Impact of Gestational Diabetes Detection Thresholds on Infant Growth and Body Composition: A Prospective Cohort Study Within a Randomized Trial

Gestational diabetes mellitus (GDM) is a major public health problem that is associated not only with increased risk of type 2 diabetes for women but also of metabolic disease for offspring, and this may manifest in later childhood with obesity (1,2) and other precursors of metabolic syndrome (3,4). Despite such associations being reported, causal mediators of the effect of GDM on offspring metabolic outcomes remain to be established. In clinical trials, treatment of GDM to improve maternal glycemic control compared with no treatment reduced the risk of infants being born large for gestational age (LGA) by 40% to 50% (5,6) but did not decrease the risk of overweight or obesity in midchildhood (7,8). Thus, additional strategies may be needed to improve offspring metabolic outcomes after GDM, including postnatal interventions, if suitable infant mediators can be found.

Infancy is recognized as a critical period of development during which homeostatic trajectories and metabolic capacity are established (9). Infant risk factors for later obesity include rapid early weight gain (10), excess adiposity and low lean mass (LM) (9,11), higher energy and protein intake, especially from formula (12,13), early introduction of solid foods (14,15), shorter duration of breastfeeding (16), and more vigorous feeding styles (17,18). In a systematic review, infants exposed to GDM compared with control infants not so exposed had increased fat mass (FM) from 1 to 6 months, higher rates of formula use, shorter duration of breastfeeding, and decreased linear growth, although the evidence was of low to very low certainty, particularly as few studies accounted for confounding, and no data were available for infant energy intake, complimentary feeding, appetitive traits, or LM (19). Moreover, the influence of maternal treatment on the effect of GDM on infant risk factors for later metabolic disease could not be determined (19). Similarly, clinical trials of GDM comparing treatment with no treatment have not reported infant outcomes beyond birth (20).

In New Zealand, the lower detection criteria for GDM recommended by the International Association of Diabetes in Pregnancy Study Group (IADPSG) were recently compared with higher criteria in use for >20 years in the randomized Gestational Diabetes Mellitus Trial of Diagnostic Detection Thresholds (GEMS, ACTRN12615000290594) (21). Women in the lower (IADPSG) criteria group (GDM diagnosed in 15%) compared with the higher (current New Zealand) criteria group (GDM diagnosed in 6%) had a 4% absolute increase in the rate of induction of labor, and their infants were more likely to be screened and treated for hypoglycemia, but there was no significant difference between groups in the rate of being born LGA or other maternal or neonatal clinical outcomes. While the primary objective of the GEMS trial was to assess at the population-level the benefits and risks of adopting the lower IADPSG detection criteria, including the economic consequences, we took the opportunity to perform a nested cohort study within the GEMS trial, the BabyGEMS study, to evaluate the effect of GDM, both treated and untreated, on infant risk factors for later metabolic disease compared with randomly selected control infants without exposure to GDM. Given previous reports of a continuous relationship between oral glucose tolerance test (OGTT) results and underlying perturbations in maternal glucose metabolism in pregnancy (22), we hypothesized that whole-body FM at 5–6 months would be higher than in control infants in infants exposed to GDM detected by higher criteria and in infants exposed to GDM by lower (IAPDSG) criteria that was untreated, but not those exposed to GDM detected by lower (IAPDGS) criteria that was treated. We also hypothesized that for other infant risk factors for later metabolic disease, a similar pattern would be observed among exposure groups.

The BabyGEMS study was a prospective cohort study of infants born to women who participated in the GEMS trial (21) and met the lower IADPSG criteria for GDM (9), irrespective of their GEMS allocated criteria group and diagnosis, and a 5% random sample of control infants whose mothers did not have GDM by the lower criteria (Table 1). In GEMS, women with a singleton pregnancy and no previous diabetes or GDM were randomly assigned after a one-step 75-g OGTT at 24–32 weeks’ gestation to be evaluated for GDM using lower or higher criteria. In the lower criterion group, GDM was diagnosed if one or more of the following criteria were met, as recommended by the IADPSG: fasting plasma glucose concentration ≥5.1 mmol/L (≥92 mg/dL), 1-h plasma glucose concentration ≥10.0 mmol/L (≥180 mg/dL) or 2-h plasma glucose concentration ≥8.5 mmol/L (≥153 mg/dL) (22). In the higher criterion group, GDM was diagnosed with a fasting plasma glucose concentration ≥5.5 mmol/L (≥99 mg/dL) or 2-h plasma glucose concentration ≥9.0 mmol/L (162 mg/dL), as recommended by the New Zealand Ministry of Health (23). Women with OGTT results that indicated GDM according to the detection criterion to which they were assigned were informed that they had the condition and received usual care for GDM, including nutritional therapy, blood glucose monitoring, and pharmacologic treatment, if required. Pharmacologic treatment could include metformin and/or insulin, according to national guidance and the preference of the woman and her physician (23).

If the OGTT results did not indicate GDM, according to the assigned detection criterion, women received routine pregnancy care. Numerical OGTT results were not disclosed, such that trial participants, caregivers, and researchers were unaware of the GEMS criteria group allocation, including at infant follow-up. Women provided written consent at trial entry for assessment of their infants, and assent for infant follow-up was confirmed after birth. Maternal ethnicity was determined by self-report, according to national protocols (24). Infants born outside the Auckland region were excluded from the BabyGEMS study. This study was approved by the Northern B Health and Disability Ethics Committee (13/NTB/18; Auckland, New Zealand).

Infants in the BabyGEMS study were assessed at birth and corrected age of 5–6, 9, and 12 months. This report describes the study assessments and prespecified outcomes up to the primary end point at 5–6 months.

Anthropometric measurements at birth and 5–6 months included weight by electronic scale to the nearest 10 g; length by neonatometer to the nearest 1 mm; head, chest, abdominal, and midarm circumference by lasso tape to the nearest 1 mm; acromion-radiale length to the nearest 1 mm; and skinfold thickness by Harpenden calipers to the nearest 0.2 mm (subscapular and triceps, and suprailiac at birth). Skinfold thickness was measured after a 3-s delay to standardize the degree of tissue compression. All skinfolds were measured twice, with a third measure taken if the difference between the first two measures was >0.4 mm. The median result was used in analysis. Left arm muscle area was calculated from the midarm circumference and triceps skinfold thickness. Two-compartment body composition was measured at birth and 5–6 months using air displacement plethysmography (ADP; PEA POD, COSMED, Chicago, IL) (25). All measurements were undertaken by trained study personnel, including a pediatrician, research dietitians, and research midwives.

At 5–6 months, feeding history was recorded by questionnaire, including the type of milk and introduction of solid food, following the World Health Organization indicators of breastfeeding and feeding practices (19). Nutritional intake was estimated using a validated 4-day Complementary Food Frequency Questionnaire (26). Appetitive traits and feeding behavior were assessed with the retrospective form of the Baby Eating Behavior Questionnaire (27).

The primary outcome of the BabyGEMS study was whole-body ADP FM at 5–6 months. The main secondary outcomes were other infant risk factors for later metabolic disease, including other measures of adiposity at 5–6 months (skinfold thickness; whole-body FM length index [kg/m²]); LM at 5–6 months (whole-body LM and length index; left arm muscle area); rapid growth in weight from birth to 5–6 months, defined as weight z score conditional gain >1; not predominantly breastfed to 5 months and introduction of solids <5 months; energy intake at 5–6 months; and food responsiveness scores. For completeness, additional secondary outcomes are reported, including body size at birth and 5–6 months, other items from the feeding questionnaire, macronutrient intake, and additional appetitive trait scores and Baby Eating Behavior Questionnaire subscales.

Statistical analysis was performed with JMP 15 and SAS 9.4 software (SAS Institute, Cary, NC). The maternal OGTT results and the treatment that the women received were used to categorize infants in the BabyGEMS study into four glycemic exposure groups: control infants, whose mothers had a nondiagnostic OGTT using lower criteria and received routine pregnancy care; group A, whose mothers had GDM by lower but not higher criteria and received usual diabetes treatment as they were allocated to the GEMS lower criteria group; group B, whose mothers had GDM by lower but not higher criteria and did not receive treatment for diabetes as they were allocated to the GEMS higher criterion group and were informed that they did not have GDM; and group C, whose mothers had GDM detected using higher criteria and received usual diabetes treatment (Table 1). The latter included women allocated to both the GEMS lower and higher criterion groups.

In the primary analysis, primary and secondary outcomes of infants in the three GDM exposure groups were compared with the control infants using generalized linear models, with the Dunnett correction of family-wise error rate. Models were adjusted for potential confounding by maternal body size (BMI at <18 weeks’ gestation), New Zealand Deprivation Index (28), ethnicity, and infant sex. Ethnicity is a measure of cultural affiliation and is self-perceived, but also reflects ancestry and is strongly associated with health status (24). For analysis, ethnicity was prioritized according to Ministry of Health protocols in the following order: Māori, Pacific, Indian, other Asian, other non-European, and European (24). Anthropometric data were converted to gestation- and sex-specific z scores (29). Rates of LGA and small for gestational age (SGA) at birth were also determined by customizing for maternal size, parity, and ethnicity (30). For body composition measures without established normative data (skinfold thickness, ADP, arm muscle area), standard scores (SS) were calculated from the residuals of regression on length (31). Conditional growth to 5–6 months was determined from z scores or SS (31).

Adjusted exposure effects are presented as risk difference or mean difference (MD) with 95% CIs. For categorical variables, adjusted relative effect (odds ratio) was also estimated, with 95% CIs. Missing outcome data were not imputed. Hypothesis testing was performed for the primary outcome with a two-tailed family-wise α of 0.05.

It was estimated that for the BabyGEMS study to have 80% power to detect differences of ≥200 g in whole-body FM at 5–6 months, 90 infants would need to undergo ADP in each group. The minimum clinically significant difference in this outcome is uncertain, but in an analysis of a mixed population of infants (38% exposed to GDM), the odds of overweight/obesity in midchildhood was increased eightfold for every 100-g gain in FM (32), while in another study, a gain of 0.67 SD, equivalent to 270 g, was found to be significantly associated with higher fat gain velocity throughout early childhood (33). The primary hypothesis was that compared with control infants, infants in groups B and C, but not A, would have increased whole-body FM at 5–6 months (19). The secondary hypothesis was that similar patterns would be observed for the main secondary outcomes.

The data sets generated and/or analyzed during the current study are not publicly available as ethical approval for this study did not include sharing of individual data.

Of the 4,061 women in GEMS, 804 infants were eligible for inclusion in the BabyGEMS study, of whom 760 participated (40 declined assessment, 4 born out of region), including 169 control infants and 591 infants exposed to GDM (Supplementary Fig. 1). At 5–6 months’ corrected age, 661 infants participated in follow-up (21 withdrew, 4 declined assessment, 51 were lost to follow-up, 17 moved out of region, and 6 could not be assessed due to the coronavirus disease 2019 pandemic and/or a measles epidemic).

At entry to GEMS, mothers who had GDM detected and treated using higher criteria (group C), compared with the control group, had higher BMI and were taller, and were more likely to be nulliparous and have a family history of diabetes (Table 2). Mothers who had GDM by lower but not higher criteria and were not treated (group B) had higher BMI than the control group, although their rates of overweight and obesity were similar to mothers who had GDM by lower but not higher criteria and were treated (group A). All women treated for GDM (groups A and C) were more likely to undergo induction of labor compared with the control group, and their infants were born ∼1 week earlier. While the prioritized ethnicity of women with GDM was not significantly different from the control group, among those with GDM detected and treated using higher criteria (group C), there was a preponderance of women of Indian ethnicity and fewer European women.

Compared with the control group, there were no significant differences among infants exposed to GDM in sex, cesarean birth, low Apgar score, and admission to a neonatal intensive care or special care baby unit (Table 2). However, infants whose mothers were treated for GDM (groups A and C) had shorter gestation length and were more likely to be born after induction of labor and at early term.

Overall, 661 of 760 infants (87%) in the BabyGEMS study were assessed at 5–6 months, and primary outcome data were available for 432 infants (57%) (Table 3). The remaining infants did not have ADP measurements because they could not be brought to one of the two research clinics with PEA POD, due to parental transport issues, time constraints, and cultural factors. Infants were assessed at a mean corrected age of 5.2 (SD 0.9) months. Whole-body FM at 5–6 months did not differ significantly between infants exposed to GDM and the control group, regardless of the detection threshold or whether women were treated: group A adjusted MD (aMD) −0.09 kg (95% CI −0.29, 0.10; P = 0.51); group B aMD −0.15 kg (95% CI −0.35, 0.06; P = 0.20); and group C aMD −0.17 kg (95% CI −0.37, 0.03; P = 0.13) (Table 3).

At birth, infants whose mothers had GDM by lower but not higher criteria that was not treated (group B), compared with the control infants, had thicker skinfolds and increased whole-body FM, but similar whole-body LM compared with the control infants (Table 3). Infants whose mothers were treated for GDM (groups A and C) had lower LM compared with control infants, and in group A, this association remained after accounting for length (index or SS) (Table 3).

At 5–6 months, infants exposed to GDM by lower but not higher criteria that was not treated (group B) had lower risk of rapid growth in weight at 5–6 months compared with control infants (Table 4). However, infants in all three GDM exposure groups were similar to control infants in measures of adiposity and LM at 5–6 months (Table 3), energy intake and food responsiveness (Supplementary Table 4), and risk of not receiving predominant breastfeeding to 5 months or starting solids early (Supplementary Table 4).

At birth, infants whose mothers were treated for GDM (groups A and C) had reduced body size (length and weight) compared with control infants, but similar risk of being born LGA or SGA, and similar gestation- and sex-specific z scores for length, weight, and head circumference (Supplementary Table 1). Infants whose mothers had GDM by lower but not higher criteria that was not treated (group B), compared with control infants, had higher z scores for birth weight (aMD 0.29; 95% CI 0.05, 0.53) and BMI (aMD 0.27; 95% CI 0.03, 0.50), and increased risk of being born LGA by the customized standard (adjusted risk difference [aRD] 11%; 95% CI 3%, 18%) (Supplementary Table 1).

Infants whose mothers had GDM detected and treated using higher thresholds (group C), compared with control infants, had an increased rate of receiving formula milk in the first week after birth (aRD 17; 95% CI 4, 31) (Supplementary Table 4). This group, along with group A, had increased risk of receiving treatment for transitional neonatal hypoglycemia (blood glucose concentration <2.6 mmol/L [<47 mg/dL]) compared with control infants (control infants 7%; group A 28% [aRD 20%; 95% CI 10, 29]; group B 9% [aRD 2%; 95% CI −5, 10]; group C 29% [aRD 21%; 95% CI 12, 29]). At 5–6 months, rates of continued breastfeeding and never breastfeeding were similar between GDM exposure groups and the control group (Supplementary Table 4).

At 5–6 months, infants in all three GDM exposure groups were similar to the control group in measures of body size (Supplementary Table 2); growth in weight and length, and whole-body FM and LM from birth (Table 4); and macronutrient intake and other appetitive traits (Supplementary Table 4).

In prespecified secondary analyses of the primary outcome, results were not altered by adjustment for the additional potential confounders of family history of diabetes and maternal height; adjustment for gestational age at diagnosis of GDM; sensitivity analysis excluding group B infants whose mothers were diagnosed with GDM after trial entry (n = 5); and sensitivity analysis excluding infants of mothers who had a 2-h OGTT plasma glucose concentration ≥15% higher than the 1-h plasma glucose concentration (indicating possible sample switching) and where this would change exposure group allocation (group A n = 5; group B n = 6; group C n = 11). In exploratory post hoc analysis, whole-body composition at birth and at 5–6 months in groups A and C did not appear to be influenced by in utero exposure to metformin (Supplementary Tables 5 and 6).

Contrary to the study hypothesis and previous reports (19), exposure to GDM, treated or untreated, was not associated at 5–6 months with increased whole-body FM or other infant risk factors for later metabolic disease. At birth, GDM detected at lower but not higher criteria was associated with increased adiposity if untreated (∼0.3 SD) but reduced whole-body LM if treated (∼0.3 SD), which appeared to be unrelated to metformin use.

The lower detection criteria for GDM recommended by the IADPSG were based on the results of the Hyperglycemia and Adverse Pregnancy Outcome (HAPO) study, which found a continuous association between maternal glycemia and risk of preeclampsia, cesarean birth, shoulder dystocia/birth trauma and being born LGA, fetal insulin concentrations, and measures of neonatal adiposity, with no obvious thresholds above which risks increased (22). The BabyGEMS study is consistent with the HAPO study in that infants who were exposed to GDM by lower but not higher criteria that was untreated (group B) had increased adiposity at birth (subcutaneous fat and whole-body FM) and increased risk of being born LGA compared with control infants. The BabyGEMS study has shown that routine treatment of GDM, whether detected at lower (group A) or higher (group C) criteria, appears to normalize fetal fat accretion. However, treatment of GDM also reduced fetal LM, especially if detected at the lower IADPSG criteria (group A), consistent with a previous study of infants of women with well-controlled GDM (34). Thus, although not evident at 5–6 months, some caution is required, as lower LM in early life could constrain metabolic capacity in the longer-term (9,34). It is also important to consider the overall balance of benefits and risks when treating milder dysglycemia in pregnancy; for example, adoption of lower criteria for GDM is likely to increase early-term birth, which has been associated with poorer long-term neurodevelopment compared with full-term birth (35).

Whole-body FM at 5–6 months was selected as the primary outcome for this study as rapid weight gain in infancy is a key risk factor for later obesity and metabolic disease (36), and this association is strongest for weight gain in the first 6 months (37–39). Furthermore, by 6 months of age, there is modest tracking of subcutaneous fat throughout the preschool years (11). In a previous systematic review, infants exposed to GDM had higher FM at 1–6 months of age, but the evidence was of very low certainty (19). Our study extends this evidence, providing robust estimates of the impact of GDM, including treatment, on body composition in infancy. Although the follow-up rate for the primary outcome was only 57%, largely due to the nonportability of ADP, type 2 error seems an unlikely explanation for the study results as sample size assumptions were met and point estimates for whole-body FM and the associated length-adjusted measures were lower for infants exposed to GDM than control infants. Moreover, anthropometric measures of body size and composition, which were obtained in most infants assessed at 5–6 months, were broadly consistent with ADP results.

Given the widely reported associations between diabetes in pregnancy and offspring obesity and precursors for metabolic syndrome in childhood (1–4), the role of infant nutritional risk factors in mediating the effects of GDM in offspring was also investigated. Although GDM was associated with increased formula supplementation in the first week, consistent with a recent review (19), exposure to GDM, treated or untreated, was not associated with reduced duration or intensity of breastfeeding or earlier complementary feeding, contrary to previous reports (19), and there was similar nutritional intake among groups at 5–6 months. Furthermore, while maternal diabetes has been shown to alter the organization of hypothalamic satiety centers and responsiveness to leptin in animals (40), in this study, GDM did not appear to influence infant appetite. It is interesting that the increase in early formula use was seen only in infants in group C and not group A, even though the proportion of infants treated for transitional hypoglycemia was similar, suggesting that lactogenesis may be more delayed in women with more severe glucose intolerance (41). Although the greater use of formula in this group did not appear to be associated with subsequent changes in body composition, the possibility of later effects cannot be entirely excluded (12,42), and research into interventions that assist women with diabetes to breastfeed remains important.

Taken together, the BabyGEMS study data suggest that GDM detected and treated at lower IADPSG, but not higher New Zealand thresholds, does not substantially alter key metabolic risk factors in early infancy, a period considered critical for the development of long-term homeostasis (43). Indeed, there was evidence of compensatory catch-down growth in FM in the exposed but untreated group (B), possibly mediated by postnatal adaptations in leptin concentrations (44). These results are somewhat discrepant with other cohort studies that have linked GDM exposure, including by lower IADPSG criteria, to excess body fat in later childhood (1,2) and lower insulin sensitivity in adolescence (4). Possible explanations for these differences include residual confounding, such as by maternal BMI (1,2) or genetic predisposition, or bias in the selection of the control group. Importantly, lack of evidence of medium-term benefit for offspring from maternal treatment of GDM does not preclude a causal association between GDM and later metabolic outcomes. While we did not identify casual mediators in early infancy, epigenetic modification of pancreatic β-cell function and insulin signaling could contribute to later metabolic disease and may not be clinically evident during infancy (45). Accordingly, some studies have suggested that metabolic disease in offspring exposed to GDM may not emerge until adolescence (4,46). Therefore, long-term follow-up of children in BabyGEMS and other well-controlled prospective cohorts will be important to quantify the relationship between GDM and offspring metabolic outcomes, independent of other perinatal risks, and to identify suitable targets for postnatal intervention.

A key strength of the BabyGEMS study is that the nested trial design meant that mothers of infants exposed to GDM at lower but not higher criteria (groups A and B) were randomly assigned to detection criteria and therefore treatment or not, and GEMS trial group allocation was blinded, thereby minimizing confounding and ascertainment bias. Moreover, the control group was randomly selected from the general obstetric population and prospectively followed within GEMS, reducing the impact of regression to the mean (47). The cohort also was demographically diverse; thus, results are likely to be broadly generalizable to other populations. Although our study sample included women of Polynesian ancestry, ∼30% of whom carry the CREBRF rs373863828 minor (A) allele, which is associated with lower risk of GDM (48), to limit confounding, we adjusted all analyses for self-identified ethnicity and indexed infant body composition measures to length to ensure that comparisons were independent of skeletal size (19).

A limitation of the BabyGEMS study is that it was not possible, due to ethical considerations, to include an untreated higher-criteria group, making interpretation of the results of group C more difficult. The detection thresholds currently used in New Zealand are similar to those of earlier randomized trials (5,49), which found that treatment of GDM reduced perinatal morbidity but had no effect on the incidence of overweight or risk factors for cardiometabolic disease in childhood (7,8). Thus, while it is reassuring that infants in group C had similar outcomes at 5–6 months of age, compared with control infants, whether this is a result of the treatment the mothers received remains unclear.

Another limitation is that body composition could only be assessed using a two-compartment model. While measurement of skeletal and regional soft tissue mass is desirable, techniques such as DEXA and MRI were not feasible for this large study. Even with ADP available at two centers, we were only able to assess whole-body composition in 65% of the infants seen at 5–6 months, as transport, time constraints, and cultural factors prevented many mothers bringing their infant back to the research clinics. We maintained a high follow-up rate for other assessments by completing home visits.

Finally, given that GEMS was a pragmatic trial, data on maternal glycemic control were not available, and thus, actual fetal glycemic exposure can only be inferred from the maternal detection and treatment thresholds.

GDM detected using lower IADPSG but not higher New Zealand criteria was not associated with increased whole-body FM at 5–6 months’ corrected age or other infant risk factors for later metabolic disease, regardless of maternal treatment. Similarly, infants exposed to GDM detected and treated using higher New Zealand criteria did not have increased whole-body FM or other infant risk factors for later metabolic disease at 5–6 months’ corrected age. Further follow-up is needed to determine the longer-term benefits and risks of lower versus higher GDM detection thresholds for offspring and to identify causal mediators suitable for postnatal intervention.

Acknowledgments. The authors thank the women and infants who participated in the BabyGEMS study. The authors also acknowledge the following members of the BabyGEMS study team as nonauthor contributors: research coordinators Brittany Morison, Florella Keen, Debbie Samuel, Jenny Rogers, and Nina Slabkevich, and research assistants Olga Ivashkova, Sarah Philipsen, Lisa Mravicich, Catherine Overfield, and Gesina Marie (all nonauthor contributors are from Liggins Institute, University of Auckland, Auckland, New Zealand).

Funding. This study was supported by project grants from the Health Research Council of New Zealand (14/104), Auckland Medical Research Foundation (1115018), Cure Kids Foundation, Nurture Foundation for Reproductive Research, Tupu Fund, Counties Manukau Health, and the University of Auckland. Lotteries Health, Auckland University of Technology, and Gravida supported the purchase of one of the PEA PODs used in this study.

The funders had no role in the analysis and interpretation of data or the decision to publish.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. K.M. designed the data collection instruments, collected data, performed the initial analyses, drafted the initial manuscript, and critically reviewed and revised the manuscript. C.A.Cr., J.E.H., M.P.M., E.C.R., J.M.A., L.M.E.M., J.A.R., R.E., and F.A. conceptualized and designed the study and critically reviewed and revised the manuscript. C.A.Co. designed the data collection instruments, conceptualized and designed the study, and critically reviewed and revised the manuscript. C.J.D.M. conceptualized and designed the study, designed the data collection instruments, collected data, performed the initial analyses, drafted the initial manuscript, and critically reviewed and revised the manuscript. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work. C.J.D.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.