The effect of mobile phone electromagnetic fields on the human resting state wake EEG and event‐related potential: A systematic review and meta‐analysis

Abstract

The rapid growth of mobile phone usage and its use of radiofrequency electromagnetic fields (RF‐EMF) have raised concerns about potential health risks. Researchers have conducted studies to examine the effects of RF‐EMF on the brain using electroencephalography (EEG). We conducted a systematic quality assessment and meta‐analysis of published research in this field to establish high‐quality studies as references for future protocols. The electronic search yielded 244 records from which a total of 51 studies were included in the review after excluding studies based on study design, and data or report availability. Of these 51 studies, 31 (61%) focused on resting state wake EEG and 20 (39%) on event‐related potentials (ERP). None of the 51 studies were free from risk of bias. From the 51 included studies, we were able to use seven studies to create three different groups for meta‐analysis for resting state wake EEG and five studies to create 10 different groups for meta‐analysis for ERP. Per group the number of studies varies from 1 to 5. Our procedure is the first systematic quality assessment in this field and revealed three important findings. First, there is evidence of an effect on the EEG of a 2G protocol using an eyes‐open condition. Second, we did not find evidence for EEG effects during task performance. This suggests that the impact of EMF during task performance is less pronounced compared to the resting state condition. Third, this meta‐analysis shows that the field is unable to create an evidence base for most comparisons due to heterogeneity. We therefore advise that all future studies are double‐blind in nature, adhere to the methodological standard of randomized experiments, and publish their protocols first.

Highlights

• Evidence exists for the presence of increased alpha band activity during 2G exposure using an eyes‐open condition.

• The impact of mobile phone electromagnetic field (EMF) measured by electroencephalography during task performance is less pronounced compared to the wake resting state condition.

• Currently the field is unable to create a conclusive evidence base for the impact of mobile phone EMF due to the diverse range of methodologies employed in research.

Abbreviations

CF

carrier frequency

EC

eyes closed

EEG

electroencephalogram

EMF

electromagnetic field

EO

eyes open

ERP

event‐related potential

MP

mobile phone

PM

pulse modulation

rEEG

resting state wake EEG

RF‐EMF

radiofrequency electromagnetic fields

RoB

risk of bias

SAR

specific absorption rate

UMTS

Universal Mobile Telecommunications System

1. INTRODUCTION

Mobile phone usage, and consequently the use of radiofrequency electromagnetic fields (RF‐EMF), has become indispensable over the past decades. In 2022, 5.4 billion people globally subscribed to a mobile service (“The Mobile Economy,” ²⁰²³) and in 2023, 78% of the world's population owned a mobile phone (“ITU,” ²⁰²³). The technology of wireless communication continuously evolves, and its growth is shaped by various generations of telecommunication protocols. Each of these generations was developed in consecutive decades and is characterized by its own frequency bands, standardization, and type of modulation.

The current deployment of the fifth generation (5G) of cellular technology has revived existing public concerns about the risks of cellular radiation on human health. Up until now, international organizations such as the World Health Organization (WHO) and the International Commission on Nonionizing Radiation Protection (ICNIRP) (WHO, ²⁰¹⁰; Ziegelberger et al., ²⁰²⁰) have been closely monitoring possible adverse effects of electromagnetic radiation.

In the past, threshold levels for power density for both the general population and workers have been identified and are still used as references by ICNIRP in determining the safety of novel communication technology (Ziegelberger et al., 2020). It is generally accepted that RF‐EMF at high levels surpassing the ICNIRP reference levels can induce thermal effects in the brain leading to adverse health effects (Barnes and Greenebaum, 2020). Still, a large body of literature exists claiming that also below the ICNIRP limits for power density physiological effects of RF‐EMF exist. Especially the presence of low frequency pulsation in communication signals has been proven to be associated with nonthermal biological effects (Belyaev et al., 2022). While consensus exists that there is no evidence that RF‐EMF is harmful below the ICNIRP limits, the debate on how to interpret its biological effects is still ongoing.

Apart from epidemiological studies and studies using animals or cell cultures, researchers have been attempting for decades to study the effects of EMF in vivo. These studies often focus on the brain because mobile phones are mostly used near the head. A noninvasive technique to measure brain activity is resting state electroencephalography (EEG), which records the electrical signals that the brain generates with electrodes that are placed at different positions on the scalp, while an individual is awake, relaxed, and not actively engaged in any specific task and/or exposed to a sensory stimulation. The resulting resting state EEG‐signal can be categorized in separate frequency ranges (Jobert et al., 2012): gamma (30–40 Hz), beta (12.5–30 Hz), alpha (8.5–12.5 Hz), theta (6–8.5 Hz), and delta (1.5–6 Hz). The power spectrum of a time series of EEG samples in a certain frequency range, also called “spectral band power,” is frequently employed to characterize the electrical oscillations in the brain. It is believed that changes in spectral band power are closely related with processes of cognition (Cacioppo et al., 2016; Ward, ²⁰⁰³).

In addition to resting state EEG, EEG studies regularly incorporate an experimental design in which subjects are requested to perform certain tasks, allowing investigation of the event‐related potential (ERP), a waveform that occurs in response to a specific event. This waveform represents the brain's cognitive response to a stimulus and varies with respect to the nature of the stimulus, for example, responses evoked by a visual stimulus are different from those evoked by an acoustic stimulus. The ERP consists of multiple components that are named after the direction of the deflection (positive [P] or negative [N]), and the time delay after the stimulus. Each component is believed to be related to a different part of a cognitive and/or sensory process (Luck, 2005). Next to the classic ERP components, event‐related resynchronization, and synchronization can be studied, representing a short and local reduction or enhancement of oscillations in a specific frequency band during the performance of the task (Pfurtscheller, 1991).

Multiple reviews discussing the effects of RF‐EMF as measured by EEG exist, either focusing on resting EEG solely (Danker‐Hopfe et al., 2019; Gjoneska et al., ²⁰¹⁵; Sofri et al., ²⁰²²; Wallace and Selmaoui, ²⁰¹⁹) or on both resting EEG and ERPs (Hamblin and Wood, 2002; Hinrikus et al., ²⁰²¹; Kwon and Hämäläinen, ²⁰¹¹; Marino and Carrubba, ²⁰⁰⁹; Valentini et al., ²⁰⁰⁷; Van Rongen et al., ²⁰⁰⁹; Zhang et al., ²⁰¹⁷). These reviews nearly all conclude that findings are inconsistent and that differences in methodology are a probable reason for this heterogeneity, underlining a strong need for standardization in this field. A systematic quality assessment and meta‐analysis of existing published research can contribute to this standardization by letting high‐quality studies form a reference for future experimental protocols. The purpose of this study is, therefore, to systematically assess the existing evidence related to the impact of mobile phone‐like radiation on the resting state wake EEG and ERP using a meta‐analytic approach.

2. METHODS

2.1. Literature search strategy and selection

The literature search and selection methodology were based on the PRISMA Statement for reporting systematic reviews (Page et al., 2021). One reviewer (A. P.) carried out the literature search in three electronic databases (PubMed, Embase, and EMF‐Portal). The used keywords were variations of the terms “electroencephalogram,” “event‐related potential,” “electromagnetic fields,” “cell phone radiation,” and “human” (see Appendix SA for the full search queries). The literature search was performed on 8 February 2023 and updated on 1 October 2023 and 15 May 2024. Two reviewers (A. P. and K. B.) then screened the title and abstract of the resulting studies and included studies if (1) participants were healthy, adult and human, (2) radiation in the frequency range related to mobile phones was used, (3) examination focused on the resting state wake EEG and/or the ERP, (4) studies were published in the English language, and (5) studies were published in peer‐reviewed journals. Additional studies were included by screening the references of relevant articles and personal knowledge. Next, full‐text articles were retrieved and reviewed by the same two reviewers for eligibility using the additional following inclusion criteria: (1) the design used was a randomized, controlled experiment with at least a single‐blind condition, and (2) sufficient data about the exposure was provided, which means that at least the power of the radiation must be reported. We also included articles if only part of the research fulfilled the criteria, for example, both healthy and nonhealthy participants were included in the study. In these cases, we only included the part of the study that fulfilled our criteria. Contradictions between the results of the reviewers were resolved after internal discussion or consulting a third reviewer (M. J.).

2.2. Assessment of risk of bias (RoB)

The Cochrane revised tool for RoB in randomized studies (Sterne et al., 2019), specifically their tool for developed for crossover trials, was used to assess RoB for the included trials regarding six domains: the randomization process, period and carryover effects, deviations from the intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. A detailed description of the RoB assessment can be found in Appendix SC. Studies that shared the same set of participants were regarded as one trial. For the RoB domain of period and carryover effects, a period of 24 h was deemed sufficient for any carryover effects to disappear before assessing outcomes in the second period. The latter was based on sleep studies in which EMF effects could be observed during sleep after prior EMF exposure (Danker‐Hopfe et al., 2019). However, since the mechanism behind carryover is still poorly understood and the effect can be accounted for in the analysis of the data such that interpreting carryover as an effect can be avoided, we decided to adjust the RoB from carryover effects to “Some Concerns” instead of “High Risk” if the time between measurements was less than 24 h. The RoB was assessed independently by two reviewers (A. P. and K. B.). Disagreements were resolved through discussion and when necessary, a third reviewer (M. J.) was consulted.

2.3. Data extraction

For every study included in the review, two reviewers (A. P. and A. J. N.) extracted its key characteristics, which included: first author and year, sample size, number of electrodes, duration of EEG recording, eyes open (EO) or eyes closed (EC) (resting state wake EEG studies), type of task, duration of EMF exposure, moment of EMF exposure, carrier frequency, pulsation, protocol type, investigation of interference between EMF exposure and EEG recording, exposure setup, specific absorption rate (SAR), power and principal results of each study. For the resting state wake EEG studies we mentioned all results reported per frequency band (either positive or negative); frequency bands not reported by the authors are not mentioned. Any disagreements were resolved by discussion. In Appendix SD, we provide an overview of the definitions of the frequency bands in the resting state wake EEG studies if mentioned.

2.4. Meta‐analysis

From the selected studies, a subset of studies could be grouped together for meta‐analysis. To maintain comparability for the meta‐analysis, studies were grouped according to their experimental design based on the key characteristics. To create a sensible meta‐analysis, studies must be sufficiently homogeneous across population, intervention, comparison, and outcome(s). Means and standard deviations could only be extracted if data was reported numerically or if values could be read visually from figures. We considered resting wake state EEG studies applicable for meta‐analysis if: (1) they reported mean spectral power for two conditions in one specific band, (2) administered RF‐EMF during measuring EEG, (3) had identical carrier frequency and pulse modulation (PM), and (4) instructed their participants to have their eyes only closed or only open during measurements. ERP studies were considered sufficiently homogenous for meta‐analysis if: (1) they reported mean amplitude and/or latency of one ERP component for two conditions, (2) administered RF‐EMF during measuring EEG, (3) had identical carrier frequency and PM, and (4) participants performed the same task.

A meta‐analysis may require substantial reordering of study results. Hence, when results in a study were divided into different subgroups, their outcomes were combined by calculating the average outcome. Subgroups of outcomes consisted of electrode locations (e.g., frontal, posterior), time (e.g., first and last 5 mins of exposure), exposure position (e.g., ipsilateral and contralateral), exposure strength (e.g., high and low SAR), and task difficulty. If the power in a specific band or ERP component was broken down into different subbands or subcomponents, a combined average value was calculated as well. When necessary, outcomes were converted such that units were µV² for power, milliseconds (ms) for latency ERP components and µV for amplitude ERP components.

It is unlikely that all studies share a common effect size considering the difference in set‐up and measurements. Hence, we applied a random‐effects mixed model, using inverse variance, to estimate summary treatment effects (Borenstein et al., 2009), and chose to report the results using the standardized mean difference (SMD) and 95% confidence intervals. The SMD was calculated for each outcome in a group using the method of Hedges and Olkin which weights every effect size by the studies' sample size (Hedges and Olkin, 1985). When necessary, standard errors were converted to standard deviations to calculate the SMD.

We used three different tests to assess the level of homogeneity: (1) with a τ ² statistic to estimate the amount of “true” heterogeneity across studies, (2) with a χ ² procedure to indicate heterogeneity in the dispersion of effect sizes, and (3) with an I ² test to portray the percentage of inconsistency attributable to heterogeneity and not chance. The I ² values ranged between 0% and 100%, with percentages of 25%, 50%, and 75% representing low, moderate, and substantial heterogeneity, respectively (Higgins et al., 2003). All analyses were conducted using the Review Manager Web software (The Cochrane Collaboration ²⁰²³).

3. RESULTS

3.1. Study selection

The electronic search yielded 244 records, and 13 articles were added through citation searching and personal knowledge of the authors (Figure 1). After subtracting 74 duplicate records, we screened the 170 remaining records from which 68 were included in the full‐text examination, see Appendix SB for an overview of the reasons for exclusion. A total of 51 studies were included in the review after excluding studies based on study design, and data or report availability. Of these 51 studies, 31 (61%) focused on resting state wake EEG and 20 (39%) on ERP.

Figure 1.

PRISMA flow diagram of the search and selection methodology for both the resting state wake electroencephalogram (rEEG) and event‐related potential (ERP) studies (Page et al., 2021).

3.2. Key study characteristics

Table 1 summarizes the key characteristics of the included studies on the effects of mobile phone‐like radiation on the resting EEG. Table 2 summarizes the key characteristics of the included studies on the effects of mobile phone‐like radiation on the ERP. Both tables display significant heterogeneity among studies in terms of the sample size, number of electrodes utilized, and the specific task being investigated.

Table 1.

Key characteristics of the included studies on the effects of mobile phone‐like radiation on the resting state wake EEG.

References	Sample size	No. of electrodesa	Ref. electrode	EEG‐duration	?G	Task	Duration of RF‐exposure	Moment of EMF‐exposure	CF (MHz)	PM	Interference	Exposure setup	SAR (W/kg)	Exposure power	Main results
Bachmann et al. (2007)	15	19	Cz	NR	n/a	EC	40 min	During EEG	450	40 and 70 Hz	NR	13 cm quarter‐wave antenna located 10 cm from skin on the left side of head	0.35	0.16 mW/cm2	40 Hz: significant changes in the EEG time variability and energy variation, 40 and 70 Hz: significant changes in energy spectral distribution. Increase of beta power, no effect on theta and alpha band
Croft et al. (2002)	24	19	Linked mastoids	60 min	2 G	EO	20 min EMF + 20 min of attenuated EMF	During EEG	900	217 Hz	NR	MP 5 cm radially from the subject's scalp, midway between Oz and Pz	NR	3–4 mW	Increased alpha and decreased right hemisphere delta power, no effect on theta, beta, and gamma bands
Croft et al. (2008)	120	58	Left mastoid	2 × 10 min	2 G	EO	30 min	During EEG	895	217 Hz	NR	MP placed in a cradle over recording cap, either over right or left temporal region, in touch position	On 10 g: 0.674	250 mW	Increased alpha power
Croft et al. (2010)	62	61	Between Cz and CPz	3 × 60 min	2G and 3G	EO	2G: 55 min, 3G: 55 min	During EEG	2G: 894.6, 3G: 1900	2G: 217 Hz, 3G: none	NR	Two MPs placed in a cradle over the subjects' EEG recording cap	On 10 g: 0.7 (2G), 1.7 (3G)	2G: 250 mW, 3G: 125 mW	2G: Increased alpha power in young adults, no effect on the elderly, 3G: no effect on alpha power for both age groups
Curcio et al. (2005)	20	7	A1 and A2	3 × 7 min	2G	EC	45 min	Before and during EEG	902.40	217 Hz	NR	MP placed 1.5 cm from the left ear using a helmet	0.5	250 mW	Increased alpha power
D'Costa et al. (2003)	10	6	FP2, C4, and O2	2 × 50 min	2G	EC	25 min full power and 25 min standby	During EEG	900	217 Hz	NR	MP antenna positioned 2 cm away from the occipital region	NR	250 mW	Decreased alpha and beta power, no effect on delta and theta bands
Dalecki et al. (2021)	36	19	Left mastoid	3 × 60 min	2G	EO and EC	2 × 30 min (Low and High SAR)	During EEG	920	217 Hz	NR	Antenna positioned 42 mm above the left auditory canal and 115 mm away from the head	On 10 g: Low: 1, High: 2	NR	Increased alpha power, larger in EO than EC
Ghosn et al. (2015)	26	29	AFz	2 × 61 min	2G	EO and EC	26 min	During EEG	900	217 Hz	Phantom measurement, no interference	MP positioned against the left ear	On 10 g: 0.93	250 mW	Decreased alpha power in eyes closed
Hietanen et al. (2000)	19	21	A1 and A2	6 × 30 min	2G	Vigilance‐controlled EC	5 × 20 min (five different MPs)	During EEG	900 and 1800	NR	No interference detected	MP located alongside a test person's head at a distance of 1.0 cm from the head	On 10 g: 0.851	Peak: 1–2 W	Decreased absolute delta power, but likely due to chance. No effect on theta, alpha, and beta bands
Hinrichs and Heinze (2006)	18	21	Linked mastoids	30 min	2G	EC	5 × 4 min with different field conditions	During EEG	1800	None, 217, 108, and 27 Hz	NR	MP attached to the chair frame and individually positioned with its antenna close to the ear (~10 mm)	On 10 g: 1.22 and 0.61	Peak: 0.25, 1, 2 and 8 W	No effect on delta, theta, alpha, and beta bands
Hinrikus et al. (2008)	13	9	Cz	40 min	n/a	EC	40 min	During EEG	450	7, 14, and 21 Hz	Phantom measurement, artefacts filtered	13 cm quarter‐wave antenna located 10 cm from skin on the left side of head	0.35	0.16 mW/cm2	Increases in alpha and beta power, no effect on theta band
Hinrikus et al. (2009)	15 and 19	9	Cz	40 min	n/a	EC	40 min	During EEG	450	40,70, 217, and 1000 Hz	NR	13 cm quarter‐wave antenna located 10 cm from skin on the left side of head	0.35	0.16 mW/cm2	Increase in EEG energy at frequencies lower or close to the modulation frequency in alpha and beta bands, no effect on theta band
Hinrikus et al. (2011)	28	9	Cz	40 min	n/a	EC	40 min	During EEG	450	7, 14, 21, 40, and 70 Hz	Phantom measurement, artefacts filtered	13 cm quarter‐wave antenna located 10 cm from skin on the left side of head	Peak SAR averaged over 1 g 0.303 W/kg	0.16 mW/cm2	Increase in the average EEG at frequencies with fixed ratios to the modulation frequency
Hinrikus et al. (2017)	15	9	Cz	40 min	n/a	EC	40 min	During EEG	450	7, 40, and 1000 Hz	Phantom measurement, artefacts filtered	13 cm quarter‐wave antenna located 10 cm from skin on the left side of head	Peak SAR averaged over 1 g 0.303 W/kg	0.16 mW/cm2	Increase in EEG power mainly for 40 Hz pulsation in alpha and beta bands, no effect on theta band
Huber et al. (2002)	16	4	A1 and A2	3 × 9 min	2G	NR	2 × 30 min (PM and not PM)	Before EEG	900	None and 2, 8, 217, and 1736 Hz	NR	Subjects' heads positioned between two planar antennas	On 10 g: 1	NR	Increased alpha power in PM‐EMF, no effect with continuous wave EMF
Jamal et al. (2023)	34	64	FCz	25 min	5G	EO and EC	1 h	During EEG	3500	217 Hz	NR	Horn antenna placed 1.2 m away and 45 degrees to the right of each participant	0.037 mW/kg	0.68 W/m2	No effects on the beta, alpha, theta, and delta bands
Kleinlogel et al. (2008a)	15	19	Cz	4 × 37.5 min	2G and 3G	Vigilance‐controlled EC	2G: 30 min, 3G: 2 × 30 min (Low and High SAR)	During EEG	2G: 900, 3G: 1950	2G: 2, 8, and 217 Hz, 3G: none	No interference detected	Antenna placed directly over the left ear	On 10 g: 0.1 (UMTS low) and 1 (GSM and UMTS high)	NR	No effect with 2G and 3G on delta, theta, alpha, and beta bands
Loughran et al. (2019)	36	19	Left mastoid	3 × 60 min	2G	EO	2 × 30 min (Low and High SAR)	During EEG	920	217 Hz	NR	Antenna positioned 42 mm above the left auditory canal and 115 mm away from the head	On 10 g: Low: 1, High: 2	NR	Increased alpha power
Lv et al. (2014)	10	32	NR	2 × 50 min	4 G	EC	30 min	During EEG	2573	none	NR	Dipole antenna was placed on the right ear with 1 cm distance	On 10 g: 1.07	NR	Modulation of synchronization likelihood patterns in theta, alpha, beta, and broadband
van der Meer et al. (2023)	32	63	FCz	2 × 15 min	2G	EO and EC	15 min	During EEG	900	217 Hz	Phantom measurement, no interference	Antenna array consisting of four antennas placed in a rectangular configuration above the brain	NR	76 mW/m2	Increased alpha power in the EO condition, no effect on alpha band with EC
Nakatani‐Enomoto et al. (2020)	38	128	Cz	2 × 60 min	4G	EC	30 min	During EEG	1950	none	NR	Antenna placed in front of the right ear, under the electrode net	On 10 g: 2	NR	No effect on alpha, beta, and theta bands
Perentos et al. (2007)	12	16	Linked mastoids	120 min	2G	EC	2 × 15 min (PM and not PM)	During EEG	900	None and 2, 8, 217, and 1736 Hz	NR	The handset device placed according to the standard “touch” ear‐to‐mouth position (left hemisphere)	On 10 g: 1.56	Mean: 276 mW, Peak: 2 W	No effect on delta, theta, alpha, and beta bands
Perentos et al. (2013)	72	19	Left mastoid	120 min	2G	EO	3 × 20 min (PM, not PM, ELF)	During EEG	900	None, RF 217 Hz, ELF 217 Hz	Phantom measurement, extra shielding, no interference	Antenna placed over the right hemisphere in a constructed cradle	On 10 g: 1.95 (not PM and PM), 0.06 (ELF)	NR	Decreased alpha power in continuous wave and PM‐EMF
Regel et al. (2007)	24	2	A2	9 × 6 min	2G	EO and EC	2 × 30 min (PM and not PM)	Before EEG	900	None and 2, 8, 217, and 1733 Hz	NR	Antenna positioned 42 mm above the left auditory canal and 115 mm away from the head	On 10 g: 1	NR	Increased alpha power in EC 30 min after exposure
Reiser et al. (1995)	36	16	Cz	2 × 60 min	2G	NR	15 min	During EEG	902.4	217 Hz	NR	MP placed 40 cm behind the head with the antenna being centered	NR	Peak: 8 W	Increased alpha and beta power, no effect on delta and theta bands
Röschke and Mann (1997)	34	4	A1 and A2	2 × 10 min	2G	EC	3.5 min	During EEG	900	217 Hz	NR	MP positioned 40 cm from the vertex of the subject	NR	Mean: 0.05 mW/cm2 Peak: 8 W	No effect on delta, theta, alpha, and beta bands
Suhhova et al. (2013)	15	9	Cz	40 min	n/a	EC	40 min	During EEG	450	40 Hz	Phantom measurement, artefacts filtered	13 cm quarter‐wave antenna located 10 cm from skin on the left side of head	0.303 and 0.003 W/kg	0.16 mW/cm2	Increase in alpha and beta bands for high SAR condition, no effect on theta band
Vecchio et al. (2010)	15	19	Linked ear	2 × 10 min	2G	EC	45 min	Between measuring EEG	902.4	8.33 and 217 Hz	NR	MP positioned on the left side of the head by a modified helmet, 1.5 cm between MP and the ear	0.5	Mean: 250 mW, Peak 2 W	No effect on spectral coherence values of delta, theta, and alpha bands
Wallace et al. (2022)	21	74	Linked mastoids	2 × 75 min	2G	EO and EC	25 min 30 s	During EEG	900	217 Hz	Phantom measurement, no interference	MP placed against the left ear using a tubular tissue bandage	On 10 g: 0.49	Mean: 250 mW, Peak 2 W	No effect on alpha band
Wallace et al. (2023)	21	74	Linked mastoids	2 × 75 min	2G	EO and EC	25 min 30 s	During EEG	900	217 Hz	Phantom measurement, no interference	MP placed against the left ear using a tubular tissue bandage	On 10 g: 0.49	NR	Decrease in theta band activity with EO, increase in theta band with EC. No effect on delta and beta bands
Yang et al. (2017)	25	19	Left earlobe	2 × 50 min	4G	EC	30 min	During EEG	2160	None	Phantom measurement, no interference	Dipole antenna placed 1 cm from the right ear	On 10 g: 1.34	NR	Decreased alpha and beta power, no effect on delta and theta bands

Abbreviations: CF, carrier frequency; EC, eyes closed; EEG, electroencephalography; ELF, extremely low frequency; EMF, electromagnetic field; EO, eyes open; EXP, experiment; GSM, Global System for Mobile Communications; MP, mobile phone; n/a, not applicable; NR, not reported; PM, pulse modulation; RF, radiofrequency; SAR, specific absorption rate; UMTS, Universal Mobile Telecommunications System.

^a This number includes the reference electrode(s).

Table 2.

Key characteristics of the included studies on the effects of mobile phone‐like radiation on the ERP.

References	Sample size	No. of electrodesa	Ref. electrode		EEG‐duration	?G	Task	Duration of RF‐exposure	Moment of EMF‐exposure	CF (MHz)	PM	Interference	Exposure setup	SAR (W/kg)	Exposure power	Main results
Dalecki et al. (2018)	36	19	Left mastoid		3 × 60 min	2G	Visual discrimination task	2 × 30 min (low SAR and high SAR)	During EEG	920	217 Hz	NR	Antenna positioned 42 mm above the left auditory canal and 115 mm away from the head	On 10 g: Low: 1, High: 2	NR	Reduced P100 amplitude and increased N100 latency, but likely due to chance
de Tommaso et al. (2009)	10	30	Linked mastoids		3 × 8 min	2G	Auditory discrimination task	8 min	During EEG	900	217 Hz	No interference detected	MP mounted on the left side of the head with a plastic helmet	0.5	NR	Decrease of early CNV amplitude
Eggert et al. (2015)	30	19	NR		31 min	n/a	Visual monitoring, self‐paced finger tapping, externally‐paced finger tapping (with warning and imperative stimulus)	2.5 h	During EEG	365 (TETRA)	17.65 Hz	Interference prevented by shielding	Exposure applied by a printed circuit board antenna with foam spacers and a textile cover	(1) on 10 g: 1.5 W/kg (2) on 10 g: 6 W/kg;	NR	No effects
Eulitz et al. (1998)	13	30	NR		NR	2G	Auditory discrimination task	NR	During EEG	916.2	217 Hz	NR	MP mounted to the subject's head over the left posterior temporal region	NR	Peak: 2.8 W	Reduced beta spectral power with task‐relevant stimuli
Freude et al. (1998)	16	30	Cz		2 × 8 min	2G	Visual monitoring preceded by index finger movement	8 min	During EEG	916.2	217 Hz	NR	MP with extended antenna was positioned in direct contact to the left ear	1.42	350 mW	SFMT: No effect, VMT: Decrease of slow brain potential (−250 ms)
Freude et al. (2000)	EXP 1: 20; EXP 2: 19	30	Cz		EXP 1: 2 × 3 min; EXP 2: 2 × 10 min	2 G	Visual Monitoring, self‐paced finger tapping, externally‐paced finger tapping (on imperative stimulus)	EXP 1: 3 min, EXP 2: 10 min	During EEG	916.2	217 Hz	NR	MP with extended antenna was positioned in direct contact to the left ear	1.42	350 mW	SFMT: No effect, VMT: Decrease of slow brain potential (−250 ms)
Hamblin et al. (2004)	12	62	Linked mastoids		30 min	2G	Auditory discrimination task	60 min	During EEG	894.6	217 Hz	No interference present in physiological range	MP cradle with a standard GSM digital phone mounted over the right temporal region	0.87	250 mW	Decrease of N100 amplitude and latency, delay of P300 latency
Hamblin et al. (2006)	120	62	Left mastoid		2 × 62 min	2G	Auditory and visual discrimination task	31 min	During EEG	895	217 Hz	No interference present in physiological range	MP attached over the temporal region using a plastic cradle‐like apparatus	On 10 g: 0.11	250 mW	No effect for both tasks
Kleinlogel et al. (2008b)	15	19	Cz		4 × 37.5 min	2G & 3G	Oddball continuous performance and auditory discrimination task	2G: 30 min, 3G: 2 × 30 min (Low and High SAR)	During EEG	2G: 900, 3G: 1950	2G: 2, 8, and 217 Hz, 3G: none	No interference detected	Antenna placed directly over the left ear	On 10 g: 0.1 (UMTS low) and 1 (GSM and UMTS high)	NR	No effect with 2G and 3G
Krause et al. (2000a)	16	20	Linked mastoids		60 min	2G	Auditory memory task	30 min	During EEG	902	217 Hz	NR	MP mounted to the subject's head positioned over the right posterior temporal region, ~20 mm from the skin	NR	Mean: 0.25 W	Altered ERD/ERS responses in all frequency bands (alpha and theta)
Krause et al. (2000b)	24	20	Linked mastoids		60 min	2G	Visual working memory task	30 min	During EEG	902	217 Hz	NR	MP mounted to the subject's head positioned over the right posterior temporal region, ~20 mm from the skin	NR	Peak: 2 W, mean: 0.25 W	Altered ERD/ERS responses in 6–8 and 8–10 Hz frequency bands
Krause et al. (2004)	24	20	Linked mastoids		60 min	2G	Auditory memory task	30 min	During EEG	902	217 Hz	NR	MP positioned in direct contact to the left cheek, 40 mm from the head	On 10 g: 0.648	Mean: 0.25 W	Decreased ERS response in 4–6 Hz frequency band
Krause et al. (2007)	36	NR	Right mastoid		EXP 1: 3 × 30 min, EXP 2: 3 × 40 min	2 G	EXP 1: Auditory memory task, EXP 2: 0, 1, 2, 3‐back (visual task)	EXP 1: 2 × 30 min EXP 2: 2 × 40 min (PM and no PM)	During EEG	902	None and 217 Hz	NR	Antenna ~20 mm from the posterior temporal lobe (left side), or over the inferior and posterior temporal lobe (right side)	On 10 g: 0.738	0.25 W	No effect for both tasks
Kruusing and Lass (2008)	30	3	NR		20 min	2G	Visual discrimination task	10 min	During EEG	450	21 Hz	NR	Antenna located 10 cm from the skin on the left side of head	On 1 g: 0.303	Mean: 160 W/cm2	Reduced N100 amplitude
Leung et al. (2011)	62	61	Between Cz and CPz		3 × 60 min	2G and 3G	EXP 1: Auditory discrimination task, EXP 2: visual task: 1, 2, and 3‐back	2G: 55 min, 3G: 55 min	During EEG	2G: 894.6, 3G: 1900	2G: 217 Hz, 3G: none	NR	Two MPs placed in a cradle over the subjects' EEG recording cap	On 10 g: 0.7 (2G), 1.7 (3G)	2G: 250 mW, 3G: 125 mW	Independent of age: EXP 1: increase of N100 with 2G, no effect with 3G; EXP 2: increased latency in ERD/ERS response in 2G and 3G
Maganioti et al. (2010)	EXP 1: 19, EXP 2: 20	15	Average		2 × 45 min for both EXPs	2G	Auditory memory task	45 min for both EXPs	During EEG	EXP 1: 900, EXP 2: 1800	None	NR	Dipole antenna fixed about 20 cm from the right ear	NR	64 mW (EXP 1), 128 mW (EXP 2)	Effect on P600 amplitude and latencies for females
Stefanics et al. (2008)	29	3	Linked mastoids		2 × 20 min	3G	Auditory discrimination task	20 min	Between measuring EEG	1947	None	NR	External patch antenna mounted on a plastic headset in a position similar to that of a normal MP handset	On 1 g: 0.39	NR	No effect
Trunk et al. (2013)	26	3	Nose		2 × 20 min	3G	Auditory discrimination task	30 min	Between measuring EEG	1947	None	NR	External patch antenna mounted on a plastic headset in a position similar to that of a normal MP handset	On 1 g: 1.75	NR	No effect
Trunk et al. (2014)	25	32	Nose		4 × 20 min	3G	Visual discrimination task	2 × 15 min (caffeine and no caffeine)	During EEG	1947	None	NR	External patch antenna mounted on a plastic headset in a position similar to that of a normal MP handset	On 10 g: 0.738	NR	No effect
Vecchio et al. (2012)	11	56	Between Afz and Fz		2 × 20 min	2G	Visual go/no go task	2 × 30 min (Low and High SAR)	During EEG	920	8.33 and 217 Hz	NR	MP positioned on the left side of the head by a modified helmet, 1.5 cm between MP and the ear	0.5	Mean: 250 mW, Peak 2 W	Lower alpha ERD

Abbreviations: CF, carrier frequency; CNV, contingent negative variation task; CPT, continuous performance test; EEG, electroencephalography; EMF, electromagnetic fields; ERD, event‐related desynchronization; ERP, event‐related potentials; ERS, event‐related synchronization; EXP, experiment; GSM, Global System for Mobile Communications; MP, mobile phone; n/a, not applicable; NR, not reported; PM, pulse modulation; RF, radiofrequency; SAR, specific absorption rate; SFMT, simple finger movement task; UMTS, Universal Mobile Telecommunications System; VMT, visual monitoring task.

^a This number includes the reference electrode(s).

3.2.1. Resting EEG studies

Studies were published between 1995 and 2023 and the majority of the 31 studies focused on the influence of EMF on the alpha rhythm. The sample size varied between 10 and 120 participants (median: 20.5; interquartile range: 15.3–31.8). Five studies used an EO set‐up, whereas 17 studies used EC and seven studies performed both EC and EO. Two studies did not report on EO or EC. A total of 21 studies examined the impact of 2G, while 2, 3, and 1 studies investigated 3G, 4G, and 5G, respectively. The number of electrodes used varied between four and 128, including reference electrode(s). Carrier frequency varied between 450 and 2573 MHz in which the majority was around 900 MHz. PM was 217 Hz for most studies. The applied EEG protocol varied both in duration and number of repetitions with most studies having two or three sessions summing to a total of 120–180 min. Exposure power was around 250 mW for most studies, if reported.

In terms of EMF effects 14, 26, 18, and 12 studies investigated the effects in the beta, alpha, theta, and delta band, respectively. From these 8, 17, 1, and 2 studies reported an effect in the beta, alpha, theta and delta band respectively, while the remaining studies reported no effect. In the studies reporting changes the direction of the effect was increase in 6, 14, 1, and 0 cases in the beta, alpha, theta, and delta band, respectively, while the remaining studies reported a decrease.

3.2.2. ERP‐studies

Studies were published between 1998 and 2018. The sample size varied between 10 and 120 participants. Four and 10 studies used a visual or auditory discrimination task, respectively. Additional tasks utilized included a simple finger movement task (two studies), a visual monitoring task (three studies), a contingent negative variation task (two studies), a continuous performance test (one study), an auditory memory task (four studies), an N‐back test (three studies), and a go/no go task (one study). Sixteen studies assessed the influence of 2G, whereas five studies looked at 3G. The number of electrodes used varied between 3 and 62, including reference electrode(s). Carrier frequency varied between 450 and 1950 MHz in which the majority was around 900 MHz. PM was 217 Hz for most studies. The applied EEG protocol varied both in duration and number of repetitions with most studies having two tasks for a total of 60 min. Exposure length and power were around 30 min for most studies and varied between 0.64 mW and 2.8 W at peak. Effects were reported in 13 studies, while seven studies reported no effects.

3.3. RoB

The studies of Dalecki et al. (2021) and Loughran et al. (2019) shared the same set of participants and were therefore regarded as one trial to assess on RoB. This was also the case for the studies of Wallace et al. (2022, ²⁰²³). No ERP studies were grouped together, resulting in 29 resting state wake EEG trials and 20 ERP trials assessed on RoB.

None of the 29 resting state wake EEG trials and 20 ERP trials were at no risk (Tables 3 and 4). Overall, we assessed 7/29 (24%) resting state wake EEG and 8/20 (40%) ERP trials to be at some or high risk for the randomization process, 8/29 (28%) resting state wake EEG trials, and 8/20 (40%) trials for period and carryover effects, 1/29 (3%) resting state wake EEG trials and no ERP trials for deviations from the intended interventions, no trials for either missing outcome data or for measurement of the outcome. Only one ERP trial (5%) and no resting wake EEG trials were assessed to be at no risk for selection of reported results.

Table 3.

Risk of bias of the included trials on the resting state wake EEG.

Trial	A	B	C	D	E	F	Overall
Bachmann et al. (2007)
Croft et al. (2002)									Low risk
Croft et al. (2008)									Some concerns
Croft et al. (2010)									High risk
Curcio et al. (2005)
D'Costa et al. (2003)								A	Randomization process
Dalecki et al. (2021) and Loughran et al. (2019)								B	Bias arising from period and carryover effects
Ghosn et al. (2015)								C	Deviations from the intended interventions
Hietanen et al. (2000)								D	Missing outcome data
Hinrichs & Heinze (2006)								E	Measurement of the outcome
Hinrikus et al. (2008)								F	Selection of the reported result
Hinrikus et al. (2009)
Hinrikus et al. (2011)
Hinrikus et al. (2017)
Huber et al. (2002)
Jamal et al. (2023)
Kleinlogel et al. (2008a)
Lv et al. (2014)
Van der Meer et al. (2023)
Nakatani‐Enomoto et al. (2020)
Perentos et al. (2007)
Perentos et al. (2013)
Regel et al. (2007)
Reiser et al. (1995)
Röschke & Mann (1997)
Suhhova et al. (2013)
Vecchio et al. (2010)
Wallace et al. (2022) and Wallace et al. (2023)
Yang et al. (2017)

Abbreviation: EEG, electroencephalography.

Table 4.

Risk of bias of the included trials on the ERP.

Trial	A	B	C	D	E	F	Overall
Dalecki et al. (2018)									Low risk
de Tommaso et al. (2009)									Some concerns
Eggert et al. (2015)									High risk
Eulitz et al. (1998)
Freude et al. (1998)
Freude et al. (2000)								A	Randomization process
Hamblin et al. (2004)								B	Bias arising from period and carryover effects
Hamblin et al. (2006)								C	Deviations from the intended interventions
Kleinlogel et al. (2008b)								D	Missing outcome data
Krause et al. (2000a)								E	Measurement of the outcome
Krause et al. (2000b)								F	Selection of the reported result
Krause et al. (2004)
Krause et al. (2007)
Kruusing & Lass (2008)
Leung et al. (2011)
Maganioti et al. (2010)
Stefanics et al. (2008)
Trunk et al. (2013)
Trunk et al. (2014)
Vecchio et al. (2012)

Abbreviation: ERP, event‐related potentials.

3.4. Meta‐analysis

From the 51 included studies, we were able to use seven studies to create three different groups for meta‐analysis for resting state wake EEG and five studies to create 10 different groups for meta‐analysis for ERP. Appendix SB and Figure 1 provide an overview for reasons of exclusion from the meta‐analysis. Of note: all studies not selected for meta‐analysis are still reported in Tables 1 and 2. For the resting state wake EEG studies, three groups were selected for the meta‐analysis consisting of studies using PM 2G radiation with EO, PM 2G radiation with EC, and 3G radiation with EO. The number of studies in the meta‐analyses for resting state wake EEG varies from 1 (3G radiation EO) to 5 (2G radiation with EO). For the ERP studies, 10 groups were selected, consisting of studies using either a visual discrimination task (four groups) or an auditory discrimination task (six groups). For the visual discrimination task, separate groups for the outcome P300 and P100 latency and amplitude were created for 2G radiation. For the auditory discrimination task, separate groups for the outcome P300 latency and amplitude were created, for both 2G and 3G radiation, as well as for N100 for 3G radiation latency and amplitude. The number of studies in the meta‐analyses for ERP varies from two for all analyses related to a visual discrimination task and 2–4 for an auditory discrimination task. Forest plots for all groups are shown in Figures 2, 3, and 4 and summary statistics can be found in Table 5. None of the meta‐analyses led to significant results, except for the PM 2G radiation with EO, which showed an increase in alpha band activity. A sensitivity analysis is provided in Appendix SE for this meta‐analysis.

Figure 2.

Forest plots comparing differences in alpha power for sham and exposure condition for (a) 2G with eyes open, (b) 2G with eyes closed, (c) 3G with eyes open. Risk of Bias legend: (A) Randomization process, (B) Period and carryover effects, (C) Deviations from the intended interventions, (D) Missing outcome data, (E) Measurement of the outcome, (F) Selection of the reported result.

Figure 3.

Forest plots comparing differences in latency and amplitude for sham and exposure condition for a visual discrimination task and 2G radiation for P300 (a and b) and P100 (c and d). Risk of Bias legend: (A) Randomization process, (B) Period and carryover effects, (C) Deviations from the intended interventions, (D) Missing outcome data, (E) Measurement of the outcome, (F) Selection of the reported result.

Figure 4.

Forest plots comparing differences in latency and amplitude for sham and exposure condition for an auditory discrimination task for (a and b) 2G and P300, (c and d) 2G and in N100, and (e and f) 3G and P300. Risk of Bias legend: (A) Randomization process, (B) Period and carryover effects, (C) Deviations from the intended interventions, (D) Missing outcome data, (E) Measurement of the outcome, (F) Selection of the reported result.

Table 5.

Summary statistics from the meta‐analysis for each comparison made.

Comparison	Outcome	Number of studies	N	SMD	95% CI	p Value	I 2 (%)
2G radiation with eyes open	Alpha power (μV2)	5	322	0.16	0.01; 0.32	0.04	0
2G radiation with eyes closed	Alpha power (μV2)	3	81	−0.04	−0.35; 0.27	0.80	0
3G radiation with eyes open	Alpha power (μV2)	1	62	−0.16	−0.51; 0.20	0.39	0
2G radiation with a visual discrimination task	P300 latency (ms)	2	156	0.12	−0.10; 0.34	0.29	0
2G radiation with a visual discrimination task	P300 amplitude (μV)	2	156	−0.08	−0.31; 0.14	0.46	0
2G radiation with a visual discrimination task	P100 latency (ms)	2	156	−0.01	−0.23; 0.21	0.91	0
2G radiation with a visual discrimination task	P100 amplitude (μV)	2	156	−0.05	−0.28; 0.17	0.63	0
2G radiation with an auditory discrimination task	P300 latency (ms)	4	209	0.20	−0.14; 0.55	0.25	54
2G radiation with an auditory discrimination task	P300 amplitude (μV)	3	197	−0.05	−0.25; 0.15	0.63	0
2G radiation with an auditory discrimination task	N100 latency (ms)	3	194	−0.06	−0.44; 0.33	0.78	61
2G radiation with an auditory discrimination task	N100 amplitude (μV)	3	194	0.01	−0.28; 0.29	0.96	34
3G radiation with an auditory discrimination task	P300 latency (ms)	2	77	0.09	−0.22; 0.41	0.57	0
3G radiation with an auditory discrimination task	P300 amplitude (μV)	2	77	−0.11	−0.43; 0.20	0.48	0

Abbreviations: CI, confidence interval; I ², percentage of inconsistency attributable to heterogeneity and not chance (lower is better); N, total sample size of comparison; SMD, standardized mean difference.

4. DISCUSSION

This meta‐analysis is the first systematic quality assessment in this field and has three important findings. First, there is evidence of an effect on the EEG of a 2G protocol using an EO condition. Still, it should be acknowledged that more studies are needed to further strengthen this outcome. Second, both for a visual and auditory discrimination task we did not find an effect on latency and amplitude components. This suggests that the impact of EMF during task performance is less pronounced compared to the resting state condition. Third, this meta‐analysis shows that the field is unable to create an evidence base for most comparisons due to heterogeneity of the methodology.

Our meta‐analysis contributes further evidence to increased alpha activity upon EMF exposure for a PM 2G protocol under EO conditions. Alpha waves typically manifest in healthy, awake adults during periods of rest with closed eyes, but they diminish during sleep and when an individual focuses on a specific task. While several studies suggest that the thalamus is the primary alpha pacemaker, the underlying neuronal network generating the alpha waves is still not fully understood (Bazanova and Vernon, 2014; Halgren et al., ²⁰¹⁹). The alpha rhythm is associated with cognitive inhibition and visual relaxation and, therefore, the suppression of the alpha amplitude is a reflection of the activation of the brain in response to visual and cognitive load (Bazanova and Vernon, 2014; Klimesch, ¹⁹⁹⁹).

It should be noted that the 2G EO meta‐analysis is based on five studies and four of them are from the same research group. In all studies in the meta‐analysis, the direction of the effect was the same: EMF invariably leads to an increase of alpha activity. For one study this differed from the direction mentioned in the original paper (Perentos et al., 2013), where the authors observed a decrease by comparing the change in alpha activity from baseline measurements to PM RF exposure. In this meta‐analysis, however, similar to the rest of the literature, we do not investigate the change in alpha power from baseline to measurement, but the difference in alpha power between the sham and the PM RF measurement, which in this case leads to a change in the direction of the effect compared to the original paper.

The increase in alpha activity was observed in many other studies not included in our meta‐analysis (see Table 1). Especially in studies by the research group from Hinrikus and co‐workers increased alpha activity was reported, but the center frequency used (450 MHz) was different from 2G and differences with regard to the time frames used to derive results exist, prohibiting inclusion of this work in our meta‐analysis. Apart from the papers from Hinrikus and co‐workers, in total four other studies reported on increased alpha activity during 2G EO exposure, but they could not be included in the meta‐analyses for several reasons: EMF exposure prior to EEG measurements (Huber et al., 2002; Regel et al., ²⁰⁰⁷), not reported whether eyes were opened or closed (Reiser et al., 1995), mean spectral power and standard deviations not reported (Croft et al., 2002). Although several papers have claimed that 2G exposure leads to increased alpha activity, our contribution is to strengthen this claim through a meta‐analysis.

There is a clear difference in results when resting state EEG studies use EO compared to EC. Most of the studies reporting no effects were with EC condition, with only a few exceptions, which was further supported by the meta‐analysis. However, it is unlikely that the effect of EMF is entirely absent during EC conditions. This is because alpha power is stronger during EC, making it more challenging to detect subtle modulations arising from EMF exposure. This line of reasoning was also provided by Dalecki et al. (2021) who found that the presence of the large alpha power increase during eyes close leads to a poor signal‐to‐noise ratio.

Most of the studies used a 2G protocol and newer protocols were less frequently studied. Only one study used a 5G protocol, but in combination with pulsation that is more typical for 2G (Jamal et al., 2023). Although meta‐analyses could be performed in this review, the number of studies involving 3G, 4G, and 5G is too small to draw conclusions.

It has been suggested that the brain does not exhibit a dose–effect relationship in response to EMF (Belyaev et al., 2022; Hinrikus et al., ²⁰²²; Suhhova et al., ²⁰¹³). This would mean that not only intensity, but also carrier frequency and modulation, could have a significant impact on the extent of the biological effect. However, the existing data mainly focusing on 2G EMF are insufficient to provide clarity on this matter.

As to the meta‐analyses for visual and auditory tasks, it should be acknowledged that the number of studies per analysis is small (2–4) and apart from the studies from Kleinlogel et al. (2008a, ^2008b), all studies come from the same research group. Interestingly, the studies from Kleinlogel et al. do report effects pointing in the same direction as the other studies. In addition to this, for the auditory discrimination task both for 2G and 3G exposure the direction of the effect for P300 latency and amplitude is similar: an increase in latency and a decrease in amplitude. These observations emphasize that further studies are needed to substantiate these trends and that no conclusions on the presence or absence of an effect can be drawn at this time.

Several observations can be made regarding the RoB analysis. All studies except one did not mention that they included a prespecified analysis plan that was finalized before unblinded outcome data were available, making it impossible to assess selective reporting or publication bias. In all studies with high RoB except one the scores originate from the randomization process. It is also interesting to observe that ERP studies with a high RoB tend to find more effects compared to studies with some concerns. Only two studies with high RoB were selected in the meta‐analyses.

An important aspect of the setup relates to the possible interference between EMF exposure and EEG recording possibly leading to false positives in the data. We noticed that only in 33% (17/51) of the studies phantom measurements were carried out to study the interference between EMF exposure and EEG recording. Especially for the 2G protocol it seems that the interference does not touch the relevant physiological range of the EEG signal, as was also suggested in an experiment by Wood et al. (2003). Nevertheless, we believe that it is imperative to carry out these measurements in individual experiments. At the same time, we agree with an observation made by Dalecki et al. (2021) who have argued that the absence of an EMF effect in the EC condition contradicts the presence of electrical interference, since one would expect the latter to be present in both EO and EC conditions.

This meta‐analysis has some limitations. First, we were not able to assess the quality of the EEG processing‐ and exposure setup. Second, it was impossible to find two studies that were identical to each other from a methodological perspective. Hence, included studies still differed methodologically in a considerable number of aspects, such as duration of the EMF exposure, SAR, and exposure protocols. Third, the majority of the meta‐analyses included a limited number of studies and all except one meta‐analysis were not significant. It is important to recognize that although the nonsignificant meta‐analyses found no evidence for the presence of an EMF effect, this is not the same as finding evidence for the absence of an effect. Last, for one paper we had to read the ERP amplitude and latency from a published bar graph, introducing further inaccuracy (Leung et al., 2011).

Our meta‐analysis reveals that the current knowledge base on the effects of EMF on EEG and ERP is weak. Every systematic review aims to collect data for a meta‐analysis. If a meta‐analysis is not possible due to a lack of combinable data, or far‐reaching methodological issues, the road toward the meta‐analysis still offers important information about the current status of the field. For instance, it is telling that out of 51 papers included, and numerous comparisons made, and outcomes measured, we were only able to include 13 comparisons. Some of these contained different aspects of the same study. Furthermore, many studies showed severe issues in key parts of experimental studies, such as the randomization process, carryover effects, and blinding. We therefore advise that all future studies are double‐blind in nature, adhere to the methodological standard of randomized experiments, and publish their protocols first.

In conclusion, this review demonstrates that the existing literature provides evidence supporting the existence of an EMF effect during EO in the alpha band of the EEG. However, a significant portion of the studies is not suitable for inclusion in a meta‐analysis, highlighting the limitations of the current body of research. Our results suggest that unless the challenge of methodological heterogeneity is resolved, the scientific foundation for substantiating the physiological effects of EMF in vivo may continue to be insufficient.

CONFLICTS OF INTEREST

The authors declare no conflict of interest.

ETHICS STATEMENT

This systematic review involved the analysis of data from previously published studies and did not involve any new data collection from human subjects by the authors. As such, ethical approval and informed consent were not required for this review.

Supporting information

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Prins, A.C. , Baas, K. , van der Meer, J.N. , Jacobs, M ., Nederveen, A.J. : The effect of mobile phone electromagnetic fields on the human resting state wake EEG and event‐related potential: A systematic review and meta‐analysis. Bioelectromagnetics, 46, e22531 (2025)