The initial search yielded 2227 records. After removing duplicates and screening titles and abstracts, 57 full-text articles were selected. One article could not be retrieved, leaving 56 articles assessed for eligibility (Fig. 1). Twenty-six studies in total met the inclusion criteria and were included in the systematic review. Table S2 (Supplementary Material) lists all full texts excluded with reasons and also notes one record that could not be retrieved. Details of the study characteristics are summarized in Table 1 and Table S3 (Supplementary Material). The included studies were published from 1993 through 2024. Sample sizes ranged from 103 to 320,479 participants (median 874, IQR 362–2111). Figure 2 shows the number of children included in the review by country. Most studies included children aged 0–5 years. Two studies included children up to 2 years,15,16 three studies17,18,19 included children up to 12 years, and one did not specify an age cut-off.20
Fig. 1
PRISMA flow diagram of study selection process.
Fig. 2: Map showing the spatial distribution of participants included in the systematic review (n = 886,144).
Map created using QGIS 3.44.3. The legend includes only the classes corresponding to those present in the image.
Table 1 Selected characteristics of the included studies (n = 26) assessing malnutrition after extreme weather events.
Although the search strategy included terms for a broad range of extreme weather events (e.g., heatwaves, storms, tropical cyclones, heavy precipitation, and wildfires), only studies evaluating floods and droughts met the eligibility criteria and were included in the final review. Thirteen studies15,17,21,22,23,24,25,26,27,28,29,30,31 examined the effects of drought, eleven16,18,19,20,32,33,34,35,36,37,38 focused on floods, and two39,40 investigated both drought and floods. Sixteen studies17,18,19,21,25,26,28,29,30,31,32,33,36,38,39,40 investigated stunting, thirteen20,23,24,25,26,28,29,30,35,37,38,39,40 examined wasting, eleven16,19,25,26,27,28,29,30,31,38,39 focused on underweight, two on MUAC,22,34 one on BMI for age,19 and none assessed overweight/obesity.
The definitions for stunting, wasting, and underweight were generally consistent across studies, typically based on minus two standard deviations (-2SD) from the median expected value. The reference charts used varied over time, including the National Center for Health Statistics (NCHS) charts of 1977,41 the Center for Disease Control and Prevention (CDC) charts,42 and the World Health Organization (WHO) charts.43 The specific chart reference for each study is provided in Table S3. The only exception to the -2SD cutoff was the study by Choudhury et al.,16 which used a value inferior to the 61st percentile of the median of the NCHS 1977 growth chart reference.
The time interval between the exposure event and outcome assessment varied widely across studies, spanning from in utero or infancy exposure to assessments conducted months or years later (up to approximately 6 years).
The risk of bias was assessed using tools appropriate to each study design. Overall, most studies demonstrated a low-to-moderate risk of bias, with the most common concerns related to exposure measurement and potential confounding. A detailed appraisal is provided in Tables S4–S6 (Supplementary Materials).
Meta-analysis of the prevalence of malnutrition
Ten studies reported the prevalence of wasting after an extreme weather event,20,23,24,26,28,30,35,37,38,40 ten reported the prevalence of stunting,15,18,26,28,30,31,33,36,38,40 and eight reported the prevalence of underweight.16,24,26,27,28,30,31,38 Meta-analyses were performed for all three outcomes.
For the prevalence of stunting, the estimate corresponding to the 5–60-month age group from ref. 18 was used, as it was the most comparable with the other studies. Moreover, two studies30,31 investigated the same population and were included only once in the pooled estimate.30 The pooled prevalence of wasting excluded ref. 20 because no clear age cut-off was specified, and ref. 37 because prevalence estimates were reported separately for children exposed to one versus two flood events, preventing inclusion of a single non-overlapping estimate in the meta-analysis. The pooled prevalence of underweight did not include the study by Choudhury,16 which was excluded after sensitivity analysis because the definition used differed from the others (being based on older reference standards and more consistent with a prevalence of severe underweight). Furthermore, to avoid duplicating data from an overlapping population, only one of the two studies by Singh et al.30 was retained in the pooled estimate.
Therefore, three meta-analyses were conducted, including eight studies for wasting (three on floods35,38,40 and five on droughts23,24,26,28,30), nine studies for stunting (five on floods18,33,36,38,40 and four on droughts15,26,28,30), and six for underweight (one on floods38 and five on droughts24,26,27,28,30).
The pooled prevalence of wasting, stunting, and underweight following an extreme weather event (either drought or flood) was 19% (95% CI: 12–28%), 39% (95% CI: 31–48%), and 42% (95% CI: 22–64%), respectively (Figs. 3–5). These estimates were based on 37,072, 32,948, and 6018 children, respectively. When the study by Choudhury et al.16 was included in the sensitivity analysis, the pooled prevalence of underweight changed to 36% (95% CI: 22–64%), based on 6906 children (Fig. S1, Supplementary Material). When the study by Haq et al.20 was included in the sensitivity analysis, the pooled prevalence of wasting changed to 18% (95% CI: 12–26%), based on 37,728 children (Fig. S2, Supplementary Material).
Fig. 3
Random-effects meta-analysis of the prevalence of wasting among children after an extreme weather event, divided by event type (drought or flood).
Fig. 4
Random-effects meta-analysis of the prevalence of stunting among children after an extreme weather event, divided by event type (drought or flood).
Fig. 5
Random-effects meta-analysis of the prevalence of underweight among children after an extreme weather event.
Subgroup analyses by type of event revealed no significant difference in stunting prevalence between drought- and flood-affected populations (p = 0.68). In contrast, the prevalence of wasting differed significantly (p = 0.04), with higher prevalence following droughts (22%) compared with floods (13%). Substantial heterogeneity was observed in all analyses (I² > 95%).
Association between extreme weather events and malnutrition
Fifteen studies16,17,19,21,22,25,27,29,32,33,36,37,38,39,40 assessed the association between malnutrition and extreme weather events. Among these six reported associations after drought exposure,17,21,22,25,27,29 seven reported associations following flood exposure,16,19,32,33,36,37,38 and two evaluated the risk of malnutrition following either drought or flood exposure.39,40
Drought exposure
Petscavage et al.40 reported that exposure to drought during infancy was associated with an increased risk of wasting, with a relative risk (RR) of 1.50 (95% CI: 1.40–1.61) at 3 months and 1.35 (95% CI: 1.26–1.44) at 12 months after exposure. In contrast, Dimitrova et al.39 found that exposure to an abnormally dry monsoon season during infancy was associated with a reduced risk of wasting among Indian children.
Regarding stunting, Petscavage et al.40 observed a decreased risk at both 3 and 12 months following drought exposure, and Dimitrova et al.39 also reported a lower risk of stunting after both in-utero and infancy exposure to abnormally dry monsoon seasons. Conversely, Shaw et al.29 and Amegbor et al.21 reported an increased risk of stunting following exposure, while Bahru et al.17 found that long-term drought exposure was associated with lower height-for-age z-scores (HAZ).
For underweight, Mason et al.27 found a higher prevalence during drought years, with increases of +3.4 percentage points in Southern Africa and +7.6 percentage points in the Horn of Africa. Similarly, Le et al.25 reported that in-utero drought exposure was associated with lower weight-for-age z-scores (WAZ).
Finally, Bauer et al.22 found that drought severity, measured by a reduction in the normalized difference vegetation index (NDVI), was associated with a 0.52 standard deviation decrease in MUAC.
Flood exposure
Petscavage et al.40 found an increased risk of wasting among children exposed to floods, both at 3 and 12 months after the event, with RRs of 1.66 (95% CI: 1.54–1.80) and 1.48 (95% CI: 1.41–1.56), respectively. Similarly, Rodriguez-Llanes et al.37 reported higher prevalence ratios (PRs) of wasting among children exposed to floods—both once and twice—compared with unexposed children, with PRs of 2.30 (95% CI: 1.86–2.85) and 1.94 (95% CI: 1.43–2.63), respectively. In contrast, Dimitrova et al.39 found no statistically significant association between monsoon flood exposure and wasting (OR = 0.91, 95% CI: 0.83–1.00).
Regarding stunting, Petscavage et al.40 found a slightly increased risk at both 3 and 12 months after flood exposure, with RRs of 1.10 (95% CI: 1.05–1.15) and 1.05 (95% CI: 1.02–1.08), respectively. Dimitrova et al.39 also reported a slight increased risk after both in-utero and infancy exposure, while Rahaman et al.36 reported nearly a twofold increase in risk (OR = 2.14; 95% CI: 1.53–2.28) among exposed children. Rodriguez-Llanes et al.38 also found a higher PR of 1.60 (95% CI: 1.05–2.44) in exposed compared with unexposed children. Consistent with these findings, Del Nino et al.32 reported lower HAZ among exposed children, whereas Rosinger et al.19 did not observe significant differences in HAZ between exposed and unexposed groups. Contrary to the general trend, Gaire et al.33 identified a reduced likelihood of severe and moderate stunting among flood-exposed children (severe: OR = 0.57; 95% CI: 0.31–0.96; moderate: OR = 0.66; 95% CI: 0.41–0.94).
For underweight, Dimitrova et al.,39 Choudhury et al.,16 and Rodriguez-Llanes et al.38 found increased prevalence or risk following flood exposure. In contrast, Rosinger et al.19 observed higher WAZ among children after flood exposure; however, this effect was not evident in communities located farther from affected areas, where post-disaster aid distribution was limited.
Meta-analysis of associations between extreme weather events and stunting
Five studies provided risk estimates for stunting: four reported OR21,33,36,39 and one reported RR.40 Among these, one study21 examined drought, two33,36 investigated floods, and two39,40 assessed both event types. Consequently, studies investigating both exposures39,40 contributed data to both analysis groups.
Separate random-effects meta-analyses were performed for the risk of stunting following floods (four studies) and droughts (three studies). Petscavage et al.40 reported results at both 3 and 12 months post-exposure; therefore, sensitivity analyses were conducted using each period.
The pooled OR for stunting was 1.08 (95% CI: 0.98–1.19) following floods and 0.96 (95% CI: 0.90–1.02) following droughts, indicating no statistically significant associations when considering the 12-month estimates. Using the 3-month data, the association with floods became statistically significant (OR = 1.12; 95% CI: 1.01–1.25), while the association with drought remained non-significant (OR = 0.95; 95% CI: 0.87–1.04). Forest plots for all analyses are presented in the Supplementary Material (Fig. S3–S5).