Topical repellents for malaria prevention

Abstract

Background

Insecticide‐based interventions, such as long‐lasting insecticide‐treated nets (LLINs) and indoor residual spraying (IRS), remain the backbone of malaria vector control. These interventions target mosquitoes that prefer to feed and rest indoors, but have limited capacity to prevent transmission that occurs outdoors or outside regular sleeping hours. In low‐endemicity areas, malaria elimination will require that these control gaps are addressed, and complementary tools are found. The use of topical repellents may be particularly useful for populations who may not benefit from programmatic malaria control measures, such as refugees, the military, or forest goers. This Cochrane Review aims to measure the effectiveness of topical repellents to prevent malaria infection among high‐ and non‐high‐risk populations living in malaria‐endemic regions.

Objectives

To assess the effect of topical repellents alone or in combination with other background interventions (long‐lasting insecticide‐treated nets, or indoor residual spraying, or both) for reducing the incidence of malaria in high‐ and non‐high‐risk populations living in endemic areas.

Search methods

We searched the following databases up to 11 January 2023: the Cochrane Infectious Diseases Group Specialised Register; CENTRAL (in the Cochrane Library); MEDLINE; Embase; CAB Abstracts; and LILACS. We also searched trial registration platforms and conference proceedings; and contacted organizations and companies for ongoing and unpublished trials.

Selection criteria

We included randomized controlled trials (RCTs) and cluster‐randomized controlled trials (cRCTs) of topical repellents proven to repel mosquitoes. We also included non‐randomized studies that complied with pre‐specified inclusion criteria: controlled before‐after studies (CBA), controlled interrupted time series (ITS), and controlled cross‐over trials.

Data collection and analysis

Four review authors independently assessed trials for inclusion, and extracted the data. Two authors independently assessed the risk of bias (RoB) using the Cochrane RoB 2 tool. A fifth review author resolved any disagreements. We analysed data by conducting a meta‐analysis, stratified by whether studies included populations considered to be at high‐risk of developing malaria infection (for example, refugees, forest goers, or deployed military troops). We combined results from cRCTs with RCTs by adjusting for clustering and presented results using forest plots. We used the GRADE framework to assess the certainty of the evidence. We only included data on Plasmodium falciparum infections in the meta‐analysis.

Main results

Thirteen articles relating to eight trials met the inclusion criteria and were qualitatively described. We included six trials in the meta‐analysis (five cRCTs and one RCT).

Effect on malaria incidence

Topical repellents may slightly reduce P falciparum infection and clinical incidence when both outcomes are considered together (incidence rate ratio (IRR) 0.74, 95% confidence interval (CI) 0.56 to 0.98; 3 cRCTs and 1 RCT, 61,651 participants; low‐certainty evidence); but not when these two outcomes were considered independently. Two cRCTs and one RCT (12,813 participants) evaluated the effect of topical repellents on infection incidence (IRR 0.76, 95% CI 0.56 to 1.02; low‐certainty evidence). One cRCT (48,838 participants) evaluated their effect on clinical case incidence (IRR 0.66, 95% CI 0.32 to 1.36; low‐certainty evidence). Three studies (2 cRCTs and 1 RCT) included participants belonging to groups considered at high‐risk of being infected, while only one cRCT did not include participants at high risk.

Adverse events

Topical repellents are considered safe. The prevalence of adverse events among participants who used topical repellents was very low (0.6%, 283/47,515) and limited to mild skin reactions.

Effect on malaria prevalence

Topical repellents may slightly reduce P falciparum prevalence (odds ratio (OR) 0.81, 95% CI 0.67 to 0.97; 3 cRCTs and 1 RCT; 55,366 participants; low‐certainty evidence). Two of these studies (1 cRCT and 1 RCT) were carried out in refugee camps, and included exclusively high‐risk populations that were not receiving any other background vector control intervention.

Authors' conclusions

There is insufficient evidence to conclude that topical repellents can prevent malaria in settings where other vector control interventions are in place. We found the certainty of evidence for all outcomes to be low, primarily due to the risk of bias. A protective effect was suggested among high‐risk populations, specially refugees, who might not have access to other standard vector control measures.

More adequately powered clinical trials carried out in refugee camps could provide further information on the potential benefit of topical repellents in this setting. Individually randomized studies are also likely necessary to understand whether topical repellents have an effect on personal protection, and the degree to which diversion to non‐protected participants affects overall transmission dynamics.

Despite this, the potential additional benefits of topical repellents are most likely limited in contexts where other interventions are available.

Plain language summary

What are the benefits and risks of using topical insect repellents to prevent malaria?

Key messages

• Topical repellents may slightly reduce the incidence and prevalence of malaria caused by Plasmodium falciparum.

• These changes seem to be particularly important in high‐risk populations, specifically in refugees living in camps where there are fewer other options.

• Topical repellents may make little or no difference in malaria prevalence and incidence in settings where insecticide‐treated nets, and other options to control the transmission of malaria are readily available.

What is malaria?

Malaria is a disease caused by at least five species of parasites from the genus Plasmodium, and spread by the bite of Anopheles mosquitoes. The disease regularly affects people in tropical areas of Central and South America, South and Southeast Asia, and particularly, Africa. Over 247 million malaria cases and 619,000 deaths occurred in 2021, mostly in Africa. The disease affects the function of red blood cells, which transport oxygen through the body. This generally causes fever, malaise (a feeling of 'just not feeling well'), and other mild symptoms. However, some people can develop complicated disease, which is associated with a severe reduction in the number of circulating red blood cells, and problems in the liver, brain, and other organs.

Malaria can be treated with different medicines, which are generally effective. Certain tools that prevent mosquito bites, like nets treated with insecticides, can protect people from getting it, and have helped to significantly reduce the number of cases around the world. Nonetheless, most of these approaches target mosquitoes that feed indoors and on humans. They are less effective against species that can feed outdoors, so do not really eliminate the disease.

What did we want to find out?

The aim of this Cochrane Review was to find out if topical insect repellents (substances applied to the skin to prevent mosquito bites) can prevent malaria in people living in regions where this disease occurs regularly. We were particularly interested in their effect on people who might not be adequately protected by other measures, which are more commonly used to prevent malaria.

We wanted to find out if topical repellents were better than a placebo, or no intervention at all, to reduce two indicators of malaria transmission:

• Malaria incidence (the number of new cases in a period of time);

• Malaria prevalence (the number of all cases at a certain moment).

We also wanted to know if topical repellents caused any adverse side effects to people who used them.

What did we do?

We searched the existing literature for studies that compared the effect of topical repellents (alone or in combination with other tools to prevent mosquito bites) with a placebo or no intervention. We compared and summarized the results of the included studies, and rated our confidence in the evidence they provided, based on the methods used in each one.

What did we find?

We included a total of eight studies, which included over 60,000 people. The studies took place in areas with low malaria transmission, mostly in Southeast Asia and South America.

The topical repellents evaluated included lotions, soaps, and cosmetics. We found evidence suggesting that topical repellents may slightly reduce the incidence and prevalence of malaria cases caused by P falciparum in settings where other tools to prevent mosquito bites are not available. Despite this, our findings suggest that repellents probably make little or no difference in places where these tools are already widely used. Topical repellents are considered safe, and the prevalence of adverse side effects was very low.

What are the limitations of the evidence?

The benefits of topical repellents were particularly clear among refugees. However, shortfalls in the design of the included studies did not allow us to generalize these observations to other contexts. We only included cases of malaria caused by the parasite P falciparum. We also recognize that studies measured and reported adherence differently, and often did not know if the participants actually used the repellent as advised.

How up to date is this review?

The evidence is up to date to 11 January 2023.

Summary of findings

Summary of findings 1. Topical repellents versus placebo or no treatment in adults and children living in malaria‐endemic areas.

Patient or population: adults and children in malaria‐endemic areas Setting: malaria‐endemic areas (November 1991 to June 2016; Bolivia, Cambodia, Ecuador, Laos PDR, Myanmar, Pakistan, Peru, Tanzania, Thailand) Intervention: topical repellents Comparison: placebo or no treatment
Outcome	Anticipated absolute effects (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with placebo or no treatment	Risk difference with topical repellents
Malaria infection incidence alone (Follow‐up: mean 6 months)	37 per 1000	9 fewer per 1000 (16 fewer to 1 more)	Rate ratio 0.76 (0.56 to 1.02)	12,813 (2 cRCTs, 1 RCT)	⨁⨁◯◯ Lowa,b	Topical repellents may make little or no difference to the incidence of P falciparum infection.
Malaria case incidence alone (Follow‐up: mean 12 months)	22 per 1000	7 fewer per 1000 (15 fewer to 8 more)	Rate ratio 0.66 (0.32 to 1.36)	48,838 (1 cRCT)	⨁⨁◯◯ Lowc,d	Topical repellents may make little or no difference to P falciparum clinical case incidence
Malaria case and infection incidence together (Follow‐up: mean 7 months)	24 per 1000	6 fewer per 1000 (10 fewer to 0 fewer)	Rate ratio 0.74 (0.56 to 0.98)	61,651 (3 cRCTs, 1 RCT)	⨁⨁◯◯ Lowb,e	Topical repellents may slightly reduce P falciparum infection and clinical case incidence when both outcomes are pooled
Adverse events Assessed with self‐reported questionnaires and in‐person interviews (Follow‐up: mean 13 months)	A total of 283 adverse events (0.6%) were observed among participants who received topical repellents.			47,515 (6 cRCTs, 1 RCT)	‐	Topical repellents are considered safe
Malaria prevalence (Follow‐up: mean 13 months)	13 per 1000	2 fewer per 1000 (4 fewer to 0 fewer)	Odds ratio 0.81 (0.67 to 0.97)	55,366 (3 cRCTs, 1 RCT)	⨁⨁◯◯ Lowb,f	Topical repellents may slightly reduce P falciparum prevalence

Abbreviations: CI: confidence interval; cRCT: cluster‐randomized controlled trial; RCT: randomized controlled trial.

The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). The denominator in the malaria incidence outcomes is person‐years, and people in the prevalence outcome.

GRADE Working Group grades of evidence
High certainty: we are very confident that the true effect lies close to that of the estimate of the effect
Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different
Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect
Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect

^aDowngraded 1 level due to risk of bias associated with the procedures used to randomize participants, conceal allocation, and imbalances in the allocation groups.
^bDowngraded 1 level due to indirectness associated with the inclusion of only pregnant women in one study.
^cDowngraded 1 level due to risk of bias associated with imbalances in the allocation groups and the lack of placebo in controls.
^dDowngraded 1 level due to imprecision, as 95% CIs include a relevant reduction in malaria incidence and no effect.
^eDowngraded 1 level due to risk of bias associated with procedures used to conceal allocation, imbalances in the allocation groups, and a large proportion of losses to follow‐up (16.6%) in one study. 
^fDowngraded 1 level due to risk of bias associated with the step‐wedged design and the lack of placebo in the control group of two studies, issues in the procedures used to blind study participants, and imbalances in allocation groups.

Background

Description of the condition

Malaria is arguably the most important parasitic disease in the world. Five species of protozoan parasites from the genus Plasmodium regularly infect humans: Plasmodium falciparum, Plasmodium vivax, Plasmodium ovale, Plasmodium malariae, and Plasmodium knowlesi. Over 247 million cases of malaria were estimated to have occurred in 2021, increasing from 245 million the previous year, most of them in Africa (WHO 2022).

All of these parasites are transmitted by the bite of female mosquitoes from over 40 species of the genus Anopheles, which are widely spread throughout tropical and subtropical regions (Sinka 2012). Malaria has a wide clinical presentation, with most cases characterized by the presence of mild non‐specific signs, such as fever, chills, headache, nausea, and malaise. However, cases caused by P falciparum, which account for 98% of all cases in Africa, can result in severe disease and death, if untreated. In 2021, 619,000 people were estimated to have died of malaria, most of them in Africa (WHO 2022). Severe disease is usually characterized by impaired consciousness, respiratory distress, hypoglycaemia, and severe anaemia. Infection with P knowlesi, the most important zoonotic species, is restricted to Southeast Asia, and is also often associated with complications that include acute kidney disease and hyperparasitaemia. Infections with P vivax, P ovale, and P malariae are generally associated with fewer complications and deaths. Simian species of malaria, such as P cynomolgi and P inui, among others, can occasionally infect humans, but are not considered to be of relevance to public health (Vythilingam 2021).

Significant advances in malaria control have been achieved in the last two decades, mostly by the development and wide distribution of commodity‐based preventive, treatment, and diagnostic measures. These measures include long‐lasting insecticide‐treated nets (LLINs), artemisinin‐based combination therapies (ACTs), and inexpensive, easily scalable rapid diagnostic tests (RDTs). The addition of these interventions to existing strategies, such as indoor residual spraying (IRS) has translated into a 27.5% reduction in global malaria incidence (measured as cases per 1000 people at risk), and a 44.4% reduction in malaria deaths between 2000 and 2019 (WHO 2020a).

Unfortunately, progress has slowed over the last six years. The World Health Organization (WHO) 2016 to 2030 Global Technical Strategy for Malaria presented the ambitious goal of reducing global incidence by 40% in 2020, and by 90% in 2030, compared to 2015 figures (WHO 2015). The world remains off‐track to meet this target, with a global incidence reduction of less than 2% between 2015 and 2019, and an increase of 2 million cases in 2021, compared to the previous year (WHO 2022). This regression is attributed to several factors, including poor access to health systems, political unrest, poor government commitment to malaria control, reduced effectiveness of traditional chemical‐based interventions, and disruption in health services caused by the COVID‐19 pandemic (WHO 2022). New interventions might bring the world closer to that goal, as long as they reach those who can benefit from them.

Insecticide‐based interventions, LLINs, and IRS, are the backbone of current malaria vector control. In settings where pyrethroid resistance is established, these interventions are thought to remain partially effective at reducing transmission (Kleinschmidt 2018), through their physical barrier and associated sublethal effects (Unwin 2023). However, their effectiveness in reducing malaria transmission depends on adequate coverage. Furthermore, these interventions predominantly target mosquito species that feed mostly indoors, on humans (Killeen 2014). While LLINs and IRS have been highly effective in most African settings, where anthropophilic and endophilic vectors are dominant, their efficacy is considerably lower in other settings, where mosquito species have different feeding and resting habits (Steinhardt 2017). For instance, this is the case in forested regions within the Greater Mekong Subregion in Southeast Asia, where outdoor malaria transmission by exophilic vectors represents a major challenge in the elimination of the disease (Chaumeau 2022; Durnez 2013); and likely influences transmission among gold miners in South America (Douine 2020). Interventions targeting outdoor malaria transmission are still considered an unmet public health need (WHO 2023a)

Another important concern is the persistence of residual malaria transmission, whereby malaria elimination cannot be achieved despite optimal coverage of effective interventions (Killeen 2014). Selection by LLINs and IRS has favoured species and strains with pre‐existing traits that limit the efficacy of these interventions, such as biting and resting outdoors, or feeding on animal hosts (Govella 2013; Killeen 2014). The expansion of Anopheles stephensi into the Horn of Africa leads to further concerns. This invasive species is already established in Djibouti (Sinka 2020), and is expanding through Ethiopia (Carter 2021), where refugees displaced by armed conflict north of the country might be particularly vulnerable to new outbreaks. It has also been detected in Sudan, Somalia, Nigeria (WHO 2022), and Kenya (Ochomo 2023). The efficacy of existing vector control strategies to curb the expansion of this vector in African settings remains unknown.

Description of the intervention

Personal protective measures that effectively prevent mosquito bites, regardless of place and time, may address current control gaps and complement existing interventions (Killeen 2014). Among these measures, topical repellents are a particularly attractive candidate, given extensive data on their safety and efficacy at reducing mosquito bites (Alpern 2016; Nguyen 2023). Topical repellents can be distributed easily among susceptible populations through co‐operation with the private sector and local governments. As an intervention tool, repellents may be particularly useful for high‐risk groups who have increased behavioural or occupational exposure to malaria vectors, and who are not as likely to be protected by either LLINs or IRS. These groups include refugees (Rowland 2001), miners (Olapeju 2021), forest‐goers, soldiers, or indigenous groups (Bevilacqua 2015), among others, who play an important role in maintaining malaria transmission.

Topical repellents include any substance that is applied directly to the skin to prevent insect bites. They represent one of the most widely‐used forms of vector control throughout human history (Herodotus 1996). They are commonly available as lotions, sprays, or gels, but can also be found in the form of soaps that leave a repellent residue on the skin (Kroeger 1997; Rowland 2004). Oils derived from plants, such as citronella (Cymbopogon), neem (Azidarachta indica), and eucalyptus (Eucalyptus maculate citriodon) have been used since antiquity for this purpose, alone or combined with petroleum jelly and similar preparations (Maia 2011). The development of modern repellents began during the 1950s. Of these, N,N‐diethyl‐m‐toluamide (DEET) is the most widely used. Other common compounds include: 2‐2‐hydroxyethyl‐1‐piperidinecarboxylic (icaridin, or picaridin in the USA), para‐menthane‐3,8‐diol (PMD), and 3‐(N‐butyl‐N‐acetyl)‐aminopropionic acid ethyl ester (IR3535). Commonly commercially available formulations include: DEET (5% to 100%), picaridin (7% to 20%), IR3535 (7.5% to 19.7%), oil of lemon eucalyptus (10% to 40%), citronella oil (0.5% to 20%), catnip oil (7% to 15%), and 2‐undecanone (1% to 2% (Nguyen 2023)).

Topical repellents are already regularly used by tourists and expatriates working in malaria‐endemic settings, and their use has been proposed as an alternative to prevent malaria in these and other groups in which LLINs and IRS are expected to be less effective (WHO 2023b). Despite this, it is unclear if the programmatic integration of repellents as an additional vector control commodity into existing control programmes in endemic areas will result in fewer malaria cases (WHO 2023b). There are important drawbacks that may influence the programmatic usefulness of repellents. First, topical repellents do not kill mosquitoes, they offer protection by preventing mosquito bites. Because mosquitoes are not killed, they may be diverted from individuals who use repellents to those who do not (Maia 2013). This raises health equity implications, as accessibility to these products may vary across the different societal strata. Second, their effect is short‐lived and requires repeated administrations. Therefore, protection is highly dependent on user compliance (Sangoro 2014). While repellents are usually well‐received by communities (Sangoro 2014), their regular and adequate use has been shown to be poor, even in trial settings where engagement is enhanced (Sluydts 2016). In some communities, many of the perceived benefits of repellents derive from non‐prescribed uses, such as applying them directly to insects or bed nets (Gryseels 2015). Finally, the large‐scale distribution of topical repellents entails further costs to already under‐funded control programmes, with an estimated incremental cost‐effectiveness ratio (ICER) ranging between USD 212 and USD 832 per infection averted (Agius 2020).

How the intervention might work

Topical repellents do not kill mosquitoes, but prevent bites by interfering with their olfaction, affecting their ability to locate and feed on a treated host. The mode of action involves complex interactions between repellent compounds and the olfactory and gustatory receptors of haematophagous insects (Dickens 2013). While the exact mechanism is unclear, DEET, the most thoroughly studied repellent, appears to act by activating and inhibiting several olfactory receptors, interfering with the mosquito's capacity to detect airborne molecules, while simultaneously triggering avoidance behaviours (Norris 2017). IR3535 and picaridin are thought to act similarly (Norris 2017). These compounds have also been shown to stimulate other receptors in the feeding apparatus of mosquitoes, acting as deterrents upon contact (Dickens 2013).

Malaria transmission is dependent on vectorial capacity. This is a concept coined during the first Global Malaria Eradication Campaign, and can be understood as the daily rate at which parasites are inoculated to susceptible hosts from an original infective case, assuming all mosquitoes biting it become infected (Garrett‐Jones 1964). Among the key determinants of vectorial capacity is the probability of a mosquito biting a person, which has an exponential effect on the number of new infections. Recent updates to the classical model have also incorporated changes to the mosquito biting rate, which not only reduces the probability of infection, but also impacts the number of eggs laid, further contributing to reductions in transmission (Brady 2016).

Therefore, topical repellents could reduce malaria transmission by reducing an individual's probability of being bitten, while inhibiting feeding, and therefore, reducing egg production in female mosquitoes. However, because vectors may be diverted from treated to untreated individuals, there is a possibility of increased transmission among unprotected or non‐compliant groups, potentially leading to a loss of effectiveness at a community level. Nevertheless, they may be useful to prevent malaria transmission among high‐risk groups, in which LLINs, IRS, and other traditional vector control interventions are unfeasible, and when effectively covering the entire susceptible population is not possible.

Why it is important to do this review

The incorporation of LLINs into malaria control programmes accounts for around 68% of the 660 million cases averted between 2000 and 2015 (Bhatt 2015). However, the recent stagnation in progress highlights the need for improved or new complementary interventions, or both, which address the gap that nets fail to cover. This is particularly true in settings in which the main vectors of the disease often feed outdoors and early in the evening, or can take blood meals from animals, as well as humans. It is widely accepted that complementary interventions will be required if elimination is to be achieved. Citing the WHO Director General: "If we continue with a business as usual approach – employing the same level of resources and the same interventions – we will face near‐certain increases in malaria cases and deaths" (WHO 2020a).

In Africa, high coverage of LLINs and IRS programmes has been linked to changes in mosquito behaviour that could limit the efficacy of intra‐domiciliary interventions (Ferreira 2017; Russell 2011). Similarly, secondary vectors, such as Anopheles arabiensis, have replaced other species more readily targeted by these interventions (Killeen 2017).

Outdoor transmission is essential in maintaining malaria among high‐risk groups living in forested areas (Carnevale 2021), including forest workers, miners, and military personnel, which in turn represents a major obstacle in malaria elimination in places like India (Ranjha 2021), the Greater Mekong Subregion (Jongdeepaisal 2022), and the Amazon basin (Fletcher 2022; Saavedra 2019). Refugees are another particularly vulnerable group, given their exposure to outdoor transmission and additional risk factors, which include limited access to health services, poor sanitation and housing, and inconsistent access to vector control interventions (Semakula 2023). These factors also increase the risk of developing severe malaria in this group. Malaria represented the second most common morbidity among refugees in 2021 (18.7%), and the second most common cause of mortality in this group in 2019 (8.3% of total reported deaths (UNHCR 2022)).

This Cochrane Review aims to measure the effectiveness of topical repellents, alone or in combination with LLINs and other background interventions, in reducing the risk of malaria infection among high‐risk and non‐high‐risk populations in endemic regions. This can also be framed within the United Nations Sustainable Development Goal 3: Good health and well‐being, which presents both a global reduction of maternal and child mortality, and the end of the malaria epidemic, as targets to meet by 2030 (WHO 2020a).

Objectives

To assess the effect of topical repellents alone or in combination with other background interventions (long‐lasting insecticide‐treated nets or indoor residual spraying, or both) for reducing the incidence of malaria in high‐ and non‐high‐risk populations living in endemic areas.

Methods

Criteria for considering studies for this review

Types of studies

We included randomized and non‐randomized controlled studies in this review.

Randomized studies

We included studies randomized at either the cluster or individual level, including:

• Randomized controlled trials (RCTs);

• Cluster‐randomized controlled trials (cRCTs) with at least two clusters per arm;

• Cluster‐randomized cross‐over studies with at least three data points both before and after the intervention was introduced; and

• Cluster‐randomized studies using a stepped‐wedge approach.

Non‐randomized studies

We also planned to include non‐randomized studies of interventions (NRSI) that met our inclusion criteria. However, we assessed these studies separately, in a secondary analysis of observational studies for adverse effects and any summary estimates of effectiveness. We included NRSI in the search because we expected to find a limited number of randomized studies addressing the research question. We included:

• Controlled before‐after studies (CBA) with baseline data, a contemporaneous control group, and at least two sites per arm, if the study had ruled out any significant baseline imbalances;

• Controlled interrupted time series (ITS) with at least three data points before and after the intervention was introduced;

• Non‐randomized controlled cross‐over trials with a clearly defined time point for when the cross‐over occurred, and monitoring of at least two transmission seasons before and after the cross‐over.

We assessed the methodological characteristics of each observational study design according to Cochrane Effective Practice and Organization of Care (EPOC) criteria for inclusion (EPOC 2017). For studies that met the Cochrane EPOC criteria, we used the ROBINS‐I signalling questions to assess their risk of bias (Sterne 2016); if we considered the study did not present a critical risk of bias, we included it and extracted the methodological characteristics.

Types of participants

Eligible participants were children and adults who lived in a malaria‐endemic area, categorized into high‐risk or non‐high‐risk populations. For the purpose of this review, we considered high‐risk populations to be populations who either did not have access to, or were less likely to benefit from programmatic vector control interventions (long‐lasting insecticide‐treated nets (LLINs) or indoor residual spraying (IRS)). Examples of these groups included, but were not limited to, refugees, miners, forest‐goers, soldiers, or indigenous groups.

Types of interventions

We included trials with or without background interventions (LLINs, IRS, or other), as long as they were balanced between trial arms.

Intervention

The interventions of interest were topical repellents, including N,N‐diethyl‐m‐toluamide (DEET), icaridin (picaridin), 3‐(N‐butyl‐N‐acetyl)‐aminopropionic acid ethyl ester (IR3535), para‐menthane‐3,8‐ diol (PMD), oil of lemon eucalyptus (OLE), N, N‐diethylbenzamide, or 2‐undecanone (methyl nonyl keton).

Control

Individuals in eligible control groups received a placebo or no treatment.

We excluded the data of participants infected with P vivax or P ovale from trials carried out in endemic areas for these parasites if the participants were not screened and cleared of parasites at the beginning of the trial. This was to prevent the inclusion of recrudescent cases in the analysis, since these cannot be prevented by topical repellents. Participants who received radical cure with an 8‐aminoquinoline (such as primaquine), and a schizonticidal drug (such as chloroquine), were considered to be clear of latent infection, following World Health Organization (WHO) guidelines (WHO 2021).

Types of outcome measures

Primary outcomes

• Malaria case incidence: new cases of clinical malaria (caused by Plasmodium spp.) confirmed through blood smears or rapid diagnostic tests (RDTs)

• Malaria infection incidence: new Plasmodium spp. infections, confirmed through thick or thin blood smears, RDTs, or polymerase chain reaction (PCR)

• Malaria case and infection incidence: new Plasmodium spp infection, or case of clinical malaria, confirmed through blood smears, RDTs, or PCR

Secondary outcomes

• Incidence of recorded adverse events (including skin irritation, local pain, eye irritation, irritation of upper airways, nausea, vomiting, headaches, dizziness or confusion, allergic or anaphylactic reactions, and systemic toxicity)

• Malaria prevalence

• Anaemia (haemoglobin < 8 g/dL)

• Time to first infection (days)

• All‐cause fever

• Incidence of severe malaria

• Malaria‐related mortality

• Adherence to regular usage of the intervention (defined based on recommendations provided by researchers to participants of individual trials)

• Human biting rate (HBR)

• Entomological inoculation rate (EIR)

• Sporozoite rate (SR)

• Human blood index (HBI)

Search methods for identification of studies

We tried to identify all relevant trials, regardless of language or publication status. We cross‐referenced this list with studies included in a previous Cochrane Review that evaluated the effect of insect repellents (including topical repellents) on malaria transmission (Maia 2018), to guarantee that we included all relevant studies from that review in this one.

Electronic searches

We searched the following databases, using the search terms and strategy described in Appendix 1:

• Cochrane Infectious Diseases Group Specialized Register (searched 11 January 2023, through Cochrane CENTRAL);

• Central Register of Controlled Trials (CENTRAL; 2023, issue 1), in the Cochrane Library (searched 11 January 2023);

• MEDLINE Ovid (1946 to 11 January 2023);

• Embase Ovid (1947 to 11 January 2023) ;

• LILACS (Latin American and Caribbean Health Science Information database; 1982 to 11 January 2023);

• CAB Abstracts (Web of Science; 1910 to 11 January 2023);

• The French Institute of Research for Development's Horizon Pleins Textes database (www.documentation.ird.fr/; searched 11 January 2023).

We also searched the WHO International Clinical Trials Registry Platform (apps.who.int/trialsearch/) and Clinicaltrials.gov (www.clinicaltrials.gov) on 11 January 2023.

Searching other resources

Conference proceedings

We searched the following conference proceedings for relevant abstracts from 2017 to 2022:

• Multilateral Initiative on Malaria conference abstract booklets;

• Annual American Society of Tropical Medicine and Hygiene conference;

• Entomological Society of America; and

• Society of Vector Ecology of America.

Organizations and pharmaceutical companies

We contacted several organizations for ongoing and unpublished trials: the WHO, Centres for Disease Control and Prevention, United States Agency for International Development, Deployed War Fighter Protection Program, London School of Hygiene and Tropical Medicine, and the Liverpool School of Tropical Medicine.

Reference lists

We also checked the reference lists of all included trials for further relevant studies.

Data collection and analysis

Selection of studies

Two review authors (MGW and JCGF) independently assessed the titles and abstracts of trials identified by the searches. Four review authors (MGW, CW, LBA, and JCGF) independently assessed full‐text copies of potentially relevant trials for inclusion, using an eligibility form based on the inclusion criteria. We compared included trials, and resolved disagreements by discussion and consensus, with arbitration by the fifth review author (MFM). We ensured that multiple publications of the same trial were only included once. Despite reporting the same trial, we included Gryseels 2015 and Sluydts 2016 separately, since they measured one secondary outcome of this review differently (adherence to intervention). We listed excluded studies, with their reasons for exclusion, in the Characteristics of excluded studies table.

Data extraction and management

Four review authors (MGW, CW, LBA, and JCGF) independently extracted information from the trials using pre‐piloted, electronic data extraction forms. In case of differences in extracted data, the four review authors discussed these differences to reach consensus. If unresolved, further discussion involved the fifth author (MFM). In case of missing data, we contacted the original trial author(s) for clarification.

For all eligible studies, we extracted data on the following.

• Trial design: type of trial; method of participant selection; unit of randomization (for RCTs); adjustment for clustering (in the case of cRCTs); sample size; method of blinding of participants and personnel; diagnostic method; method used for ascertainment of adherence; primary vector; vector biting time; vector biting preference; malaria endemicity (prevalence); presence of different Plasmodium species; clearance of P vivax and P ovale parasites at start

• Participants: trial settings, population characteristics, whether participants were considered a high‐risk population, and if so, in which category they would fit (for example: displaced populations, such as refugees, forest goers, or deployed military); whether participants were likely to have had no previous exposure to malaria (for example, displaced populations, deployed troops, or other); recruitment rates; withdrawal; and loss to follow‐up

• Intervention: description of intervention; co‐interventions; description of controls

• Outcomes: definition of outcome; number of participants; power; unit of analysis; incomplete outcomes/missing data; time of follow‐up; passive or active case detection; compliance

For dichotomous outcomes, we extracted the number of participants experiencing each outcome and the number of participants in each treatment group, cluster‐adjusted risk, or odd ratios (OR), and standard errors (SE). For count data outcomes, we extracted the number of events in the treatment and control groups, the total person‐time at risk in each group, or the rate ratio, and a measure of variance (for example, standard error). For continuos outcomes (time to first infection in months), we extracted the mean and a measure of variance, if available (range).

RCTs randomized by the individual

We extracted information on the number of participants randomized to each treatment arm, and the number of events in each of the treatment arms.

cRCTs

For cRCTs we recorded the number of clusters randomized; number of clusters analysed; measure of effect (such as rate ratio, risk ratio (RR), odds ratio (OR), or mean difference (MD)) with confidence intervals (CI) or standard deviations; number of participants; and the intra‐cluster correlation coefficient (ICC) value.

Other studies

For NRSI that we considered eligible for inclusion according to EPOC 2017 criteria, and not considered to be at critical risk of bias (Sterne 2016), we extracted data on estimates of effectiveness and adverse events.

Assessment of risk of bias in included studies

We quantified the effect of assignment to the interventions, regardless of whether the interventions were adhered to as intended (intention‐to‐treat).

Randomized studies

Two review authors (MGW and JCGF) independently assessed the risk of bias of each study using the Cochrane RoB 2 (Higgins 2022a; Sterne 2019).

We assessed the risk of bias across the different studies that measured our primary outcomes, and the incidence of adverse events.

The two review authors (MGW and JCGF) resolved any discrepancies through discussion or by consulting a third review author (MFM). We assessed the risk of bias according to the following domains (Higgins 2022b):

• Bias arising from the randomization process;

• Bias due to deviations from intended interventions;

• Bias due to missing outcome data;

• Bias in measurement of the outcome; and

• Bias in the selection of the reported result.

For cluster‐randomized clinical trials, we added RoB 2 Domain 1b, Bias arising from the timing of identification and recruitment of participants, with its corresponding signalling questions, in order to assess identification/recruitment bias (Higgins 2022c).

We used the Cochrane ROB 2 tool to evaluate and determine the risk of bias of each domain across the different studies. The overall judgement of risk of bias depended on the classification of each individual domain. We classified studies in which we determined the risk of bias of at least one domain to be high, or presented some concerns, as presenting that same risk of bias.

We used the risk of bias Excel tool (available from www.riskofbias.info/), and made a summary of the risk of bias by each outcome within and across studies (Higgins 2022b).

Non‐randomized studies

For NRSI, two review authors (MGW and JCGF) independently assessed the risk of bias using the ROBINS‐I tool (Sterne 2016). For each outcome, we answered signalling questions to systematically judge the risk of bias and provide the basis for an overall risk of bias judgement. The signalling questions assessed bias according to seven different domains.

• Bias due to confounding

• Bias in selection of participants into the study

• Bias in classification of interventions

• Bias due to deviations from intended interventions

• Bias due to missing data

• Bias in measurement of outcomes

• Bias in selection of the reported result

Domains one and two cover bias pre‐intervention, the third domain is bias at the stage of intervention, and domains four to seven represent bias post‐intervention.

We judged the risk of bias to be low, moderate, serious, or critical. If any domain reached critical risk of bias, we stopped the assessment and excluded the study.

For the evaluation of bias due to confounding, we also assessed the following subdomains.

• Socioeconomic status: lower socioeconomic status is considered a prognostic factor for increased risk of malaria transmission.

• Geographical location: malaria transmission is heterogenous across different geographical regions, and therefore, can be a predictor of malaria risk.

In the review, we presented the risk of bias assessments for RCTs and NRSI using outcome‐level traffic light plots created using RevMan Web 2023.

Measures of treatment effect

We compared intervention and control data using incidence rate ratios (IRR) for count data, or OR for dichotomous outcomes, and presented them with their associated 95% CIs.

Unit of analysis issues

We combined results from cRCTs with individual RCTs, accounting for cluster adjustment. For count data, we extracted cluster‐adjusted IRRs directly from the studies and included them in the meta‐analysis, using the general inverse variance method (Higgins 2022c).

For dichotomous data, we extracted the cluster‐adjusted ORs reported by individual studies and included them in the meta‐analysis, as described above. For studies in which adjusted ORs were not reported, we extracted the number of participants and events in both treatment arms and adjusted for the cluster effect using existing methods, and based on the reported ICC (Higgins 2022c). If the trial did not report the ICC value, we used one from a similar, methodologically‐sound study. If cluster adjustment was not possible, we excluded the cRCT from the meta‐analysis and presented its results narratively, or in a separate table. We presented results using forest plots.

Dealing with missing data

In cases of missing data, we only included data on the known results. For outcomes with no missing data, we carried out an intention‐to‐treat analysis by analysing all recruited participants in the group to which they were randomized.

Assessment of heterogeneity

We inspected forest plots for overlapping CIs and assessed statistical heterogeneity in each meta‐analysis using I² and Chi² statistics. We classified heterogeneity as moderate if I² values were between 30% and 60%; substantial if they were between 61% and 75%; and considerable if they were between 76% and 100%. We interpreted a Chi² test statistic with P ≤ 0.10 as indicative of statistically significant heterogeneity. We explored clinical and methodological heterogeneity accounting for the trial population, methods, and interventions.

Assessment of reporting biases

We did not assess reporting bias in this review, as the number of studies included in the meta‐analysis (6 trials) was too low to adequately do so (Harbord 2006).

Data synthesis

We analysed data using RevMan Web 2023 software. We used a fixed‐effect, or a random‐effects model to calculate pooled measurements, based on the clinical and methodological heterogeneity observed in the meta‐analysis. We presented data on estimates of effectiveness and adverse events from eligible NRSI, narratively.

Subgroup analysis and investigation of heterogeneity

We grouped trials and analysed them by the method used to measure incidence (whether passively detected cases of clinical malaria, or actively detected infection with Plasmodium sp). We undertook other subgroup analyses aggregating studies based on whether they included background interventions (LLINS or IRS), and high‐risk populations, and whether participants were randomized at the cluster or individual level. For the malaria prevalence outcome, we further disaggregated studies based on whether they were carried out in refugee camps.

Sensitivity analysis

For the primary outcomes, we performed the following sensitivity analyses:

• Exclusion of trials at high risk of bias;

• Exclusion of cRCTs;

• Exclusion of trials that were not placebo‐controlled;

• For cRCTs with an estimated ICC, the impact of varying the ICC; and

• The impact of adherence (i.e. including only participants who reported that they adhered, as compared to our primary ITT analysis, which assumes that all participants adhered equally to the intervention).

Summary of findings and assessment of the certainty of the evidence

We assessed the certainty of the evidence for each primary outcome and the malaria prevalence outcome using the GRADE approach (Guyatt 2011), as described in Balshem 2011.

• High: we are very confident that the true effect lies close to that of the estimate of the effect

• Moderate: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect

• Low: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect

• Very low: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect

Evidence from RCTs start as high certainty, but can be downgraded, provided valid reasons exist within the following five categories: risk of bias, imprecision, inconsistency, indirectness, and publication bias (Balshem 2011). Evidence from NRSI assessed with the ROBINS‐I tool also starts as high‐certainty evidence, but is automatically downgraded by two levels due to the inherent risk of lacking randomization. Evidence from NRSI, and rarely RCTs, can also be upgraded if there is a large effect; a dose‐response effect; and if all plausible residual confounding would reduce a demonstrated effect, or would suggest a spurious effect if no effect were observed

We summarized our findings in a summary of findings table.

Results

Description of studies

Results of the search

We searched the existing literature to 11 January 2023. We identified 888 records through electronic datasets, and one additional article was directly provided by its authors (Figure 1). After removing 326 duplicates, we performed title and abstract screening on 563 studies. Twenty‐one records underwent full‐text screening, of which we excluded eight. The remaining 13 records related to eight different trials, which we included in the qualitative synthesis; we included six trials in the meta‐analysis.

1.

PRISMA study selection flow diagram

Included studies

One randomized controlled trial (RCT (McGready 2001)), and seven cluster‐RCTs (cRCTs) met the inclusion criteria of the review (Agius 2020; Chen‐Hussey 2013; Hill 2007; Kroeger 1997; Rowland 2004; Sangoro 2014a; Sluydts 2016). Sluydts 2016 and Gryseels 2015 reported on the same study, but both described and analysed the outcome, adherence to treatment, differently; therefore, we presented the results independently in the qualitative synthesis. Sluydts 2016 outcomes were reported in three other publications; Agius 2020 outcomes were reported in one other publication. One of the studies was provided directly by its author, who also provided data on the number of individual participants diagnosed with malaria in each study arm, which was not reported in the published paper (McGready 2001).

For measurement of all primary outcomes (malaria infection incidence, malaria clinical case incidence, and malaria infection and case incidence together), and the secondary outcome of all‐cause fever, we directly extracted cluster‐adjusted incidence rate ratios (IRRs) from the published papers, and pooled them. We used a similar approach for the secondary outcome, malaria prevalence, for which we extracted reported cluster‐adjusted odd ratios (ORs), and included them in the meta‐analysis, using the generalized inverse variance method. For the remaining secondary outcomes, we either included the reported prevalence in the intervention and control groups to calculate odds ratios (ORs; anaemia), or included the reported results in the qualitative synthesis and corresponding tables (adverse events and adherence to treatment).

None of the studies that took place in areas where P vivax was endemic cleared the presence of hypnozoites prior to starting. Therefore, we excluded these studies, and only included data from P falciparum infections in the meta‐analysis. Five studies were carried out in South or Southeast Asia (Agius 2020 in Myanmar; Chen‐Hussey 2013 in Laos PDR; McGready 2001 in Thailand; Rowland 2004 in Pakistan; and Sluydts 2016 in Cambodia), two in South America (Hill 2007 in Bolivia; Kroeger 1997 in Peru and Ecuador), and one in Africa (Sangoro 2014a in Tanzania). Five studies included high‐risk populations (Agius 2020; Chen‐Hussey 2013; McGready 2001; Rowland 2004; Sluydts 2016), two of which were carried out in refugee camps (McGready 2001; Rowland 2004). All studies included adults and children, except McGready 2001, which only included pregnant women. In total, 61,651 participants (31,488 in the intervention arm and 30,163 in the controls) and approximately 54,657 person‐years (27,967 in the intervention arm, and 26,690 in the control group) were included in the assessment of the malaria infection and clinical case incidence outcome. We included 55,366 participants (25,429 in the intervention arm, and 29,937 in controls) in the assessment of the effect on malaria prevalence.

The type and concentration of the interventions used varied, and included: 12% N‐dietylbenzamide (Agius 2020), 15% DEET (Chen‐Hussey 2013), 30% PMD (Hill 2007), 0.5% DEET + 20% permethrin (Kroeger 1997), 20% DEET (McGready 2001), 20% DEET + 0.5% permethrin (Rowland 2004), 15% DEET (Sangoro 2014a), and 10% or 20% picaridin (Sluydts 2016).

Excluded studies

We excluded eight studies upon full‐text screening. We excluded seven due to inadequate study design (Constantini 2004; Dadzie 2013; Lindsay 1998; Maia 2012; Moore 2007; Rowland 2004b; Uzzan 2009), and one for not providing enough information on the applied intervention (Deressa 2014). We listed the detailed reasons for excluding these individual studies in the Characteristics of excluded studies table.

Ongoing studies

We identified two ongoing studies whose results have not been yet published (NCT02938975; NCT05117567).

Risk of bias in included studies

We assessed the risk of bias for our primary outcomes (malaria case and infection incidence) in four studies (clinical case incidence in Sluydts 2016; infection incidence in Chen‐Hussey 2013, Hill 2007, and McGready 2001). Similarly, despite being secondary outcomes, we assessed the risk of bias for adverse events in seven studies (Agius 2020; Chen‐Hussey 2013; Hill 2007; Kroeger 1997; McGready 2001; Rowland 2004; Sluydts 2016); and for the malaria prevalence outcome in four studies (Agius 2020; McGready 2001; Rowland 2004; Sluydts 2016).

Malaria case and infection incidence

Only one study included for the assessment was classified as having a low risk of bias (Chen‐Hussey 2013). Two studies were classified as having a high risk of bias (Hill 2007; McGready 2001), and one as having some concerns (Sluydts 2016). The reasons for this were mostly related to issues in the randomization process, the distribution or concealment of placebos, or the presence of unexplained losses to follow‐up.

Adverse events

Four studies were classified as presenting some concerns (Agius 2020; Chen‐Hussey 2013; Rowland 2004; Sluydts 2016), and three presented a high risk of bias (Hill 2007; Kroeger 1997; McGready 2001). The reasons for this were mostly related to issues in the definition of adverse events, and the lack of description of the methods used to measure and report their incidence, as well as issues in the randomization process.

Malaria prevalence

Three studies were classified as having some concerns (Agius 2020; Rowland 2004; Sluydts 2016), and one was classified at high risk of bias (McGready 2001). In general, most studies presented issues with the randomization process, as well as the distribution and concealment of placebos among participants in the control groups.

A complete risk of bias assessment can be found on Analysis 1.1 and Analysis 1.5.

1.1. Analysis.

Comparison 1: Topical repellents versus placebo or no intervention, Outcome 1: Malaria incidence (case and infection)

1.5. Analysis.

Comparison 1: Topical repellents versus placebo or no intervention, Outcome 5: Malaria prevalence

Effects of interventions

See: Table 1

Primary outcomes

Malaria case incidence

Only one study measured clinical case incidence (Sluydts 2016), and failed to find an effect following the use of topical repellents (IRR 0.66, 95% CI 0.32 to 1.36; 1 study, 48,838 participants; low‐certainty evidence; Analysis 1.1). While Sangoro 2014a also reported data on this outcome, we were unable to extract a cluster‐adjusted IRR; thus we excluded this study from the meta‐analysis. This study failed to demonstrate an effect of topical repellents on malaria incidence (68.29 malaria cases per 1000 people (95% CI 37.05 to 99.53) in the control group, and 60.45 cases per 1000 people (95% CI 48.30 to 72.60) in the intervention group). Kroeger 1997 also reported malaria case incidence, but the outcome was not confirmed via microscopy, rapid diagnostic tests (RDTs), or polymerase chain reaction (PCR), so we excluded it from the meta‐analysis.

Malaria infection incidence

Three studies investigated the effect of topical repellents on malaria infection incidence, regardless of the development of symptomatology (Chen‐Hussey 2013; Hill 2007; McGready 2001). Topical repellents failed to reduce infection incidence (IRR 0.76, 95% CI 0.56 to 1.02; 12,813 participants, low‐certainty evidence; Analysis 1.1).

However, when we pooled data from these two subgroups, we observed a reduction of malaria incidence in the topical repellent groups (IRR 0.74, 95% CI 0.56 to 0.98; 4 studies, 61,651 participants, low‐certainty evidence; Analysis 1.1).

Subgroup analysis

We carried out further subgroup analysis, disaggregating studies by whether other background vector control interventions (LLINS or IRS, or both) were used; and by whether they included high‐risk groups. The only study in which nets were not provided as a background intervention was McGready 2001. Eliminating this study from the meta‐analysis translated into the lack of effect of repellents on malaria incidence (IRR 0.78, 95% CI 0.51 to 1.18; 3 studies, 60,754 participants; Analysis 1.2).

1.2. Analysis.

Comparison 1: Topical repellents versus placebo or no intervention, Outcome 2: Malaria incidence (use of other interventions)

In the three studies that included at least one group of individuals classified as high‐risk, repellents failed to cause a reduction in malaria incidence (IRR 0.76, 95% CI 0.58 to 1.01; 3 studies, 57,643 participants; Analysis 1.3). Only Hill 2007 did not explicitly indicate the inclusion of any participants belonging to high‐risk groups, and also failed to see a reduction in malaria (IRR 0.18, 95% CI 0.02 to 1.4; 1 study, 4008 participants).

1.3. Analysis.

Comparison 1: Topical repellents versus placebo or no intervention, Outcome 3: Malaria incidence (by risk population)

Sensitivity analysis

We performed a series of sensitivity analyses to evaluate the effect of excluding studies with a high risk of bias, cRCTs, and studies that did not include a placebo in the control group, from the meta‐analysis.

When we excluded studies with a high risk of bias (Hill 2007; McGready 2001), we observed no reduction in malaria case and infection incidence (IRR 0.83, 95% CI 0.54 to 1.27; 56,746 participants).

We observed similar results after excluding cRCTs and leaving only McGready 2001 (IRR 0.71, 95% CI 0.49 to 1.04; 897 participants); and those without placebo control (excluding Sluydts 2016, IRR 0.76, 95% CI 0.56 to 1.02; 12,813 participants).

We intended to study the effect of adherence on the effect of repellents, by carrying out a meta‐analysis with data from reported per‐protocol analysis (PPA). While Chen‐Hussey 2013 and Hill 2007 reported a PPA, a sensitivity analysis including only PPA data was not possible, since neither cluster‐adjusted IRRs nor the number of cases were aggregated by the different levels of compliance in Chen‐Hussey 2013. Hill 2007 exclusively provided PPA data, but this was treated as an ITT analysis and included in the overall meta‐analysis, given the low number of participants excluded (1.5% in both groups).

Secondary outcomes

Adverse events

Seven studies reported adverse events (Agius 2020 used 12% N, N‐diethylbenzamide, Chen‐Hussey 2013: 15% DEET, Hill 2007: PMD: 30% + eucalyptus extract, Kroeger 1997: 0.5% DEET and 20% permethrin, McGready 2001: 20% DEET, Rowland 2004: 20% DEET + 0.5% permethrin and Sluydts 2016 used picaridin).

Only Chen‐Hussey 2013 reported the incidence of adverse events in both study arms, while the remaining studies either reported no adverse events, or just indicated those that occurred in the intervention arm.

In three studies, adverse events were passively reported by participants to community health workers or study volunteers (Agius 2020; Rowland 2004; Sluydts 2016). In two studies, participants were actively asked about the occurrence of adverse events during each study visit (Chen‐Hussey 2013; Hill 2007). McGready 2001 did not provide information about how adverse events were reported. Kroeger 1997 reported adverse events associated with the intervention, but later indicated that a large proportion were apparently caused by ectoparasite infestations.

In general, the prevalence of adverse events was very low, with 283 adverse events reported among 47,515 participants who received interventions (0.6%, Table 2Analysis 1.4), and limited to mild skin reactions. The prevalence of adverse reactions in all studies remained below 0.1%, except for McGready 2001 (11.4%), which included only pregnant women; Kroeger 1997 (3%), and Chen‐Hussey 2013, in which prevalence was also higher in the intervention group (3.8% in the intervention arm and 3.2% in controls, P = 0.029).

1. Reported adverse events in the topical repellents arm of included studies (Analysis 1.4).

Study	Total adverse events reported in the topical repellents arm	Study population in the topical repellents arm	%
Agius 2020	0	13809	0
Chen‐Hussey 2013a	150	3947	3.8
Hill 2007	0	2041	0
Kroeger 1997	48	1600	3
McGready 2001	51	449	11.4
Rowland 2004	1	618	0.16
Sluydts 2016	33	25051	0.13
Total	283	47515	0.6

^aStudies that reported adverse events in both intervention arms

1.4. Analysis.

Comparison 1: Topical repellents versus placebo or no intervention, Outcome 4: Reported adverse events in the topical repellents arm of included studies

Reported adverse events in the topical repellents arm of included studies
Study	Total adverse events reported in the topical repellents arm	Population in the topical repellents arm	%
Agius 2020	0	13809	0%
Chen‐Hussey 2013	150	3947	3.8
Hill 2007	0	2041	0%
Kroeger 1997	48	1600	3%
McGready 2001	51	449	11.4%
Rowland 2004	1	618	0.16%
Sluydts 2016	33	25051	0.13%

Malaria prevalence

Four studies evaluated the effect of topical repellents on malaria prevalence (Agius 2020; McGready 2001; Rowland 2004; Sluydts 2016). For Sluydts 2016, four independent surveys, using PCR to diagnose infection, were carried out between 2012 and 2013. These surveys were individually included in the meta‐analysis. For all the cRCTs (Agius 2020; Rowland 2004); and the four individual surveys in Sluydts 2016, we used the cluster‐adjusted OR reported by the study authors.

In general, repellents caused a 19% reduction in malaria prevalence (OR 0.81, 95% CI 0.67 to 0.97; 4 studies, 55,366 participants; low‐certainty evidence; Analysis 1.5). While all studies included at least some participants belonging to a high‐risk group, a subgroup analysis revealed that the effect was mostly driven by two studies carried out in refugee camps (McGready 2001; Rowland 2004), in which all participants were classified as high‐risk (OR 0.61, 95% CI 0.44 to 0.86; 2045 participants). These were also the only studies in which nets were not used as a background intervention. This effect was not observed if these studies were excluded from the meta‐analysis (OR 0.90, 95% CI 0.73 to 1.11; 2 studies, 53,321 participants).

Differences between both groups were maintained upon exclusion of studies that were not placebo‐controlled (the four surveys from Sluydts 2016 (OR 0.73, 95% CI 0.58 to 0.91; 34,239 participants)), but not after excluding those with a high risk of bias (McGready 2001; OR 0.82, 95% CI 0.67 to 1.00; 54,469 participants).

Only Agius 2020 and Sluydts 2016 reported PPA data, with Agius 2020 further dividing participants based on whether they used the repellent on a monthly, weekly, or daily basis. For the PPA sensitivity analysis, reported ORs from Agius 2020 were pooled with the reported PPA OR from Sluydts 2016. This evidenced no effect on malaria prevalence (OR 0.81, 95% CI 0.42 to 1.56; 32,824 participants; Analysis 1.6), regardless of whether the data from participants who used the repellent on a monthly (OR 0.86, 95% CI 0.44 to 1.69), weekly (OR 0.80, 95% CI 0.40 to 1.60), or daily basis (OR 0.78, 95% CI 0.39 to 1.57) were incorporated into the meta‐analysis.

1.6. Analysis.

Comparison 1: Topical repellents versus placebo or no intervention, Outcome 6: Malaria prevalence (per‐protocol analysis)

Anaemia

One included study investigated the impact on anaemia (McGready 2001). Topical repellents had no impact on this outcome (OR 1.12, 95% CI 0.81 to 1.55; 1 study, 587 participants; Analysis 1.7).

1.7. Analysis.

Comparison 1: Topical repellents versus placebo or no intervention, Outcome 7: Anaemia

Time to first infection

One included study reported this outcome (Chen‐Hussey 2013). They found no effect of topical repellents (intervention group: mean time to first infection: 4 months, range 0.9 to 7.5 months; control group: 3.9 months, range 0.7 to 7.5 months).

All‐cause fever

One study investigated the impact on all‐cause fever (Hill 2007). Participants who used topical repellents were less than half as likely to develop a fever when compared to participant in the control arm (IRR 0.42, 95% CI 0.32 to 0.56; 1 study, 4008 participants; Analysis 1.8).

1.8. Analysis.

Comparison 1: Topical repellents versus placebo or no intervention, Outcome 8: All‐cause fever

Incidence of severe malaria

No results were found for this outcome.

Malaria‐related mortality

No results were found for this outcome.

Adherence to the regular usage of the intervention

All included studies measured the adherence to the intervention (Table 3; Analysis 1.9). However, the way this was measured varied substantially and results were diverse.

2. Adherence to the intervention (Analysis 1.9).

Study	Regular use % (self‐reported)	Regular use % (observed)a
Agius 2020	NA	56
Chen‐Hussey 2013	61.3	47.4
Gryseels 2015b	34	7.9
Hill 2007	NA	98.5
Kroeger 1997	81.9 (Ecuador), 91.3 (Peru)	NA
McGready 2001	90.5	84.6
Rowland 2004	95	NA
Sangoro 2014a	NA	99.2c
Sluydts 2016b	72 and 69 (successive surveys)	NA

^aMeasured using different strategies:
Agius 2020 reported by individual village health workers at the village level; details not provided
Chen‐Hussey 2013 observed by carrying out random sniff‐checks
Hill 2007 observed by carrying out sniff‐checks
McGready 2001 observed via random house visits, not specified

^bGryseels 2015 and Sluydts 2016 report results from the same study, but Sluydts 2016 only measured adherence based on usage the night before the survey, while Gryseels 2015 evaluated usage throughout the whole study period. The determination of observed adherence in this study also included unannounced visits to randomly selected houses to perform sniff checks and observe the correct application of the repellent.

^cSangoro 2014a estimated based on the number of participants included in the per‐protocol analysis (PPA), compared to the intention‐to‐treat (ITT).

1.9. Analysis.

Comparison 1: Topical repellents versus placebo or no intervention, Outcome 9: Adherence to the intervention

Adherence to the intervention
Study	Use % (self‐reported)	Use % (Confirmed)
Agius 2020	NA	56%
Chen‐Hussey 2013	61.3%	47.4%
Gryseels 2015	34%	7.9%
Hill 2007	NA	98.5%
Kroeger 1997	81.9% coverage household members in Ecudaor; 91.3% in Peru
McGready 2001	90.5%	84.6%
Rowland 2004	95%	NA
Sangoro 2014a	NA	99.2%
Sluydts 2016	72% and 69% in successive surveys	NA

In three studies, adherence was exclusively self‐reported by participants (Kroeger 1997; Rowland 2004; Sluydts 2016). In another three studies, it was exclusively reported through observations by study workers during field visits (Agius 2020; Hill 2007; Sangoro 2014). In another three studies, authors reported both self‐reported, and observed adherence (Chen‐Hussey 2013; Gryseels 2015; McGready 2001). The methods used to measure observed adherence were: estimating the weight of returned repellent bottles (Chen‐Hussey 2013; Hill 2007), random sniff checks to smell the arms of participants to determine if the repellent had been applied (Chen‐Hussey 2013; Gryseels 2015; Hill 2007; McGready 2001), or counting the number of bottles issued to individual households (Sangoro 2014).

Four studies reported a very high adherence in the intervention group: Hill 2007 (98.5%), Kroeger 1997 (81.9% in Ecuador and 91.3% in Peru), McGready 2001 (90.5%), and Rowland 2004 (95%). Adherence was not directly reported by Sangoro 2014a, but was estimated to be of approximately 99.2%, based on differences between the samples used in the ITT and PPA analyses. There was a difference between self‐reported and observed adherence in Chen‐Hussey 2013 (self‐reported 61.3%, observed 47.4%); Gryseels 2015 (self‐reported 34%, observed 7.9%); and McGready 2001 (self‐reported 90.5%, observed 84.6%).

The adherence reported by Sluydts 2016 (72% and 69% in two successive self‐reported surveys) was also different from that reported by Gryseels 2015 (self‐reported 34%, observed 7.9%), even though both papers refer to the same clinical trial. These differences are likely explained by the way the question was asked in the two questionnaires; Sluydts 2016 asked if participants had used the repellent the day before, and Gryseels 2015 asked about the continuous use of the lotion throughout the entire study period.

Human biting rate

This was only reported by Kroeger 1997, as 9.5 bites per hour in participants who received the intervention and remained resting, and 15.4 in those who exercised; compared to 40 bites per hour in controls (P < 0.01).

Entomological inoculation rate

No results were found for this outcome.

Sporozoite rate

No results were found for this outcome.

Human blood index

No results were found for this outcome.

Discussion

Summary of main results

See Table 1.

Our results suggest that the effect of topical repellents on malaria incidence and prevalence is being driven by reductions in participants from high‐risk groups without access to other vector control interventions. These results are based on five cluster‐randomized controlled trials (cRCTs) and one RCT. None of the studies included in this review cleared the presence of hypnozoites before starting. Therefore, the meta‐analysis only accounted for cases and infections caused by P falciparum.

The effect of topical repellents on malaria incidence

Topical repellents may make little or no difference to the incidence of clinical malaria and infection with P falciparum when these outcomes are evaluated separately. However, these results were based on only two cRCTs and one RCT measuring infection incidence, and one cRCT measuring clinical case incidence. When data from all these studies were pooled and treated as a single malaria incidence outcome, we observed a slight reduction in malaria incidence (6 fewer events per 1000 person‐years in the intervention arm, incidence rate ratios (IRR) 0.74, 95% confidence interval (CI) 0.56 to 0.98), but the certainty of evidence for this outcome was low.

We observed no effect of the intervention when we disaggregated studies based on the use of other background interventions, although the only study in which these were not used contributed to 56% of the overall weight of the meta‐analysis (McGready 2001).

A subgroup analysis, grouping studies by whether they included participants who belonged to high‐risk groups, failed to show an effect on either group. It is important to note that Hill 2007 was the only study we considered did not include high‐risk populations, since these were not explicitly described by the authors. Nonetheless, it is possible that some of the participants included in the study participated in high‐risk activities, given the location of the studied communities near forest areas.

The effect of topical repellents on malaria prevalence

Topical repellents may also slightly reduce P falciparum prevalence, with two fewer events per 1000 participants in the intervention arm (odds ratio (OR) 0.81, 95% CI 0.67 to 0.97). However, the certainty of evidence for this outcome was also low.

A subgroup analysis revealed that this effect was mostly driven by two studies carried out in refugee camps (OR 0.61, 95% CI 0.44 to 0.86), which accounted for 27.7% of the weight in the meta‐analysis, and in which no long‐lasting insecticide treated nets (LLINs) were being used at the time (McGready 2001; Rowland 2004).

We are unable to determine if this beneficial effect would still be observed with the concurrent use of LLINs, which currently represents a standard malaria vector control intervention.

Adverse events associated with topical repellents

Adverse events associated with the use of topical repellents were rare, with an overall prevalence of 0.6% among 47,515 participants receiving interventions. In all these cases, adverse events were limited to mild skin reactions. Three studies reported a prevalence of adverse events above 1%. Participants from McGready 2001 were pregnant women living in a refugee camp, and might have been exposed to certain confounding factors that explained their higher prevalence of adverse events. Similarly, an unspecified proportion of skin reactions described by Kroeger 1997 were attributed to ectoparasite infestation. Data on adverse events were not included in the meta‐analysis, since only Chen‐Hussey 2013 reported adverse events in both intervention arms. Results from this study suggest that these were linked to the use of repellents (prevalence of 3.8% in the intervention arm and 3.2% in controls, P = 0.029).

The effect of adherence to the use of topical repellents

The influence of adherence to the intervention on the effect of topical repellents on malaria incidence could not be demonstrated in this review, as data from most studies that reported a (per‐protocol analysis) PPA could not be included in the meta‐analysis. Nonetheless, PPA results reported by Agius 2020 and Sluydts 2016 indicated that the regular use of repellents did not lead to a reduction of malaria prevalence in those studies.

The methods used to measure adherence varied across studies, and might not reflect the real compliance of participants. Measuring the adherence to the correct use of topical repellents is complex and logistically difficult, considering the lack of a standardized method to do so, and the varying degrees to which a participant can comply with instructions provided by study workers. This is demonstrated by the diverse estimates derived from different inter‐ and even intra‐study analysis.

In general, adherence observed by study workers was lower than that self‐reported by participants. A particularly large difference between these two outcomes was reported in Gryseels 2015 (self‐reported adherence of 34% versus observed adherence of 7.9%), suggesting that self‐reported adherence is likely inflated due to respondent bias. This publication also thoroughly described the methods used to measure observed adherence, which included unannounced visits to randomly selected houses to perform sniff checks and observe the correct application of the repellent. These are likely to provide the most objective ascertainment of the real level of adherence in a cohort.

The effect of topical repellents on other outcomes

An effect of topical repellents was not observed for anaemia (McGready 2001), or the time to first malaria infection (Chen‐Hussey 2013). However, each one of these outcomes was reported by a single study. A 58% reduction in all‐cause fever was reported by Hill 2007. Reductions in the human biting rate of mosquitoes were reported by Kroeger 1997, and associated with the use of topical repellents. None of the included studies reported data for the effect of the intervention on the entomological inoculation rate, the sporozoite rate, or the human blood index of mosquitoes.

Overall completeness and applicability of evidence

Most studies included participants from all age groups. However, as previously noted, one study only included pregnant women living in a refugee camp (McGready 2001). Given the significant contribution of this study to the results of our meta‐analysis, it is possible that our findings are not generalizable to other groups.

The results from studies that included participants in high‐risk groups accounted for 98.2% of the weight in the meta‐analysis of malaria incidence.

Similarly, all the studies that evaluated the effects on malaria prevalence included participants belonging to at least one high‐risk group. The two studies carried out in refugee camps contributed to 28% of the weight of the effect on this outcome, despite having a disproportionally smaller population than the rest (McGready 2001; Rowland 2004). These results suggest that the observed reduction in malaria transmission is likely more important among these high‐risk groups, and not generalizable to the rest of the population.

Studies included in the meta‐analyses were carried out in diverse geographical settings, including Bolivia (Hill 2007), Cambodia (Sluydts 2016), Laos DPR (Chen‐Hussey 2013), Myanmar (Agius 2020), Pakistan (Rowland 2004), and Thailand (McGready 2001). Only Sangoro 2014a was carried out in Africa (Tanzania), but we did not include its results in the meta‐analysis because we were unable to extract the necessary information on cluster‐adjusted estimates of effect.

The baseline malaria incidence (calculated from the number of cases/100 person‐years in the control arm) of most studies was low (39.8 cases per 100 person‐years in McGready 2001; 2.18 in Sluydts 2016; 1.5 in Chen‐Hussey 2013; and 0.96 in Hill 2007). Furthermore, these incidence rates were lower than those used to estimate corresponding sample sizes (5.0 in Sluydts 2016; 2.0 in Chen‐Hussey 2013; 10.0 in Hill 2007), indicating that most of these studies were underpowered to detect changes in the outcomes of interest.

A similar situation was observed with studies reporting prevalence; most studies reported prevalences below those used to estimate their sample sizes: reported prevalence of 8.9% versus 12% used for sample size calculations (Rowland 2004), and 1.62% versus 5% (Sluydts 2016). The reported malaria prevalence in Agius 2020 was 0.05% when using rapid diagnostic tests (RDTs) to measure the outcome (versus 1% used for sample size calculation), but rose to 1.74% when using polymerase chain reactions (PCRs), highlighting the advantages of this technique to measure infection outcomes in trials taking place in areas of low endemicity. McGready 2001 did not report the assumed transmission estimate used to calculate the study sample size, but is less likely to be underpowered, considering the high baseline incidence (39.8 cases per 100 person‐years) and prevalence (11.38%) reported by the trial authors.

The amount by which studies were underpowered was worsened by the use of large clusters, as was seen in two studies: Agius 2020 (average cluster size: 2230 participants, design effect: 345.8) and Sluydts 2016 (average cluster size: 498 participants, design effect: 15.62). While this was partially addressed by including the cluster‐adjusted OR reported by authors in the meta‐analysis, rather than using the design effect to manually adjust the reported prevalences, it illustrates the problems derived from using this experimental design in low‐endemicity areas. In both studies, this was a direct consequence of randomizing entire villages, as already noticed in a previous review (Maia 2018). We believe that randomizing households could pose a feasible alternative to limit the effect of cluster adjustment on effective sample size estimations, particularly in settings of low endemicity.

Most included studies in this review were cRCTs. This design results in the effect being measured at the cluster level (community protection) rather than at the individual level (personal protection). To understand whether topical repellents provide protection at the individual level, adherence can be evaluated and considered when measuring protective effect. However, at the cluster‐level, it is virtually impossible to differentiate the lack of effect derived from poor adherence, from that caused by a poor protective effect by the intervention itself. In addition, individually randomized studies with adequate measurement of observed adherence would be required to determine whether repellents do offer personal protection. Nonetheless, these studies must consider the potential diversion of mosquitoes to unprotected participants, and take measures to quantify this phenomenon in their experimental design.

Certainty of the evidence

We assessed the evidence supporting the results of the primary outcomes to be of low‐certainty. We downgraded the certainty of the evidence for incidence of malaria infection by two levels due to the high risk of bias of two studies (Hill 2007; McGready 2001), and the indirectness of the population of one (McGready 2001 exclusively included pregnant women), which contributed to 55.6% of the weight in the meta‐analysis.

We downgraded the certainty of the evidence for the incidence of malaria cases by two levels due to high risk of bias and imprecision. The only cRCT included in this outcome failed to administer a placebo to the control group, and presented few events and CIs that crossed the point of no effect (Sluydts 2016).

We also graded the evidence for the malaria case and infection incidence outcome, which pooled data from the two previous ones, as low‐certainty.

Given its relevance to the conclusions of the review, we assessed the certainty of the evidence for malaria prevalence (secondary outcome), and also classified it as low‐certainty evidence. We downgraded the evidence by two levels due to the risk of bias of a study that failed to provide a placebo intervention, despite using a stepped‐wedge design that allowed participants to know when the intervention was applied (Agius 2020); and the indirectness of the population of McGready 2001.

Potential biases in the review process

We attempted to minimize bias in the review process by conducting a comprehensive search of published and unpublished literature, without language restrictions. Two review authors, who were not part of the author team of the included papers, independently screened abstracts, and assessed the risk of bias, and four extracted data. We resolved discrepancies by involving a fifth review author. We were unable to create funnel plots to assess reporting biases, since fewer than 10 RCTs/cRCTs met the inclusion criteria.

Agreements and disagreements with other studies or reviews

A systematic review and meta‐analysis carried out by Wilson 2014 suggested that topical repellents are unlikely to provide protection against malaria, but that better designed studies were necessary before reaching a conclusion. A similar work by Maia 2018 concluded that there was insufficient evidence to determine if the use of topical repellents had a beneficial effect on malaria transmission, but that this intervention could have advantages in refugee camps and disaster settings. Our results agree with these, as well as with recently published data, indicating a benefit of topical repellents in refugee camps (Messenger 2023), based on a meta‐analysis that also included McGready 2001 and Rowland 2004.

Authors' conclusions

Implications for practice.

There is insufficient evidence to conclude that topical repellents can prevent malaria in settings where other vector control interventions are in place.

Our results suggest that repellents could have a modest beneficial effect among certain groups with poor access to these interventions, particularly refugees. This result agrees with another recently published systematic review, suggesting their potential usefulness in the context of humanitarian crises (Messenger 2023). Nonetheless, reductions in malaria incidence and prevalence were marginal, and we failed to see them in settings where nets or indoor residual spraying were widely available. The real effect of the intervention might also be different from our estimate, since the certainty of evidence for all the outcomes included in this Cochrane Review was low.

Implications for research.

Similar to previous reviews (Maia 2018; Wilson 2014), we conclude that more well‐designed trials that evaluate the use of topical repellents could provide further insight into their real effect on reducing malaria transmission. For instance, while cluster‐randomized controlled trials (cRCTs) can provide estimates of the effect of topical repellents on malaria prevention at the community‐level, their implementation in areas of low transmission is challenging, given the large sample sizes required to design adequately powered studies. Studies like Rowland 2004, which used randomization at the household level, rather than larger clusters at the village/hamlet‐level, could help reduce design effects, potentially contributing to more adequately powered trials. Individually‐randomized studies could also be an alternative, as they are usually less resource‐demanding and provide information on personal protection. However, this is likely to produce other problems, such as contamination between the trial arms and diversion of vectors to unprotected participants, which should not be overlooked.

All trials evaluating the use of topical repellents should be placebo‐controlled, due to the high risk of performance bias; and include the use of long‐lasting insecticide‐treated nets as a background intervention. These trials should focus on groups categorized as presenting a high risk of malaria infection, according to the criteria outlined in this review. Similarly, clearance of hypnozoites prior to study start should be included in the protocols of studies measuring the effect of topical repellents on infection caused by P vivax and P ovale, which are the predominant species in many regions where this intervention might be particularly useful. If studies include participants from groups considered high‐risk, the results should be disaggregated and reported with independent effect size estimations for high‐ and non‐high risk groups.

Despite this, existing evidence indicates that topical repellents are unlikely to provide substantial protection to groups already benefiting from other malaria vector control interventions.

History

Protocol first published: Issue 1, 2022

Risk of bias

Risk of bias for analysis 1.3 Malaria incidence (by risk population).

Study

Bias

Randomisation process

Deviations from intended interventions

Missing outcome data

Measurement of the outcome

Selection of the reported results

Overall

Authors' judgement

Support for judgement

Authors' judgement

Support for judgement

Authors' judgement

Support for judgement

Authors' judgement

Support for judgement

Authors' judgement

Support for judgement

Authors' judgement

Support for judgement

Subgroup 1.3.1 High‐risk population

Chen‐Hussey 2013

Low risk of bias

The randomization process seems adequate and while baseline malaria rates were slightly higher in the placebo group, this was non‐significant. Overall, the households in the two treatment groups were very similar despite wide variation between households.

Low risk of bias

It is likely that despite using an apparently adequate placebo, participants might have been aware of their treatment arm. However, outcome assesors, as well as those delivering interventions remained blinded and it is unlikely that there were any major deviations from the intended interventions

Low risk of bias

The proportion of withdrawals remained balanced between the study arms, and it did not affect baseline characteristics of the cohort.

Low risk of bias

Infection incidence was measured using RCTs and molecular mehtod performed in a single institution

Low risk of bias

The data analysis plan was defined before the randomisation of participants. Nonetheless, the low malaria incidende meant that the originally planned Poisson analysis was substitued by a survival analysis. The method, however, seems to be appropriate

Low risk of bias

Despite participants probably being capable of determining whin intervention group they were in, this is unlikely to have had a major effect on the way the intervention was delivered. Changes in the originally planned analysis were well justified and appropriate.

McGready 2001

Some concerns

The authors do not provide any information on how participants were allocated to either group, making it difficult to understand if any bias was introduced.

Low risk of bias

There does not seem to be important deviations from the protocol

High risk of bias

A significant proportion of women abandoned the study, either because they gave birth, or because the presented health problems, albeit probably not related tot he formulation. Furthermore, the number of withdrawals per group is not provided, making it difficult to interpret its effect.

Low risk of bias

Malaria diagnosis was performed via RDTs

Some concerns

The study report is extremely short and no information on pre‐specified analysis is provided.

High risk of bias

No details are provided regarding the nature of the randomization process, or the effect that a significant number of withdrawals might have had on the reported results.

Sluydts 2016

Some concerns

There was a higher proportion of participants with high risk of malaria exposure in the intervention group (15 vs 9%) although researchers indicate this difference was non‐significant.

Some concerns

Control groups received no placebo lotion. Since villages were located close to each other, it is likely that participants were aware of being in a control group. Similarly, field workers were probably aware too, since they carried out visits to confirm the use of the repellent solution.

Low risk of bias

This was malaria case incidence, which means that the outcome was only measured in participants that sought medical care. The passive case detection was balanced in both groups.

Low risk of bias

Malaria case incidence was reported by local health authorities, but the methods used for the diagnosis or confirmation of infection are not reported. Since both groups received the same medical care, it is unlikely that there were differences in the way the outcome was measured.

Low risk of bias

The protocol of the study is available upon request and matches the analysis carried out.

Some concerns

There are concerns regarding the presence of imbalances in the study groups, as well as the lack of a placebo in the control group.

Subgroup 1.3.2 Non‐high risk population

Hill 2007

High risk of bias

Participants used a sequential allocation procedure in which they were sequentially allocated to either A or B group. There is no information on whether the allocation sequence was concealed until all clusters were assigned to the intervention. Participants in the repellent group were probably of a higher socioeconomic status than those in the placebo group.

Low risk of bias

Participants were provided lotions that looked and felt the same. However, since the tubes were only marked as A, B, it seems to have been relatively easy to determine which one was the intervention, and which was the control. Non‐adherence to the protocol was low (1.5%), and similar between the different groups.

Low risk of bias

Although not disclosed, the proportion of data lost to follow up seems to be balanced between both groups.

Low risk of bias

Researchers assesed malaria infection incidence monthly, using RDTs.

Low risk of bias

Authors present results for the outcomes specified in the methods section. There is no available protocol.

High risk of bias

There are significant concerns regarding the randomization process used in this study.

Appendices

Appendix 1. Detailed search strategies for electronic databases

CENTRAL

Issue 1 of 12, January 2023

#1 malaria

#2 (vector or vectors or mosquito or anopheles):ti,ab,kw

#3 MeSH descriptor: [Insect Vectors] explode all trees

#4 MeSH descriptor: [Insect Bites and Stings] explode all trees

#5 #2 or #3 or #4

#6 #1 and #5

#7 MeSH descriptor: [Mosquito Control] explode all trees

#8 #6 or #7

#9 MeSH descriptor: [Insect Repellents] explode all trees

#10 (repel*):ti,ab,kw

#11 (("Insecticide treated clothing" or ITC)):ti,ab,kw

#12 lotion* or gel or gels

#13 wipe* or soap* or cream*

#14 roll‐on

#15 #9 or #10 or #11 or #12 or #13

#16 #15 and #8

Ovid MEDLINE(R) and In‐Process, In‐Data‐Review & Other Non‐Indexed Citations (1946 to present)
1 Malaria/
2 malaria.tw.
3 1 or 2
4 Insect Vectors/
5 vector*.tw.
6 mosquito*.mp. or Mosquito Vectors/
7 Anopheles/ or anopheles.tw.
8 Insect Bites and Stings/
9 4 or 5 or 6 or 7 or 8
10 3 and 9
11 Mosquito Control/
12 10 or 11
13 Insect Repellents/
14 topical repel*.tw.
15 repellent*.tw.
16 (lotion* or gel or gels or roll‐on* or wipe* or soap* or cream*).tw.
17 (Spray* and skin).mp.
18 personal protection.mp.
19 13 or 14 or 15 or 16 or 17 or 18
20 12 and 19
21 randomized controlled trial/
22 Controlled Clinical Trial/
23 (randomized or placebo or randomly or trial or groups).tw.
24 Interrupted Time Series Analysis/
25 Controlled Before‐After Studies/
26 Cross‐Over Studies/
27 21 or 22 or 23 or 24 or 25 or 26
28 20 and 27

Embase 1947‐Present, updated daily

1 Malaria/

2 malaria.tw.

3 1 or 2

4 insect vector/

5 vector*.tw.

6 mosquito*.mp. or mosquito vector/

7 Anopheles/ or anopheles.tw.

8 Insect Bite/ or insect sting/

9 4 or 5 or 6 or 7 or 8

10 3 and 9

11 Mosquito Control/

12 10 or 11

13 insect repellent/

14 topical repel*.tw.

15 repellent*.tw.

16 ("Insecticide treated clothing" or ITC).tw.

17 (lotion* or gel or gels or roll‐on* or wipe* or soap* or cream*).tw.

18 (Spray* and skin).mp.

19 personal protection.mp.

20 13 or 14 or 15 or 16 or 17 or 18 or 19

21 12 and 20

22 (random* or factorial* or placebo* or assign* or allocat* or crossover*).tw.

23 (control* adj group*).tw.

24 (trial* and (control* or comparative)).tw.

25 ((blind* or mask*) and (single or double or triple or treble)).tw.

26 (randomized or randomly or RTC).ab.

27 crossover procedure/ or double blind procedure/ or single blind procedure/

28 placebo/ or clinical trial/ or parallel design/ or Latin square design/

29 time series analysis/

30 22 or 23 or 24 or 25 or 26 or 27 or 28 or 29

31 21 and 30

LILACS

Search on: malaria vector$ or mosquito$ or anopheles [Words] and topical repellent$ or personal repellent$ or soap or lotion$ or cream$ or cloth$ [Words]

CAB Abstracts (Web of Science)

#20 #19 AND #14

#19 #18 OR #17 OR #16 OR #15

#18 ("Controlled Before‐After" OR "Interrupted Time Series") (Abstract)

#17 randomized or placebo or randomly or trial or groups (Abstract)

#16 controlled clinical trial (Topic)

#15 randomized controlled trial (Topic)

#14 #13 AND #7

#13 #12 OR #11 OR #10 OR #9 OR #8

#12 personal near/2 protection (Topic)

#11 Spray* near/2 skin (Topic)

#10 (lotion* or gel or gels or roll‐on* or wipe* or soap* or cream*) (Topic)

#9 ("Insecticide treated clothing" or ITC) (Topic)

#8 insect repellent* or topical repel* (Topic)

#7 #1 OR #6

#6 #5 AND #4

#5 malaria (Topic)

#4 #2 OR #3

#3 insect near/2 bite* (Topic)

#2 mosquito* or mosquito near/2 vector or anopheles (Topic)

#1 malaria near/2 vector* (Topic)

Clinicaltrials.gov; WHO ICTRP

mosquito* and repel*

US Armed Forces Pest Management Board

mosquito repellent*

French Institute of Research for Development's Horizon Pleins Textes

mosquito repellents

Data and analyses

Comparison 1. Topical repellents versus placebo or no intervention.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Malaria incidence (case and infection)	4	61651	Rate Ratio (IV, Fixed, 95% CI)	0.74 [0.56, 0.98]
1.1.1 Case incidence	1	48838	Rate Ratio (IV, Fixed, 95% CI)	0.66 [0.32, 1.36]
1.1.2 Infection incidence	3	12813	Rate Ratio (IV, Fixed, 95% CI)	0.76 [0.56, 1.02]
1.2 Malaria incidence (use of other interventions)	4	61651	Rate Ratio (IV, Fixed, 95% CI)	0.74 [0.56, 0.98]
1.2.1 Use of bednets and other interventions	3	60754	Rate Ratio (IV, Fixed, 95% CI)	0.78 [0.51, 1.18]
1.2.2 Without use of bednets and other interventions	1	897	Rate Ratio (IV, Fixed, 95% CI)	0.71 [0.49, 1.04]
1.3 Malaria incidence (by risk population)	4	61651	Rate Ratio (IV, Fixed, 95% CI)	0.74 [0.56, 0.98]
1.3.1 High‐risk population	3	57643	Rate Ratio (IV, Fixed, 95% CI)	0.76 [0.58, 1.01]
1.3.2 Non‐high risk population	1	4008	Rate Ratio (IV, Fixed, 95% CI)	0.18 [0.02, 1.40]
1.4 Reported adverse events in the topical repellents arm of included studies	0		Other data	No numeric data
1.5 Malaria prevalence	4	55366	Odds Ratio (IV, Fixed, 95% CI)	0.81 [0.67, 0.97]
1.5.1 No refugees	2	53321	Odds Ratio (IV, Fixed, 95% CI)	0.90 [0.73, 1.11]
1.5.2 Refugees	2	2045	Odds Ratio (IV, Fixed, 95% CI)	0.61 [0.44, 0.86]
1.6 Malaria prevalence (per‐protocol analysis)	2		Odds Ratio (IV, Fixed, 95% CI)	0.81 [0.42, 1.56]
1.7 Anaemia	1	587	Odds Ratio (M‐H, Fixed, 95% CI)	1.12 [0.81, 1.55]
1.8 All‐cause fever	1	4008	Rate Ratio (IV, Fixed, 95% CI)	0.42 [0.32, 0.56]
1.9 Adherence to the intervention	0		Other data	No numeric data

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Agius 2020.

Study characteristics
Methods	cRCT (stepped‐wedge) Unit of randomization: clusters of villages ICC: 0.15 Trial duration: approximately 15 months
Participants	Adults or children living in malaria‐endemic regions High‐risk population (forest workers) Endemic region for P vivax Participants not screened for hypnozoites
Interventions	Topical repellent: 12% N, N‐diethylbenzamide cream Controls: no intervention Co‐interventions: LLINs Treatment arms: since this was a stepped wedge trial, all clusters moved from control to intervention. There were 14 clusters (116 villages, approx. 31,016 people)
Outcomes	Malaria prevalence Incidence of recorded adverse events (including skin irritation, local pain, eye irritation, irritation of upper airways, nausea, vomiting, headaches, dizziness or confusion, allergic or anaphylactic reactions, and systemic toxicity) Adherence to regular usage of the intervention (defined based on recommendations provided by researchers to participants of individual trials)
Notes	Malaria case incidence is mentioned as an outcome in the paper, but not reported Carried out in Myanmar Funded by: Three Millennium Development Goal, Australian Research Council, Australian Centre for Research Excellence in Malaria Elimination

Chen‐Hussey 2013.

Study characteristics
Methods	cRCT Unit of randomization: cluster of houses ICC: not reported Trial duration: 1.5 years (average of 6.4 months follow‐up per cluster)
Participants	Adults or children living in malaria‐endemic regions High risk population (agricultural workers sleeping outdoors) Endemic region for P vivax Participants not screened for hypnozoites
Interventions	Topical repellent: DEET 15% Controls: placebo lotion Co‐interventions: LLINs Treatment arms: Repellent arm: 795 households; 3972 participants Placebo arm: 802 households; 4008 participants
Outcomes	Malaria infection incidence: new Plasmodium spp. infections confirmed through thick or thin blood smears, RDTs, or polymerase chain reaction (PCR) Incidence of recorded adverse events (including skin irritation, local pain, eye irritation, irritation of upper airways, nausea, vomiting, headaches, dizziness or confusion, allergic or anaphylactic reactions, and systemic toxicity) Time to first infection (days) Adherence to regular usage of the intervention (defined based on recommendations provided by researchers to participants of individual trials)
Notes	Conducted in Laos PDR Funded by the Department of Homeland Security, and Fogarty International Center, National Institutes of Health

Gryseels 2015.

Study characteristics
Methods	Sub‐analysis of Sluydts 2016 (see main study for methods)
Participants	Sub‐analysis of Sluydts 2016 (see main study for participant description)
Interventions	Sub‐analysis of Sluydts 2016 (see main study for intervention description)
Outcomes	Adherence to regular usage of the intervention (defined based on recommendations provided by researchers to participants of individual trials)
Notes	This paper reports the adherence to repellent use from participants recruited to the Sluydts 2016 trial. Carried out in Cambodia Funded by: Bill and Melinda Gates foundation, Belgian Cooperation

Hill 2007.

Study characteristics
Methods	cRCT Unit of randomization: cluster of houses ICC: not reported Trial duration: 7 months (treatment period of 4 months)
Participants	Adults or children living in malaria‐endemic regions Non‐high risk population Endemic region for P vivax Participants not screened for hypnozoites
Interventions	Topical repellent: PMD: 30% + eucalyptus extract Controls: placebo lotion (0.1% clove oil used as placebo) Co‐interventions: LLINs Treatment arms: Intervention: 424 households (1967 individuals) Control: 436 households (2041 individuals)
Outcomes	Malaria infection incidence: new Plasmodium spp. infections confirmed through thick or thin blood smears, RDTs, or polymerase chain reaction (PCR) Incidence of recorded adverse events (including skin irritation, local pain, eye irritation, irritation of upper airways, nausea, vomiting, headaches, dizziness or confusion, allergic or anaphylactic reactions, and systemic toxicity). All‐cause fever Adherence to regular usage of the intervention (defined based on recommendations provided by researchers to participants of individual trials)
Notes	Carried out in Bolivia Funded by Gates Malaria Partnership grant from London School of Hygiene and Tropical Medicine

Kroeger 1997.

Study characteristics
Methods	cRCT Unit of randomization: villages ICC: not reported Trial duration: 7 months (treatment period of 6 months)
Participants	Adults or children living in malaria‐endemic regions Non‐high risk population Endemic region for P vivax Participants not screened for hypnozoites
Interventions	Topical repellent: soap with 0.5% DEET and 20% permethrin Controls: no intervention Co‐interventions: none Treatment arms: Intervention: 9 communities (total population not reported) Control: 9 villages (total population not reported)
Outcomes	Incidence of recorded adverse events (including skin irritation, local pain, eye irritation, irritation of upper airways, nausea, vomiting, headaches, dizziness or confusion, allergic or anaphylactic reactions, and systemic toxicity) Adherence to regular usage of the intervention (defined based on recommendations provided by researchers to participants of individual trials) Human biting rate (HBR)
Notes	Malaria case incidence was listed as an outcome, but it was self‐reported, and raw data for this outcome is not presented Carried out in Ecuador and Peru Funded by: Commission of European Communities, Catholic Organization for Development

McGready 2001.

Study characteristics
Methods	RCT Trial duration: 17 months (treatment period: 18 weeks)
Participants	Adults or children living in malaria‐endemic regions (women 3 to 7 months pregnant) High‐risk population (refugees) Endemic region for P vivax Participants not screened for hypnozoites
Interventions	Topical repellent: 20% DEET added to Thanaka (popular local cosmetic) Controls: Thanaka alone Co‐intervention: none Treatment arms: Intervention: 449 Controls: 448
Outcomes	Malaria infection incidence: new Plasmodium spp. infections confirmed through thick or thin blood smears, RDTs, or polymerase chain reaction (PCR) Incidence of recorded adverse events (including skin irritation, local pain, eye irritation, irritation of upper airways, nausea, vomiting, headaches, dizziness or confusion, allergic or anaphylactic reactions, and systemic toxicity) Malaria prevalence Anaemia (haemoglobin < 8 g/dL) Adherence to regular usage of the intervention (defined based on recommendations provided by researchers to participants of individual trials)
Notes	Carried out in camps for displaced people of the Karen ethnic minority in Thailand Funded by the Danish Bilharziasis Laboratory and was part of the Wellcome‐Mahidol University of Oxford Tropical Medicine Research Programme funded by the Wellcome Trust

Rowland 2004.

Study characteristics
Methods	cRCT Unit of randomization: cluster of households ICC: 0.04 Trial duration: 7 months (treatment period of 6 months)
Participants	Adults or children living in malaria‐endemic regions High‐risk population (refugees) Endemic region for P vivax Participants not screened for hypnozoites
Interventions	Topical repellent: Mosbar soap (20% DEET + 0.5% permethrin) Controls: placebo lotion Co‐interventions: none Treatment arms: Intervention: 67 households (618 participants) Controls: 60 households (530 participants)
Outcomes	Incidence of recorded adverse events (including skin irritation, local pain, eye irritation, irritation of upper airways, nausea, vomiting, headaches, dizziness or confusion, allergic or anaphylactic reactions, and systemic toxicity) Malaria prevalence Adherence to regular usage of the intervention (defined based on recommendations provided by researchers to participants of individual trials)
Notes	Carried out in a refugee camp with Afghan refugees in malaria‐endemic region of Pakistan Funded by HealthNet International’s Malaria and Leishmaniasis control and research programme

Sangoro 2014a.

Study characteristics
Methods	cRCT Unit of randomization: cluster of households ICC: not reported Trial duration: 14 months
Participants	Adults or children living in malaria endemic regions Non‐high risk population
Interventions	Topical repellent: 15% DEET lotion Controls: placebo lotion Co‐interventions: LLINs Treatment arms: ‐ Intervention: 10 clusters, 468 households and 2224 participants ‐ Controls: 10 clusters, 469 households and 2202 participants
Outcomes	‐ Malaria case incidence: new cases of clinical malaria (caused by Plasmodium spp.) confirmed through blood smears or RDTs ‐ Adherence to regular usage of the intervention (defined based on recommendations provided by researchers to participants of individual trials)
Notes	Carried out in Tanzania. Funded by Population Services International.

Sluydts 2016.

Study characteristics
Methods	cRCT Unit of randomization: cluster made up by individual villages or a group of closely located villages ICC: not reported, retrospectively calculated: 0.0294 Trial duration: 20 months
Participants	Adults or children living in malaria endemic regions High risk population (forest workers) Endemic region for P vivax Participants not screened for hypnozoites
Interventions	Topical repellent: picaridin (KBR3023) Controls: no treatment Picaridin 10% for children < 10 years, and Picaridin 20% in individuals ≥ 10 years Co‐interventions: LLINs Treatment arms: Intervention: 49 clusters, 57 villages (5642 households, 25,051 individuals) Controls: 49 clusters, 56 villages (5287 households, 23,787 individuals)
Outcomes	Malaria case incidence: new cases of clinical malaria (caused by Plasmodium spp.) confirmed through blood smears or RDTs Incidence of recorded adverse events (including skin irritation, local pain, eye irritation, irritation of upper airways, nausea, vomiting, headaches, dizziness or confusion, allergic or anaphylactic reactions, and systemic toxicity) Malaria prevalence Adherence to regular usage of the intervention (defined based on recommendations provided by researchers to participants of individual trials)
Notes	Carried out in Cambodia Funded by: Bill and Melinda Gates Foundation, Belgian Cooperation

cRCT: cluster‐randomized controlled trial; ICC: intra‐class correlation coefficient; LLIN: long‐lasting insecticide‐treated nets; RDTs: rapid diagnostic tests

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Constantini 2004	Inadequate study design; this study is an entomologic trial and does not evaluate the effect of repellents on epidemiological outcomes
Dadzie 2013	The study does not comply with the inclusion criteria for CBA (only one site per arm)
Deressa 2014	The characteristics of the repellent were not properly described
Lindsay 1998	This study is an entomological and acceptability study not measuring malaria epidemiological outcomes.
Maia 2012	The study is an entomological evaluation and does not report outcomes included in this review.
Moore 2007	This study is an entomological evaluation not measuring malaria epidemiological outcomes
Rowland 2004b	Inadequate study design; the paper reports a case‐control study.
Uzzan 2009	This study is an entomological evaluation, and does not measure malaria epidemiological outcomes

Abbreviations: CBA: controlled before‐after study; RCT: randomized controlled trial

Characteristics of ongoing studies [ordered by study ID]

NCT02938975.

Study name	Field efficacy of insecticide treated uniforms and skin repellents for malaria prevention (URCT)
Methods	Cluster‐RCT using a 4‐arm non‐inferiority design with 12 months of follow‐up
Participants	Healthy recruits of the Tanzanian National Service Program JKT Mgambo Camp
Interventions	Ultra 30 insect repellent lotion (30% Lipo DEET) in combination or not with permethrin factory‐treated army combat uniforms
Outcomes	The primary epidemiological outcome will be the incidence of P falciparum malaria through monthly measurement of malaria positivity by direct polymerase chain reaction (PCR) to detect parasite DNA.
Starting date	November 2017
Contact information	Sarah Moore (smoore@ihi.or.tz)
Notes	clinicaltrials.gov/ct2/show/record/NCT02938975

NCT05117567.

Study name	Reducing malaria transmission in forest‐going mobile and migrant populations in Lao PDR and Cambodia
Methods	Stepped‐wedge cluster‐randomized controlled trial (one‐way crossover) with nested mixed methods study. Open label study
Participants	Mobile and migrant individuals aged 18 years and over in selected villages including: Traditional slash‐and‐burn and paddy field farming communities visiting their forest farms (commonly ethnic minority groups) Seasonal agricultural labourers Forest workers in the informal sector (hunters, small‐scale gem/gold miners, people gathering forest products (precious timber, construction timber, rattan/bamboo) Transient or mobile camp residents associated with commercial projects (road/pipeline construction, large‐scale logging, deep seaport projects, etc.) Formal and informal cross‐border migrant workers For qualitative research component, local health stakeholders meeting the following criteria will be eligible: Aged 18 years and over The local health stakeholders, such as health centre staff, Operational District Malaria Supervisor (ODMS), and Provincial Malaria Supervisor (PMS), and basic health staff, such as malaria unit staff in health centres, midwives, health assistants, district health officers and district focal person from CMPE and CNM Health staff members from HPA and Lao malaria community service organizations Estimated enrolment: 5868 participants
Interventions	Combination product: personal protection package A personal protection package that includes long‐lasting insecticidal hammock net (LLIHN), insect repellent (Icaridin), and MMP‐tailored behavioural change communication (BCC) package
Outcomes	Primary outcomes Plasmodium spp. infection diagnosed by RDT (time frame: assessed weekly, longitudinally over 12 months) Change in the number of Plasmodium spp. infections detected by RDT per week per village Secondary outcomes Symptomatic malaria diagnosed by RDT (time frame: assessed weekly, longitudinally over 12 months) Change in the number of symptomatic Plasmodium spp. infections detected by RDT per week per village. Plasmodium spp. infection, determined by polymerase chain reaction (PCR) on RDT cassette samples (time frame: assessed weekly, longitudinally over 12 months) Change in the prevalence of Plasmodium spp. infection, determined by polymerase chain reaction (PCR) from RDT cassette samples Plasmodium spp. infection, determined by polymerase chain reaction (PCR) on dried blood spot samples (time frame: assessed weekly, longitudinally over 12 months) Change in the prevalence of Plasmodium spp. infection, determined by polymerase chain reaction (PCR) from dried blood spot (DBS) samples Plasmodium spp. infections with drug resistance mutations (time frame: assessed weekly, longitudinally over 12 months) Change in the prevalence of Plasmodium spp. infection with drug resistance mutations Prevalence of antibodies to Plasmodium spp. (time frame: assessed weekly, longitudinally over 12 months) Prevalence of antibodies to Plasmodium spp., determined by enzyme‐linked immunosorbent assay (ELISA), from RDT and DBS samples Levels of antibodies to Plasmodium spp. (time frame: assessed weekly, longitudinally over 12 months) Levels of antibodies to Plasmodium spp., determined by ELISA, from RDT and DBS samples Prevalence of antibodies to vector salivary antigens (time frame: assessed weekly, longitudinally over 12 months) Levels of antibody biomarkers of vector exposure Levels of antibodies to vector salivary antigens (time frame: assessed weekly, longitudinally over 12 months) Levels of antibody biomarkers of vector exposure Levels of knowledge, attitude, and practice regarding malaria prevention among MMPs (time frame: at approximately 12 months) Focus group discussions Proportion of survey respondents (MMPs) who accept and are willing to use/did use the personal protection package according to the protocol (time frame: at approximately 12 months)
Starting date	01 December 2021
Contact information	Freya JI Fowkes: +613 8506 2310 freya.fowkes@burnet.edu.au
Notes

Abbreviations: PCR: polymerase chain reaction; RCT: randomized controlled trial.

Differences between protocol and review

In the protocol, we presented malaria infection incidence and malaria case incidence exclusively as individual outcomes (Wagah 2022). Given the small number of studies included in the review, both outcomes were ultimately pooled and treated as a single outcome: malaria case and infection incidence. We performed a subgroup analysis separating both outcomes, and reported both the independent and pooled results in the summary of findings table. Since cluster‐adjusted IRRs and ORs were directly extracted from studies reporting them, we did not use the design effect of the studies to estimate cluster effects. For the same reason, a sensitivity analysis using different estimated ICCs was not necessary.

Contributions of authors

JCGF, MGW, and MFM contributed to the design of the study.

All review authors contributed to the data extraction and analysis of the results.

All review authors contributed to the writing of the draft and definitive versions of the manuscript.

All review authors read and approved the final manuscript version prior to publication.

Sources of support

Internal sources

• Liverpool School of Tropical Medicine, UK

External sources

• Foreign, Commonwealth, and Development Office (FCDO), UK

  Project number 300342‐104

• La Caixa Inphinit Fellowship, Spain

  ID 100010434, fellowship code: LCF/BQ/DI21/11860037

Declarations of interest

JCGF has no known conflicts of interest.

MGW has no known conflicts of interest.

LBA has no known conflicts of interest.

CW has no known conflicts of interest.

MFM has no known conflicts of interest.

New