Does Dietitian Involvement During Pregnancy Improve Birth Outcomes? A Systematic Review

Comprehensive literature searches of PubMed (1966–2021), CINAHL (1937–2021), Embase (1947–2021), and Web of Science (1970–2021) were conducted independently on December 18 2021 by two authors (WL and MH). A search strategy was developed using the key search terms in Supplementary File A¹. Search string, MeSH terms, and subject headings were then created for each database. The search strategy was developed by the authors and a health sciences librarian at Brescia University College. Search results from the four databases were imported into Covidence (Covidence systematic review software, Veritas Health Innovation, Melbourne, Australia. 2022), and duplicates were removed. Covidence was used to screen titles, abstracts, and full texts. Reference lists of retrieved articles were assessed to identify additional studies not obtained in the original search.

Studies needed to be original research, written in English, conducted in Canada and (or) the United States, and include pregnant individuals. The search was limited to the two countries to ensure consistency in the legal practice of RDs. Each study should have observed at least one of the six birth outcomes: LBW (<2500g), SGA (birth weight <10th percentile for gestational age or infants smaller in size than normal for their gestational age) [18, 19], LGA (a birth weight >90th percentile) [20], macrosomia (birth weight >4000g) [21], PTB (<37 weeks’ gestation), and infant mortality (death of the fetus or infant death following a live birth; i.e., stillbirth, neonatal death, and perinatal death). Eligible studies must have had direct involvement from an RD or those with equivalent legal titles (Supplementary File B¹). Direct involvement was defined as nutrition practice (i.e., nutrition interventions, dietary changes/plan, nutrition counselling, or medical nutrition therapy) by an RD alone or as a part of a multidisciplinary team. Reviews, editorials, letters, and grey literature were excluded. There were no restrictions on study design or publication dates.

Title, abstract, and full-text screening were conducted independently by two authors (WL and MH), with conflicts resolved by consensus or adjudication with the senior author (JAS).

A standardized coding guide was developed using Microsoft Excel (Microsoft Corporation, Redmond, Washington, USA, 2019). Information included were publication year, author name(s), objectives, sample characteristics (e.g., health status, maternal age, race/ethnicity, socioeconomic status), geographic location, sample size, study design, types of RD involvement, key findings, and study limitations. Two authors (WL and MH) independently extracted data and collectively resolved discrepancies. The senior author (JAS) adjudicated any unresolved discrepancies.

The methodological quality of studies was assessed by two independent authors (WL and MH) using the JBI critical appraisal tool [22, 23]. This tool determined the risk of bias through a customized checklist created for each study. A specific JBI checklist was used for cross-sectional studies, case-control studies, cohort studies, and RCTs [22, 23]. The checklist consisted of four possible answers: “yes,” “no,” “unclear,” and “not applicable.” Answers that were “not applicable” were excluded from the final count towards determining the overall quality rating of each paper. The quality rating of each study was determined using a three-range scale (≥ 80% “yes” is good; 60-79% “yes” is fair; ≤ 59% “yes” is poor). The three-range scale and its ratings closely align with other studies using the JBI [24, 25]. Checklist questions and a summary of the assessment can be found in Supplementary Tables 1–4¹.

The search strategy identified 310 studies, 114 were removed as duplicates, and 196 required screening. From the 196 articles, 188 did not meet inclusion criteria. Six additional articles were retrieved from reference lists screened. Therefore, 14 studies were included in this systematic review (Figure 1).

Thirteen studies were conducted in the United States, and 1 study was from Canada. The publication years ranged from 1989 to 2018. Study designs consisted of 2 cross-sectional studies, 2 case-control studies, 3 retrospective cohort studies, 5 RCTs, 1 implementation trial, and 1 pilot study. Low birth weight was examined in 7 studies, SGA in 3 studies, LGA in 4 studies, macrosomia in 6 studies, PTB in 9 studies, and infant mortality in 2 studies. The average age of pregnant women was 29.3 years among 9 studies, with 1 study specifying the common age as 20 to 24 years [26] and another study specifying the median age as 26.8 years in its study group and 37.3 years in its control group [27]. Three studies did not report maternal age [28–30]. Seven studies described participants’ gestational age. Most women entered the studies before the 28^th week of gestation, although participants in 3 studies joined before 16 weeks of gestation [12, 28, 31]. Sample sizes ranged from 87 to 2,377 participants. Eleven of the 14 studies included diverse ethnic groups. Registered dietitian involvement during pregnancy included nutrition care, education, and counselling. A description of the studies (e.g., study design, sample characteristics, type of RD involvement) can be found in Table 1.

Four of the 5 studies that compared RD involvement to no RD involvement found a lower prevalence of LBW in groups with RD involvement [26, 31–33]. When comparing standard RD care to advanced RD care, an RCT [34] found no difference between groups receiving standard nutrition counselling (control) compared to advanced nutrition care (intervention), although very few patients had LBW overall. For example, among women identified as normal weight at the first prenatal visit, 5/92 women had a LBW infant in the control group versus 4/90 in the intervention group. Similarly, only 4/86 overweight women had a LBW infant in the control group compared to 5/81 in the intervention group. One study found that 9.7% of all participants had a LBW infant, although LBW was reduced to 6.7% among women with adequate gestational weight gain, compared to 17.2% among those with inadequate weight gain (p < 0.001) [30]. In an implementation trial for clinical practice guidelines [35], both the intervention and control groups included RDs. Registered dietitians in the intervention group (n = 130) provided prenatal care as described in nutrition practice guidelines for gestational diabetes mellitus (GDM), whereas RDs in the control group (n = 85) provided usual prenatal care. No statistically significant difference in LBW was found between the nutrition practice guidelines group and the usual care group (2.4% vs. 8.4%, respectively, p = 0.27) [35].

In an RCT by Peccei et al. assessing the effect of a culturally appropriate nutritional intervention delivered to overweight and obese patients in a community health setting, there was no significant difference in SGA prevalence between the RD involvement intervention group and the control group (6.1% vs. 3.3%; odds ratio [OR] = 1.9, 95% confidence interval [CI]: 0.5–7.1) [28]. In the RCT by Vesco et al., where the intervention group received individualized nutrition counselling and the control group received one-time dietary advice, no significant difference in SGA was found between the intervention and control group (5% vs. 7%; OR = 0.76, 95% CI: 0.11−4.76) [12]. In Artal et al.’s RCT, both groups received medical nutrition therapy (MNT) administrated by RDs, and there was no difference in the rate of SGA in the group receiving MNT and exercise (2/39; 5.9%) to those receiving the diet intervention alone (1/57; 2.2%) [36].

Fassett et al. found no significant difference in LGA between the RD involvement group and the non-RD involvement group (14% vs. 14%; p = 0.94) [37]. Peccei et al. examined LGA among all participants and separately among obese participants [28]. The intervention group, which involved intensive prenatal nutrition counselling by an RD, was not significantly different from the control group (6% vs. 13%; OR = 0.4, 95% CI: 0.2-1.0; p = 0.058) for LGA. Among obese participants however, those in the intervention group had 70% lower odds of LGA infants than in the control group, and the difference was statistically significant (7% vs. 17%; OR = 0.3, 95% CI: 0.1–0.99; p = 0.048) [28]. An RCT by Vesco et al. also found a significantly lower prevalence of LGA in the intervention group than the control group (9% vs. 26%; OR = 0.28, 95% CI: 0.09–0.84) [12]. In the Artal et al. RCT, both groups received a diet intervention, but there was no significant different in LGA between the group receiving MNT alone (7/57) and the group receiving an additional exercise intervention (4/39) (p = 0.64) [36].

The retrospective case-control study by Fassett et al. revealed no difference in macrosomia between the RD involvement group and the non-RD involvement comparator (14% vs. 14%; p = 0.94) [37]. The pilot study by Weiderman et al. also found no significant difference between groups with and without RD involvement (p = 0.40) [29]. Phelan et al. found that, among those with a prepregnancy BMI indicating normal weight, 3/92 birthed children with fetal macrosomia in the standard nutrition care group compared to 6/90 in the intensive nutrition care group [34]. For women with a prepregnancy BMI indicating overweight, the rates of macrosomia were 14/86 (16.3%) and 14/81 (17.3%) for the standard and intensive nutrition care groups, respectively. In studies by Reader et al. [35], Vesco et al. [12], and Thornton et al. [27], there were no statistically significant differences in macrosomia in the control groups receiving basic nutrition care (i.e., nutrition counselling) and the groups receiving extensive nutrition care (i.e., dietary advice, weight gain discussions, dietitian follow-ups).

Among 5 studies examining PTB that compared groups with RD involvement to those without RD involvement, 4 studies found a lower prevalence of PTB in groups with RD involvement [26, 31–33], with 2 studies indicating a significant difference [31, 32]. One study found no significant difference between groups [38]. Among the remaining 4 studies, all participants received nutrition care by RDs, with 1 study indicating a lower prevalence of PTB in groups receiving extensive nutrition care [34], and 3 studies finding no significant differences in PTB between groups receiving standard and in-depth nutrition care [12, 27, 35].

Perinatal death and stillbirth were reported as infant mortality in 2 studies [26, 32]. Both found a lower prevalence of infant mortality with RD involvement compared to no RD involvement.

This systematic review demonstrates that pregnant individuals have a lower prevalence of LBW infants when RDs are involved during prenatal care. This is an important finding because LBW increases the risk for infant mortality, poor cognitive development, respiratory distress, and asthma during childhood, and cardiovascular disease, type 2 diabetes, and hypertension during adulthood [2]. While the etiology of LBW is multifactorial [39, 40], poor maternal nutrition is a major contributing factor [41, 42]. Two of the 4 studies which found a lower prevalence of LBW with RD involvement compared to no involvement were of good quality, 1 was fair quality, and 1 was poor quality. RD involvement was also associated with a lower prevalence of PTB, particularly when RD involvement was compared to no involvement (4/5 studies). Furthermore, while RD involvement during pregnancy was not associated with macrosomia, more research is needed to elucidate its correlation with SGA, LGA, and infant mortality. All 3 studies that assessed SGA as an outcome were not of good methodological quality. Additionally, 2/4 studies found a reduction in LGA infants with RD involvement compared to no involvement, and only 2 studies assessed infant mortality, both of which found a lower rate of mortality with RD involvement.

Although this systematic review investigates the association between RD involvement during pregnancy and birth outcomes in Canada and the United States, our findings are fairly consistent with studies from other countries. In a quasi-experimental design from Mexico City, Perichart-Perera et al. [43] examined the relationship between dietitian-initiated MNT (i.e., counselling, education, and capillary glucose monitoring) and birth outcomes. Among women with GDM, the MNT group were significantly less likely to have a LBW infant than the control group (5.1 % vs 20.5%, p = 0.03), although there were no differences in LBW between groups among women with type 2 diabetes. The study found no significant differences in rates of prematurity and macrosomia between the two groups. In an RCT from Australia examining whether treatment of women with GDM reduced the risk of perinatal complications, Crowther et al. [13] found that the intervention group (n = 506, receiving dietary advice from an RD, blood glucose monitoring, insulin therapy) was significantly less likely than the routine-care group (n = 524) to have an LGA (13% vs. 22%, p < 0.001) or fetal macrosomia (10% vs 21%, p < 0.001) outcome, although SGA was similar between the two groups. Noteworthy is that the current review includes various population types, such as twin pregnancies, adolescent pregnancies, obese pregnant women, and women diagnosed with GDM, all of which increase the risk for adverse birth outcomes [44–47]. It is plausible that pregnant individuals may have received different nutrition care from RDs depending on their health concerns and conditions, adding to greater heterogeneity in the exposure.

This systematic review provides a comprehensive assessment of the association between RD involvement during pregnancy and birth outcomes in Canada and the United States. Given that many prenatal health care providers report insufficient education and training in nutrition [48, 49], and that dietitians are regulated health professionals who undergo rigorous training, RD advice to pregnant individuals (e.g., recommended foods and those to avoid) may enhance maternal health and fetal development, if the advice is followed. However, there are some limitations to consider when interpreting these findings. First, there was limited sociodemographic information reported in most studies with respect to race, ethnicity, and socioeconomic status. This is unfortunate because research shows that pregnant women who are non-Hispanic Black, Hispanic, and those with low educational attainment have poorer diet quality than those who are non-Hispanic Whites [50]. Future research should address the extent to which RD involvement during pregnancy can attenuate disparities in birth outcomes by factors such as socioeconomic status, race, and ethnicity. Second, several studies had inadequate statistical power to detect clinically meaningful results in birth outcomes. Underpowered studies are problematic because they increase the risk of a type II error/false negative [51]. While underpowered studies can sometimes be mitigated by pooling the data when conducting a meta-analysis, the heterogeneity would have been very large between the included studies since the definition of RD involvement varied considerably. Future interventions that involve RDs during pregnancy need to accurately specify the content and frequency of their involvement.