The research team searched PubMed, Embase, Scopus, the Cochrane Central Register of Controlled Trials, the World Health Organization International Clinical Trials Registry Platform, Google Scholar, and reference lists from inception to January 2026. They included only randomized controlled trials testing plant-derived therapies for nausea and vomiting of pregnancy up to 20 weeks of gestation. The search ultimately identified 28 randomized controlled trials encompassing 2,798 participants, of which 23 contributed to at least one outcome-specific network. Rather than pooling everything into a single estimate, the authors synthesized three distinct symptom instruments in separate frequentist network meta-analyses: the Pregnancy-Unique Quantification of Emesis and Nausea score, Rhodes-derived scales, and a 0 to 10 visual analog scale. Random-effects models were used for the PUQE and VAS networks, while a fixed-effect model was applied to the Rhodes network.

The technical architecture of the analysis matters because network meta-analysis allows indirect comparisons between treatments that have never been tested head-to-head in the same trial. If ginger has been compared with placebo and lavender has been compared with placebo, the network can estimate how ginger might compare with lavender, even though no trial has directly pitted them against each other. But this statistical power depends on the network being coherent, meaning that direct and indirect evidence agree. When the underlying trials are small, use different outcome instruments, and measure symptoms at different time points, the network can become inconsistent, and the resulting estimates lose their reliability. That is precisely the problem the Thai team encountered in one of their three networks.

The PUQE network, which included 8 trials and 655 participants, showed marked global inconsistency with a chi-squared statistic of 117.38 on 3 degrees of freedom and a p value below 0.001, alongside substantial heterogeneity with a tau-squared of 4.25. In plain terms, the trials feeding this network disagreed with each other so fundamentally that the pooled estimates could not be interpreted reliably. Confidence in every comparison within this network was rated very low using the Confidence in Network Meta-Analysis framework, a structured tool for grading how much trust can be placed in network estimates. The authors concluded that the PUQE findings represent comparative efficacy signals rather than a basis for clinical recommendations, a deliberately cautious formulation that resists the temptation to crown a single winner.

The Rhodes network told a more favorable story, at least on the surface. Drawing on 8 trials and 835 participants, the placebo-referenced estimates favored lavender with a standardized mean difference of minus 1.82, with a 95 percent confidence interval from minus 2.25 to minus 1.39, and orange blossom with a standardized mean difference of minus 1.44, with a 95 percent confidence interval from minus 1.85 to minus 1.03. Several other interventions also showed favorable point estimates. However, confidence was rated low for both lavender and orange blossom, and each of these estimates was informed by a single trial. Confidence for ginger within this network was rated very low because of local inconsistency, meaning that the direct and indirect evidence pathways for ginger disagreed within this specific network.

The visual analog scale network, comprising 7 trials and 594 participants, produced its own pattern. Ginger was associated with a mean difference of minus 2.06 on the 0 to 10 nausea scale, with a 95 percent confidence interval from minus 2.91 to minus 1.20, and vitamin B6 showed a mean difference of minus 1.89, with a 95 percent confidence interval from minus 3.04 to minus 0.74, both favoring the active intervention over placebo. Yet confidence in these estimates was rated low, and the two treatments showed little or no clinically important difference from each other. In other words, ginger and vitamin B6 may both reduce nausea scores compared with placebo on this instrument, but the evidence does not support declaring one superior to the other.

Safety data emerged as another weak point across the evidence base. Of the 28 included trials, only 11 reported at least one adverse event. In the remaining trials, adverse-event reporting ranged from explicit statements that no events occurred to incomplete or entirely unavailable safety information. This patchy reporting matters because pregnant women and their clinicians need to weigh not only whether a remedy works but also whether it carries risks to the developing fetus. The absence of systematic safety monitoring in more than half of the trials means that the overall safety profile of these plant-derived therapies remains incompletely characterized, even for widely consumed remedies like ginger.

The authors were explicit about the limits of their findings. The available randomized evidence does not establish a single most effective plant-derived therapy for nausea and vomiting of pregnancy, they concluded. Favorable estimates in the Rhodes and VAS networks were supported by low or very low confidence evidence, and marked inconsistency precluded reliable interpretation of the PUQE network. The study was registered prospectively on PROSPERO under identifier CRD420251116302, and risk of bias was assessed using the Cochrane Risk of Bias 2 tool, both markers of methodological rigor that strengthen the review’s credibility even as its findings remain cautious.

Why does this matter for the millions of women who experience morning sickness each year? The answer lies in the gap between popular belief and clinical evidence. Ginger in particular has accumulated a reputation as a go-to natural remedy, and aromatherapy with lavender or orange blossom is frequently recommended in pregnancy wellness circles. This analysis does not invalidate those choices, but it does make clear that the evidence supporting them is fragile, built on small trials, single-study estimates, and outcome measures that resist coherent synthesis. Women considering these therapies should discuss them with their healthcare providers rather than relying on confident claims drawn from incomplete data.

The study also highlights a broader challenge in complementary medicine research. When trials are small, use different symptom instruments, and rarely report adverse events, even the most sophisticated statistical tools cannot manufacture certainty. Network meta-analysis can stretch limited data across a wider comparative landscape, but it cannot conjure reliability from inconsistency. The Thai team’s work, supported by publication assistance from HRH Princess Mahachakri Sirindhorn Medical Center at Srinakharinwirot University and conducted without external funding, offers a template for honest appraisal: report the signals, grade the confidence, and resist the urge to recommend. For now, the search for the single best plant-derived remedy for morning sickness remains open, and the strongest conclusion this analysis supports is that better, larger, and more standardized trials are urgently needed.

Subject of Research: Health

Article Title: Plant-derived complementary therapies for nausea and vomiting of pregnancy: a systematic review and network meta-analysis of randomized controlled trials

Article References: Thitinan, N., Nontasuwan, P.-E., Meyprakhon, P., Sakorntanagorn, L., Teerakantrakorn, P., Nontaprom, K., Kassayanan, P., & Suntipap, M. (2026). Plant-derived complementary therapies for nausea and vomiting of pregnancy: a systematic review and network meta-analysis of randomized controlled trials. BMC Complementary Medicine and Therapies. https://doi.org/10.1186/s12906-026-05647-1

Image Credits: AI Generated

DOI: 10.1186/s12906-026-05647-1

Keywords: Plant-derived, complementary, therapies, nausea, vomiting, pregnancy, systematic, review, network, meta-analysis, randomized, controlled