Gestational diabetes mellitus screening and diagnosis: a prospective randomised controlled trial comparing costs of one-step and two-step methods


  • NCT00295659 Registered Study Title: ‘What is the optimal method for screening and diagnosis of gestational diabetes? A randomised clinical trial to evaluate incidence of GDM, cost-effectiveness and clinical outcomes using three methods.’

Dr SJ Meltzer, Division of Endocrinology and Metabolism, Royal Victoria Hospital, McGill University Health Center, 687 Pine Avenue West, M9, Montreal, QC, Canada, H3A 1A1. Email


Please cite this paper as: Meltzer S, Snyder J, Penrod J, Nudi M, Morin L. Gestational diabetes mellitus screening and diagnosis: a prospective randomised controlled trial comparing costs of one-step and two-step methods. BJOG 2010;117:407–415.

Objective  To conduct a cost minimisation analysis of three methods of gestational diabetes mellitus (GDM) screening and diagnosis.

Design  Prospective randomised controlled trial.

Setting  University teaching hospital.

Population  Pregnant women (n = 1594) presenting for GDM screening.

Methods  Women presenting for GDM screening, who consented to participate, were randomised to GR1 [1-hour, 50-g glucose screen (GS) ± 3-hour, 100-g oral glucose tolerance test (OGTT)], GR2 (50-g GS ± 2-hour, 75-g OGTT) or GR3 (2-hour, 75-g OGTT). Demographics, health and time/travel cost information were assessed for each glucose testing visit.

Main outcome measures  Costs (direct and indirect) and prevalence of GDM diagnosis.

Results  The direct sampling costs of the glucose tests per woman were as follows: GS, CAN$12.57; 75-g OGTT, $36.10; 100-g OGTT, CAN$48.13. Among women in the two-step method groups diagnosed with GDM, 39% of the GR1 and 61% of the GR2 groups were diagnosed at the first step by GS ≥ 10.3 mmol/l, according to the Canadian Diabetes Association recommendations, contributing to a lower total cost in these groups. The total costs per woman screened were as follows: GR1, CAN$91.61; GR2, CAN$89.03; GR3, CAN$108.38. The GDM prevalence was similar (3.7%, 3.7% and 3.6%, respectively). The higher costs of GR3 were related to more blood draws and the time required for all women to undergo the 2-hour OGTT.

Conclusions  Careful consideration should be given to an internationally recommended method of universal screening for GDM which minimises the burden and cost for individual women and the healthcare system, yet provides diagnostic efficacy. The two-step method (GS ± OGTT) accomplished this better than the one-step method (75-g OGTT).


Gestational diabetes mellitus (GDM) is one of the most common medical complications of pregnancy, with an overall prevalence of 3–5%, but as high as 18% depending on the population and diagnostic criteria.1–3 Previous studies involving cost analyses have suggested that the diagnosis and treatment of GDM can improve pregnancy outcomes and be cost-effective.4,5 Although the screening of pregnant women imposes direct and indirect costs to the healthcare system, the individual and society, the treatment of this condition represents cost savings, primarily as a result of a reduction in the high costs of neonatal care.6,7 As the cost of managing such complications can be greater than the cost of screening for and treatment of the condition, many countries recommend universal screening, as selective screening has been shown to underdiagnose GDM and may represent a cost saving of only 5%.4,7–10

Worldwide controversy exists with regard to the best method and criteria for GDM screening and diagnosis (Table 1).8,9,11–14 Although the recent Hyperglycemia and Adverse Pregnancy Outcomes (HAPO) publication, of the largest study evaluating 2-hour, 75-g oral glucose tolerance test (OGTT) values and outcomes, has generated urgent momentum for the development of new criteria for GDM screening and diagnosis, the costs implicated in the one-step and two-step methods must be considered.15 Cost minimisation compares the costs of alternative strategies with equivalence clinically, and it has been recommended that such analyses should incorporate the indirect or productivity costs associated with the time burden placed on women receiving medical care.16,17 Despite this recommendation, few previous studies have evaluated the actual direct and indirect costs incurred by women, and none have been based on randomised trials.7,10,18,19 As part of a prospective randomised controlled trial of 5487 women designed to compare three methods of GDM screening and diagnosis, we undertook a cost minimisation analysis involving the first 1594 women enrolled by evaluating actual direct medical costs, direct nonmedical costs, indirect costs and GDM prevalence.

Table 1.   Selected international recommendations for gestational diabetes mellitus (GDM) testing*
 Plasma glucose (mmol/l)**
50-g glucose screen***Oral glucose tolerance test (OGTT)
Glucose load (g)Fasting1-hour2-hour3-hour
  1. *ADA, American Diabetes Association; CDA, Canadian Diabetes Association; NDDG, National Diabetes Diagnosis Group; WHO, World Health Organisation.

  2. **Plasma glucose is considered to be normal if less than the value shown; to obtain mg/dl, multiply mmol/l by 18.

  3. ***Glucose screen criteria for GDM diagnosis: ≥10.3 mmol/l for CDA; ≥11.1 mmol/l for WHO.

  4. ****ADA recommends one-step approach with OGTT alone in clinics with a high prevalence of GDM.

CDA: two-step137.8755.310.68.9
NDDG: two-step117.81005.810.69.28.1
ADA: one- or two-step8****7.2 or 7.81005.310.08.67.8
WHO: one-step12None755.8None7.8


The study took place between January 2001 and December 2004 at the Royal Victoria Hospital site of the McGill University Health Center (MUHC) in Montreal, Quebec, Canada. This tertiary care university hospital serves a population with a broad ethnic and socioeconomic base and manages approximately 3500 deliveries per year. This cost analysis was undertaken on a subset of the first 1594 women recruited between January 2001 and January 2002, 94 of whom were excluded, as shown in Figure 1. The study protocol was approved by the MUHC Research Ethics Board and registered through the identifier: NCT00295659.

Figure 1.

 Consort flow chart. 1GS, glucose screen; 2OGTT, oral glucose tolerance test; 3AC/PC, ante-cebum and post-cebum breakfast.

As recommended by the Canadian Diabetes Association (CDA), universal screening for GDM is generally performed between 24 and 28 weeks of gestation, although it may be earlier in the presence of multiple risk factors.13 A glucose screen (GS) or OGTT may be repeated if an early test is negative or if risk factors become evident after a routine test is negative. During the study period, all pregnant women referred for GDM screening were informed about the study by their obstetricians and asked to come to the test centre in the fasting state. On arrival, the study coordinator explained the study and, when informed consent had been obtained, women were randomly assigned to one of three study groups, as illustrated in Figure 1: Group 1 (GR1) using the two-step method (50-g GS ± 3-hour, 100-g OGTT); Group 2 (GR2) using the two-step method (50-g GS ± 2-hour, 75-g OGTT); Group 3 (GR3) using the one-step method (2-hour, 75-g OGTT alone).

A block randomisation method was used and opaque envelopes were prepared by research staff not involved in participant recruitment. As a result of the nature of the test, both the participant and study coordinator were nonblinded. Women were excluded who arrived in a nonfasting state, had a significant language barrier or refused to participate.

All obstetricians were provided with reminders of CDA clinical practice guidelines for GDM screening and diagnostic criteria, but the interpretation of the glucose test results and referral for further testing or GDM clinic care were their responsibility as usual. If the study participants were referred for further testing, their study group was verified at the test centre to ensure that the appropriate test was conducted. Glucose testing results were evaluated as recommended by the CDA, including the GS cut-off for GDM diagnosis of 10.3 mmol/l or 185 mg/dl (Table 1).13 Diagnoses of GDM (two or more abnormal OGTT values) and impaired glucose tolerance (IGT) (one abnormal OGTT value) were made according to the 1979 National Diabetes Data Group criteria for the 3-hour, 100-g OGTT and the CDA criteria for the 2-hour, 75-g OGTT.11,13 The usual multidisciplinary and intensive management of GDM continued throughout the study period, with the initiation of insulin when the fasting plasma glucose (FPG) exceeded 5.3 mmol/l or 1-hour plasma glucose (PG) exceeded 7.8 mmol/l despite dietary therapy. Using the same glycaemia goals and a less intensive protocol, management of IGT by the nutritionist also remained consistent during the study.

Information about participant demographics, obstetric and medical history, medication use, family history of diabetes, and time and transportation expenditures for each visit was collected prior to glucose testing, and entered into a computerised database which automatically uploaded the GS and OGTT results. Maternal data were obtained from the McGill Obstetric and Neonatal Database.20

Direct medical costs were determined for each aspect of the glucose test, including test materials, technician time and laboratory analysis. The indirect costs of the women’s time for glucose testing and transportation, as well as direct nonmedical costs for transportation to the test centre, were derived from self-reported questionnaires completed at each visit. Time costs, which reflected the time a woman spent in transit to the test centre and the time she spent at the clinic from her arrival to completion of the test, were allocated at the rate of $15.68/hour, an estimated value of paid and unpaid productive time of Canadian women, adjusted for inflation to 2002 Canadian dollars.21 The transportation costs for taxi, parking and public transportation were self-reported. Vehicle operation costs were based on estimates of one-way distance and an allowable expense of CAN$0.41/km obtained from the Canada Revenue Agency 2002 guidelines.22

The interpretation of test results and scheduling of further tests, if needed, were performed by a secretary, nurse, nutritionist and/or physician, or any combination of these individuals. The costs of their time were not included, as it would not have been feasible logistically for each individual involved to collect data on the name and minutes spent discussing/scheduling glucose test results for 1594 study participants.

The sample size calculation for the entire study population of 5758 women was powered to detect a difference of 2% between study groups in the prevalence of GDM with 80% power and α = 0.05, allowing for 5% attrition. Within our budget constraints for cost analyses, the initial 500 women in each group with complete cost data, who had completed the allocated test procedures, were included, as this sample was felt to be representative. Statistical analysis of maternal characteristics, GDM prevalence and costs were performed using SPSS v13.0 (SPSS Inc., Chicago, IL, USA). The results are reported as percentages, means and standard deviations or as 95% confidence intervals. Statistical evaluation was performed using analysis of variance (ANOVA) for comparison of means and chi-squared analysis for comparison of proportions. The results were considered to be significant if P < 0.05 or confidence intervals did not overlap.


The population studied represents a subset of a large randomised controlled trial, which had an 83.5% recruitment rate (Figure 1).23 Women were excluded from this subset as a result of attrition, allocation errors and missing data. Attrition accounted for a small proportion of women (1.3%, or seven of 525, in GR1; 0.6%, or three of 533, in GR2) who had abnormal GS, but did not return to the blood test centre for OGTT, probably because it was performed elsewhere. We excluded 1.2% of the 1594 women studied who, secondary to allocation errors, did not undergo the test assigned through randomisation: three in GR1, 11 in GR2 and five in GR3; this permitted more precise cost comparisons of the three methods. These allocation errors may have been related to the transition from usual GDM screening and diagnosis practice to the study protocol, as such errors decreased in subsequent years of the study. Missing laboratory data prevented the determination of diagnosis, and missing cost data prevented accurate cost analyses.

Maternal characteristics and GDM prevalence rates for this subsample of 1500 pregnant women are presented in Table 2. The majority of participants were Caucasian, with the remainder from multiethnic backgrounds. Most women presented for initial glucose testing between 24 and 28 weeks of gestation, as recommended.1 Although height and weight data were not collected from the initial 1300 women entered into the study, among the remaining main study population for whom mean pre-pregnancy body mass index (BMI) could be calculated, BMI (kg/m2 ± standard deviation) was similar between the groups: GR1, 23.2 ± 4.5 (n = 1396); GR2, 23.4 ± 4.5 (n = 1431); GR3, 23.6 ± 4.7 (n = 1445). The BMI distribution among this sample was 6.8% underweight (<18.5 kg/m2), 66.9% normal weight (18.5–24.9 kg/m2), 17.4% overweight (25.0–29.9 kg/m2) and 8.9% obese (≥30.0 kg/m2), according to World Health Organisation categories, which have been adopted by Health Canada.24 There were no statistically significant differences in the known risk factors for GDM (maternal age, ethnicity, pre-pregnancy BMI, parity, family history or previous GDM) between the groups. The prevalence of GDM and IGT was also similar among the three study groups; although the rate of IGT appeared to be higher in GR3, it was not statistically significant (P = 0.7).

Table 2.   Maternal characteristics and diagnostic status*
 Group 1**Group 2**Group 3**
  1. *No statistically significant differences were found between the study groups for all characteristics, including the diagnosis of IGT (P = 0.7).

  2. **Sample size was 500 per group, unless indicated otherwise as a result of missing data.

Age (years) (SD)30.9 (5.0)31.1 (5.0)31.3 (5.3)
Ethnicity (%) (n)n = 494n = 494n = 493
Caucasian61.9 (306)61.1 (302)61.5 (303)
Black5.9 (29)8.3 (41)6.7 (33)
South Asian7.3 (36)4.5 (22)4.5 (22)
Hispanic1.8 (9)2.2 (11)2.8 (14)
Middle Eastern12.1 (60)13.4 (66)11.0 (54)
East Asian9.7 (48)8.7 (43)12.4 (61)
Other1.2 (6)1.8 (9)1.2 (6)
Previous gestational diabetes mellitus (GDM) (%) (n)2.4 (12)1.8 (9)1.4 (7)
Family history of diabetes mellitus (%) (n)43.2 (216)40.6 (203)43.2 (216)
Parity (%) (n)
055.8 (279)56.4 (282)62.4 (312)
1–240.6 (203)40.8 (204)35.0 (175)
3+3.6 (18)2.8 (14)2.6 (13)
Gestational age 1st glucose test (%) (n)n = 474n = 473n = 467
<24 weeks3.6 (17)4.0 (19)3.4 (16)
24–28 weeks81.6 (387)83.1 (393)86.3 (403)
>28 weeks14.8 (70)12.9 (61)10.3 (48)
Diagnosis (%) (n)n = 486n = 485n = 497
GDM3.7 (18)3.7 (18)3.6 (18)
Impaired glucose tolerance (IGT)3.7 (18)4.3 (21)5.6 (28)

The cost (in Canadian dollars) of the Glucola™ drink was $1.60 per 75-g bottle, i.e. $1.07 for 50-g GS, $1.60 for 75-g OGTT and $2.13 for 100-g OGTT. The Glucola drink was measured for 50-g GS and the unused portions were used for subsequent tests. The cost of a single blood draw, including consumable tubing, needles and blood technician time, was $10.00 per sample. Laboratory analysis of a glucose sample cost $1.50. Thus, the total sampling costs were CAN$12.57 for 50-g GS, CAN$36.10 for the three-sample, 75-g OGTT and CAN$48.13 for the four-sample, 100-g OGTT.

Glucose testing was also analysed per study group according to the amount of visits to the test centre, blood samples drawn for each woman and the time the women spent in hours, as shown in Table 3. Most women in this study visited the clinic once: 81.2% in GR1; 78.8% in GR2 for GS (one blood draw); 97.8% in GR3 for 75-g OGTT (three blood draws). A total of 17.6% of GR1, 20.4% of GR2 and 2.2% of GR3 participants came back for a second visit. Of those in GR1 and GR2 who underwent GS as a first test, about 1% returned for a third visit. The majority of women in GR3 had one visit for the 75-g, 2-hour OGTT, with 1.74 and 1.87 times more blood draws than women in GR1 and GR2, respectively. Thus, the number of actual blood draws was 43% less in GR1 and 47% less in GR2 than in GR3.

Table 3.   Glucose testing: amount of visits, blood draws and time spent per study group
 Group 1 (GR1)Group 2 (GR2)Group 3 (GR3)
nVisitsBlood drawsnVisitsBlood drawsnVisitsBlood draws
  1. *Glucose screen (GS) and oral glucose tolerance test (OGTT) were performed according to randomisation to GR1/2 or GR3.

  2. **One GR2 woman had repeat GS and 11 GR3 women had repeat OGTT.

  3. ***All women had repeat GS or OGTT.

  4. ****Statistical significance is indicated when confidence intervals between groups do not overlap.

1 visit for glucose testing*4064064063943943944894891467
2 visits**88176440102204404112266
3 visits***6183641220000
Total visits5006008825006108185005111533
Visits/woman, mean [95% CI]1.20 [1.16–1.24]1.22 [1.18–1.26]1.02 [1.01–1.03]
Total hours, mean [95% CI]****3.52 [3.33–3.70]3.42 [3.26–3.57]3.80 [3.72–3.88]

Among the 1000 women in the two-step method groups who were diagnosed with GDM, 39% of GR1 (n = 7/18) and 61% of GR2 (n = 11/18) were diagnosed at the first step by GS ≥ 10.3 mmol/l, according to CDA recommendations. Women in GR1 and GR2 still visited the test centre more often than did women in GR3, as needed for OGTT (the second step). However, GR3 women sustained higher indirect costs for glucose testing related to the time spent in transit and in the clinic (Table 3). The method of transportation was not statistically significantly different between the study groups (P > 0.05), with mean rates as follows: 69.5% by car, 17.7% by public transport, 6.3% by taxi, 2.7% by walking and the remaining 3.8% by a combination of public transport plus car or taxi.

For the women diagnosed with GDM, the mean gestational age at the time of diagnosis was significantly earlier for the one-step method (GR3, 189 ± 19 days) than for either of the two-step methods (GR1, 202 ± 29 days; GR2, 200 ± 20 days) (P = 0.023). Among the entire subset, there was more pre-eclampsia (GR1, 3.5%; GR2, 3.3%; GR3, 5.4%) and more neonatal hypoglycaemia (GR1, 3.5%; GR2, 4.2%; GR3, 6.5%) for GR3 (P < 0.05).

The direct medical costs of screening, including the glucose drink, blood draw and laboratory analysis, were similar statistically for GR1 and GR2 and significantly lower than for GR3 (Table 4a). The two-step methods of GR1 and GR2 resulted in higher transportation costs, as approximately one in five women returned for a second test. Compared with GR3, the combined costs of transportation and productivity time lost were also lower for GR1 and GR2 by $1.65 and $2.61, respectively. Similarly the total costs per woman screened for GDM were significantly lower for women in GR1 and GR2. Thus, the one-step, 75-g OGTT method for both screening and diagnosis was 1.18 and 1.22 times more expensive than the two-step methods.

Table 4.   Costs of gestational diabetes mellitus (GDM) screening and diagnosis by study group
 Costs per woman in CAN$ [95% CI]*
Group 1 (n = 500)Group 2 (n = 500)Group 3 (n = 500)
(a) For the cost subset population
Direct medical costs**21.77 [20.08–23.45]20.16 [18.85–21.46]36.89 [36.43–37.36]
Direct transportation costs***14.66 [13.59–15.73]15.32 [14.26–16.37]11.92 [11.13–12.71]
Indirect time costs****55.18 [52.28–58.08]53.56 [51.13–55.98]59.57 [58.31–60.82]
Total cost per woman screened91.61 [86.49–96.73]89.03 [84.83–93.23]108.38 [106.44–110.32]
Average cost per woman tested to obtain diagnosis*****1145.13 [1081.13–1209.13]1112.88 [1060.38–1165.38]1354.75 [1329.75–1379.00]
Actual cost per woman to diagnose IGT or GDM1272.36 [1201.25–1343.47]1085.85 [1034.51–1136.95]1389.49 [1364.62–1414.36]
 Costs per Asian woman in CAN$ [95% CI]*
Group 1 (n = 84)Group 2 (n = 66)Group 3 (n = 82)
  1. IGT, impaired glucose tolerance.

  2. *Statistical significance is indicated when confidence intervals between groups do not overlap.

  3. **Direct medical costs include the drink, the blood draw and laboratory analysis.

  4. ***Direct transportation costs include one-way transport from home to the hospital.

  5. ****Indirect time costs refer to the time women spent in one-way transit and at the blood test centre for glucose testing.

  6. *****Calculations were made taking an average per group divided by the average number diagnosed overall.

  7. ******Asian women include women originally from the Orient, South-East Asia or East India.

(b) For the Asian subset population******
Direct medical costs*25.18 [20.56–29.80]27.34 [22.94–31.73]36.54 [35.66–37.42]
Direct transportation costs**12.82 [10.58–15.06]13.33 [10.63–16.04]8.84 [7.33–10.35]
Indirect time costs***57.88 [51.00–64.75]64.57 [55.67–73.47]58.76 [55.75–61.78]
Total cost per woman screened95.87 [83.64–108.10]105.24 [91.35–119.14]104.15 [100.40–107.90]

The actual costs incurred for diagnoses (IGT or GDM) were higher for women in GR3, as shown in Table 4. However, as the diagnostic rates were similar, the average cost per woman tested to obtain a diagnosis was determined to provide a cost per method with greater generalisability. The average was calculated by taking the total cost of all women screened (500 × cost per woman screened per group) and dividing that by the average number of women diagnosed per group (121/3 = about 40).

Costs were also assessed in our high-risk population of Asians (South Asians and East Asians), who represented 15.5% (232/1500) of the population (Table 4b). Their prevalence of GDM/IGT was 22.0% in GR1, 24.2% in GR2 and 16.7% in GR3 versus 6.4%, 8.0% and 9.2%, respectively, of all women. There were 58% (7/12) in GR1 and 40% (4/10) in GR2 whose GDM diagnosis was made on the basis of GS alone. The total visits in this higher risk Asian population were as follows: GR1, 1.26 (1.17–1.36); GR2, 1.41 (1.29–1.53); GR3, 1.01 (1.00–1.04); however, the total costs per woman screened were similar among the three groups. The average costs per woman diagnosed in the Asian subset were $455.38 ($397.29–$513.48) for GR1, $499.89 ($433.91–$565.92) for GR2 and $494.71 ($476.90–$512.53) for GR3, which were not statistically significantly different.


The importance of the treatment of GDM has been clarified by the Australian Carbohydrate Intolerance Study in Pregnant Women (ACHOIS) and HAPO groups in landmark studies, which have demonstrated the relationship between glycaemia and perinatal morbidity.15,25 A two-step method with selective screening was used by Crowther et al.25 in their large randomised trial of the treatment of GDM. A universal, one-step, 75-g OGTT was used in an observational investigation of over 23 000 women in the HAPO study. An international committee is presently working towards a consensus on which method should be used to diagnose GDM and minimise any related perinatal morbidity. Given the increasing prevalence of GDM, in-depth cost evaluations of the one-step and two-step methods are of paramount interest. This is the only prospective randomised controlled cost analysis study of GDM diagnosis with a sample as large as 1500 women. Other GDM diagnosis cost assessments, which were not based on models, represent sample sizes of under 200 women,10,19,26 except for a nonrandomised retrospective study comparing universal (n = 1338) and selective (n = 4035) screening methods over two time periods.4 Our cost analysis of GDM screening and diagnostic methods was conducted within the publicly funded Canadian health system, where the medical costs of universal screening are covered. Cost minimisation analyses assume clinical equivalence of the compared strategies, such as that seen in the population characteristics and rates of GDM prevalence.5,16

The two-step approach, using GS and either 2-hour, 75-g or 3 hour, 100-g OGTT, was found to be less expensive with equivalent diagnostic power to the one-step approach (2-hour, 75-g OGTT alone). In the 1500 women, the average cost for diagnosis was significantly higher with the one-step universal 75-g OGTT test, although the difference was not significant in the Asian group from this population. Among the high-risk Asian women, because more women screened were diagnosed, the average costs of screening were less ($455–$500) relative to the rest of the population ($1113–$1354). The 1-hour, 50-g GS step of the two-step method significantly lowered the direct medical and time expenditure costs relative to those of the one-step method.

Several studies evaluating the cost of GDM screening have also demonstrated that a two-step method using 50-g GS is the least costly.10,18,19 Lavin18 found that the two-step protocol was associated with lower direct costs and less time than a one-step protocol, employing a model based on a Medline search of the prevalence of positive glucose screening tests and catalogue estimates of the costs involved in the screening process. Poncet et al.19 established the costs involved in the diagnosis and treatment of GDM among 120 women using the French public health system and extrapolated the outcome measures from a 25-year Medline search. These values were combined in a decision and cost-effective analysis model comparing three screening strategies: selective screening with a 50-g GS + 100-g, 3-hour OGTT (American Diabetes Association); universal screening with a 50-g GS + 3-hour, 100-g OGTT; and universal screening with a 75-g OGTT [World Health Organisation (WHO) criteria]. As in our study, they found that the one-step 75-g OGTT was the most expensive; however, unlike our study, they found it to be more efficient, possibly as a result of the different diagnostic criteria used. Nicholson et al.10 reported that the 2-hour, 75-g OGTT one-step method was more costly and less effective than the two-step method in a statistical modelling study based on a literature review and including the derived costs for diagnosis, treatment and perinatal outcomes. Evidence that high-risk women do better with universal screening using a one-step method is limited; however, it has been suggested that a one-step OGTT may be more effective for high-risk populations.8,10,18 The assessment of our highest risk group suggests that this may not be the case, as GS diagnosed a large proportion of the Asian women studied (58% in GR1, 40% GR2), whereas the overall prevalence of GDM/IGT was lower in GR3 using one-step OGTT (22% GR1, 24% GR2, 17% GR3).

Universal or selective administration of GDM testing remains a contentious issue. In North America, universal screening is the standard of care; thus, our study was not designed to evaluate this issue. By definition, a screening test should be well defined, easy to administer, reproducible and inexpensive; it is not meant to be diagnostic, but rather to identify a subgroup of individuals at risk who require further testing for diagnosis. In addition, the sensitivity and specificity of GS vary with the ethnicity of the population, as reported by Esakoff and colleagues;27,28 a goal to maintain false positives below 10% would require a GS cut-off of 7.8 mmol/l for Caucasians and 7.5 mmol/l for African-Americans; however, the threshold would need to be higher for Asians. Poncet et al.19 found that selective screening had a more favourable cost-effectiveness ratio. This may be explained by the use of the more sensitive 2-hour diagnostic criteria of WHO compared with the CDA criteria used in our study. In their nonrandomised retrospective study, Di Cianni et al.4 compared the estimated costs of screening and management strategies from two different periods and reported that universal screening was most cost-effective, whereas selective screening allowed a cost saving of only 5% per GDM diagnosis. Their study affirmed that, although GDM diagnosis and intensive management invoke costs to the healthcare system, they result in significant monetary savings related to reduced perinatal morbidity, which has also been reported for the treatment of mild GDM.4,29

Most cost studies for GDM testing are based on projected not actual costs, and employ mathematical modelling to extrapolate the findings.10,18,19 As recommended for economic evaluations in obstetrics, this study assessed in-depth the actual costs involved.17 Lavin’s study26 included numerous direct and indirect costs; however, the data were based on modeling with an estimation that 12–17% of women would return for a second test; a value lower than our actual rates. This would lead to lower estimates of cost per case diagnosed. The direct medical costs per woman for the one-step method (75-g OGTT) in our study with actual sampling was almost double that for the two-step method. The direct cost of a 2-hour, 75-g OGTT of CAN$36.89 can be compared with the £17.58 (CAN$31.59) determined in the National Institute for Health and Clinical Excellence (NICE) Antenatal Care Diabetes in Pregnancy Costing Report.30 In addition, the women’s time and transportation costs (indirect costs) represented a large proportion of the total costs, highlighting the need to incorporate the cost burden for women when evaluating the costs for GDM diagnostic methods. Such in-depth cost comparison of GDM testing methods provides important insights into the impact of choosing one method over another, whilst demonstrating that total actual individual direct and indirect costs are greatest for the one-step, universal, 2-hour OGTT.

In terms of cost, our findings support the preferential use of a 50-g GS (with a diagnostic level of ≥10.3 mmol/l), followed by OGTT, relative to a one-step, 2-hour, 75-g OGTT, as reported in other studies.4,10,18,31,32 The lower Canadian diagnostic value for GS (10.3 mmol/l) may limit the comparability of this study with others, although only four women avoided an OGTT by the use of the lower GS cut-off, lower than 11.1 mmol/l, which would be considered as diagnostic by many. Potentially, the universal use of a lower 50-g screen cut-off value, such as ≥10.3 mmol/l, could lead to a more cost-efficient diagnostic test, as not only are the costs lower, but preliminary outcome data suggest outcomes at least as good as those of a universal 75-g OGTT. A study from Thailand suggested that the predictive 50-g GS threshold value for GDM in their population would be 9.8 mmol/l, even lower than the Canadian criteria that we employed.31 Cheng et al.27,31 suggested that the treatment of women at risk of perinatal morbidity, by virtue of a GS value of 10.0 or 11.1 mmol/l, without the confirmatory OGTT, may actually lead to better outcomes, which is also suggested by our results.

The prevalence rates of GDM and IGT were not statistically significantly different in this analysis; however, there was a trend in GR3 for the diagnosis of more IGT than in GR1 (3.7% versus 5.6%; P = 0.07). Pre-eclampsia and neonatal hypoglycaemia were higher in the one-step group despite consistent management teams and protocols for all women throughout the study period. The gestational age of diagnosis was significantly earlier in the one-step group (an average of 3–5 days difference between groups); however, our cost analysis study was not designed to determine the impact of this lag time. Although the two-step method, on average, diagnosed women later, the clinical importance of this would depend on the difference in severity and duration of the women’s disease and the effectiveness of the therapy received. A next step would be to conduct a well-designed, cost-effectiveness study to address these issues.

Some limitations of our study have been identified. All women were advised to arrive in the fasting state as a criterion of inclusion, such that they could complete the 75-g OGTT if randomised into GR3. The conduct of a fasting GS, rather than at any time of day as generally prescribed, may have altered the sensitivity of the test to identify true positives,33,34 which may have had implications on the cost and prevalence, which could not be assessed with our study design. In addition, if GS was combined with an obstetrician visit, the cost associated with the two-step methods may have been even lower as a result of a decrease in indirect patient travel and time costs; however, we did not gather this information. Among the multiethnic population studied in the urban setting of Montreal, many women worked, and the dominant form of transportation was the car. Our use of one-way transportation costs provides a conservative estimate of the costs borne by the women. The inherent characteristics of our study population may render some of the findings difficult to extrapolate to other populations.

This study design has provided high-quality evidence among a large sample of women with similar characteristics and prevalence of GDM, permitting a detailed economic evaluation of the cost advantage of the two-step method.


This large randomised controlled trial of the methods employed for GDM screening and diagnosis has substantiated previous modelled cost assessments using actual direct and indirect cost analyses. The two-step method of diagnosis of GDM – with GS and OGTT (75 or 100 g) – involved the lowest costs; conversely, a 2-hour OGTT alone was substantially more costly as a result of the use of more time and blood testing resources and a full OGTT in the fasting state for all. This study underlines the use of GS as an inexpensive, easy to administer tool for GDM screening, especially with the use of a lower diagnostic cut-off. Careful consideration should be given to international recommendations for a method which minimises the burden of cost and inconvenience for individual women and the healthcare system, whilst providing diagnostic efficacy.

Disclosure of interests


Contribution to authorship

SM, JS and LM conceived and designed the study, led and interpreted the data analysis, and co-wrote the paper. JP designed and conducted the cost analysis, and co-wrote the paper. MN enrolled the participants, collected and analysed the data, and co-wrote the paper.

Details of ethical approval

Ethics approval of the study protocol was granted by the McGill University Health Center Research Ethics Board on 22 September 2000 (Reference number: Med-A 00-893).


This study was funded as an operating grant from the Canadian Diabetes Association.


We greatly appreciate the participation of the women studied and the cooperation of the obstetric staff.