Prenatal exposure to non-ionizing electromagnetic fields and adverse perinatal outcomes: a systematic review

¹Doctoral Study Program of Medical and Health Sciences, Universitas Diponegoro, Semarang, Indonesia, ²Anatomy-Histology Department, Faculty of Medicine, Universitas Diponegoro, Semarang, Indonesia, ³Department of Physics, Faculty of Science and Mathematics, Universitas Diponegoro, Semarang, Indonesia, ⁴Department of Biology, Faculty of Science and Mathematics, Universitas Diponegoro, Semarang, Indonesia

^&Corresponding author
Desy Armalina, Doctoral Study Program of Medical and Health Sciences, Universitas Diponegoro, Semarang, Indonesia

Introduction    

Non-ionizing electromagnetic fields (EMF) are ubiquitous in modern environments, encompassing the ELF, IF, and RF spectra. Common sources of exposure include power lines and household appliances (ELF), consumer electronics (IF), and wireless communication systems, such as mobile and cordless phones, as well as wifi (RF). These frequency bands differ in their physical interactions with tissues and the metrics used for assessment-for example, magnetic flux density is commonly used for ELF, while specific absorption rate is used for RF-), complicating comparisons across studies and outcomes [1]. The widespread adoption of wireless technologies and antennas in handheld devices has further increased exposure opportunities during routine activities, including during pregnancy [2].

Pregnancy is a period of heightened biological sensitivity, characterized by rapid cellular proliferation and modulation of the endocrine and immunological systems. Epidemiological research on prenatal exposure to electromagnetic fields (EMF) has yielded mixed results. Certain studies have identified associations between elevated magnetic field levels or proxies for mobile-phone-related radiofrequency (RF) exposure and adverse outcomes, such as miscarriage, reduced gestational duration, or indicators of fetal growth (e.g., low birth weight and small-for-gestational-age). On the other hand, several studies have shown conflicting or nonexistent results. On the other hand, several studies have shown conflicting or nonexistent results [3-8].

In particular, a prospective cohort study reported an elevated risk of miscarriage with higher magnetic field exposure [3]. In contrast, multi-cohort studies have linked greater maternal mobile phone use to shorter pregnancy duration [4,8]. Conversely, a recent meta-analysis for ELF found no correlation with miscarriage, stillbirth, neonatal defects, or preterm delivery, with uncertainty remaining for SGA and low birthweight [7].

Mechanistic considerations include thermal and non-thermal pathways. RF exposure is commonly described in terms of SAR and may result in small temperature increases under specific configurations. Experimental work suggests that context-dependent SAR peaks around mid-gestation and emphasizes the spatial relationships among the device, placenta, and fetus [9]. Non-thermal hypotheses suggest oxidative stress and alterations in cellular signaling. Reviews have noted changes in antioxidant defense and inflammatory markers under specific exposure conditions [10,11]. Given the heterogeneity in exposure metrics (ELF versus RF; spot versus personal measurements; behavioral proxies), timing (trimester-specific capture), outcome definitions, and confounder adjustment, a narrative rather than quantitative synthesis is appropriate for integrating this literature [6,7].

In accordance with the PRISMA 2020 guidelines [12], this systematic review adhered to a registered protocol (PROSPERO CRD42023475665) to ensure transparent identification, selection, and appraisal of observational human studies on prenatal EMF exposure and perinatal outcomes [12]. Our aim was to (1) map human epidemiological evidence on prenatal EMF exposure and adverse perinatal outcomes, (2) summarize results by outcome domain using structured narrative synthesis, and (3) evaluate study-level risk of bias to prioritize methodological improvements for future research.

Methods     

Reporting framework and registration

We followed the PRISMA 2020 reporting guidelines and aligned our search write-up with PRISMA-S. The protocol was registered in PROSPERO (CRD42023475665). Ethical approval was not required for the systematic review of the published literature.

Eligibility criteria (PEOS)

Population/exposure: researchers have found evidence of non-ionizing electromagnetic fields (EMF) in the ELF, IF, and RF spectrums, including magnetic fields and wireless sources (such as cell phones, Wi-Fi, and cordless routers), in human prenatal investigations. The mode of conception (spontaneous/assisted) was not a restriction; if reported, it was recorded as a characteristic of the study.

Outcomes: the study focuses on one or more predetermined perinatal outcomes, including miscarriage or spontaneous abortion, stillbirth, gestational duration or preterm birth, fetal growth indicators such as birth weight, length, and small-for-gestational age, congenital anomalies or birth defects, Apgar score, neonatal anthropometry, and maternal pregnancy outcomes, such as gestation length.

Study designs: observational human studies, including cohort, case-control, and cross-sectional designs, were examined. Previous systematic reviews and meta-analyses were utilized solely for contextual purposes and were not included in the pooled analysis. Animal/in vitro evidence-informed biological discussions, but were not synthesized quantitatively.

Exclusions: non-research items (editorials, commentaries, letters), non-peer-reviewed proceedings, non-English records, unavailable full texts, non-human studies, and articles not reporting prenatal EMF exposure were excluded.

Information sources

We conducted a comprehensive search of the PubMed/MEDLINE (NLM), Scopus (Elsevier), Cochrane Library (Wiley), ScienceDirect (Elsevier), and Google Scholar databases from their inception through September 12, 2025. Additionally, we performed backward citation chasing of the reference lists of the included studies.

Search strategy (database-specific, reproducible)

Exact Boolean strings (with parentheses and field tags) and final run dates are provided below. The default database sorting was retained. The interface filters were explicitly stated.

PubMed (final search: September 12, 2025): (("Pregnancy"[Mesh] OR pregnan*[tiab] OR gestat*[tiab] OR prenatal[tiab] OR antenatal[tiab]) AND ("Electromagnetic Fields"[Mesh] OR "Radio Waves"[Mesh] OR "Microwaves"[Mesh] OR electromagnet*[tiab] OR EMF*[tiab] OR radiofrequenc*[tiab] OR RF[tiab] OR "magnetic field*"[tiab] OR "extremely low frequency"[tiab] OR ELF[tiab] OR "non-ionizing"[tiab] OR non-ionizing[tiab] OR "Wi-Fi"[tiab] OR wifi[tiab] OR "cell phone*"[tiab] OR "mobile phone*"[tiab] OR smartphone*[tiab])) AND ("Humans"[tiab]).

Scopus (final search: September 12, 2025): TITLE-ABS-KEY( (pregnan* OR gestat* OR prenatal OR antenatal) AND ("electromagnetic field*" OR EMF* OR radiofrequenc* OR RF OR "radio waves" OR microwav* OR "non-ionizing" OR non-ionizing OR "magnetic field*" OR "extremely low frequency" OR ELF OR "Wifi" OR wifi OR "cell phone*" OR "mobile phone*" OR smartphone*)) AND PUBYEAR > 1989 AND (LIMIT-TO (LANGUAGE, "English"))

Cochrane Library - all text (final search: September 12, 2025): ([mh Pregnancy] OR pregnan*:ti,ab,kw OR gestat*:ti,ab,kw OR prenatal:ti,ab,kw OR antenatal:ti,ab,kw) AND ([mh "Electromagnetic Fields"] OR electromagnet*:ti,ab,kw OR EMF*:ti,ab,kw OR radiofrequenc*:ti,ab,kw OR RF:ti,ab,kw OR "magnetic field*":ti,ab,kw OR "extremely low frequency":ti,ab,kw OR ELF:ti,ab,kw OR microwav*:ti,ab,kw OR "non-ionizing":ti,ab,kw OR wifi:ti,ab,kw OR "cell phone*":ti,ab,kw OR "mobile phone*":ti,ab,kw OR smartphone*:ti,ab,kw).

ScienceDirect (final search: September 12, 2025): (pregnan* OR gestat* OR prenatal OR antenatal) AND (electromagnetic OR "electromagnetic field*" OR EMF* OR radiofrequenc* OR RF OR "magnetic field*" OR microwav* OR "extremely low frequency" OR ELF OR "non-ionizing" OR non-ionizing OR wifi OR "Wifi" OR "cell phone*" OR "mobile phone*" OR smartphone*).

Google Scholar (final search: September 12, 2025): "pregnancy" AND ("electromagnetic fields" OR EMF OR radiofrequency OR RF OR "magnetic fields" OR "extremely low frequency" OR ELF OR microwaves OR "non-ionizing" OR non-ionizing OR wifi OR "mobile phone" OR smartphone). Field-tag notes: ti=title, ab=abstract, kw=keywords (Cochrane); PubMed uses [tiab] and [Mesh], and Scopus uses TITLE-ABS-KEY-.

Citation management and deduplication

All records were exported with full metadata and managed using a reference manager. Duplicates were removed using automated matching (title/DOI/PMID/author/year), and the results were manually verified before screening.

Screening process

Following a thorough evaluation of the entire texts of possibly relevant records, two reviewers independently assessed the titles and abstracts according to the qualifying criteria. Talks or consultations with a third-party reviewer settled any disputes. Study numbers are presented in the Results/PRISMA Figure, and the screening procedure was conducted in accordance with the PRISMA 2020 recommendations.

Data items and extraction

From each included study, we extracted the following: study setting and design; population characteristics; exposure assessment (frequency band, instrument/proxy, metric-e.g., mG/µT for ELF; SAR/proxy for RF-and timing/trimester); outcome definitions; effect measures (OR, RR, HR, MD/SMD) and 95% CIs; adjustment sets (maternal age, smoking, parity, SES, co-exposures); and notes on selection, missingness, and sensitivity analyses. Extraction forms were piloted on a subset of studies and refined before complete extraction.

Risk of bias assessment

The Newcastle-Ottawa Scale [13], was used to assess observational research, and the ROBINS-I tool [14] was used where relevant, especially for complicated non-randomized designs. In the event of a disagreement, the studies were reviewed by two separate individuals, and the matter was finally settled by consensus. We did not use the STROBE principles for quality assessment, but rather as a platform for reporting.

Evidence synthesis

Given the non-comparable exposure metrics (ELF vs. RF; instrumented vs. proxy-based), non-aligned time windows (trimester capture), and diverse effect measures/contrasts, we conducted a structured narrative synthesis only. Findings were grouped by outcome domain (miscarriage, gestational duration/preterm, fetal growth, congenital anomalies, and neonatal measures) and, within each domain, by exposure type/metric. In future research, where ≥10 sufficiently comparable studies exist, small-study effects (e.g., Egger´s test) and quantitative pooling could be considered; however, these were not performed in this study.

Deviations from protocol

Any deviations from the registered protocol (if any) are reported transparently in the Results/Discussion section; none affected the core eligibility or synthesis approach.

Current status of knowledge     

Overview of the evidence base

This systematic review synthesizes observational evidence on prenatal exposure to non-ionizing electromagnetic fields (EMF) and perinatal outcomes, without conducting a meta-analysis, due to insufficient comparability of exposure metrics, timing windows, and effect measures across studies [6,7]. Following PRISMA 2020 guidelines and a registered protocol (PROSPERO CRD42023475665), the literature search identified a total of 107,847 records (23,687 from databases and 84,160 from other sources). After removing 94,658 duplicates, 13,189 records were screened based on titles and abstracts, resulting in the exclusion of 12,721 records. Subsequently, 468 full-text articles were assessed for eligibility, of which 437 were excluded, leading to the inclusion of 31 studies in the final analysis, as illustrated in the PRISMA 2020 flow diagram [12] (Figure 2). In addition, the risk of bias for each included study was assessed using the Newcastle-Ottawa Scale (NOS) following Aromataris (2015) [15], and the results were visualized using the Risk of Bias Visualization (robvis) tool (McGuinness, 2020) [14], as presented in Figure 2.

The included body of evidence spanned cohort, case-control, cross-sectional, and review designs conducted across diverse regions and healthcare settings (Table 1, Table 2, Table 3). The literature collectively examines a variety of outcome domains, including miscarriage and spontaneous abortion, stillbirth, gestational duration and preterm birth, fetal growth (e.g., birth weight and small-for-gestational age), congenital anomalies, neonatal anthropometry, and early neurodevelopment. Although this offers a comprehensive scope, it also introduces considerable methodological variability, complicating quantitative synthesis [1,4,5,7,8,16-28].

Findings from cohort and cross-sectional studies

Cohort and cross-sectional studies show that EMF may emanate from many different places and have different effects in the main epidemiological literature. Statistically significant relationships between exposure to high-frequency radiation (300 kHz to 300 MHz) and congenital abnormalities or other harmful pregnancy outcomes were not found in early research on physiotherapists from Denmark [29].

Similarly, a study that compared the mean birth weight of babies born to mothers exposed to common sources of ionizing and non-ionizing radiation, such as mobile phones, cordless phones, cathode-ray tube monitors, and dental and non-dental radiography, found no statistically significant differences [30]. Abad et al. (2016) in Tehran examined 462 pregnant women exposed to electromagnetic fields (EMFs) in the 27 MHz-3 GHz range; they found no evidence that EMFs induce miscarriage, citing a small sample size as a major limitation [31]. When it came to cohort studies, ELF-MF was the main emphasis. No apparent link with fetal growth was shown by Eskelinen et al. (2016), who assessed spot ELF-MF in 373 pregnant women and found tiny, imprecise mean changes in birth weight, -45 grams (95% CI: -547, 456) for all spot measures and -142 grams (95% CI: -505, 221) for bedroom readings [32].

Evidence from studies of everyday RF-EMF sources

Several more recent studies have examined everyday RF-EMF sources, including television, mobile phones, computers, Wi-Fi, and proximity to base stations. In a cohort of 400 pregnant women, exposure categories based on television viewing, mobile phone use, computer use, and Wi-Fi duration were associated with adverse anthropometric indicators at birth, including negative correlations between maternal mobile phone use and newborn birth weight, as well as between multiple mobile phone use and birth week, birth weight, and birth length [33]. A 24-hour personal dosimetry study of 128 pregnant women found that higher prenatal ELF-MF exposure in the 40-1000 Hz band was associated with decreased fetal growth among girls, with no significant association observed among boys, based on birth weight, body measurements, skinfold thickness, and circumferences [34]. Keterlambatan pertumbuhan saat lahir, yang didefinisikan sebagai skor AUDIPOG di bawah persentil ke-10, lebih umum terjadi pada kelompok yang lebih besar yang terdiri dari 1.368 ibu; hasil sekunder meliputi berat badan lahir, lingkar kepala, skor APGAR, dan malformasi; dan rata-rata penggunaan telepon harian selama kehamilan adalah 29,8 menit (kisaran 0,0-240,0) [6]. By contrast, a longitudinal cohort of 119 women seeking fertility treatment reported that personal power-frequency magnetic field exposures (with reported levels of 1.10 mG (2.14 mG) and 15.54 mG (58.73 mG)) were not associated with pregnancy outcomes [35]. Together, these cohort studies provide examples of both positive and null findings for fetal growth, anthropometry, and pregnancy loss under varying exposure conditions [6,29-35]

Narrative reviews and early syntheses

Beyond individual cohort studies, numerous narrative reviews, systematic reviews, and meta-analyses have examined non-ionizing EMF and reproductive outcomes, and these are summarized in Table 2 and Table 3. Early narrative and epidemiologic reviews highlighted concerns around spontaneous abortion and video display terminal (VDT) use, but generally did not find a strong or consistent association between maternal VDT use and spontaneous abortion Shawet et al. (1993) [24]. Later, Gye et al. (2012) described a positive relationship between occupational monitor work during pregnancy and natural abortion, but noted limitations in exposure characterization and potential confounding [25]. Aghaei et al. (2014) reported that pregnancy loss appeared to increase with higher maximum magnetic field exposure levels in daily life, while also combining multiple EMF sources and exposure metrics [26]. Mahmoudabadi et al. (2015) concluded that mobile phone use could be related to early spontaneous abortions, although study design and residual confounding limited causal interpretation [27]. Wdowiak et al. (2017) reported worse APGAR scores among newborns of women more heavily exposed to GSM-band emissions [28].

Broad overviews of EMF health effects

Broad overviews of health effects from non-ionizing radiation have also discussed reproductive endpoints. Belpomme et al. (2018) argued that EMFs-spanning RF-EMF (30 kHz-300 GHz) and ELF-EMF (3 Hz-3 kHz)-may increase the risk of brain cancer, adversely affect male and female reproduction, induce neurobehavioral changes, and contribute to electro-hypersensitivity, with concerns regarding in utero exposure, fetal programming, impaired learning and memory, and oxidative stress [1]. Pall et al. (2018) described Wi-Fi-including routers and connected devices as an important potential threat to human health and suggested that chromosomal mutations induced by EMF could have a role in spontaneous abortion [36]. Kumar et al. (2019) reviewed radiofrequency (100 kHz-300 GHz) and intermediate-frequency (300 Hz-100 kHz) exposures and concluded that health risks, especially during pregnancy, justified precautionary behavior such as avoiding carrying mobile phones close to the body [37]. At the same time, Saliev (2019) emphasized that, despite extensive experimental and mechanistic work, no firm evidence had been provided for mutagenic or teratogenic effects of EMF in humans [16].

Systematic reviews and meta-analyses (specific exposures and outcomes)

More focused systematic reviews and meta-analyses have evaluated specific exposure bands and outcomes. Tsarna et al. (2019) examined 55,507 pregnant women and found that using a cell phone during pregnancy might shorten the duration of the pregnancy and increase the risk of having a baby prematurely. However, they did note that these findings could be due to stress or other residual factors rather than being directly caused by mobile phone use [4]. Ebadi et al. (2020) examined research on analogue, digital, and third-generation mobile phone frequencies (450-900 MHz, 850-1900 MHz, and 2000 MHz, respectively) and found that high levels of electromagnetic fields (EMF) significantly increased the risk of miscarriage in pregnant women. The researchers noted that the risk varied with distance from the source as well as the strength and frequency of the waves [17].

Researchers Khojastehfard et al. (2020) looked at radio frequency (RF) exposures between 300 MHz and 300 GHz. They discovered that moms whose cell phone use was more than an hour per day had shorter pregnancies than moms whose use was less than an hour per day. The researchers also compiled evidence on complications like preterm birth, low birth weight, small for gestational age, and short birth length [18]. In their assessment of very low-frequency magnetic fields (1-300 Hz), Karimi et al. (2020) concluded that the strength of the magnetic field, the time of exposure, and the current intensity seemed to determine the connections with miscarriage [38]. Exposure to electromagnetic fields (EMF) over 50 Hz or 16 mG was linked to a 1.27-fold higher risk of abortion, according to a meta-analysis conducted by Ghazanfarpour et al. (2021) [19].

Additional recent syntheses and protocols

Complementary protocols and reviews have further scoped reproductive outcomes. According to the protocols outlined by Kenny et al. (2021), two meta-analyses and systematic reviews were conducted on the topic of RF-EMF exposure (100 kHz-300 GHz) and its effects on various birth outcomes, including preterm birth, small-for-gestational age, stillbirth, congenital anomalies, spontaneous miscarriage (25-25%), preterm birth (10-10%), stillbirth (2%), congenital anomalies (up to 5-5% of newborns), and low birth weight (14.6% [20]. In their analysis of electromagnetic fields (EMFs) emitted by smartphones (100 kHz-300 GHz), Li et al. (2017) found a correlation between prenatal EMF exposure and an increased risk of miscarriages, changes in fetal temperature and heart rate variability, and alterations in newborn anthropometric measurements [21]. Ben et al. (2022) summarized evidence on RF exposures in the 800-1600 MHz range and highlighted reported impacts on miscarriage, fetal temperature, fetal heart rate variability, and infant anthropometry, while emphasizing wide variability in exposure assessment and the types and durations of exposures across cohort studies [22]. Preterm delivery, stillbirth, neonatal congenital disabilities, or miscarriage were not linked to maternal exposure to 1-300 Hz fields, according to an ELF-EMF meta-analysis by Zhou et al. (2022). However, the researchers did note that there was still uncertainty regarding evidence for small-for-gestational age and low birth weight, and that huge studies in diverse regions were necessary [7].

A more recent systematic review and meta-analysis by Kashani et al. 2023) focusing on exposures above 50 Hz or 0.82 mG pointed to growing concern that daily radio-wave exposure could be linked with infertility, stillbirth, congenital anomalies, and abortions, although methodological heterogeneity limited firm conclusions [5]. In their review of the literature on radiofrequency electromagnetic fields (RF-EMF; 100 kHz to 300 GHz) and their effects on children's health, Lim et al. (2023) found that mothers whose mobile phone use was higher during their pregnancies had shorter pregnancies and higher rates of inattention and behavioral problems. However, the results regarding the effects of RF-EMF on foetal development, birth weight, gestational age, premature delivery, spontaneous abortion, craniosynostosis, and other outcomes were inconsistent [8]. Finally, Davis et al. (2023) discussed prenatal exposure to 2.45 GHz (a common Wi-Fi frequency) and reported associations with poorer psychomotor development and lower mental development indices in children [23].

Summary of patterns across studies

Across this literature, overall findings are mixed. Signals of association have been reported for miscarriage and abortion [5,7,17,19,26,27,38], shorter gestational duration, and preterm birth [4,8,18], and indicators of impaired fetal growth and adverse anthropometry in subsets of cohort and cross-sectional studies using higher magnetic-field categories or proxies for mobile-phone-related RF exposure [6,21-23,28,33,34]. At the same time, other comparably designed studies and reviews have reported null or inconsistent results for miscarriage, stillbirth, congenital anomalies, birth weight, and preterm delivery [7,16,24,29,30,32,35]. The divergence of findings across outcomes, exposure bands (ELF vs RF), and assessment methods suggests that study design and measurement choices meaningfully influence observed associations and that chance, bias, and residual confounding cannot be excluded as explanations for both positive and null results in many instances [1,4-8,16-26].

Methodological factors underlying discordant results

Three methodological themes likely underlie this discordance. First, exposure misclassification is a primary concern. Personal dosimetry (e.g., 24-hour measurements of ELF-MF in Ren et al. 2019 [34] and Eskelinen et al. 2016 [32], fixed-site or spot measurements in homes and workplaces, proxy indicators based on behaviors (e.g., phone-time logs in Boileau et al. (2020) [6], self-reported mobile phone or Wi-Fi use and proximity to base stations in Özge Kömürcü Karuserci et al. 2019 [33], Abad et al. 2016 [31], and studies summarized by Li et al. (2017) [21], Ben et al. (2022) [8,22], and job-exposure matrices for occupational settings (e.g., physiotherapists in Larsen et al. (1991) [29], monitor users in Gye et al. (2012) [25] capture non-identical constructs of EMF exposure-intensity, duration, and peak versus average levels, and are differentially affected by behavior and context [1,25,36,37]. These methods are sensitive to microenvironments (home, work, transport) and time-activity patterns that change throughout pregnancy, and the timing of exposure capture was frequently limited to a single time point. Relatively few studies repeatedly measured exposure across trimesters despite a plausible trimester-specific susceptibility [6-8, 18, 21-23, 32-34].

Second, heterogeneity in outcome definitions and covariate adjustment was substantial. Definitions of preterm birth and small-for-gestational age [18,30] classification of congenital anomalies[5,20,29], and the composition of adjustment sets (e.g., maternal age, smoking, parity, socioeconomic status, occupational factors, co-exposures) varied across studies, complicating cross-study comparisons and introducing opportunities for residual confounding [4-6,8,17-19,21,22,27,28,30,32,33]. For example, some studies focused exclusively on spontaneous abortion or miscarriage [17,19,24-27,31,38], others on birth weight, small-for-gestational age, and anthropometric measures [18,21-23,28,30,32,33], and still others on gestational duration, preterm birth, or stillbirth [4,5,7,8,18,20].

Third, residual bias and unmeasured confounding cannot be ruled out. Selection mechanisms and correlated behaviors, such as technology use patterns, occupational context, sleep, or stress, may confound associations; several authors have explicitly cautioned that observed associations with mobile phone use or Wi-Fi exposure could reflect such correlated behaviors rather than EMF itself [4,5,7,8,16,20-22]. Nonthermal biological pathways, including oxidative stress, inflammatory responses, and potential interactions with other environmental contaminants (e.g., lead or air pollutants), provide plausible mechanistic hypotheses, but translation to typical environmental exposure levels and clinically meaningful perinatal outcomes remains uncertain [1,5,16,22,23,25,26,36-38].

Mechanistic plausibility and interpretation

The mechanistic and experimental literature cited in the included reviews indicates that RF energy deposition (described by the specific absorption rate, SAR) and related maternal/placental temperature changes are configuration- and timing-dependent, with some models showing mid-gestation peaks, underscoring the importance of precise exposure characterization and trimester-specific assessment [9,16,22,23,25,26,36-38]. In parallel, animal and in vitro studies have suggested nonthermal effects on oxidative stress, endocrine regulation, neurodevelopment, and fetal programming, and these themes feature prominently in broader reviews of EMF health effects [1,12,14,15,21-23,29,32]. There was no connection found in the ELF-focused meta-analysis by Zhou et al. (2022) between maternal ELF-EMF exposure and preterm delivery, small-for-gestational age, stillbirth, newborn congenital impairments, or miscarriage. However, the authors did note that there was confusion surrounding low birth weight and small-for-gestational age [7]. Several contemporary reviews and meta-analyses have similarly emphasized null or inconsistent findings for selected outcomes and highlighted the variability of exposure assessment across the literature [4,5,7,8,16-22,24,26] (Table 2 and Table 3). In short, biological plausibility remains an active area of investigation, but existing observational data, taken as a whole, do not yield a consistent pattern of association.

Implications for future research

The implications for research follow directly from these patterns (Table 3). Future studies should adopt standardized exposure metrics (e.g., mG/µT for ELF and SAR or validated behavioral and environmental proxies for RF), incorporate repeated measurements across trimesters, and harmonize outcome definitions and prespecified confounder sets to improve comparability and reduce bias [6-8,16,18,19,29-33,38] Large prospective cohorts employing personal dosimetry and detailed time-activity logs, ideally paired with concurrent assessment of co-exposures (air pollution, heat, noise), would help address exposure misclassification and residual confounding. Preregistration and transparent reporting should continue, using PRISMA 2020 for reviews and validated risk-of-bias tools for observational designs [12,14,15,20] with the important reminder that STROBE is a reporting guideline rather than a quality appraisal instrument. Protocol papers, such as Kenny et al. (2021), illustrate how clearly predefined exposures, outcomes, and analysis plans can facilitate future pooling and meta-analysis [20].

Clinical interpretation and conclusion

Interpretation and practice should remain cautious. While signals exist, particularly for miscarriage, spontaneous abortion, and shorter gestation in some ELF and RF exposure scenarios, null and inconsistent findings elsewhere prevent firm causal conclusions [3-7,16-23,25-33]. Clinicians may offer precautionary guidance, especially regarding heavy or unnecessary exposure to mobile devices carried close to the abdomen or prolonged Wi-Fi-intensive use during pregnancy, without alarmism, while emphasizing that the current evidence does not justify strong causal claims or major changes in clinical practice [5,7,8,21-23]. Overall, most studies in this review were judged to have low selection bias, while concerns remained regarding confounding (D2) and outcome assessment (D3). Several studies with small samples and limited precision in exposure measurement, such as Larsen et al. (1991) [29] and some of the miscarriage-focused reviews [17,19,26,27,38], demonstrated a high risk of bias in outcome assessment and limited ability to detect modest effects. Only a few studies achieved a low risk across all domains (Figure 2).

The strengths of this review include a registered protocol (PROSPERO CRD42023475665), multi-database searches using reproducible search strings, independent duplicate screening, and structured narrative synthesis. Limitations include the English-only restriction and between-study heterogeneity, which precluded statistical pooling of the data [6,7]. Taken together, the literature does not show a consistent pattern linking prenatal EMF exposure with adverse perinatal outcomes; reducing uncertainty will require coordinated, standardized measurements and reporting so that future syntheses-potentially including meta-analyses, can be conducted with greater validity.

Conclusion     

Current observational evidence is insufficient for causal inference regarding prenatal EMF exposure and adverse perinatal outcomes. The results across studies are mixed and limited by heterogeneous exposure metrics, timing, and confounding factors. Future research should prioritize standardized trimester-specific exposure assessments, harmonized outcomes, and improved control of confounders to enable more definitive synthesis.

Competing interests     

The authors declare no competing interests.

Authors' contributions     

All authors contributed significantly to the development of this study. Desy Armalina and Sunarno jointly developed the study concept. Armalina and Sutanto formulated the research design. Neni Susilaningsih and Sunarno performed data collection and processing. Heri Sutanto and Sunarno performed data analysis and interpretation. Neni Susilaningsih and Desy Armalina conducted a comprehensive literature review. Desy Armalina drafted the manuscript, while Neni Susilaningsih, Heri Sutanto, and Sunarno contributed to critical revisions and final editing. Every single author has taken full responsibility for their work and has reviewed and approved the final document.

Acknowledgments     

The authors express their sincere gratitude to the Faculty of Medicine, Universitas Diponegoro, for the research facilities and academic guidance, and to the Ministry of Education, Culture, Research, and Technology (Kemdikbudristek) of Indonesia for their continued support in advancing scientific publication. Special thanks to the Systematic Review Methodology Unit, Universitas Diponegoro, for the consultation and protocol validation assistance.

Tables and figures     

Table 1: cohort studies on non-ionizing electromagnetic exposure during pregnancy

Table 2: systematic review studies on non-ionizing electromagnetic exposure in pregnancy (studies 1-11)

Table 3: systematic review studies on non-ionizing electromagnetic exposure in pregnancy (studies 12-21)

Figure 1: PRISMA flowchart of selected study

Figure 2: the risk of bias for each study was assessed using the Newcastle-Ottawa Scale (NOS); visual representations of the risk of bias were generated utilizing the risk of bias visualization (robvis) tool

References