Structural and community‐level interventions for increasing condom use to prevent the transmission of HIV and other sexually transmitted infections

Description of the condition

HIV is a virus that attacks the immune system, weakening it progressively and leading some years later to the onset of opportunistic infections, a stage that is known as acquired immunodeficiency syndrome (AIDS). If not treated, the natural progression of this syndrome winds up in death (Weiss 1993). HIV's routes of transmission are sexual intercourse through contact of mucosae with semen or vaginal fluids, exchange of intravenous syringes, blood transfusion, hemodialysis, and mother‐to‐child transmission via blood transfer and breastfeeding.

In 2012 the World Health Organization (WHO) and the Joint United Nations Programme on HIV/AIDS (UNAIDS) estimated that globally 2.3 million people were newly infected with HIV, 35.3 million people were already living with the virus, and 1.6 million people died from AIDS‐related causes (UNAIDS 2013). Moreover, the HIV/AIDS pandemic affects disproportionally low‐ and middle‐income countries (LMICs), with several regions in Sub‐Saharan Africa facing the highest burden of the disease (Ortblad 2013).

Since the onset of the pandemic, condoms have been considered one of the most effective biological mechanisms for reducing the transmission of HIV (Holmes 2004). Laboratory and epidemiological studies have demonstrated that condoms are effective against a number of sexually transmitted infections (STIs), such as gonorrhoea, non‐gonococcal urethritis, trichomoniasis, genital herpes (da Silveira 2005), and HIV (Weller 2002). Of the many strategies to reduce infection, promotion of condom use continues to be the keystone of HIV and STI prevention worldwide (Matson 2010). However, the positive impact of condoms may be limited by their inconsistent or low use (Hearst 2004). Several factors are known to be associated with condom non‐use during sexual intercourse, including perception of reduced pleasure (Sarkar 2008; Sunmola 2005), discomfort (Crosby 2005), partner trust (Abdullah 2002; Jadack 1997), use of other contraceptives (Abdullah 2002), fear of revealing HIV status to partners (Medley 2004), and religious beliefs (Thomsen 2004). It has also been suggested that the spread of information about highly active antiretroviral therapy (HAART) reducing the infectiousness of HIV may be a potential hindrance to condom use (Akinyemi 2010); nonetheless, evidence supporting this theory is mixed (Berhan 2012; Crepaz 2004).

The availability of and accessibility to condoms is often a major obstacle to condom use. In most low‐ and middle‐income countries cost is a barrier to purchase condoms among the poor (Essien 2005). Even when cost is not an issue, non‐availability of condoms often results in unprotected sexual practices (Kumar 2006). This problem affects not only LMICs, but also high‐income countries such as the United States, where many African‐Americans are unable to purchase condoms due to financial barriers (Essien 2005). Other authors have also mentioned that in certain societies, the setting in which condoms are provided, commercialized and purchased may hinder their acquisition from possible buyers/consumers (e.g. buying them in a public place (Roth 2001)). Also, in many societies moral values and ethnic and religious factors play an important role in health behaviors and health promotion policies (Koenig 2012). For example, the Catholic Church considers any contraceptive method as a transgression of divine law (HUMANAE VITAE 1968), a mandate that some devout Catholics observe (Lefkowitz 2004).

Moreover, lack of condom usage is also a consequence of limited knowledge about HIV prevention and a low level of education (Dandona 2005). Other social determinants such as women's lack of empowerment, scarce dialogue within couples, domestic violence, and prejudice related to condom use could increase the risk of unsafe sexual practices (Sarkar 2008).

Among female sex workers (FSW), limited knowledge about HIV transmission, a client's refusal to use condoms, or coercion and control by pimps or brothel owners act as barriers to condom use (Xia 2005). For men who have sex with men (MSM), negative community attitudes towards homosexuality combined with legal restrictions on their sexual behavior and official relationships contribute to the marginalization and stigmatization of this community and drive their relationships underground, hindering an easy acquisition of condoms and resulting in risky sexual behavior (Adimora 2010).

It is also important to mention that depending on the local context, prevalence of condom use tends to differ between different sex‐risk‐behavior groups. Several studies have described a tendency of high‐risk groups such as individuals with multiple sexual partners and frequent casual sexual encounters (Maticka‐Tyndale 2012), MSM (Adam 2009; Lau 2008; Reece 2010) and FSWs (Maticka‐Tyndale 2012; Rojanapithayakorn 1996; Suleman 2012) to report higher condom use figures than non‐high‐risk groups (Abdullah 2002; Jones 2012; Reece 2010). The general population tends to present a lower condom use prevalence for reasons such as use of other contraceptive methods (Abdullah 2002; Jones 2012) and a negative perception of condoms, which symbolize promiscuity and unfaithfulness (Maticka‐Tyndale 2012). Consequently, condom promotion interventions may also have a different impact on each of these populations.

Description of the intervention

It is widely understood in the field of public health that social determinants of health, or the circumstances in which individuals are born, live, work and age, play an important role in their health (WHO 2012). Those contexts are shaped by structural determinants such as physical, social, cultural, legal, and organizational factors, and several other aspects of the environment that hinder or promote HIV prevention (Gupta 2008; Sumartojo 2000).

A structural intervention is defined as any public health intervention that aims to improve society's health status by modifying the context in which health‐related risk behavior takes place (Blankenship 2000). When addressing condom use through structural interventions, there are three key concepts to consider (Blankenship 2000). The first one is availability, which focuses on the attitudes, materials, or settings that are essential to reduce exposure to a specific health problem (Blankenship 2000). For example, community campaigns may install condom machines in public places, or implement a policy of 100% condom use in brothels (Blankenship 2000; Charania 2011). The second factor is acceptability, which sets out to modify the preconceptions in a society, allowing people to consider and accept new ideas (Blankenship 2000) mainly through social learning theory (Bandura 1986). Programs to prevent the stigmatization of prostitution, sexual education, and marketing of condoms are examples (Blankenship 2000; Charania 2011). The last factor is accessibility, which aims to improve access to interventions, services and goods in health for those who face unfavourable social or economic conditions (Blankenship 2000). Examples of this concept include distribution of free or low‐cost condoms and the creation of female controlled prevention strategies, such as female condom use (Blankenship 2000; Charania 2011). Therefore, structural interventions usually include changes to policies, environment, or social beliefs, and in many cases also involve the empowerment of minorities and other disenfranchised groups (Adimora 2010).

Structural interventions can be implemented at three levels: individual, group, and community (Charania 2011). Individual and group level interventions target knowledge, attitudes, and behaviors associated with the condom use of the participant in one‐on‐one settings and in existing (e.g. a couple) or recently created groups, respectively (Charania 2011). Community‐level interventions have typically been conceptualized as geographic units (e.g. cities, counties, villages), although alternative conceptualizations of smaller units also exist (e.g. workplaces and schools) (Atienza 2002); the present review was developed based on the former concept. Interventions at a community level directly and indirectly concentrate on knowledge, attitudes and behaviors in relation to the entire community and strongly address the alteration of social norms (Charania 2011). This strategy might have the capacity to modify social determinants of health at all levels in which it is implemented (Swerissen 2004).

How the intervention might work

Previous research on HIV and STI prevention strategies shows that structural interventions have the capacity to initiate significant change in behaviors and attitudes.

For example, Thailand's renowned 100% Condom Program modified availability and accessibility of condoms by empowering sex workers to consistently engage in safe sexual practices via government policies that enforced mandatory condom use (Rojanapithayakorn 1996). Likewise, in the Dominican Republic two non‐governmental organizations (NGOs) investigated in 1995 the adoption of Thailand's initiative into the Dominican setting (Kerrigan 2006). This intervention addressed the three different components of structural interventions at a group and community level. Assessment of this program revealed that condom use did increase after the introduction of this intervention (Kerrigan 2006). Another program, the Sonagachi project in India, focused on community mobilization, aiming to reduce discrimination and stigma against FSWs (Basu 2004).

Dealing with inequitable gender norms, especially those that define masculinity, was an important component of HIV prevention strategies in Brazil (Pulerwitz 2006). In this intervention, the aim was to modify acceptability at a community level. In South Africa an HIV prevention strategy was implemented to modify both accessibility through a local micro‐finance program addressed at women and acceptability by initiating community activism in opposition to women mistreatment (Gupta 2008). Such interventions aim to empower women, decrease domestic violence and enable women to negotiate safe sex (Gupta 2008).

Finally, media campaigns promoting condom use also have the potential to influence norms and change values, expectations, and behaviors by modifying acceptability at a community level (Cohen 2006).

In spite of the cost, financial analyses demonstrate that structural interventions may have the most impact in HIV prevention, and may also generate other social advantages, such as stimulating the economy and making gains in human rights (Hecht 2009). Given the large amount of resources required for individual‐based interventions to be successfully implemented on a broader scale, it has been hypothesized that structural interventions might be an adequate solution since they are more ample in their scope (Cohen 2006).

Why it is important to do this review

Despite the implementation of extensive HIV prevention programs over the past decades, the number of people newly infected with HIV continues to rise in some regions across the globe (UNAIDS 2013). Some studies argue that one of the reasons for this phenomenon is the excessive attention to individual behaviors at the expense of changes to the context in which those behaviors occur (Barnett 2006).

The consensus in recent years that social factors affect HIV transmission as well as risky sexual behaviors has prompted an increasing attention on structural interventions for HIV and other STIs (Gupta 2008). However, few strategies to prevent HIV transmission have implemented structural interventions, mainly due to a lack of clear definition, a poor understanding of technical implementation, and insufficient data to support the effectiveness of such interventions in the context of people living with HIV (Gupta 2008).

Behavioral interventions that aim to encourage a change in individual attitudes toward safer sexual practices such as condom use or reduction of sexual partners play an important role in overall HIV prevention efforts. However, these interventions alone may be inadequate to produce substantial and long‐term reductions in HIV transmission (Coates 2008). As a result, taking into account the limitations of these strategies, structural interventions offer a potentially powerful shift in the manner HIV prevention is implemented (Blankenship 2010).

Despite a growing interest in the topic, there is little evidence on structural interventions at a community level that aim to increase condom use to prevent HIV and other STIs. Policy makers require high quality quantitative information in order to develop and implement prevention programes, as well as to set priorities in research and funding (Johnson 2008). Determining if structural interventions for promoting condom use are effective in halting HIV transmission at a community level is central to the current strategies against HIV, given the number of people whose lives may be improved by this policy. In addition,deciding what the most effective approaches to reducing HIV transmission are will allow wise investment in public health resources.

Criteria for considering studies for this review

Search methods for identification of studies

Data collection and analysis

Description of studies

Risk of bias in included studies

Risk of bias assessment is illustrated in Figure 3 and Figure 4.

Effects of interventions

See: Table 1

Summary of main results

This review addressed the question of whether structural interventions to promote condom use at the community level reduced the incidence or prevalence of HIV and other STIs, assessed through laboratory test, as well as other secondary outcomes such as knowledge, self efficacy and sexual behaviors.

Nine studies meeting all inclusion criteria were obtained. Loss to follow‐up in three studies was the only relevant source of bias. The studies featured condom acceptability by imparting sexual education and promoting social marketing of condoms; accessibility was addressed through provision of free condoms and implementation of micro‐credits. After adjusting for clustering effects, the meta‐analysis demonstrated the intervention(s) to have a non‐significant effect in reducing incidence or prevalence of HIV, nor did they reduce prevalence or incidence of HSV‐2, syphilis, gonorrhea, chlamydia or trichomonas. One individual study (Garcia 2012) proved to be effective in reducing the prevalence of trichomoniasis in the general population when addressing both genders.

In each trial authors reported the possible reasons for the intervention(s) to have not had a significant effect on biological outcomes, including: a 'diluted' intervention in which the results were assessed in some people who were not directly exposed to the intervention (Cowan 2010); social ambivalence toward condoms, thus hindering their use (Cowan 2010; Doyle 2010; Feldblum 2001); attrition bias since some people who received the intervention then left the study, consequently reducing the effect of the intervention (Cowan 2010; Doyle 2010; Garcia 2012); 'secular trend', defined as the exposure of the control groups to other sources of HIV and condom awareness such as media condom campaigns or sex education programs, thus reducing the relative impact of the intervention(s) in the experimental group (Doyle 2010; Gregson 2007; Kamali 2003; NIMH 2010); difficulty in changing behaviors (Doyle 2010); insufficient statistical power (Gregson 2007; Jewkes 2008; Kamali 2003); insufficient follow‐up (Gregson 2007); and self reporting bias related to sexual behaviors (Jewkes 2008; Kamali 2003; Ross 2007).

The results varied substantially regarding secondary outcomes. There was no significant change in the prevalence of respondents with two or more sexual partners in the last year. Although individual trials did not demonstrate any statistically significant change in unprotected sex, the meta‐analysis proved the intervention(s) to be effective; individual trials exploring consistent condom use showed non‐significant results. All meta‐analyses regarding change in knowledge yielded statistically significant results in favor of the intervention(s). Meta‐analysis addressing self efficacy was not feasible, but all three individual studies reported no significant results.

After conducting sensitivity analysis with different ICC values, we found that biological and knowledge‐related outcomes were robust, whereas behavioral outcomes such as condom use and number of sexual partners tended to fluctuate between significant and non‐significant depending on the magnitude of the ICC.

Overall completeness and applicability of evidence

A potential limitation to the external validity of this review is that seven out of the nine trials were carried out in Sub‐Saharan Africa, a region that in turn is widely diverse. For example, male circumcision is practiced in most of West Africa, rather infrequent in several southern African countries, and variable in central and eastern Africa (UNAIDS 2007); Islam is the predominant religion in the countries located in the Sahel and the African horn, whereas Christianity is widely extended in central and southern Africa (Britannica 2003); also, concurrency has been reported as more prevalent in western and central Africa when compared to eastern and southern Africa (Reniers 2010). In addition, the multi‐country trial (NIMH 2010) did not provide individual figures for each disease, leaving only one study (Garcia 2012) with specific results conducted outside Sub‐Saharan Africa. Although the results from this review may not be applicable in other settings, most of the global HIV burden is concentrated in Sub‐Saharan Africa, which underlines the relevance of the results to the current global HIV prevention context.

Furthermore, given both our definition of community and the population to which the interventions were addressed, the participants included in each trial were mainly drawn from the general population rather than high‐risk populations such as men who have sex with men (MSM), injecting drug users (IDUs) or female sex workers (FSWs). Had the trials included and identified large segments of high‐risk populations, our meta‐analysis may have yielded different results.

Each individual trial proposed a specific set of STIs to be measured, thus making it unfeasible to assess the same STIs in all studies. This was also the case with the proposed secondary outcomes.

It is important to mention that although most studies reported on condom use, their type of assessment was not always comparable across trials; thus preventing a larger pooling of studies in the meta‐analysis.

Quality of the evidence

Although only nine trials were retrieved, the number of total participants was large at 75,891. To the best of our knowledge, this meta‐analysis covers the largest total sample size related to this topic. Moreover, all meta‐analyses were consistent and had low heterogeneity.

An important finding of this review is that none of the studies implemented promotion of condoms as a unique intervention. In fact, all trials promoted it through education, distribution, or both, in combination with one or several of the following: training of medical staff, providing accessible and open health services to adolescents, educating parents on sexuality and encouraging dialogue about sex with teenagers, task‐shifting through community members, income generation and strengthened treatment of STIs. Therefore, these results reflect the synergy of promoting condom use along with other strategies as opposed to an outcome derived solely from condom promotion. Nevertheless, conclusions can still be inferred based on these outcomes. Incidence or prevalence of HIV and most biological results did not change after the interventions, implying that condom use combined with other strategies enacted at a community level may not work effectively.

Potential biases in the review process

This review only included biological outcomes confirmed by laboratory tests as primary results and thus excluded the possibility of self reporting bias. Moreover, all studies were RCTs, reducing the likelihood of imbalance of known and unknown baseline characteristics and confounders.

In order to conduct meta‐analysis, each set of results should be comparable. Due to the limited number of studies that were retrieved and a large variation in methodology, performing meta‐analysis was not always possible. When meta‐analyses were feasible, only a few studies were included with an average of three trials per analysis.

Most studies reported blinding as unfeasible, or did not mention it at all, leading to possible bias in self reported outcomes. This may be the case with measures of condom use, in which a significant result was obtained. Bias arising from lack of blinding is unlikely in the other self reported outcome (two or more sexual partners in the last year) since it did not yield a significant result.

Attrition bias was an important finding in this review. Since the present study focused only on community interventions, entire populations were the subject of research. Under these circumstances, it was frequent that large numbers of participants left the study due to either migration or death, leading to incomplete data. This situation may have undermined to some extent the validity of our findings (Higgins 2011).

In addition, the secular trend phenomenon that tends to take place in community trials may have biased to some extent the results in this review by diminishing the apparent effect that education could have on HIV/STIs prevention.

Moreover, It is possible that HIV/STIs prevention programs have different levels of impact depending on the actor leading the initiative. For example, the success of the100% condom use program in Thailand's brothels is likely to have worked since there was government enforcement at a national level (Rojanapithayakorn 1996). In Uganda, the early commitment of the government coupled with intersectoral collaboration (i.e. civil organizations, NGOs, private sector) is thought to have hampered the HIV epidemic (USAID 2002). In most of the included studies for this revision, the interventions were led by an international organization/NGO, and some through a partnership between these institutions and local governments. Unfortunately, it was not possible to obtain studies in which a policy change led primarily by the local government took place under the experimental conditions proposed in this review.

Finally, the primary purpose of this review was to explore the effect of condom promotion on the transmission of HIV and other STIs; thus, only studies reporting biological outcomes were included. This approach, in consequence, may have limited the validity of the secondary outcomes reported in this revision. Depending on the scope and purpose of further updates of this document, the inclusion of studies might not be restricted by the type of outcomes so that all results are free of bias.

Agreements and disagreements with other studies or reviews

This review found that the intervention(s) had a significant impact in increasing both condom use at last sex and knowledge that condoms can prevent the transmission of HIV and other STIs. However, the incidence and prevalence for HIV and HSV‐2 as well as the prevalence of gonorrhea, chlamydia, syphilis and trichomonas did not yield a significant change after the interventions were implemented. Similarly, number of sexual partners did not show a significant decrease.

The present findings are in agreement with other systematic reviews that found the continued transmission of HIV despite increased condom use (Ng 2011; Ota 2011; Shahmanesh 2008) One systematic review featuring behavioral interventions (i.e acceptability) at group and community level in FSWs (Wariki 2012) found that HIV transmission declined when results were assessed within six months immediately after the intervention was conducted; however, a non‐significant effect was found at 12 months’ follow‐up. Sangani 2004 found that population‐based interventions implementing education for preventing STIs did not reduce HIV transmission. It should be also noted that few publications have directly assessed HIV sero‐prevalence as an outcome, so many reviews could not be compared.

Systematic reviews have found contradictory results regarding condom use along with behavioral interventions and their impact in transmission of other STIs. We found that the intervention(s) described in this paper do not reduce transmission of most STIs. Thus, our results were consistent with other systematic reviews on this topic: Medley 2009 and Ota 2011 found a non‐significant effect in STI reduction, Underhill 2008 did not find any statistical effect on self reported STIs, and Michielsen 2010 found only one study reporting a decrease in HSV‐2. In contrast, studies with pooled results of several STIs such as Charania 2011; Herbst 2007a; Lyles 2007; Shahmanesh 2008, and with individual STI outcomes such as Ng 2011 and Wariki 2012 found statistical differences between intervention and control groups after the interventions were conducted. Sangani 2004 found mixed results (individual STI results). It is, however, difficult to establish a direct comparison with several of those analyses, since many of them did not provide independent figures for each disease, but aggregated all STIs into a unique composite score.

This review yielded a non‐significant effect in changing the number of sexual partners. This was consistent with previous studies by Ojo 2011, Michielsen 2010 (who found heterogeneous results, highlighting an overall non‐significant result), and partially by Townsend 2013 who found mixed or little effect. The opposite result was reported by Paul‐Ebhohimhen 2008, Lyles 2007, Herbst 2007a, and Noar 2008. However, in the majority of systematic reviews we found, the number of sexual partners is usually either aggregated into a composite score (risky behaviors) or not assessed at all.

This review found that condom use increased after intervention. This was also reported by Charania 2011, Medley 2009, Herbst 2007a, Herbst 2007b, Lyles 2007, Shahmanesh 2008, Johnson 2008, Underhill 2008 and Wariki 2012; studies in which the target was mainly high risk populations (i.e. FSWs, MSM, IDUs). Wang 2011, Bertrand 2006 and Sangani 2004 found a small increase in condom use; with the exception of Wang 2011, they were conducted in the general population. Conversely, Michielsen 2010, Paul‐Ebhohimhen 2008 and Ota 2011 found no significant change; apart from Ota 2011 which targeted FSWs, these trials were carried out in general populations. Foss 2007 reported that interventions were not effective in increasing condom use in steady couples and evidence for casual partners was scarce, whereas the intervention worked for FSWs and their clients.

Findings in this review demonstrated increased knowledge about HIV and STIs. The same result was reported by Medley 2009, Paul‐Ebhohimhen 2008, Underhill 2008, Ota 2011 and Bertrand 2006. No systematic review was found stating that knowledge did not increase after intervention.

Among the previous reviews Charania 2011, Herbst 2007a, Johnson 2008, Underhill 2008, Wariki 2012, Ng 2011, Sangani 2004 and Bertrand 2006 included studies that were conducted at the community level.

Although several of these systematic reviews have directly or indirectly assessed the impact of condom promotion, this is the first review reporting on individual biological outcomes and including only community‐based interventions. For example, Charania 2011 focused on reported condom use and only provided a unique pooled figure for STIs incidence in which all three levels of interventions (i.e. individual, groups and community) were combined; moreover, it did not mention what particular STIs were included in the analysis. Herbst 2007a did not include the general population, and Ng 2011 focused on STI control and did not use structural approaches. Moreover, despite the different methodologies used in the trials included in this review, a quantitative estimate of the intervention's effect (i.e. meta‐analysis) was possible.

Characteristics of included studies [ordered by study ID]

Characteristics of excluded studies [ordered by study ID]

Characteristics of studies awaiting assessment [ordered by study ID]

Characteristics of ongoing studies [ordered by study ID]

Title in the Review

The title slightly changed since the Protocol was published, from "Structural and community‐level interventions for increasing condom use to prevent HIV and other sexually transmitted infections" to " Structural and community‐level interventions for increasing condom use to prevent the transmission of HIV and other sexually transmitted infections."

Authors

The authors changed from “Nababan H, Ota E, Wariki WMV, Koyanagi A, Ezoe S, Shibuya K, Tobe‐Gai R” in the last published protocol, to “Moreno R, Nababan HY, Ota E, Wariki WMV, Ezoe S, Gilmour S, Shibuya K” in this document.

Type of interventions

When a trial did not explicitly mention condom promotion but the intervention suggested it, authors were contacted (via email) in order to clarify whether condom promotion occurred or not.

Population

The protocol mentioned examples of 'communities', but without specifying the definition used during study selection. Nonetheless, this concept was crucial during the selection of trials as each study defined the term 'community' in a different manner. Therefore, the selection criterion was based on the most accepted and common concept of community: “a geographic unit, such as counties, cities or villages” (Atienza 2002). Thus, trials conducted exclusively in 'smaller communities' such as schools, workplaces, brothels or others alike were ruled out.

Outcomes

The original primary outcomes were: reported condom use, HIV incidence or prevalence and STI incidence or prevalence. However, during the review process it was decided that in order to give strength to the review, intermediate outcomes (i.e. reported condom use) should be avoided as primary outcomes and as selection criteria. Instead, only those studies reporting biological data were included. Studies that solely contained self reported STIs without laboratory confirmation were excluded as well.

The initial secondary outcomes for this review were: knowledge about condom use (both self reported and tested), attitudes towards condom use (both self reported and tested), condom acquisition/procurement, stated intention to use condoms, negotiation skills around condom use (both self reported and tested), changes of other STI risk‐related practices, including number of partners, process indicators reporting on satisfaction with and/or acceptability of aspects of the intervention (including the mode of condom distribution as well as the structure and/or content of the condom promotion strategy), reported harms related to condom use, such as domestic violence, abuse or allergic reactions to latex, and comparative costs of interventions. However, for several reasons not all these outcomes were included in this review: first, most of these results were not reported in the original papers; second, if they were reported, it was difficult to include these results in a meta‐analysis given the different presentation of the data; third, in this review 26 different results were already presented, so the addition of extra outcomes would have made the document difficult to follow given its extension.

Subgroup analysis and investigation of heterogeneity

The protocol stated that when possible, stratified analysis would be undertaken among participant subgroups. Subgroups included age, gender, and primary motivation for condom use (contraception and prophylaxis versus prophylaxis only). Additional stratified analyses divided studies according to methodological quality, when sufficient numbers were available. However, since outcomes were homogeneous for all results, subanalysis was deemed unnecessary.

Correction for cluster effect

In the original protocol it was stated that different statistical methods would be used when needed, without mentioning a specific strategy to calculate cluster effect and ICCs. Since it was not possible to find ICCs from authors, and a unique ICC value applied to all results taken from a trial with the same total population is inaccurate, we decided to use a linear regression model to find the missing ICCs.

Internal sources

• Department of Global Health Policy, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan.

External sources

• Ministry of Health, Labour and Welfare, Health Labour Sciences Research Grant (No.13800128), Japan.