Knowledge Mediates the Effects of Game Changers for Cervical Cancer Prevention (GC-CCP) Intervention on Increased VIA Screening Advocacy in Uganda

Cervical cancer is the most common cancer and accounts for about a quarter of all cancer-related mortality among women in Uganda, which has one of the highest incidence rates in the world (1–3). Cervical cancer screening via visual inspection of the cervix with acetic acid (VIA), and thermal ablation for precancerous lesions, are available for free or a low cost in Uganda. In contrast, treatment for advanced disease is scarcely available and too expensive for most women, highlighting the importance of timely cervical cancer screening to prevent onset of cancer. However, lifetime prevalence of cervical cancer screening is as low as 5% among Ugandan women (4–6), and most (80%) have advanced disease when initiating care (4). Barriers to cervical cancer screening include structural (poor access; few trained providers) and individual (poor cervical cancer knowledge; stigma associated with fears and misconceptions) factors (4, 7, 8).

One approach to addressing the individual-level barriers to increased cervical cancer screening is to empower women who have ever been screened for cervical cancer to encourage other women to get screened. Peer advocacy interventions have been effective at increasing prevention behaviors in the context of human immunodeficiency virus (HIV; refs. 9–12). However, we are unaware of any interventions designed to diffuse information through social networks to improve uptake of cervical cancer screening—this despite evidence suggesting that encouragement from others to get screened, and knowing someone who has screened for or been diagnosed with cervical cancer are facilitators of cervical cancer screening uptake (13). Women who have been screened (and treated) for cervical cancer can play a crucial role in cervical cancer prevention, as their experience positions them to be influential and credible at conveying information about the importance of screening, and to exemplify the benefits of cervical cancer screening and treatment.

Building on theories of social diffusion (14), cognitive consistency (15), and social influence (16), which posit that behavior change can be initiated by a few and diffused to others through modeling, advocacy, and shifts in social norms, we developed a social network–based advocacy group intervention—Game Changers for Cervical Cancer Prevention (GC-CCP)—to promote cervical cancer screening. As depicted in Fig. 1, GC-CCP seeks to empower and mobilize women who have ever been screened for cervical cancer to act as change agents for cervical cancer screening within their social networks by directly targeting stigma reduction, sharing of cervical cancer screening experience, knowledge of cervical cancer facts and myths, cervical cancer risk management, and advocacy skills building.

We conducted a randomized controlled trial of the GC-CCP intervention in which women who had recently screened for cervical cancer (including many who received treatment for precancerous lesions) received the intervention. We also recruited previously unscreened women from within the participants’ social network (referred to as “alters”) to evaluate the effects of the intervention on alter uptake of cervical cancer screening. Our findings revealed a dramatic effect on alter screening with over half of alters of women in the intervention group being screened for cervical cancer during the 6-month study period, compared with less than one-seventh of those in the control group (17). Furthermore, the intervention achieved its most direct goal, which was to increase engagement in cervical cancer prevention advocacy.

To identify the mechanisms that drive the intervention effects on increased advocacy, we examined whether the key components of GC-CCP mediate its effects on increased cervical cancer prevention advocacy. Drawing on our conceptual model, our hypothesis was that internalized cervical cancer stigma, sharing of cervical cancer screening experience, cervical cancer risk management, cervical cancer knowledge, and self-efficacy for engagement in cervical cancer prevention advocacy, are all catalysts (i.e., mediators) to increased advocacy. Furthermore, to understand the context in which the intervention is most effective, we explored characteristics of the individual and their social network that may moderate the effect of the intervention on engagement in prevention advocacy.

We conducted a randomized controlled trial of GC-CPP, a peer-led, group advocacy training intervention for women who had screened for cervical cancer (referred to as index participants; clinicaltrials.gov: NCT04960748; registered June 25, 2021). These women were randomly assigned to either the intervention or a wait-list control group; randomization was stratified by age (under and over age 35) and whether the participant had screened positive. Participants were not blind to study arm assignment; only the data analyst was blinded. Participants completed assessments at baseline and month 6 and received 30,000 Uganda shillings (∼$8 USD) for each. The study protocol was constructed according to the CONSORT guidelines, and approved by the Makerere University School of Public Health Research and Ethics Committee and cleared by the Uganda National Council for Science and Technology. Further details of the study protocol are available in a prior publication (18). The study was conducted in compliance with the ethical guidelines of the Declaration of Helsinki and Belmont Report.

The study took place at Buyinja Health Center IV in Namayingo, a rural district in the Busoga region of Uganda. VIA screening and thermal ablation for precancerous lesions are available at this clinic. Recruitment was conducted in September 2021. Eligibility criteria for index participants included being at least 18 years of age, screened for cervical cancer within the past year (this recency of experience was thought to enhance motivation to participate in advocacy), stable health (i.e., no advanced staged cervical cancer disease), and had told at least one female social network member (who was perceived to be unscreened for cervical cancer) about their cervical cancer screening experience. Providers informed eligible clients about the study, and they were asked to approach a balance of women who screened positive for signs of cervical cancer risk (precancerous or cancerous lesions), and women who screened negative, so that we could assess whether this factor was associated with study outcomes. Women who expressed interest were referred to the study coordinator for formal eligibility screening and obtaining written informed consent. After enrollment, women were administered the baseline assessment and then randomly assigned to the intervention or control arm.

The intervention was comprised of seven weekly group sessions. Session 1 addressed fears and concerns related to cervical cancer risk and use of self-compassion and peer support to reduce internalized stigma. Session 2 focused on building skills and decision making for sharing one's personal cervical cancer screening experience. Session 3 emphasized cervical cancer risk management (recognizing signs of cervical cancer risk and seeking health services), so that women were modeling the behaviors they encouraged other women to practice; education on cervical cancer–related facts and myths was provided to facilitate accurate cervical cancer screening advocacy. Session 4 provided instruction on how one's network can be a tool for cervical cancer prevention advocacy and dissemination of cervical cancer–related information. Sessions 5 and 6 focused on the building skills and confidence for effective cervical cancer prevention advocacy, including strategies for initiating and maintaining discussions about cervical cancer (e.g., use of reflective listening, paraphrasing, and open-ended questions). Session 7 sought to inspire commitment to ongoing cervical cancer advocacy through promotion of peer solidarity. The sessions used sharing of experiences, group problem solving, role playing, setting of personal goals, and take home practice assignments to build skills and self-efficacy. Further details on the intervention can be found elsewhere (18). Each session lasted 120–150 minutes. Participants received 30,000 Uganda shillings (∼$8 USD) for attending each session to cover transport costs.

A structured facilitator manual was developed to aid the implementation of the intervention and ensure fidelity to the planned content. The sessions were implemented in the predominant local language of Samia, by two peer facilitators who had been trained and supervised by the study team over 3 days. The supervisor observed the implementation of each session and provided feedback during weekly supervision.

The survey was interviewer-administered in Samia using Network Canvas computer software. Measures were translated using standard translation/backtranslation methodology. All measures were developed by the study team, except those in which an attribution is cited. For measures developed by the study team that included at least three items, we cite internal reliability statistics (Cronbach alpha).

The survey included two measures, one being a general measure of engagement in cervical cancer prevention advocacy; this was a 6-item measure that used a 5-point Likert response scale to measure the frequency in which specific cervical cancer–related topics (e.g., importance of cervical cancer screening, how and where to get screened, importance of getting treatment if signs of cervical cancer risk are present) were discussed by the respondent with other women, in the past 6 months. A mean item score was calculated (Cronbach alpha = 0.95). The second measure assessed advocacy for cervical cancer screening in particular, and was specific to advocacy targeted at the alters named in the respondent's social network assessment (described below). For each alter named, was assessed whether the participant had talked with the alter about the importance of cervical cancer screening, and if so, whether they had (i) encouraged the alter to get screened, (ii) provided information about how to access screening, and (iii) provided direct support to help the alter get screened (e.g., going with her to the clinic), using yes/no responses; the sum of positive (yes) responses was calculated for each alter, as well as the mean across all alters named by the participant.

Internalized cervical cancer stigma was a 5-item measures adapted from a scale used in the context of HIV [e.g., My cervical cancer screening makes me feel ashamed of myself (19)]; a higher mean score represents higher stigma. To measure sharing of cervical cancer screening experience, participants reported how much they had shared their cervical cancer screening result with sexual partners, family, and friends, in separate items; higher mean item score represents more sharing (Cronbach alpha = 0.74). Cervical cancer knowledge was assessed with 16 true/false questions related to the etiology, prevention, and treatment of cervical cancer; a sum of correct responses was calculated (Cronbach alpha = 0.75). Cervical cancer risk management self-efficacy was measured with three items that assessed confidence to notice a symptom of cervical cancer risk, seek health services for such symptoms, and obtain treatment if needed; a higher mean item score represents greater self-efficacy (Cronbach alpha = 0.64). Cervical cancer prevention advocacy self-efficacy was measured with three items assessing confidence to initiate discussions related to cervical cancer (i.e., cervical cancer screening, treatment for signs of cervical cancer risk, and sharing their cervical cancer screening experience); a higher mean item score represents greater self-efficacy (Cronbach alpha = 0.85).

These included age, education, relationship status, HIV status, and cervical cancer screening result. Binary indicators of age (≥ or <35 years) and education (any secondary education) were used in analysis.

To assess social network characteristics, index participants were first asked to list up to 12 women in their social network (referred to as “alters”) with whom they interact most. For each alter, index participants were asked to report age, HIV status, relationship to index, frequency of contact with, level of trust in the alter, perception of whether the alter had ever been screened for cervical cancer and received treatment, and knowledge of index's cervical cancer screening and treatment history. Percent of named alters who were above age 35, who were family, and who were friends were included in the analysis. To measure social support, the index participant was asked to rate how likely each alter would accompany them to the doctor, and provide advice, on a scale of 0 “not likely” to 2 “very likely”; mean item score was calculated for each alter, as well as the mean across all named alters. To assess perceived enacted cervical cancer–related stigma, for each named alter, the index participant indicated whether these two statements were true, “She has said that a person with cervical cancer must have done something wrong” and “She looks down on you because of your experience with screening for cervical cancer.” For analysis, a binary variable was created to indicate whether either of the statements was true for the alter, and then the mean across all named alters was calculated. To assess network density (i.e., connectedness between alters), participants indicated whether each unique pair of alters knew each other and how often they interacted.

A construct was examined as a mediator of the intervention effect on cervical cancer prevention advocacy whether the intervention had a significant effect on the measure of that construct. We conducted multiple linear regression analysis to examine intervention effects on potential mediators. In each model, the month 6 measure of the mediator was the dependent variable, while independent variables included the baseline measure of the mediator and an indicator of study arm. All but one index participant had completed data at the month 6 assessment; for the one participant who missed month 6, values from baseline were used to replace the missing value. Although an association between the mediator and the measure of cervical cancer prevention advocacy was not required to evaluate mediation, we nonetheless computed Pearson correlation coefficients between the potential mediators and each of the two measures of cervical cancer prevention advocacy. Structured equation modeling was used to examine mediation (STATA v16.1), along with bootstrap resampling to compute SEs and confidence intervals (CI). First, the dependent variable was the month 6 measure of cervical cancer prevention advocacy, while independent variables included an indicator of study arm and the baseline measure of the mediator. The month 6 measure of the mediator was then added to the model, resulting in the model assessing the mediating effect of change in the mediator from baseline to month 6.

To test for moderation of the intervention effect on cervical cancer prevention advocacy, we used multiple linear regression models; cervical cancer prevention advocacy at month 6 was the dependent variable in each model, while independent variables included an indicator for study arm, the potential moderator measured at baseline, and their interaction. In postestimation contrasts for continuous measures, we used the margins command to calculate the coefficients for the intervention indicator and the interaction with the moderator at its mean, one SD above the mean, and one SD below the mean to evaluate the effect of the intervention at low, medium, and high levels of the moderator. Separate analyses were conducted for each of the two measures of cervical cancer prevention advocacy.

A deidentified dataset and statistical code are available to researchers upon submission of proposal and review by the study team.

Forty women who had screened with VIA within the past year were screened for eligibility, all of whom was confirmed to be eligible and agreed to enroll in the study as index participants; 20 were randomly assigned to the intervention group and 20 to the wait-list control group. Most were age 36 years or older (58%), were in a committed relationship (85%), did not have any secondary education (23%), and were not HIV positive (93%); 24 (60%) had screened positive for signs of precancerous lesions, and each received treatment (17 received thermal therapy, 7 received cryotherapy), while the remaining 16 screened negative for any sign of cervical cancer risk. These characteristics did not differ significantly between those in the intervention versus those in the control group. The month 6 assessment was completed by all but one (98%) participant. The 20 participants in the intervention arm were divided into two groups of 10 to receive the 7-session intervention; 19 (95%) attended all seven sessions.

As reported previously, women in the intervention group reported significantly greater cervical cancer prevention advocacy at month 6, as measured by both the general and alter-specific measures of advocacy, compared with the control group, after controlling for baseline measures of advocacy (Table 1). In similar models examining intervention effects on each of the potential mediators, the intervention group reported greater increases in sharing of cervical cancer screening experience, cervical cancer knowledge, cervical cancer risk management self-efficacy, and cervical cancer prevention advocacy self-efficacy at month 6, compared with the control group (Table 1). These four constructs were then examined as potential mediators of the intervention effect on increased cervical cancer prevention advocacy. Although the correlations between the measures of cervical cancer prevention advocacy and the constructs in our theoretical framework did not influence whether to examine the presence of mediation, all constructs were found to be correlated with the general measure of cervical cancer prevention advocacy at month 6, while only cervical cancer knowledge and sharing of cervical cancer screening experience was associated with the alter-specific measure (Table 2).

Table 3 lists the results of the regression analyses that examined the mediators of the intervention effect on the general measure of cervical cancer prevention advocacy. Change in cervical cancer knowledge fully mediated the effects of the intervention on prevention advocacy, as reflected in a significant indirect effect [beta (bootstrap 95% CI) = 1.46 (0.37–2.80); P = 0.02] and the nonsignificant direct effect. Cervical cancer risk management self-efficacy partially mediated the intervention effect, with a significant indirect effect [beta (bootstrap 95% CI) = 0.81 (0.44–1.16); P < 0.0001] and significant direct effect. None of the variables were found to mediate the intervention effect on the alter-specific measure of cervical cancer screening advocacy (Table 4).

No variables were found to significantly moderate the intervention effect on the general measure of cervical cancer prevention advocacy (Table 5). In analysis of the alter-specific measure of cervical cancer screening advocacy, individual characteristics of the index participant that moderated the intervention effect were secondary education [beta (SE) = −0.30 (0.07); P <0.001] and partner status [beta (SE) = 0.18 (0.06); P = 0.006] (Table 5). The scatterplot depictions of the interactions between these variables and the intervention indicator variable reveal that those with no secondary education and those with a main sex partner reported greater engagement in cervical cancer screening advocacy in the intervention group compared with the control group, while for those with any secondary education and no main sex partner reported advocacy levels did not differ between the groups (Fig. 2).

Among social network characteristics, mean social support received across all alters [beta (SE) = 0.08 (0.02); P < 0.001] and percentage of alters exhibiting cervical cancer stigma [beta (SE) = −2.08 (0.52); P < 0.001] were significant moderators, while mean level of trust across all alters [beta (SE) = 0.20 (0.11); P = 0.07] was marginal. Regression analyses that estimated the effect of the intervention at differing levels (low, medium, high) of the moderator, revealed that the intervention effect on greater cervical cancer screening advocacy was statistically significant when trust in and support from alters was medium or high (magnitude of association was greater at the high level); the intervention effect on advocacy was significant and positive at low and medium levels of cervical cancer stigma from alters (with greater magnitude of effect at low levels of stigma), and significant and negative at high levels of stigma (Table 5).

In perhaps the first study evaluating a peer advocacy intervention to promote cervical cancer screening and prevention, GC-CCP was shown to increase engagement in cervical cancer prevention advocacy. The findings reported in this article improve our understanding of how the intervention achieved this goal, and for whom the intervention was most effective. The intervention effect on increased advocacy was mediated fully or in part by changes in cervical cancer knowledge and cervical cancer risk management self-efficacy, and moderated by several individual and social network characteristics.

As depicted in our conceptual framework (Fig. 1), the intervention targeted internalized cervical cancer stigma, comfort with sharing one's cervical cancer screening experience, knowledge of cervical cancer facts and myths, and skills and self-efficacy related to cervical cancer risk management and prevention advocacy, all of which were hypothesized to be mechanisms by which the intervention would increase advocacy. Our findings reveal that the intervention had its desired effect on each of these constructs, except for internalized cervical cancer stigma, which showed no difference between the intervention and control groups. Furthermore, each of these constructs, including internalized cervical cancer stigma, was significantly associated with the general measure of cervical cancer prevention advocacy. However, only change in cervical cancer knowledge fully mediated the intervention effect on cervical cancer prevention advocacy (as measured by the general measure of advocacy), while change in cervical cancer risk management self-efficacy partially mediated the effect on advocacy. Good knowledge of cervical cancer disease, prevention, and treatment is important not only for facilitating increased engagement in advocacy, but also enables the advocate to provide accurate information and to dispel myths and misconceptions during advocacy. Unlike the measure of cervical cancer knowledge used in this study, future research should include a measure of self-perceived knowledge of cervical cancer, which could plausibly have an even greater influence on comfort with and decision to engage in advocacy. Cervical cancer risk management self-efficacy—that is, confidence in being able to recognize signs of cervical cancer risk and to seek care and treated as warranted—may reflect a level of empowerment for combating cervical cancer risk, which may motivate engagement in cervical cancer prevention advocacy.

From our moderation analysis, the intervention had a demonstrated effect on increased advocacy (as assessed by the alter-specific measure) among women with no secondary education, and who had a main sex partner. The stigma reduction, disclosure skills, increased knowledge, and increased advocacy skills and confidence invoked by the intervention content may be particularly impactful toward advocacy among women with less education, and who may consider themselves at greater risk for cervical cancer (as a result of having a main sex partner), compared with more educated women and those at possibly lower cervical cancer risk.

The intervention was associated with greater engagement in prevention advocacy in social networks characterized as being more trustworthy, supportive, and not stigmatizing. This suggests that women are more comfortable sharing their cervical cancer screening experience, initiating cervical cancer–related discussions, and encouraging cervical cancer screening with peers they consider trustworthy and supportive. This is also a key reason why GC-CCP focuses on promoting advocacy within personal social networks, as opposed to the larger community—the latter presents distinct advocacy challenges that may require specific training components and a more select group of trainees. These social network–based moderators were evident in the analysis of the alter-specific measure of advocacy, not the analysis of the general advocacy measure, perhaps because both the advocacy and moderator variables were specific to the alters named by the index participant in their social network assessment.

There are several limitations to our analysis. A selection bias was likely present in the recruitment of the index participants, as they had decided to enroll in a study that would train them to engage in cervical cancer prevention advocacy; motivation to be such an advocate is likely associated with greater cervical cancer knowledge and other constructs (e.g., positive experiencing sharing cervical cancer screening experience) we measured, and not representative of the general population of women who had recently screened for cervical cancer. Other factors that could contribute to a selection bias are the provision of participation incentives for attending intervention sessions and completing assessments, and willingness to attend multiple lengthy group sessions, the latter of which could also impact the sustainability of implementing such an intervention and the generalizability of our findings.

Validated measures or measures used by other research groups were not available for most constructs, resulting in the need for our team to develop many of the measures used. Internal reliability statistics were moderate or good for most of these measures, but further psychometric evaluation is needed in future studies. Some of the measures also had intrinsic limitations, such as the measure of personal sharing of cervical cancer screening experience, which assessed the amount of people shared with but not the nature of the response to the sharing of experience. As alluded to above, our findings could be biased by inclusion of only women who received positive responses when sharing their screening experience. Other limitations include the small sample size and limited statistical power, although many statistically significant associations were identified, and the lack of participant blinding which could contribute to responses being influenced by social desirability.

The effects of the GC-CCP intervention on increased engagement in cervical cancer prevention advocacy was fully mediated by increased cervical cancer knowledge, highlighting the importance of the advocacy training providing information about cervical cancer–related fact and myths, as this was a key mechanism to help women feel motivated and comfortable to engage in advocacy. The findings also revealed that the intervention was most effective in facilitating increased advocacy among women with less education, who had a main sex partner, and who had received cervical cancer–related treatment, and among women whose social networks were more trustworthy and supportive, which supports the intervention's focus on promotion of advocacy within personal networks.

G.J. Wagner reports grants from Fogarty Center/NIH during the conduct of the study. No disclosures were reported by the other authors.

G.J. Wagner: Conceptualization, formal analysis, methodology, writing–original draft. J.K.B. Matovu: Conceptualization, supervision, methodology, writing–review and editing. M. Juncker: Conceptualization, supervision, methodology, writing–review and editing. E. Namisango: Supervision, methodology, writing–review and editing. J. Beyeza-Kashesya: Methodology, writing–review and editing. R.K. Wanyenze: Conceptualization, funding acquisition, methodology, writing–review and editing.

We would like to acknowledge the contribution of our study coordinators, Grace Namisi and Ishita Ghai, who were responsible for participation recruitment, data collection, and data management. We thank the women who agreed to participate in the study. This research is funded by a grant from the Fogarty International Center, NIH (R21TW011728; PI: R.K. Wanyenze).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.