|Home | About | Journals | Submit | Contact Us | Français|
The conventional screening test for gestational diabetes mellitus is measurement of plasma glucose 1 hour after 50 g glucose by mouth. The sensitivity and specificity of this test are lower than desirable; we therefore developed an index including other plasma constituents. In a preliminary study, 138 pregnant women had the standard oral glucose load screening test, and plasma fructosamine and total proteins were measured, in addition to glucose, in the 1-hour samples. An index value (I) was calculated as [fructosamine (μmol/L)÷total proteins (g/L)]×[glucose (mmol/L)÷100]. Cut-off values for I were then assessed in a second prospective study, of 642 pregnant women. Definitive diagnosis of gestational diabetes was by oral glucose tolerance test (OGTT). The index was also assessed in terms of fetal macrosomia (birthweight≥4000 g).
With a cut-off value of I=27.2, sensitivity was 98%, specificity 89%, diagnostic efficiency 90%, positive likelihood ratio 8.76. Application of the index would have avoided 42% of the OGTTs demanded by the standard screening test, reducing false positives from about 24% to 10%. Predictive efficacy for macrosomia was 10.3% versus 7.9%.
Our index offers an efficient screening test for gestational diabetes, and with more stringent cut-off points may be applicable as a single-step diagnostic procedure.
Gestational diabetes mellitus (GDM)1, a common metabolic alteration in pregnancy, is important because of the obstetric repercussions—fetal macrosomia, hypoglycaemia, polycythaemia and hyperbilirubinaemia, pre-eclampsia in the mother, and neonatal death2. GDM was first described in 1952, when Jackson and co-workers observed an association between the grade of glucose intolerance and perinatal morbidity and mortality. In 1973 O'Sullivan et al.3 proposed a screening test (ST) for the early detection of GDM—plasma glucose measurement 1 hour after oral administration of 50 g glucose. The second (1985) and third (1991) International Workshop Conferences on Gestational Diabetes Mellitus4 suggested that this test should be performed between weeks 24 and 28 of gestation; and, if the result was positive, the patient should then undergo a glucose tolerance test (OGTT) with 100 g oral glucose and plasma glucose measurements at baseline and 1, 2 and 3 hours post-load. This was considered to provide the definitive diagnosis of GDM.
According to the published work the ST has a sensitivity of about 78% and a specificity of 83%5,6. In the present study we sought to devise an index that would improve on these figures, offering better screening capability and possibly greater ease of diagnosis.
We conducted a preliminary prospective study with 138 pregnant women (excluding those who were expecting multiple births) who attended our routine antenatal clinic. They were told that the study, which had been approved by the governmental health authority, would involve additional inconvenience. All who gave their consent were entered consecutively into the study and all, irrespective of risk-factor status for GDM, underwent, between weeks 24 and 28 of gestation, the O'Sullivan ST3 with 50g oral glucose load followed by determination of plasma glucose at 1 hour. This is the standard ST in our hospital. For the purposes of the present study we measured, in addition to glucose, the concentrations of fructosamine, total proteins, albumin and HbA1c. A positive ST was registered if plasma glucose was ≥ 7.8 mmol/L. Irrespective of whether the ST was positive or negative, all women had a full OGTT. For this, 100g glucose was given after a fast of 8-14 hours. The women were advised to take a carbohydrate-rich diet (≥ 150 g/d) and be moderately active during the three days before the test. During the test they rested and refrained from eating or smoking. The concentrations of glucose were measured at baseline (fasting) and 1, 2 and 3 hours. As recommended by the third Workshop4, GDM was diagnosed when at least two of the following glucose concentrations were observed: ≥ 5.8 mmol/L at baseline; ≥ 10.5 mmol/L at 1 hour, ≥ 9.1 mmol/L at 2 hours and ≥ 8.0 mmol/L at 3 hours.
From this initial study we generated an index (I) that statistically discriminated between those women who were subsequently identified (on the results of the OGTT) as having GDM and those who did not. The index was then prospectively applied to a further 642 women attending the antenatal clinic. These women fulfilled the same criteria as those taking part in the preliminary study. Out of the initial 642, we were able to obtain outcome information—birthweight, caesarean sections, dystocia scores, and complications—in 578.
The oral glucose load (50 g or 100 g) was a commercial preparation (Biomedics SL, Spain), administered chilled to minimize nausea, vomiting and abdominal discomfort7,8. Plasma glucose was determined by the hexokinase method with UV absorbance at 340 nm. The imprecision of the glucose measurement was 1.1% at a concentration of 3 mmol/L and 1.2% at a concentration of 16.6 mmol/L. These values were obtained on 30 replicate measurements on quality control sera Precinorm and Precipath (Roche Diagnostics, USA) performed over 30 consecutive days. Fructosamien was measured by the tetrazolium blue colorimetric method at an absorbance of 552 nm and with a calibrator normalized with glycosylated poly-L-lysine. The imprecision of the fructosamine measurements was 3.0% at a concentration of 268 μmol/L and was derived from 30 replicate measurements of control sera (Precinorm) measured over 30 consecutive days. Total proteins were measured by the biuret method. The imprecision was 1.4% at a concentration of 5.14 g/dL and 1.2% at a concentration of 5.12 g/dL. The values were derived from 30 replicate measurements on control sera (Precinorm and Precipath) performed over 30 consecutive days.
All the measurements were made in a Hitachi 717 auto-analyzer (Hitachi, Japan) with commercial reagents from Roche Diagnostics. The blood samples were taken in the hospital's outpatient antenatal clinic or at the primary healthcare centre if that was the patient's choice. In the latter case, the samples were immediately centrifuged on site and transported on ice to the hospital laboratory.
The index ‘I’ derived from the analyte measurements conducted on the 1 hour post-load blood sample was defined as:
We introduced the 100 as a denominator so as to bring the values of I within a practicable range. In the above formula, I can be converted to conventional units by expressing total proteins in g/dL, glucose in mg/dL and fructosamine in μmol/L.
The ST measurements of glucose, fructosamine, total proteins, albumin, HbA1c, together with birthweight, type of birth, Apgar score, I index and the OGTT definitive GDM diagnosis were entered into a data spreadsheet (Excel, Microsoft). To obtain the cut-off point of I, we initially applied the Shapiro-Wilk test to the calculated values of I generated from the preliminary study, to assess the normality of distribution. The distribution was found to be non-Gaussian; therefore non-parametric methods (2.5th percentile) were used to select the cut-off point.
To assess the diagnostic value of I we used the GraphROC 2.0 program for Windows to calculate sensitivity, specificity, positive predictive value, negative predictive value, diagnostic efficiency, positive likelihood ratio and negative likelihood ratio. The exact 95% confidence intervals (95% CI) were calculated from the standard error of proportions or according to Miettinen's test.
The 2.5th percentile (n=12 of the 138 women) value of I was 27.2, or 49.0 with conventional units. This was adopted as the statistical cut-off point to differentiate between those with and those without GDM in the second part of the study.
A value of I < 27.2 (index negative) was used to reject the diagnosis of GDM, and I ≥ 27.2 (index positive) was used to indicate the requirement for definitive diagnosis. Of the 642 women studied, 524 were index negative. Of the 118 who were index positive, 52 had GDM according to the OGTT. Just one women had a false-negative result with the I index. Table 1 summarizes the results.
Table 2 shows the number of false positives for I in relation to the standard ST. If the I index had been used for screening, this would have avoided about 42% of the OGTTs performed because of suspected GDM.
Of the 642 women initially included in the study, we were able to assess 578 with respect to birthweight. 34 babies were macrosomic (≥4000g). I≥27.2 detected 32% (11/34) and the standard test detected 44% (15/34); however efficacy in detection of macrosomia was somewhat better with I (Table 3). The collective results for macrosomia detection are shown in Table 4.
Using three values (<27.2, 27.2-30.5, >30.5) we explored the value of I for one-stage diagnosis.
Of 524 women with a value of I<27.2, only one was subsequently diagnosed as having GDM—i.e. the sensitivity was 98%.
Of 62 women with an I value of 27.2-30.5, 20 were diagnosed as having GDM, 34 were completely normal on OGTT criteria, while the other 8 were ‘indefinite’ in having only one abnormal glucose value on the OGTT.
56 women had an I value of >30.5, of whom 32 were positive for GDM; 4 had only one abnormal glucose value on OGTT and the other 20 had a completely normal result. Hence, the specificity of this cut-off value is 96% (Table 5).
As a screening test, the sensitivity for I=27.2 was 98%—identical to that of the standard ST. This high sensitivity ensures that practically all the women with GDM were detected with this index. Incidentally, in our study the sensitivity of the ST was higher than that reported by Naylor et al.9 in a study including more cases.
The specificity achieved for I as a screening tool is one of the strong points of our study: at 89% it substantially exceeded the 75% recorded for the standard ST. A negative index discounts GDM and a positive index indicates the need for an OGTT. The low specificity of the ST implies that many women with a positive result will prove to have normal glucose tolerance. This difficulty is well recognized, and the Toronto TriHospital9 study recommended confining the ST to women of medium-to-high risk, so as to reduce the number of false positives. This strategy, however, could mean that about 4% of cases were missed because the women seemed at low risk10. These recommendations have been adopted by the American Diabetic Association and are adhered to by several other countries including Spain. The selective policy is partly driven by cost, since the standard ST is inefficient and often demands repeated investigations which are not only costly but also unpleasant for the patient11,12,13,14,15.
The use of I reduced false-positive screening from about 24% to 10% in the current study. Discrepancies such as negative OGTT with strongly positive ST have been noted by several research groups14, 15. The data in Table 1 show that I is much superior to the ST. An I-positive woman has a nearly 8-fold risk of having GDM. A positive I was also somewhat superior to the ST for prediction of macrosomia. The number of macrosomias in non-diabetic patients who were positive for I or ST deserves attention. It raises doubt as to the value of the OGTT for predicting this complication. An alternative to universal screening16,17, recommended by the fourth Workshop on Gestational Diabetes Mellitus, is to perform an OGTT directly on women of medium-to-high risk, or to screen selectively. We do not think the 4% failure rate of selective screening is acceptable. An OGTT for women at medium-to-high risk (or the diagnostic strategy based on the 2-hour oral load18) does have its advocates, despite the inconvenience and discomfort.
It does not seem logical that, to achieve a diagnosis of GDM, the patient has to be diagnosed in two stages, and we propose that the I index could offer a single test—albeit with certain cautions. A value of I<27.2 definitely excludes the necessity for any further testing for GDM, and a value of I>30.5 indicates GDM with a specificity of 96%. A value of I between 27.2 and 30.5 is borderline with respect to GDM diagnosis. In this situation there are two options. One is to consider the patient as diabetic and to proceed with established protocols for its treatment; or, alternatively, perform the OGTT in all of these women (n=62 in our study) and await confirmation of the diagnosis. The present study indicates that if the I index was used as a diagnostic tool, only 62 OGTTs would be needed for diagnostic confirmation, compared with 203 with the standard ST or 118 with the I index used as a screening tool.
The measurement of fructosamine in screening for GDM is not new. It has its proponents6 and a greater number of detractors because of insensitivity19,20,21. The ratios of fructosamine to total protein or of fructosamine to albumin22 have been evaluated and rejected for the same reasons. However, if we multiply these ratios by the concentration of glucose, the product value differentiates reasonably well between diabetic and normal pregnant women.
The response to the glucose load in the diabetic patient is very variable. Therefore it is necessary, both in screening and in diagnosis, to incorporate a measure that is less susceptible to fluctuations. Fructosamine corrects for variability in response to the oral glucose load by representing the mean concentration of glucose over 1-3 weeks, while total proteins correct for the haemodilution associated with pregnancy. As a continuous variable, glucose cannot be used with much confidence for either screening or diagnosis. The calculated index I is likewise a continuous variable, but a positive has good sensitivity and specificity. Although there were 66 false positives in our study, we suspect that some of these gestations were abnormal with respect to carbohydrate metabolism. For screening purposes, the positive and negative predictive values for I>27.2 are highly acceptable.
As a single-stage diagnostic tool, the technique offers further attractions, not least convenience. However, more work with a greater number of patients is needed to confirm the validity of the index in relation to OGTT screening and to show how it compares with the ST in prediction of complications other than macrosomia.
We thank Dr Peter R Turner, of t-SciMed (Reus, Spain) for constructive criticism and editorial assistance. The study was funded in part by the Andalucia Health Authority.