The effect on BP was not consistent across trial populations. There were variations in the types of dietary intervention and the effect of specific nutrient components. There was no heterogeneity for dietary intervention components, but considerate heterogeneity for all dietary interventions combined. Dietary counselling interventions shown to lower BP varied in frequency from one to 10 sessions. Target populations ranged from adolescent girls (Chan et al. 2006) to obese women (Wolff et al. 2008; Vinter et al. 2011), limiting the generalisability of the findings to a broader group of women of child-bearing age. Those interventions shown to be effective in reducing BP included a balanced diet complying with National recommendations (Vinter et al. 2011), energy intake individualised for the needs of the mother (Wolff et al. 2008) and modifying calcium intake (Chan et al. 2006). Identifying specific components of successful interventions can assist in understanding how the intervention exerts its effect (Michie et al. 2011). Calcium has been studied in large supplemental trials (Palacios & Pena-Rosas 2011), demonstrating a reduction in the risk of hypertensive disorders during pregnancy, particularly for women at high-risk or with low calcium intakes (Hofmeyr et al. 2010; Palacios & Pena-Rosas 2011; Imdad & Bhutta 2012). Our review identified no effect of dietary intervention on hypertensive disorders. However, the pooled trials involved whole diet recommendations, or the modification of fat intake, and there were no trials on hypertensive disorders specifically targeting calcium intake.
Like the results in our review, calcium intake has been shown to lower BP among pregnant women (Carroli et al. 1994; Van Mierlo et al. 2006; Hofmeyr et al. 2010; Palacios & Pena-Rosas 2011), and the effects appear stronger in women with low calcium intakes prior to intervention (Carroli et al. 1994; Van Mierlo et al. 2006; Hofmeyr et al. 2010; Palacios & Pena-Rosas 2011). Calcium intakes were not analysed for each included study as part of this review. The effect of low calcium during pregnancy is thought to exert its effect via an increase in parathyroid hormone secretion, which increases intracellular calcium, smooth muscle contractibility and/or releases renin from the kidney, leading to vasoconstriction and retention of sodium (Hacker et al. 2012). These physiological changes (from low calcium intakes during pregnancy) increase BP and potentially contribute to the development of hypertensive disorders (Hacker et al. 2012). Women beginning pregnancy with adequate intakes of at least 1000 mg calcium per day may not need additional amounts, while those with suboptimal intakes (<500 mg per day) may benefit from intervention (Hacker et al. 2012).
Our review also found that women consuming a balanced diet, including enough energy based on their individual requirements, had lower BP, paralleling current lifestyle recommendations for individuals with high BP (National Institutes of Health 2003). The effect size of 1–3 mmHg was very small compared with using antihypertensive agents (Patel et al. 2012). However, the effect was evident in normotensive rather than hypertensive women. Furthermore, treatment with antihypertensive drugs during pregnancy carries known and unknown risks to the fetus, because these drugs cross the placenta (e.g. nifedapine is Category C) (Department of Health, 2014). Therefore, any reductions in BP gained from dietary intervention offers clear advantages when BP is of clinical concern.
Macronutrient interventions demonstrating a reduction in the incidence of preterm delivery varied in frequency and included supplemental beverages (Rush et al. 1980; McDonald et al. 1981), dietary counselling on the nutritional needs during pregnancy (Kafatos et al. 1989; Briley et al. 2002; Quinlivan et al. 2011), limiting cholesterol and reducing saturated fat (Khoury et al. 2005), GDM-specific dietary recommendations (Thornton et al. 2009), providing DHA-fortified eggs (Smuts et al. 2003a,b), and a range of energy- and protein-based foods (Mora et al. 1978). Observational studies (Kramer 1987; Institute of Medicine 1990; Rush 2001) have reported that energy intake may be strongly and positively associated with a reduced risk of preterm birth. Our review confirms these findings, demonstrating a reduction in the incidence of preterm birth with macronutrient or whole diet intervention.
The pooled studies on preterm delivery showed little evidence of heterogeneity, with a narrow spread of data and overlapping CIs. Six of the macronutrient dietary interventions were conducted in low-income populations (Mora et al. 1978; Rush et al. 1980; McDonald et al. 1981; Kafatos et al. 1989; Briley et al. 2002; Quinlivan et al. 2011). Four interventions contained small sample sizes of less than 125 pregnant women (Briley et al. 2002; Smuts et al. 2003a; de Groot et al. 2004; Quinlivan et al. 2011) meaning these trials were individually underpowered. Of the 13 RCTs contributing data, only two trials (Kafatos et al. 1989; Khoury et al. 2005) demonstrated statistically significant effects in their respective publications. Based on the pooled dietary interventions, there were particular nutrients impacting on the result rather than the type of intervention. Modifying fat intake, particularly long-chain polyunsaturated fatty acids (LC-PUFA) were shown to reduce the incidence of preterm delivery (Smuts et al. 2003a,b; Khoury et al. 2005). Paralleling these findings, Horvath et al. (2007), Szajewska et al. (2006) and a recent Cochrane Review (Makrides et al. 2012) concluded that women allocated to LC-PUFA supplementation had longer gestation than women receiving placebo or no treatment, which remained true for both low- and high-risk pregnancies. Observational studies, mainly in populations with high consumption of seafood, have also suggested that marine LC-PUFA intake during pregnancy promotes longer gestation (Olsen et al. 1986, 1990, 1993). DHA and arachidonic acid are essential nutrients that are supplied during pregnancy to the fetus by preferential placental transfer (Al et al. 1995; Otto et al. 1997; Berghaus et al. 2000; Larquè et al. 2003). The mechanism behind their role in increasing gestational length may be an imbalance between DHA and arachidonic acid, which is associated with disturbances in the production of prostacyclin and thromboxane involved in the initiation of labour (Herrera 2002). Therefore, adequate dietary intake or supplementation of LC-PUFA may prolong gestational length, and in turn, decrease the risk of preterm delivery (Horvath et al. 2007).
The conversion of each publication's results to SMD for statistical inference effectively creates a common scale that would otherwise not be possible because of the differences in variance between publications. The SMD express differences as units of standard deviations, and thus cannot be used to interpret absolute or relative differences in clinically meaningful outcomes (e.g. for BP, the absolute reduction in mmHg). Therefore, we have also provided the RMD to give some indication of effect size in the standard units for each outcome, including odds ratios for outcomes like pre- and post-term delivery.
This systematic review is broader in scope when compared with other published systematic reviews and meta-analyses (Dodd et al. 2010; Streuling et al. 2010; Tanentsapf et al. 2011; Oteng-Ntim et al. 2012; Thangaratinam et al. 2012), and the quality of the included studies was mostly positive. Gestational weight gain was not included as an outcome in this review, as others have focused on this (Dodd et al. 2010; Streuling et al. 2010; Tanentsapf et al. 2011; Oteng-Ntim et al. 2012; Thangaratinam et al. 2012). Despite the broad scope of this review, very few trials contributed data to each pregnancy outcome, with the exception of length of gestation (n = 12). For this reason, some of the outcomes were not reported and others are underpowered, particularly with the subgroup analysis. Dietary intervention trials should measure and report on a range of pregnancy outcomes so the effects of diet on pregnancy outcomes can be determined. There was heterogeneity for BP, GDM and length of gestation only, which was not explained by subgroup analyses, but may be due to the varying intensity and duration of included trials.