be tested based on age criteria alone. Using age as a risk factor for screening has a high sensitivity (i.e. it will identify the majority of women with gestational diabetes). Owing to data limitations it was not possible to evaluate the cost-effectiveness of using a combination of the individual risk factors considered in the model (see Appendix F). In the absence of this possibility, an analysis of the cost-effectiveness of each individual risk factor followed by an OGTT was conducted, with each strategy being compared with a strategy of no screening or treatment. The results are presented in Table 11.7.

Table 11.7. ICER for single risk factor strategies followed by a diagnostic test when compared with a strategy of no screening or treatment.

Table 11.7

ICER for single risk factor strategies followed by a diagnostic test when compared with a strategy of no screening or treatment.

All of these ICERs are below the £20,000 per QALY threshold used by NICE as a willingness to pay for cost-effectiveness, and so each one could be regarded as being cost-effective.

GDG interpretation of evidence

Currently an unselected pregnant population will have the risk of gestational diabetes assessed using risk factors such as:

  • BMI >30 kg/m2
  • previous macrosomic baby ≥ 4.5 kg for Caucasian and black women, ≥ 4.0 kg for other women
  • previous gestational diabetes (see Diabetes in Pregnancy guideline636)
  • family history of diabetes (first-degree relative with type 1 or type 2 diabetes)
  • women from a high-risk ethnic group, which would include:856

    South Asian (Indian, Pakistani, Bangladeshi)

    black Caribbean

    Middle Eastern (Saudi Arabia, United Arab Emirates, Iraq, Jordan, Syria, Oman, Qatar, Kuwait, Lebanon or Egypt).

This last group of women has been identified from the Diabetes Atlas (third edition) 2006 using prevalance of type 2 diabetes as a surrogate for gestational diabetes.

Approximately 20–50% of women will have a positive screen using these risk factors, with proportions varying considerably from one area to another.

According to a 1999 survey,825 67% of UK maternity service providers currently screen using a combination of these factors.

Evidence from health economic modelling carried out by the NCC-WCH shows that screening and treating gestational diabetes is cost-effective using identification of risk factors as the screening method followed by an OGTT.

While screening using risk factors is less sensitive than performing a glucose challenge or glucose tolerance test, it is more practical and less disruptive for women. The biochemical tests considered (GCT, FPG, RBG and urine testing) perform only moderately well in terms of diagnostic value.

Approximately 10–20% of women diagnosed as having gestational diabetes will go on to need oral hypoglycaemic agents or insulin therapy where diet and exercise have not afforded adequate control of blood glucose levels (refer to Diabetes in Pregnancy (NICE clinical guideline 63)636 for further details).

Evidence from one large trial (the ACHOIS study) has shown treatment of gestational diabetes to be effective in reducing poor outcomes (refer to Diabetes in Pregnancy (NICE clinical guideline 63)636 for further details).

Recommendations on screening for gestational diabetes

Screening for gestational diabetes using risk factors is recommended in a healthy population. At the booking appointment, the following risk factors for gestational diabetes should be determined:

  • body mass index above 30 kg/m2
  • previous macrosomic baby weighing 4.5 kg or above
  • previous gestational diabetes (refer to ‘Diabetes in pregnancy’ [NICE clinical guideline 63], available from www.nice.org.uk/CG063)
  • family history of diabetes (first-degree relative with diabetes)
  • family origin with a high prevalence of diabetes:

    South Asian (specifically women whose country of family origin is India, Pakistan or Bangladesh)

    black Caribbean

    Middle Eastern (specifically women whose country of family origin is Saudi Arabia, United Arab Emirates, Iraq, Jordan, Syria, Oman, Qatar, Kuwait, Lebanon or Egypt).

Women with any one of these risk factors should be offered testing for gestational diabetes (refer to ‘Diabetes in pregnancy’ [NICE clinical guideline 63], available from www.nice.org.uk/CG063).

In order to make an informed decision about screening and testing for gestational diabetes, women should be informed that:

  • in most women, gestational diabetes will respond to changes in diet and exercise
  • some women (between 10% and 20%) will need oral hypoglycaemic agents or insulin therapy if diet and exercise are not effective in controlling gestational diabetes
  • if gestational diabetes is not detected and controlled there is a small risk of birth complications such as shoulder dystocia
  • a diagnosis of gestational diabetes may lead to increased monitoring and interventions during both pregnancy and labour.

Screening for gestational diabetes using fasting plasma glucose, random blood glucose, glucose challenge test and urinalysis for glucose should not be undertaken.

Research recommendation on screening for gestational diabetes

Is screening for gestational diabetes based on expected local prevalence, with or without modification by risk factors, clinically effective and cost-effective?

11.2. Pre-eclampsia

Clinical question

What is the diagnostic value of different screening methods in identifying women at risk of developing pre-eclampsia?

Previous NICE guidance (for the updated recommendations see below)

At first contact, a woman’s level of risk for pre-eclampsia should be evaluated so that a plan for her subsequent schedule of antenatal appointments can be formulated. The likelihood of developing pre-eclampsia during a pregnancy is increased in women who:

  • are nulliparous
  • are age 40 years or older
  • have a family history of pre-eclampsia (e.g., pre-eclampsia in a mother or sister)
  • have a prior history of pre-eclampsia
  • have a BMI at or above 35 kg/m2 at first contact
  • have a multiple pregnancy or pre-existing vascular disease (for example, hypertension or diabetes). [C]

Whenever blood pressure is measured in pregnancy, a urine sample should be tested at the same time for proteinuria. [C]

Standardised equipment, techniques and conditions for blood-pressure measurement should be used by all personnel whenever blood pressure is measured in the antenatal period, so that valid comparisons can be made. [C]

Pregnant women should be informed of the symptoms of advanced pre-eclampsia because these may be associated with poorer pregnancy outcomes for the mother or baby. Symptoms include headache, problems with vision, such as blurring or flashing before the eyes, bad pain just below the ribs, vomiting, and sudden swelling of face, hands or feet. [D]

Future research

Research is needed to determine the optimal frequency and timing of blood pressure measurement and on the role of screening for proteinuria.

11.2.1. Introduction and background

Pre-eclampsia is a condition usually associated with hypertension and proteinuria, and occurring in the second half of pregnancy. Hypertension is defined as a single diastolic blood pressure of 110 mmHg or any consecutive readings of 90 mmHg on more than one occasion at least 4 hours apart. Proteinuria is defined as 300 mg excretion of protein in a 24 hour collected urine, two clean catch urine specimens at least 4 hours apart with 2+ proteinuria by dipstick.535

Pre-eclampsia and eclampsia remain among the major causes of maternal mortality in the UK (CEMACH 2004) although the reduction in the number of deaths since the 1950s may have been at least in part due to the monitoring of blood pressure during pregnancy. Current knowledge on the pathophysiology of pre-eclampsia has identified that it is a complex disorder with widespread endothelial damage which can involve every organ of the body. Therefore presenting signs and symptoms may be more varied than a rising blood pressure and proteinuria. However, the antenatal care of all pregnant women is an opportunity to screen for rising blood pressure, especially in groups who are at increased risk, and to educate them about the symptoms which might signal fulminating disease.

11.2.2. Accuracy of screening tests

The quality of studies included in this review was variable, with deficiencies in many areas of methodology. In particular, studies suffered from a lack of blinding and relatively small sample sizes. There was heterogeneity regarding the reference standard used in each study.

Only a few tests reached specificity above 90%. These were AFP, β-hCG and uterine artery Doppler (bilateral notching). The sensitivities of these tests were variable and generally low.

It was often not possible to be certain about the definition of pre-eclampsia used in studies. There was a lack of information on the exact technique of blood pressure measurement and Korotkoff threshold for abnormality, whether the proteinuria was in the absence of urinary tract infection and pre-existing renal disease, or whether there was normalisation of blood pressure within 6 weeks of giving birth.

Alpha-fetoprotein (AFP)

Two studies have been identified in this section (Table 11.8).

Table 11.8. Alpha-fetoprotein (AFP).

Table 11.8

Alpha-fetoprotein (AFP).

Description of included studies

An American prospective cohort study (1999)857 [EL = II] evaluated the value of AFP as a predictor of pregnancy outcomes. Maternal serum markers were analysed over a 5 year period (March 1991 to May 1996) from 60 040 women who underwent serum marker screening at 14–22 weeks of gestation. All women had maternal serum AFP measurements. A value of at least 2.5 MoM was used for calculation.

A population-based cohort study in Finland (1998)858 [EL = II] sought to determine whether maternal mid-trimester AFP can predict pre-eclampsia. 1037 nulliparous women were included, of whom 637 were analysed. Measurement of AFP was made from maternal serum collected at 15–19 weeks of gestation. Sensitivity, specificity and predictive values were calculated for elevated AFP (at least 2.0 MoM).

Findings

The American study gave a very low sensitivity of 4.3% but a high specificity of 97.4% for AFP measurement. The overall incidence of pre-eclampsia was 3.2%.

The Finland-based study calculated a poor sensitivity of 3% and a specificity of 98%. The incidence of pre-eclampsia was reported as 5.3%.

Both these studies used slightly different reference standards.

Fetal DNA

A total of two studies were included in this section (Table 11.9).

Table 11.9. Fetal DNA.

Table 11.9

Fetal DNA.

Description of included studies

A case–control study in Ireland (2004)859 [EL = II] investigated whether the presence of fetal DNA in the maternal circulation in early pregnancy might be a marker for the prediction of pre-eclampsia. A total of 264 women (88 cases and 176 controls) were analysed in the study. Blood was obtained from women attending for a first antenatal clinic. Cases were asymptomatic women who subsequently developed pre-eclampsia matched to control women for parity and gestational age. Fetal DNA was quantified by real-time PCR using TaqMan primers and probes directed against SRY gene sequences.

A Hong Kong-based case–control study (2001)860 [EL = II] aimed to test whether the abnormal increase in circulating DNA concentrations could be detected in susceptible subjects before onset of the clinical disease. A total of 51 women (18 cases and 33 controls) were analysed in this study. The gestational age at testing was 11–22 weeks.

Findings

The Ireland study found that the presence of fetal DNA in the maternal circulation is associated with an eight-fold increased risk of developing pre-eclampsia. In this study, SRY copies/ml< 10 000 gave a sensitivity of 94.32% and specificity of 32.39%. SRY copies/ml< 50 000 gave a sensitivity of 81.82% and specificity 64.77%. SRY copies/ml >50 000 gave a sensitivity of 38.64% and a specificity of 90.34%.

In the Hong Kong-based study an SRY value of ≥ 33.5 Genome equivalents/ml was found to be significant and this gave a sensitivity of 67% and a specificity of 82%.

β-hCG

A total of three studies were included (Table 11.10).

Table 11.10. β-hCG.

Table 11.10

β-hCG.

Description of included studies

A US-based prospective cohort study (1999)857 [EL = II] evaluated the value of β-hCG as predictor of pregnancy outcomes. Maternal serum markers were analysed over a 5 year period (March 1991 to May 1996) from 60 040 women who underwent serum marker screening at 14–22 weeks of gestation. 45 565 women had maternal serum β-hCG measurements. A value of at least 2.5 MoM was used for calculation.

A US-based case–control study (2000)861 [EL = II] sought to determine whether second-trimester (15–21 weeks) serum levels of human chorionic gonadotrophin is predictive of the later onset of pre-eclampsia in pregnancy. A total of 359 women (60 cases and 299 controls) were included. Levels of each analyte were compared in women with pre-eclampsia and controls using matched rank analysis.

A prospective cohort study (1997)862 [EL = II] in the USA investigated the association of elevated second-trimester (15–22 weeks) β-hCG with the subsequent development of hypertension in pregnancy and evaluated its utility as a screening test for later development of pre-eclampsia. A total of 6138 women were analysed in the study. A value of 2.0 MoM was used as the cut-off for the index test.

Findings

The first study found a 3% incidence of pre-eclampsia. The sensitivity at 2.5 MoM cut-off was found to be 5.5% and specificity was 96%.

The second study used 2.0 MoM cut-off and found a 3.2% incidence of pre-eclampsia. With 95% specificity, a modelled sensitivity of 15% was found.

The third study found a 3.2% incidence of pre-eclampsia. The sensitivity was 17.5% whereas the specificity was 89.8%.

Urinary calcium excretion

Two studies were included in this section (Table 11.11).

Table 11.11. Urinary calcium excretion.

Table 11.11

Urinary calcium excretion.

Description of included studies

A US-based prospective longitudinal study (1991)863 [EL = II] was designed to determine whether an alteration in calcium excretion precedes the signs and symptoms of pre-eclampsia and therefore would be a useful early marker for this disease. A total of 99 women were analysed in this study. The index test was administered at between 10 and 24 weeks of gestation and a value of ≤ 195 mg/24 hours was considered significant.

A UK-based prospective non-interventional study (1994)864 [EL = II] assessed the potential of urinary calcium/creatinine as screening tests for pregnancy-induced hypertension in a white population. A total of 500 women were included in the study who provided a urine sample at 19 weeks of gestation.

Findings

The American study found am 8.1% incidence of pre-eclampsia. The index cut-off found a sensitivity of 86%, specificity of 84%, PPV of 46% and NPV of 98%.

The UK study found a sensitivity of 31% and a specificity of 72%. The overall incidence of pre-eclampsia was 2.6%.

Calcium/creatinine ratio

A total of four studies were included (Table 11.12).

Table 11.12. Calcium/creatinine ratio.

Table 11.12

Calcium/creatinine ratio.

Description of included studies

A Hong Kong-based cohort study (1994)865 [EL = II] attempted to clarify some of the changes that occur in enzyme and electrolyte excretion in pregnancy, before onset of clinical signs, and to relate these changes to the antenatal development of pre-eclampsia or gestational hypertension. A total of 199 women were included and the gestational age at test was in the range 18–26 weeks. A cut-off value of 0.3 was used.

One Argentina-based prospective cohort study (1994)866 [EL = II] investigated the usefulness of calcium/creatinine ratio and other laboratory tests as predictors in the development of hypertensive disorders of pregnancy. The study included 387 women and the test was administered at 20 weeks of gestation. A value of 0.07 was considered significant.

A prospective cross-sectional study (2003)867 [EL = II] in Iran determined the relationship between pre-eclampsia and calcium/creatinine ratio. A total of 102 women were included and the test was administered at 20–24 weeks of gestation. A value of ≤ 0.229 was found to be significant.

A UK-based prospective non-interventional study (1994)864 [EL = II] assessed the potential of urinary calcium/creatinine as screening tests for pregnancy-induced hypertension in a white population. A total of 500 women were included in the study who provided a urine sample at 19 weeks of gestation.

Findings

The Hong Kong study found a sensitivity of 49% and specificity of 90%. The overall incidence was 4%.

The Argentina study found an overall incidence of 3.4%. The study gave a sensitivity of 33%, specificity of 78%, PPV of 5% and NPV of 97%.

The Iran study found an incidence of 7.8%. The test showed a sensitivity of 75%, specificity of 77.7%, PPV of 20.7%, and NPV of 97%.

The UK study reported an incidence of 2.6%. The test sensitivity was 31% and specificity was 55%.

Bilateral uterine artery notching

A total of four studies were included (Table 11.13).

Table 11.13. Bilateral uterine artery notching.

Table 11.13

Bilateral uterine artery notching.

Description of included studies

A multicentre cohort study (2001)868 [EL = II] conducted in the UK examined the value of transvaginal uterine artery Doppler velocimetry at 23 weeks of gestation in the prediction of pre-eclampsia in singleton pregnancies. A total of 7851 women were analysed at 22–24 weeks of gestation. The presence of an early diastolic notch in the waveform was noted, and the mean pulsatility index of the two arteries was calculated. Screening characteristics in the prediction of pre-eclampsia were calculated.

A cohort study conducted in the UK (1997)869 [EL = II] aimed to establish the predictive value of transvaginal uterine artery Doppler studies in early pregnancy for the prediction of pre-eclampsia. A total of 626 women were included and the test was administered at between 12 and 16 weeks of gestation.

A case–control study in the UK (2003)870 [EL = II] aimed to evaluate the clinical usefulness of the Doppler velocimetry test used to screen pre-eclampsia in the period 2000–2001. A total of 895 women were included and the test was conducted at 20 weeks of gestation and then at 24 weeks.

A prospective study conducted in Germany (2005)871 [EL = II] examined the use of uterine artery Doppler at 19–22 weeks and 23–26 weeks of gestation in a low-risk population as a screening test for the prediction of pre-eclampsia. A total of 346 women were included.

Findings

The first study found a sensitivity of 25.4%, specificity of 90.9%, PPV of 2.5% and NPV of 99.3%. The overall incidence reported was 1.4%.

The second study reported an incidence of 4.8%. The sensitivity of the test was 92.9%, specificity was 85.1%, PPV was 23.6% and NPV was 99.5%.

An incidence of 2.9% was reported in the third study. The test sensitivity was 72%, specificity 94%, PPV 26% and NPV 99%.

The German study compared the results at 19–22 weeks versus 23–26 weeks of gestation. A sensitivity of 40% versus 67%, specificity of 82% versus 84%, PPV of 10% versus 17% and NPV of 97% versus 98% was reported for the two periods of gestation, respectively.

Integrated Doppler test with serum markers

A total of two studies were identified (Table 11.14).

Table 11.14. Integrated Doppler test with serum markers.

Table 11.14

Integrated Doppler test with serum markers.

Description of included studies

A prospective study in Turkey (2005)872 [EL = II] aimed to analyse the predictive power of maternal serum inhibin A, activin A, hCG, uE3, AFP levels and uterine artery Doppler, either alone or in combination, in the second trimester of pregnancy in screening for pre-eclampsia. A total of 178 women were included in whom serum samples were collected at between 16 and 18 weeks of gestation and Doppler investigation was performed at between 24 and 26 weeks of gestation.

A cohort study in France (2005)873 [EL = II] assessed the performance of early screening for pre-eclampsia and fetal growth restriction (FGR) by combining MSS with uterine Doppler ultrasound. A total of 2615 women were analysed in whom both a double test at between 14 and 18 weeks of gestation (by maternal serum AFP and total serum hCG assay) and a uterine Doppler ultrasound at between 18 and 26 weeks were performed.

Findings

The Turkish study found a 7.9% incidence of pre-eclampsia. The presence of a notch on Doppler investigation reported a sensitivity of 85.7% and specificity of 97.6%. The addition of high serum activin to the presence of a notch decreased the sensitivity to 78.6% and increased the specificity to 100%. The addition of high serum inhibin to the presence of a notch decreased the sensitivity to 71.4% and increased the specificity to 100%. The integrated test of presence of a notch or high serum activin increased the sensitivity to 100% and decreased the specificity to 86%.

In the French study, the bilateral notch test reported a sensitivity of 21.6% whereas the specificity was 95.9%. An integrated test-history of pre-eclampsia or bilateral notch or hCG > 2.5 MoM increased the sensitivity to 41.1% and reduced the specificity to 91.6%.

Time interval between pregnancies

Description of included studies

A Norwegian study (2002)531 [EL = 2+] used a large registry in Norway to evaluate the effects on the risk of pre-eclampsia of both the interbirth interval and a change of partner. A total of 551 478 women who had two or more singleton deliveries and 209 423 women who had three or more singleton deliveries were studied.

A retrospective cross-sectional study from Uruguay (2000)874 [EL = 3] studied the impact of interpregnancy interval on maternal morbidity and mortality. A total of 456 889 parous women delivering singleton infants were studied.

A Danish cohort study (2001)875 [EL = 2+] evaluated whether the interpregnancy interval may confound or modify the paternal effect on pre-eclampsia. The outcome of the second birth in a cohort of Danish women with pre-eclampsia in the previous birth (8401 women) and in all women with pre-eclampsia in second (but not first) birth together with a sample of women with two births (26 596 women) was studied.

Findings

The results from the Norwegian study showed that the risk in a second or third pregnancy was directly related to the time elapsed since the previous delivery. The association between risk of pre-eclampsia and interval was more significant than the association between risk and change of partner. When the interval was 10 years or more the risk of pre-eclampsia was about the same as that in nulliparous women. After adjustment for the presence or absence of a change of partner, maternal age and year of delivery, the probability of pre-eclampsia was increased by 1.12 for each year increase in the interval (OR 1.12, 95% CI 1.11 to 1.13).

The Uruguay study showed that women with more than 59 months between pregnancies had significantly increased risks of pre-eclampsia (RR 1.83, 95% CI 1.72 to 1.94) compared with women with intervals of 18–23 months. The authors concluded that interpregnancy intervals below 6 months and above 59 months are associated with an increased risk of adverse maternal outcomes.

The Danish study found that a long interval between pregnancies was associated with a significantly higher risk of pre-eclampsia in a second pregnancy when pre-eclampsia had not been present in the first pregnancy and paternity had not changed.

Blood pressure at booking

Description of included studies

A US-based study (1987)876 [EL = 2−] reviewed the outpatient charts of all patients with pre-eclampsia who received prenatal care at their clinics during the past 3 years. Thirty patients met their criteria for pre-eclampsia and were matched for age, race and parity with normotensive control subjects.

A US-based large clinical trial (1995)877 [EL = 1+] sought to determine whether any maternal demographic or clinical characteristics are predictive of pre-eclampsia. A total of 2947 healthy women with a single fetus were prospectively followed up from randomisation at 13–27 weeks of gestation to the end of pregnancy.

A population-based nested case–control Norwegian study (2000)878 [EL = 2+] studied the associations between established risk factors for pre-eclampsia and different clinical manifestations of the disease. A total of 323 cases of pre-eclampsia and 650 healthy controls were selected.

A US-based retrospective cohort study (2000)530 [EL = 2−] was undertaken to develop a clinical prediction rule for severe pre-eclampsia that was based on clinical risk factors and biochemical factors. Cases with severe pre-eclampsia were compared with control subjects with respect to clinical data and multiple-marker screening test results. Women were assigned a predictive score according to the presence or absence of predictive factors.

Findings

The first study found that both systolic and diastolic blood pressures were significantly higher (P < 0.05) in the first trimester for women with pre-eclampsia than for normal control subjects beginning in the first trimester. This difference persisted throughout pregnancy and was also present at the 6 week postpartum visit (P < 0.025).

The second study showed that higher systolic and diastolic blood pressures at the first visit were associated with an increased incidence of pre-eclampsia (3.8% in women with diastolic blood pressure of < 55 mmHg, 7.4% in those with diastolic blood pressure 70–84 mmHg). However, their recruitment was limited to women with a first blood pressure reading of ≤ 135/85 mmHg.

The third Norwegian study found that a systolic blood pressure ≥ 130 mmHg compared with < 110 mmHg at the first visit before 18 weeks was significantly associated with the development of pre-eclampsia later in pregnancy (adjusted OR 3.6, 95% CI 2.0 to 6.6). The association with a diastolic pressure ≥ 80 mmHg compared with < 60 mmHg was similar but not significant (adjusted OR 1.8, 95% CI 0.7 to 4.6).

The fourth study results showed that the only variables that remained significantly associated with severe pre-eclampsia were nulliparity (RR 3.8, 95% CI 1.7 to 8.3), history of pre-eclampsia (RR 5.0, 95% CI 1.7 to 17.2), elevated screening mean arterial pressure (RR 3.5, 95% CI 1.7 to 7.2), and low unconjugated estriol concentration (RR 1.7, 95% CI 0.9 to 3.4). This predictive model for severe pre-eclampsia, which included only these four variables, had a sensitivity of 76% and a specificity of 46%.

Proteinuria

Description of included studies

A US-based retrospective study (1992)879 [EL = 2−] evaluated varying degrees of chronic proteinuria as a predictor of pregnancy outcome. Their purpose was to determine the significance of otherwise ‘asymptomatic’ proteinuria identified during pregnancy. Perinatal outcomes of 65 pregnancies in 53 women with the following criteria were studied: proteinuria exceeding 500 mg per day, no previously known renal disease, no reversible renal dysfunction, and no evidence for pre-eclampsia at discovery.

Findings

The results showed that 58% of the women with proteinuria combined with renal insufficiency developed pre-eclampsia. 100% of women with proteinuria combined with chronic hypertension developed pre-eclampsia whereas 77% of women with with all three together developed pre-eclampsia.

Evidence summary for Section 11.2

Given the quality, level and precision of the evidence, no single test has emerged as a front runner in the quest to predict and prevent pre-eclampsia. Tests that offer high specificity, such as AFP, β-hCG and uterine artery Doppler (bilateral notching), have the potential to minimise unwarranted inconvenience, expense and morbidity associated with false positive results. There is evidence to show that when the interval between two pregnancies is 10 years or more the risk of pre-eclampsia is about the same as that in nulliparous women.

GDG interpretation of evidence

None of the current screening tests offer a high enough diagnostic value, all being EL II, to be used in routine care. In addition, the purpose of screening for pre-eclampsia is only to identify those women who require additional care since there is no effective intervention. However, the following risk factors for the development of pre-eclampsia should be noted:

  • age 40 years or over
  • nulliparity
  • pregnancy interval of more than 10 years
  • family history of pre-eclampsia
  • previous history of pre-eclampsia
  • BMI of 30 kg/m2 or over
  • pre-existing vascular disease such as hypertension
  • pre-existing renal disease
  • multiple pregnancy.

The routine measurement of blood pressure and of proteinuria should be undertaken on the schedule outlined in the algorithm.

Recommendations on screening for pre-eclampsia

Blood pressure measurement and urinalysis for protein should be carried out at each antenatal visit to screen for pre-eclampsia.

At the booking appointment, the following risk factors for pre-eclampsia should be determined:

  • age 40 years or older
  • nulliparity
  • pregnancy interval of more than 10 years
  • family history of pre-eclampsia
  • previous history of pre-eclampsia
  • body mass index 30 kg/m2 or above
  • pre-existing vascular disease such as hypertension
  • pre-existing renal disease
  • multiple pregnancy.

More frequent blood pressure measurements should be considered for pregnant women who have any of the above risk factors.

The presence of significant hypertension and/or proteinuria should alert the healthcare professional to the need for increased surveillance.

Blood pressure should be measured as outlined below:

  • remove tight clothing, ensure arm is relaxed and supported at heart level
  • use cuff of appropriate size
  • inflate cuff to 20–30 mmHg above palpated systolic blood pressure
  • lower column slowly, by 2 mmHg per second or per beat
  • read blood pressure to the nearest 2 mmHg
  • measure diastolic blood pressure as disappearance of sounds (phase V).

Hypertension in which there is a single diastolic blood pressure of 110 mmHg or two consecutive readings of 90 mmHg at least 4 hours apart and/or significant proteinuria (1+) should prompt increased surveillance.

If the systolic blood pressure is above 160 mmHg on two consecutive readings at least 4 hours apart, treatment should be considered.

All pregnant women should be made aware of the need to seek immediate advice from a healthcare professional if they experience symptoms of pre-eclampsia. Symptoms include:

  • severe headache
  • problems with vision, such as blurring or flashing before the eyes
  • severe pain just below the ribs
  • vomiting
  • sudden swelling of the face, hands or feet.

Although there is a great deal of material published on alternative screening methods for pre-eclampsia, none of these has satisfactory sensitivity and specificity, and therefore they are not recommended.

Research recommendation on screening for pre-eclampsia

Further research using large prospective studies should be conducted into the effectiveness and cost-effectiveness of using alpha-fetoprotein, beta-human chorionic gonadotrophin, fetal DNA in maternal blood and uterine artery Dopplers, or potentially a combination of these, to detect women at risk of developing pre-eclampsia. Testing should focus particularly on the prediction of early-onset pre-eclampsia, with priority given to blood tests.

11.3. Preterm birth

Clinical question

What is the diagnostic value of the following screening methods in identifying women at risk of preterm labour?

  • history
  • vaginal examinations
  • ulrasound scan – cervical length up to 22 weeks of pregnancy
  • oral health/dental health
  • swabs for bacterial vaginosis.

Previous NICE guidance (for the updated recommendations see below)

Routine vaginal examination to assess the cervix is not an effective method of predicting preterm birth and should not be offered. [A]

Although cervical shortening identified by transvaginal sonography (TVS) and increased levels of fetal fibronectin (FFN) are associated with an increased risk of preterm birth, the evidence does not indicate that this information improves outcomes; therefore neither TVS nor FFN should be used to predict preterm birth in healthy pregnant women. [B]

11.3.1. Introduction and background information

In the UK approximately 7% of births occur prior to 36 completed weeks of gestation and 1.4% prior to 31 completed weeks (figures for England, NHS Maternity Statistics 2003–2004). According to CEMACH, more than 70% of all neonatal deaths occur in preterm babies, that is, birth of a baby before 37 weeks of completed gestational age (Perinatal mortality surveillance, 2004, England, Wales and Northern Ireland, CEMACH, www.cemach.org.uk/publications.aspx). It is an important cause of major and minor morbidity such as necrotising enterocolitis, bronchopulmonary dysplasia, intraventricular haemorrhage, cerebral palsy and cognitive impairment during the early years of life. Even after infancy, these babies are at increased risk of developing chronic diseases in adult life.

Forty-four papers from 38 studies have been included in this review for evaluating diagnostic accuracy of the following twelve screening tests:

  1. previous history of spontaneous preterm birth (SPTB)
  2. clinical/digital examination
  3. cervico-vaginal fetal fibronectin (FFN) levels
  4. cervico-vaginal interleukin-6 (IL-6) levels
  5. cervico-vaginal interleukin-8 (IL-8) levels
  6. maternal serum alpha-fetoprotein (MSAFP) levels
  7. maternal serum beta-human chorionic gonadotrophin (MSHCG) levels
  8. maternal serum C-reactive protein (CRP) levels
  9. asymptomatic bacteriuria
  10. bacterial vaginosis (BV)
  11. transvaginal sonography (TVS) for cervical length
  12. transvaginal sonography for funnelling of cervix.

Most of the studies included for this review are prospective cohort studies. High-quality studies with EL 1 were identified and included for evaluating diagnostic accuracy of the following screening tests: previous history of SPTB, cervico-vaginal FFN levels, BV using Nugent’s criteria for Gram staining, and TVS for cervical length and funnelling. For other screening tests, the evidence level of included studies was predominantly 2 or 3 owing to two main reasons: absence of blinding and/or study population not being representative of the reference population. Only studies conducted on asymptomatic women (with no signs and symptoms of preterm labour) were considered for this review. Since most of the studies identified for cervico-vaginal IL-6 and IL-8 and serum CRP tests were conducted in symptomatic women (with threatened preterm labour), only a few quality studies remained for these tests for asymptomatic women.

Details of the screening tests including timing, frequency and thresholds have been specified where possible. The outcome assessed was spontaneous preterm delivery at less than 37 weeks (SPTB < 37 weeks), and efforts were made to calculate the diagnostic value of the tests after excluding cases of induced preterm delivery (PTD). Many studies had evaluated screening performance of various tests for outcome with different gestational age (for example < 32, 33 or 35 weeks), but for the sake of comparison results have been provided for commonly used thresholds and SPTB < 37 weeks as the outcome. Wherever possible, the incidence of SPTB and prevalence of test positive have also been calculated.

Studies included in the review of each screening test have been tabulated in decreasing order of their evidence level. In case of those with similar evidence levels, priority is given to the study with a larger sample size.

11.3.2. History of previous spontaneous preterm birth (SPTB)

Description of included studies

Three studies were included – two prospective cohort [EL = Ib] and one retrospective cohort. [EL = II] All were multicentre studies with good sample sizes. Although the thresholds of the screening tests were different in these studies and outcomes other than SPTB < 37 weeks were also evaluated, the results have been given for history of previous SPTB > 20 weeks as the screening test and outcome SPTB < 37 weeks only (Table 11.15).

Table 11.15. Characteristics of included studies on diagnostic value of maternal history of previous SPTB.

Table 11.15

Characteristics of included studies on diagnostic value of maternal history of previous SPTB.

Findings

In the three studies sensitivity and specificity ranged from 19% to 67% and 73% to 97%, respectively. The test had high LR+ of 5.78 (95% CI 4.47 to 7.46) in one study,882 but LR− was 0.84 (95% CI 0.80 to 0.89) and it was a study with EL 2. For the studies with EL 1, values of LR+ ranged from 2.26 to 2.74 and LR− from 0.45 to 0.77. On meta-analysis, significant statistical heterogeneity (P < 0.00001) was observed for both the positive and negative LR. The summary LR+ was 2.83 (95% CI 2.53 to 3.16) and summary LR− was 0.76 (95% CI 0.72 to 0.80) (Figure 11.1).

Figure 11.1. Likelihood ratios for history of previous SPTB for screening for preterm birth.

Figure 11.1

Likelihood ratios for history of previous SPTB for screening for preterm birth.

Evidence summary

Evidence indicates that history of previous SPTB does not seem to have high diagnostic value in predicting and ruling out SPTB in the current pregnancy.

11.3.3. Clinical examination

Description of included studies

Five prospective cohort studies were included – one with EL Ib and four with EL II, the reason being absence of blinding in these studies. In the study with EL Ib, Bishop score was used for screening and clinical examination carried out four times in each woman. In studies with EL II, difference was observed in the frequency, timing and threshold of the screening test used. Owing to existing heterogeneity, meta-analysis was not performed. Values for positive and negative LR have been presented separately for the two most commonly used signs at clinical examination – cervical dilatation (in four studies) and short cervix (in two studies) (Table 11.16).

Table 11.16. Characteristics of included studies on diagnostic value of vaginal digital examination.

Table 11.16

Characteristics of included studies on diagnostic value of vaginal digital examination.

Findings

For cervical dilatation, sensitivity and specificity ranged from 13% to 57% and 57% to 98%, respectively. The study by Leveno et al.887 had a high LR+ of 9.25 (95% CI 3.91 to 21.85), but LR− was 0.46 (95% CI 0.19 to 1.08). Chambers et al.885 had moderate values for LR+ and LR− of 2.16 and 0.76, respectively. LRs for the other two studies were not as good as those of the two studies just mentioned (Figure 11.2).

Figure 11.2. Likelihood ratios for cervical dilatation for screening for preterm birth.

Figure 11.2

Likelihood ratios for cervical dilatation for screening for preterm birth.

Sensitivity for a short cervix diagnosed clinically ranged from 11% to 21% and specificity from 89% to 95%. Chambers et al.885 had a better LR− of 0.88 (95% CI 0.81 to 0.97) among the included studies, but LR+ was 1.96 (95% CI 1.41 to 2.74) (Figure 11.3).

Figure 11.3. Likelihood ratios for length of cervix for screening for preterm birth.

Figure 11.3

Likelihood ratios for length of cervix for screening for preterm birth.

Evidence summary

A wide variation in results of screening accuracy is observed for different clinical methods for predicting SPTB. Evidence shows that clinical examination has poor diagnostic value in predicting and ruling out SPTB.

11.3.4. Cervico-vaginal fetal fibronectin levels

Description of included studies

The six studies concerning this test were prospective cohort studies and blinding was specified in all. In two EL II studies the dropout rate was more than 40% while the rest were classified EL Ib. The population was low-risk singleton pregnancies in all studies. A single swab in the second trimester at different gestational ages was taken usually from the posterior vaginal fornix, and the threshold used for a positive test was FFN levels ≥ 50 ng/ml. Meta-analysis was performed for the predictive accuracy of a single test in the second trimester with outcome SPTB < 37 weeks. One good-quality study was excluded from meta-analysis as it evaluated SPTB < 33 weeks as the outcome (Table 11.17).

Table 11.17. Characteristics of included studies on diagnostic value of cervico-vaginal fetal fibronectin levels.

Table 11.17

Characteristics of included studies on diagnostic value of cervico-vaginal fetal fibronectin levels.

Findings

Sensitivity ranged from 13% to 55% and specificity from 83% to 99% for the test in predicting SPTB < 37 weeks. In the study that used < 33 weeks as the time for the outcome, sensitivity and specificity were 33% and 97%, respectively.

For the individual studies, LR+ ranged from 2.19 (95% CI 1.08 to 4.47) to as high as 18.00 (95% CI 3.21 to 100.86), and LR− from 0.92 (95% CI 0.83 to 1.02) to a low of 0.53 (95% CI 0.26 to 1.11). The study with the highest LR+ (Chang et al.889) had an LR− of 0.84, but the confidence interval (CI) crossed unity. Similarly, Crane et al.892 had the best value for LR− but again the CI crossed unity.

No statistically significant heterogeneity was observed for either LR+ and LR− on performing meta-analysis. The summary LR values for a positive test was 3.53 (95% CI 2.78 to 4.49) and for the negative test was 0.86 (95% CI 0.82 to 0.90) (Figure 11.4).

Figure 11.4. Likelihood ratios for fetal fibronectin for screening for preterm birth.

Figure 11.4

Likelihood ratios for fetal fibronectin for screening for preterm birth.

Evidence summary

There is high-quality evidence to show that a single second-trimester cervico-vaginal swab with a positive result for fibronectin levels has moderate value in predicting SPTB < 37 weeks, but a negative result decreases the probability of SPTB only minimally.

11.3.5. Cervico-vaginal IL-6 levels

Description of included studies

There were three studies included for this test, all with EL II – a prospective cohort study and two nested case–control studies. All had a small sample size. The timing and frequency of screening tests, thresholds used for a positive test, and outcomes assessed were different in all the three studies. Meta-analysis was not conducted and results have been presented separately for each study (Table 11.18).

Table 11.18. Characteristics of included studies on diagnostic value of cervico-vaginal IL-6 levels.

Table 11.18

Characteristics of included studies on diagnostic value of cervico-vaginal IL-6 levels.

Findings

In the three studies, sensitivity ranged from 9% to 50% while specificity ranged from 84% to 90%. Best values for the LRs were obtained for the prospective cohort study (Lockwood et al.893). For the threshold > 250 pg/ml, it showed an LR+ of 3.34 (95% CI 1.96 to 5.70) and LR− of 0.59 (95% CI 0.42 to 0.83). Results from the other prospective cohort study (Inglis et al.894) were in complete contrast. Values obtained in the study for LR+ and LR− were poor: 0.56 (95% CI 0.08 to 3.97) for LR+ and 1.08 (95% CI 0.87 to 1.35) for the LR−. In the nested case–control study, LR for a positive test was 2.08 (95% CI 1.10 to 3.96) and for a negative test was 0.88 (95% CI 0.80 to 0.98) (Figure 11.5).

Figure 11.5. Likelihood ratios for cervico-vaginal IL-6 levels for screening for preterm birth.

Figure 11.5

Likelihood ratios for cervico-vaginal IL-6 levels for screening for preterm birth.

Evidence summary

Although studies on diagnostic performance of cervico-vaginal IL-6 levels in asymptomatic women are limited, the available evidence shows that it has poor screening accuracy for SPTB.

11.3.6. Cervico-vaginal IL-8 levels

Description of included studies

Two prospective cohorts were included – both with EL II and carried out by the same principal author in Japan. Blinding was not specified in either study. In the study with a bigger sample size, IL-8 was measured serially in the cervico-vaginal fluid – initially once at 20–23 weeks and then biweekly at 24–28 weeks. The threshold for a positive test was also different in the two studies. Owing to heterogeneity of the test timing, frequency and threshold values, meta-analysis was not performed (Table 11.19).

Table 11.19. Characteristics of included studies on diagnostic value of cervico-vaginal IL-8 levels.

Table 11.19

Characteristics of included studies on diagnostic value of cervico-vaginal IL-8 levels.

Findings

The larger study with serial testing showed sensitivity and specificity of 27% and 80%, respectively. It had an LR+ of 1.38 (95% CI 1.04 to 1.82) and LR− of 0.91 (95% CI 0.82 to 1.01) for predicting SPTB < 37 weeks. Another study with a smaller sample size showed better results for all values. Sensitivity was 42%, specificity 85%, LR+ 2.75 (95% CI 1.68 to 4.52) and LR− 0.67 (95% CI 0.30 to 1.15). In both studies, the confidence interval for the LR− crossed unity (Figure 11.6).

Figure 11.6. Likelihood ratios for cervico-vaginal IL-8 levels for screening for preterm birth.

Figure 11.6

Likelihood ratios for cervico-vaginal IL-8 levels for screening for preterm birth.

Evidence summary

Although the evidence is limited, it shows that the likelihood of SPTB < 37 weeks is increased minimally with a positive test for cervico-vaginal IL-8 levels.

11.3.7. Maternal serum alpha-fetoprotein (MSAFP) levels

Description of included studies

Three prospective cohort studies were included for this test but there was no blinding in two studies [EL = II] where retrospective analysis of data was done. In all studies, the screening test was performed at 15–20 weeks as part of routine screening for Down’s syndrome and NTD. AFP levels ≥ 2.0 MoM was the threshold used in two studies. In two studies outcome was defined as SPTB < 37 weeks while the third looked at SPTB < 32 weeks. As studies had different thresholds and outcome, they were not combined and the results are presented individually (Table 11.20).

Table 11.20. Characteristics of included studies on diagnostic value of maternal serum AFP levels.

Table 11.20

Characteristics of included studies on diagnostic value of maternal serum AFP levels.

Findings

The range of sensitivity was from 2% to 19% and for specificity from 80% to 99%. The study with the highest level of evidence had poor values for both LR+ (0.97 (95% CI 0.51 to 1.85)) and LR− (1.01 (95% CI 0.86 to 1.17)). The study by Dugoff et al.899 had a high LR+ of 6.80 (95% CI 4.75 to 9.74) but the LR− was only 0.91 (95% CI 0.87 to 0.95) for outcome SPTB < 32 weeks (Figure 11.7).

Figure 11.7. Likelihood ratios for maternal serum alpha-fetoprotein levels for screening for preterm birth.

Figure 11.7

Likelihood ratios for maternal serum alpha-fetoprotein levels for screening for preterm birth.

Evidence summary

Positive and negative results of MSAFP at 15–20 weeks seem to have poor predictive accuracy for SPTB, although the evidence is limited.

11.3.8. Maternal serum beta-human chorionic gonadotrophin (MSHCG) levels

Description of included studies

The three studies included were prospective cohort studies [EL = II] without blinding. Data were analysed retrospectively in two studies. In two studies the screening test was performed in the first trimester, while in the third it was done in the second trimester. The study population was low risk in all. The threshold of a positive test and outcome were different in all studies (Table 11.21).

Table 11.21. Characteristics of included studies on diagnostic value of maternal serum beta-hCG levels.

Table 11.21

Characteristics of included studies on diagnostic value of maternal serum beta-hCG levels.

Findings

In the study with the largest sample size (Dugoff et al.899), carried out in the second trimester for predicting SPTB < 32 weeks, values for sensitivity, specificity, LR+ and LR− were 17%, 94%, 2.87 (95% CI 2.18 to 3.78) and 0.89 (95% CI 0.84 to 0.94), respectively. In the other two first-trimester studies, wide variation was observed in all the results. Sensitivity and specificity ranged from 5% to 73% and from 21% to 95%, respectively. The confidence intervals of both the LR+ and LR− included value of 1 and gave poor probability for the test results (Figure 11.8).

Figure 11.8. Likelihood ratios for maternal serum beta-human chorionic gonadotrophin levels for screening for preterm birth.

Figure 11.8

Likelihood ratios for maternal serum beta-human chorionic gonadotrophin levels for screening for preterm birth.

Evidence summary

A positive test for a second-trimester MSHCG is more useful in predicting SPTB < 32 weeks than a negative test in ruling it out, but the evidence is poor. The screening performance of a first-trimester MSHCG test is poor.

11.3.9. Maternal serum C-reactive protein (CRP) levels

Description of included studies

Two nested case–control studies without blinding [EL = III] were identified. One study was conducted in the first trimester and used CRP levels greater than 4.3 ng/ml as the threshold for a positive test, while the other carried out in the second trimester used 7.6 ng/ml as the cut-off. Both evaluated SPTB < 37 weeks as outcome (Table 11.22).

Table 11.22. Characteristics of included studies on diagnostic value of maternal serum CRP levels.

Table 11.22

Characteristics of included studies on diagnostic value of maternal serum CRP levels.

Findings

The first-trimester study showed sensitivity of 35% and specificity of 78%. LR+ was 1.55 (95% CI 1.12 to 2.13) and LR− was 0.84 (95% CI 0.73 to 0.98). In the second-trimester study sensitivity and specificity was 26% and 86%, and values for LR+ and LR− were 1.81 (95% CI 1.12 to 2.13) and 0.86 (95% CI 0.76 to 0.99), respectively (Figure 11.9).

Figure 11.9. Likelihood ratios for maternal serum CRP levels for screening for preterm birth.

Figure 11.9

Likelihood ratios for maternal serum CRP levels for screening for preterm birth.

Evidence summary

There is a lack of good-quality studies on the diagnostic value of maternal serum CRP levels. Evidence from level III studies shows that positive and negative results of maternal serum CRP have poor predictive accuracy for SPTB < 37 weeks.

11.3.10. Asymptomatic bacteriuria

Description of included studies

All the four prospective cohort studies with EL II included for this test did not specify blinding as a study criterion. Three of these studies were conducted in the 1960s. All of them used culture of midstream urine sample (MSU) as the screening test, and in two studies it was repeated after the first positive test to confirm asymptomatic bacteriuria. The outcome evaluated was SPTB < 37 weeks in all. In two studies the sample size was very small compared with the study population as treatment was started later during the study and that population was excluded. Meta-analysis was performed to calculate summary LRs for a positive and negative test taking results from the firstly performed urine analysis only where possible (Table 11.23).

Table 11.23. Characteristics of included studies on diagnostic value of asymptomatic bacteriuria by midstream urine testing.

Table 11.23

Characteristics of included studies on diagnostic value of asymptomatic bacteriuria by midstream urine testing.

Findings

Sensitivity ranged from 7% to 30% and specificity from 65% to 97%. Statistically no significant heterogeneity was observed for either the LR+ or the LR−. The summary value of LR+ was 1.97 (95% CI 1.45 to 2.68) and the range in individual studies was from 0.89 to 2.63. LR− had a summary value of 0.46 (95% CI 0.31 to 0.67) and range of 1.19 to 0.31 (Figure 11.10).

Figure 11.10. Likelihood ratios for midstream urine testing for screening for preterm birth.

Figure 11.10

Likelihood ratios for midstream urine testing for screening for preterm birth.

Evidence summary

A negative result for an MSU sample for asymptomatic bacteriuria has good diagnostic value in ruling out SPTB < 37 weeks compared with a positive result for predicting it, but the evidence is not of high quality.

11.3.11. Bacterial vaginosis

Description of included studies

Five studies were included – all prospective cohort studies with EL Ib and II were conducted in more than one centre. The study population was low risk in four studies and risk status was not specified in the last study. In all studies, swab (usually single) was taken from the posterior vaginal fornix in the late first or second trimester, and Gram staining with Nugent’s criterion was used to diagnose BV. In one study (Hillier et al.910) results were calculated only for those women who did not receive antibiotics. All the studies used SPTB < 37 weeks as the outcome. Meta-analysis was performed for LR of a single test in the second trimester for predicting SPTB < 37 weeks (Table 11.24).

Table 11.24. Characteristics of included studies on diagnostic value of Gram staining (Nugent’s criteria) for bacterial vaginosis.

Table 11.24

Characteristics of included studies on diagnostic value of Gram staining (Nugent’s criteria) for bacterial vaginosis.

Review findings

In the studies, BV had a sensitivity ranging from 15% to 44% and specificity from 76% to 93%. For the LRs of individual studies, Purwar et al.911 had the best results, with a high LR+ value of 5.31 (95% CI 3.84 to 7.33) and a low LR− of 0.54 (95% CI 0.42 to 0.71). When the results of all the included studies were combined, significant statistical heterogeneity was observed for both LR+ and LR−, and the summary values obtained were not as good as those for individual studies. The summary LR+ was 1.70 (95% CI 1.49 to 1.94) and the summary LR− was 0.88 (95% CI 0.85 to 0.92) (Figure 11.11).

Figure 11.11. Likelihood ratios for swabbing for bacterial vaginosis for screening for preterm birth.

Figure 11.11

Likelihood ratios for swabbing for bacterial vaginosis for screening for preterm birth.

Evidence summary

There is high-quality evidence that a single second-trimester vaginal swab for BV (using Nugent’s criterion on Gram staining) has poor diagnostic value as a screening test for SPTB < 37 weeks.

11.3.12. Transvaginal sonography (TVS) for cervical length

Description of included studies

Of the five prospective cohort studies included for reviewing this test, four had EL Ib and one EL II because blinding was not a study criterion. In three studies the population was made up of both low- and high-risk pregnant women, while the other two studies had only a low-risk population. TVS for measuring cervical length was carried out in all studies in the second trimester. The critical length used for labelling a cervix as ‘short’ was calculated by ROC curve in two studies, while in others the length varied. However, all studies used a cervical length of ≤ 20 or 25 mm, and this length was used to conduct the meta-analysis. The outcome evaluated was SPTB < 37 weeks for all but one study which assessed SPTB < 34 weeks (Table 11.25).

Table 11.25. Characteristics of included studies on diagnostic value of cervical length by TVS.

Table 11.25

Characteristics of included studies on diagnostic value of cervical length by TVS.

Findings

Sensitivity ranged from 5% to 26% and specificity from 93% to 100%. Fukami et al.914 had the best LRs for positive and negative test results compared with other studies, but it was a study with EL II. Its LR+ was 34.34 (95% CI 16.18 to 72.88) and LR− was 0.51 (95% CI 0.25 to 1.01). On conducting meta-analysis of studies using data for common thresholds, significant statistical heterogeneity was observed for both LR+ and LR−. The summary LR+ was 3.84 (95% CI 3.12 to 4.17) and LR− was 0.85 (95% CI 0.82 to 0.89) (Figure 11.12).

Figure 11.12. Likelihood ratios for TVS for cervical length for screening for preterm birth.

Figure 11.12

Likelihood ratios for TVS for cervical length for screening for preterm birth.

Evidence summary

High-quality evidence shows that a shortened cervix (length ≤ 25 mm) on TVS in the second trimester increases the likelihood of SPTB < 37 weeks by a moderate value, but a cervical length of greater than 25 mm is poor at ruling it out.

11.3.13. Funnelling by TVS

Description of included studies

All the included studies were prospective cohorts (three with EL Ib, one with EL II). The population was low risk in one study, both low and high risk in two studies, and not specified in the fourth one. TVS was carried out in all studies in the second trimester, but different thresholds were used to define ‘funnelling’. The outcome evaluated was not the same in all studies. Owing to heterogeneity in thresholds and outcome, meta-analysis was not performed (Table 11.26).

Table 11.26. Characteristics of included studies on diagnostic value of cervical funnelling by TVS.

Table 11.26

Characteristics of included studies on diagnostic value of cervical funnelling by TVS.

Findings

For the EL Ib studies, sensitivity ranged from 9% to 32% and specificity from 94% to 96%. The only study with EL II had a sensitivity of 27% and specificity of 97%. On calculating the LR for positive and negative test results, all the studies showed better results for LR+ than for LR−. Among EL I studies, Leung et al.913 had the best results, with an LR+ of 5.32 (95% CI 2.70 to 10.48) and LR− of 0.73 (95% CI 0.54 to 0.99). The other two studies with EL I had lower LR+ and higher LR− values than the Leung study. In the To et al. study915 [EL = II], values for LR+ and LR− were 7.91 (95% CI 5.11 to 12.27) and 0.75 (95% CI 0.65 to 0.88), respectively (Figure 11.13).

Figure 11.13. Likelihood ratios for funnelling by transvaginal sonography for screening for preterm birth.

Figure 11.13

Likelihood ratios for funnelling by transvaginal sonography for screening for preterm birth.

Evidence summary

Funnelling detected by TVS in the second trimester seems to have moderate diagnostic value in predicting SPTB, but interpretation of the evidence is made difficult by variation in thresholds and outcome.

GDG interpretation of evidence for Section 11.3

The evidence does not justify the routine screening of low-risk women for preterm labour with clinical examination, asymptomatic bacteriuria, vaginal swabs or ultrasound to assess cervical change. The evidence shows possible moderate specificity but very poor sensitivity.

Recommendation on screening for preterm birth

Routine screening for preterm labour should not be offered.

Research recommendation on screening for preterm birth

There is need for future research investigating the value of tests that are cheap and easy to perform such as maternal serum human chorionic gonadotrophin (MSHCG), serum C-reactive protein (CRP) and cervico-vaginal fetal fibrinonectin levels. The diagnostic accuracy and cost-effectiveness of transvaginal ultrasound to measure cervical length and funnelling to identify women at risk of preterm labour should also be investigated.

11.4. Placenta praevia

Placenta praevia occurs when the placenta covers the internal os and obstructs vaginal delivery of the fetus. A higher rate of pregnancy complications, including abruption placenta, antepartum haemorrhage and intrauterine growth restriction has been reported in women with low-lying placentas identified in the second trimester, despite apparent ‘resolution’ by the time of delivery.547 [EL = 3]

Evaluation of transvaginal sonography for placental localisation has been shown to be safe in observational studies549–550 [EL = 3] and more accurate than transabdominal sonography in one RCT.551 [EL = 1b] Reported sensitivities range from 88% to 100% and false positives and false negatives are rare.549,552 [EL = 3]

Using ultrasonography, placenta praevia may be detected early in pregnancy. However, many placentas that appear to cover the cervical os in the second trimester will not cover the os at term. In one cohort study (n = 6428 women), 4.5% of women were identified with a placenta extending over the internal os at 12 to 16 weeks of gestation with transvaginal sonographic screening and only 0.16% (10/6428) of these women had placenta praevia at birth. Eight of the ten women with placenta praevia had been identified prior to delivery and, in all eight of these women, the placenta extended 15 mm or more over the internal os at the initial scan.553 [EL = 2b]

In another cohort study, among women scanned transvaginally at 18 to 23 weeks of gestation (n = 3696 women), 1.5% had a placenta extending over the internal os.554 At delivery, 0.14% of women had placenta praevia and, again, the placenta covered the internal os by 15 mm or more at the time of the first scan for all five of the women. With a cut-off of 15 mm, 0.7% (27/3696) of women would have screened ‘positive’ and all five cases of praevia at delivery would have been identified (i.e., positive predictive value 19% and sensitivity 100%). [EL = 2b]

Similarly, a cross-sectional study which examined 1252 women who underwent ultrasound examination from 9 to 13 weeks of gestation found that although 6.2% (77/1252) of women had a placenta extending over the internal cervical os at initial examination, only 0.32% (4/1252) of the cases persisted to delivery.555 In all four cases, the edge of the placenta extended over the os by more than 15 mm during the first-trimester ultrasound examination. [EL = 3]

With regard to gestational age at the time of detection, later detection appears to be related to likelihood of persisting until delivery. A retrospective study demonstrated that, among women with placenta praevia at 15 to 19 weeks of gestation, 12% persisted until delivery compared with 73% among women in whom placenta praevia was identified at 32 to 35 weeks of gestation.556 [EL = 3]

Symptomatic placenta praevia is associated with the sudden onset of painless bleeding in the second or third trimester. Women with placenta praevia are reported to be 14 times more likely to bleed in the antenatal period compared with women without placenta praevia.557 Risk factors for symptomatic placenta praevia include prior history of placenta praevia, advancing maternal age, increasing parity, smoking, cocaine use, previous caesarean section and prior spontaneous or induced abortion.558,559 [EL = 2a]

In the case of symptomatic placenta praevia, inpatient management has been recommended560 [EL = 4] and no conclusive evidence contrary to this recommendation was located. A Cochrane review of interventions for the management of placenta praevia compared home with hospitalisation and cervical cerclage with no cerclage.561 Only three trials with a total of 114 women were identified and although a reduction of length of stay in hospital was observed no other significant differences were found to support inpatient or outpatient management. [EL = 1a] Three trials of such small size were considered insufficient evidence to support a change in practice.

GDG interpretation of evidence

There is evidence that detection of a placenta extending over the internal os early in pregnancy (studies range from 9–19 weeks) is associated with persistent placenta praevia in 6–12% of women. If a placenta is found to extend over the internal cervical os at 32 to 35 weeks this persists until birth in 73% of women. The GDG noted that in clinical practice a number of women were presenting with vaginal bleeding between 32 and 36 weeks. The GDG expressed concern that the previous (2003) recommendation of repeating an ultrasound scan at 36 weeks following earlier detection of placenta praevia was too late in pregnancy and that the number of women experiencing antepartum haemorrhage would be reduced by bringing forward the timing of the repeat scan to 32 weeks. For this reason the following recommendation has been updated.

Recommendation on screening for placenta praevia

Because most low-lying placentas detected at the routine anomaly scan will have resolved by the time the baby is born, only a woman whose placenta extends over the internal cervical os should be offered another transabdominal scan at 32 weeks. If the transabdominal scan is unclear, a transvaginal scan should be offered.