Article Text

Download PDFPDF

Original article
Acceptability and response to a postal survey using self-taken samples for HPV vaccine impact monitoring
  1. Katy Sinka1,
  2. Michelle Lacey1,
  3. Chris Robertson2,
  4. Kim Kavanagh2,
  5. Kate Cuschieri3,
  6. Donna Nicholson4,
  7. Martin Donaghy1
  1. 1Immunisation, NSS Health Protection Scotland, Glasgow, UK
  2. 2Department of Mathematics and Statistics, University of Strathclyde, Glasgow, UK
  3. 3Scottish HPV reference Laboratory, Department of Laboratory Medicine, Royal Infirmary of Edinburgh, Edinburgh, UK
  4. 4Epidemiology and Statistics Group, NSS Information Services Division, Paisley, UK
  1. Correspondence to Dr Katy Sinka, NSS Health Protection Scotland, Meridian Court, 5 Cadogan Street, G2 6QE Glasgow, UK; k.sinka{at}nhs.net

Abstract

Objective To assess the feasibility and acceptance of a postal survey to measure human papillomavirus (HPV) prevalence and monitor vaccine impact, using self-taken specimens from young women who do not attend their first cervical screening appointment.

Methods Focus groups informed the survey design identifying factors that would influence acceptability. Postal testing kits were sent to a nationally representative sample of unscreened women. Overall response rate, the influence of different specimen types (urine or vaginal swab) and the receipt of a reminder letter on participation were calculated. Specimens were tested anonymously for HPV. Individual test results were not provided.

Results Of 5500 kits sent, 725 were returned (13.2%). Fifty-two women actively opted out. There was a higher return rate for urine kits (13.7% vs 12%) and from those who received a reminder letter (15.5% vs 12.2%). Response was influenced by deprivation (10.3% in the most deprived quintile vs 16.2% in the least). Overall weighted HPV prevalence was 35.9% (40.0% from swab specimens and 31.9% from urine).

Conclusions Some women were willing to participate in anonymised postal testing. However, the low uptake means that HPV prevalence results are difficult to interpret for ongoing surveillance. Monitoring HPV vaccine impact outwith the cervical screening programme remains challenging.

  • HPV vaccines
  • prevalence study
  • epidemiology
  • surveillance methods
  • vaccination
  • epidemiology (general)
  • surveillance
  • HPV
  • adolescent
  • cervical cytology
  • communication skills
  • cervical cancer
  • virology

Statistics from Altmetric.com

Background

Human papillomavirus (HPV) vaccination to prevent cervical cancer was introduced into the childhood immunisation schedule in Scotland in 2008. Vaccination is offered routinely to 12–13-year-old girls at school. For an introductory period until August 2011, catch-up vaccination was also offered to older teenage girls who were under 18 at the programme start. High vaccine uptake has been achieved to date with 90% course completion within the school-based routine and catch-up programmes. Uptake in school leavers has been lower with around 30% course completion.1 As with all vaccination programmes, monitoring public health impact is an essential component of the intervention. Since the eventual cervical cancer disease end point is typically decades after exposure, current monitoring activity focuses on measuring the anticipated earlier reduction of vaccine-type HPV infection in the population. There is no routine generalised HPV testing in Scotland at present and so a specific HPV surveillance strategy was required from 2010 to measure and monitor HPV prevalence in the target population.

In Scotland, as elsewhere, the mainstay of early vaccine impact monitoring will be carried out passively by anonymised HPV testing of residual cervical screening samples. The Scottish cervical screening programme covers women aged 20 to 60 years. The first vaccinated cohorts entered the programme, aged 20 years, at the end of 2010. However, in 2006/2007, it was estimated that only 37% of 20-year-olds in Scotland attended their first cervical screening appointment (unpublished data from the Scottish Cervical Call-Recall System, SCCRS). For a more complete picture of prevalent HPV infection in an unvaccinated population, we aimed to capture additional information from a sample of those who had ‘defaulted’ from their first appointment representing the majority of young women of this age.

Postal testing using self-taken specimens has previously been piloted for Chlamydia screening.2–5 These studies guided the design of our survey and also the expectation of response rates, which ranged from 23% to 34.5%. Acceptability and quality of different self-taken samples for HPV testing is well documented.6–17 Where women are offered different self-taken specimen types, most options are acceptable but there is a preference for less invasive less ‘personal’ tests with urine often preferred to vaginal swabs in comparison studies.2 3 11–13 Although urine samples are less robust biospecimens for HPV detection compared with vulvo-vaginal swabs, their use for surveillance purposes is still a valid option.11 Clear illustrated instructions for unsupervised self-taken samples for HPV testing have been developed and field tested.15–17

Unlike previously published studies using postal testing kits, the laboratory testing in our surveillance pilot was unlinked and anonymous and no results would be returned to participants. This factor was likely to influence response rates, and qualitative research was carried out prior to the survey to explore the acceptability of this approach.

Pre-survey qualitative research

A preliminary market research study rapidly assessed the likelihood of a low response rate and identified characteristics that would improve acceptability and uptake (Margaret Reid Research and Planning, unpublished report, 2008). Informants were asked to comment on the study design, preference of self-taken specimen type (urine or vaginal swab) and whether an incentive would improve participation. Three focus groups, with six to eight participants, and two in-depth interviews were held. The participants were women aged 21–22 years who had declined their first screening call. Participants were recruited using a standardised recruitment questionnaire in urban and rural locations. Sampling was purposive to ensure a mix of women from different socioeconomic backgrounds, representation from minority ethnic and faith groups, those in full-time education, working and non-working; those married, cohabiting and single. Two participants had children. Fieldwork took place in November 2008, shortly after the high profile launch of the Scottish HPV vaccination programme.

The market research concluded that response was likely to be low and that the most effective method of increasing uptake would be to give women their HPV results. However, it was reported that by being absolutely clear in any communications that this was a statistical study and that the information collected was intended to benefit the health of future women in Scotland, some women would participate. The urine sample test kit was preferred but not outright. In particular, it was seen as messy and there were concerns about returning specimens by post. The swab test kit was perceived as more ‘scientific’.

Participants were positive about the use of an advance communication to reduce the surprise of receiving the pack without notice, to provide an opportunity to explain the study and personalise the request to participate. They did not expect that opt-out mechanisms would be much used. Respondents also rejected the use of a reminder as unnecessary since they felt that the decision to take part would be made quickly and a reminder would not have an influence. The use of incentives was perceived as devaluing the study and was not pursued further. There was some indication from the market research that the response rate would be greater with increasing socioeconomic status.

Our study piloted a postal survey using self-taken specimens to assess the feasibility and acceptance of this approach to measure HPV prevalence and monitor vaccine impact among young women who do not attend their first cervical screening appointment.

Methods

Identification of survey sample

The Scottish Cervical Call-Recall System (SCCRS), a population-based national Information Technology system, which supports the Scottish Cervical Screening Programme, was used to identify young women aged 21 years who had defaulted from their first screening appointment in 2008. Women are classified as defaulters from a screening round if they have not responded to one invitation and two reminders. Postal address information on SCCRS is from the Community Health Index and is based on General Practice (GP) registration. SCCRS automatically alerts GP practices before inviting women to a screening round in order to check that correspondence details are current.

Sampling strategy

From the anonymised data file of 15 367 women, who were eligible for screening and who had defaulted age of 21 in 2008, an initial sample of 5500 was chosen aiming for a response rate of 20% and a final sample size of 1100. A study number was attributed to each record to link any HPV test results to the anonymised data for analysis. Any information that could potentially identify an individual, for example a postal address, was managed by the relevant local health board and not visible to the study team. The data file was stratified by NHS health board administration area. Within each stratum, the sample size was in proportion to the number of defaulters in the board and was selected by simple random sampling. Socioeconomic status was defined using the Scottish Index of Multiple Deprivation (SIMD), which is derived from postcode of residence and attributed to records on the SCCRS database.18 As the sampling frame contained a greater proportion of women from higher deprivation groups, no attempt was made to over-sample these women to compensate for a poorer anticipated response rate. The sample was effectively self-weighting in terms of SIMD accounting for any influence of socioeconomic status on non-attendance at first screening appointment. Individuals were randomly selected to receive urine or vaginal swab kits. Similarly, 20% of the sample was randomly selected to receive a reminder letter.

Postal testing kits and information for participants

An advance mailing using a colourful postcard was sent, in a plain envelope, to alert women that they would be receiving a postal testing kit. This gave a study web address, a freephone number and an SMS text number so that individuals could find out more about the study or opt-out of further participation. Two weeks later, 50% were sent urine kits and 50% self-taken swab kits. The survey packs were posted in plain packaging. Each contained a self-taken specimen kit (standard 25 ml specimen tube for urine and a dacron swab with M4RT medium in a transport tube for vaginal swabs). For return of samples the packs included, a universal sample transporter and self-seal plastic envelope addressed with prepaid postage to the Scottish HPV Reference Laboratory in Edinburgh. The information leaflet gave general details of the study, diagrammatic and written instructions on how to take part and links to further information about the study and to the Scottish Cervical Screening Programme. A blank page was provided to encourage feedback from participants about any aspect of the study.

Four weeks after the postal kits were sent, a reminder letter was sent to 20% of the sample. Although our market research suggested that this would not have an effect, previous postal surveys have shown that this improved take-up rates.3 19 The period allowed for return of specimens was 6 weeks in total, between March and April 2009. Return of a specimen was considered consent to participate in the study.

Laboratory testing

A 5 ml aliquot of urine samples was centrifuged and the pellet washed twice with phosphate-buffered saline before being reconstituted in 1 ml of phosphate-buffered saline prior to extraction. Swabs were vortexed, and a 1 ml aliquot was used for extraction. Automated extraction for both sample types used the MDX Media Kit (Qiagen, Gaithersburg, MD, USA). HPV amplification and genotyping was performed using biotinylated GP5+6+ primers, prior to downstream genotyping using the Digene HPV Genotyping RH test (Qiagen). This assay is capable of detecting 18 high-risk or putatively high-risk HPV types, namely 16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73 and 82. Other HPV types, including types 6 and 11, are not differentiated but identified as ‘HPV other’.

Statistical methods

Response rates and 95% CIs were modelled using multivariable logistic regression considering the effect of SIMD, specimen type and reminder status on response. HPV prevalence rates were weighted back to the sampling frame by post stratification to adjust for the different responses by each of these variables. Analysis of the results used the statistical package R V.2.10.1.

Results

Overall response

Of the initial 5500 individuals sampled, 725 returned the kits (figure 1)—giving a crude response rate of 13.2%. Seven of the returned urine specimens and nine of the self-taken swabs were excluded from further analysis due to missing or unlinkable study id. Response rates are summarised in table 1. There was evidence of a linear response with socioeconomic status with women living in less deprived SIMD areas more likely to respond (p value of the linear SIMD trend <0.001). More urine specimens were returned than swabs, OR 1.18 (95% CI 1.00 to 1.38, p value=0.045), and women were more likely to respond if sent a reminder letter, OR 1.33 (95% CI 1.1 to 1.6, p value=0.003).

Figure 1

Human papillomavirus (HPV) postal testing process and response (survey method adapted from Low et al3 4).

Table 1

Response rates and likelihood of response adjusted by specimen type, SIMD quintile and reminder status (n=709)

Table 2 shows the weighted HPV prevalence results. Overall, the weighted prevalence of any type HPV was 35.9% (40.0% from swab specimens and 31.9% from urine). All high-risk types were detected with HPV 16 and 18 being the most prevalent individual types. HPV 16 was detected in 10.6% of specimens returned, 95% CI 8.2% to 13.0%, HPV 18 in 3.8% of specimens, 95% CI 2.3% to 5.3%, and both types were detected in 0.9% of samples, 95% CI 0.2% to 1.7%.

Table 2

Weighted HPV prevalence: overall (any type) and by SIMD, specimen type and reminder status (n=709)

Feedback from participants

There were 22 feedback forms received, approximately 3% of all returns. Eleven women would have preferred to receive their results, or did not think the study should be anonymous, eight noted that they were happy to take part and/or that they thought it was a good study to be involved in, three stated that they had found the instructions clear and two commented that they had found taking the specimen uncomfortable. One comment was a complaint about being approached to participate.

The telephone helpline took 64 calls mostly about opting out, but a third of callers had phoned to find out more about HPV, immunisation, cervical cancer and screening. The SMS service mainly recorded opt-outs but also received four queries asking about receipt of results.

Discussion

The postal survey achieved a lower than hoped for uptake but obtained a sample from which a baseline estimate of HPV prevalence for surveillance could be obtained from non-attenders of cervical screening. The pre-survey qualitative research was accurate in predicting that uptake would be low, that some women would take part on a purely altruistic basis, that urine specimens were likely to be preferable to self-taken vaginal swabs and that response would vary by socioeconomic status. Use of a reminder to improve uptake rates was the only area where the results in practice differed from the qualitative assessment.

Strengths and weaknesses

The study benefited from the use of a national IT system to accurately identify the entire cohort of non-attenders and for each record provide the SIMD as an approximate of socioeconomic status. The study design was simple and had the advantage of qualitative input from the target audience to guide, in particular, communication materials and expectation of results. The lower than expected uptake rate raises questions of response bias in the study sample and whether the HPV prevalence results calculated are valid. It is safe to assume that no woman in the sample will know her HPV status so there can be no direct bias. Any bias in HPV prevalence would have to be indirect. If women who were more sexually active were less likely to participate, then an underestimate in HPV prevalence is likely, while if women who were less sexually active were less likely to participate, then an over estimate in HPV prevalence is likely. From previous studies, we anticipated poorer responses from women living in more deprived areas and used SIMD for responders and non-responders to obtain sampling weights to attempt to correct for non-response bias.3 19 Women with a high level of health awareness or those who had heard about HPV and the vaccination programme may have been more likely to participate. We cannot predict whether this will bias our results. Assuming that an individual's decision not to respond is independent of her HPV status, the adjusted analysis should compensate for these biases but it is an approximate and imperfect adjustment assuming that the decision not to respond is based upon factors associated with deprivation. Ideally, an accompanying questionnaire to collect sexual behaviour information would have enabled better adjustment; however, the likely cost of this would have been even lower participation.

Ongoing monitoring of vaccine impact

Although HPV testing is being considered as a primary screening modality in some countries, the timescale, the target age group and the uptake of any implementation in the UK are unknown. Consequently, active surveillance methods will continue to be required to monitor the expected decline in HPV prevalence in young women in vaccinated age groups. Although this can be achieved in part through observation of the cervical screening population, there is still a requirement to monitor the impact of the vaccination programme across the population as a whole and to do so at the earliest opportunity. Options for monitoring HPV prevalence in unscreened women can either target this group specifically, as in our study, or can recruit young women attending health services, which will include a mix of screened and unscreened individuals. For the former approach to be valid, a good response rate is required. In our study, both the pre-survey qualitative research and comments sent by study participants suggest that to return individual HPV results would probably improve uptake. The decision not to return individual results rested on a number of factors, principally the young age group targeted. High prevalence and transience of HPV infections in this young age group is well documented, and the utility and validity of returning either a positive or a negative HPV result from a single urine or swab specimen was considered to be counterproductive since it would not guide any clinical management and could be distressing, while there was negligible harm from not doing so.20–23 Recent studies of HPV testing as an adjunct to existing screening for non-attenders aged 30+ years, achieved uptakes of 10%–27%24 25 indicating that provision of results is not a guarantee of good uptake.

Our study suggests that a general approach targeting unscreened women for surveillance of vaccine impact is unlikely to be effective. An alternative approach would be to engage young women already attending services where a biological specimen could be obtained for HPV testing. These include family planning, sexual health and antenatal clinics. These service users will on the whole be more sexually active and have a higher degree of exposure to HPV than the population in general. More detailed behavioural information will need to be collected by these studies to allow this bias to be characterised and in addition to record the screening and vaccination history of individuals.

HPV prevalence

Overall HPV prevalence in cervical screening samples from unvaccinated 20-year-old women in Scotland has been estimated at 49%.26 HPV was detected in 32% of urine samples and 40% of self-taken vaginal swabs provided by 21-year-old women participating in our study. These lower prevalences are assumed to be primarily attributable to the specimen type rather than differences between the populations sampled. Studies comparing self-taken swabs with clinician taken specimens for HPV detection support this assumption.9 27 The still lower prevalence measured from the urine samples is attributed to lower cellularity in the urine specimens, and while there was a preference by participants for urine specimens, its marginally higher appeal to survey participants should be offset against its quality as a biospecimen for HPV detection. Countries with newly introduced vaccination programmes are actively considering and developing strategies to monitor the early impact on circulating HPV infection. A variety of population samples, biospecimens and age groups are involved. Despite the difficulties in accessing a non-screening population for surveillance a consistent methodology will allow relative trends over time to be monitored, but absolute comparisons of HPV prevalence between different surveillance approaches may be challenging.

Conclusions

We have shown that self-taken specimens collected as part of a postal survey can be obtained for HPV surveillance and that some women will participate for altruistic reasons. While this approach allowed a representative sample of unscreened women to be identified, the relatively low uptake rate somewhat limits extrapolation of the findings. Optimal systems to measure and monitor HPV prevalence and vaccine impact in the general population should (and will) continue to evolve.

Key messages

  • Comprehensive human papillomavirus surveillance remains challenging, particularly in those who do not attend cervical screening among whom the benefits of vaccination may be greater.

  • Anonymised human papillomavirus testing of self-taken specimens from unscreened women has limited use for vaccine impact monitoring due to low participation.

  • Urine specimens were marginally preferred by respondents, but self-taken swabs are a better biospecimen for human papillomavirus testing.

  • Women's opinions indicated that the preference for one specimen type over the other was not clear-cut.

Acknowledgments

The authors would like to thank the Cancer Research UK researchers from UCL and the Wolfson Institute for the use of their self-taken HPV test instructions, the NHS Board Screening coordinators and screening officers who maintained study confidentiality, Margaret Reid for conducting the market research and the participants who remain anonymous.

References

Footnotes

  • Funding This work was supported by Scottish Government Health Directorate.

  • Competing interests None.

  • Ethics approval The study was given ethics approval by West Glasgow Ethics Committee (09/S0703/13).

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.