Guideline groups

WHO formed a Guideline Development Group (GDG) for the screen-and-treat strategies to prevent cervical cancer, chaired by Joanna Cain. The 17 selected members provided expert clinical guidance and support throughout the guideline development process. WHO also selected an External Review Group (ERG) comprising 33 professionals, including health-care providers with experience in screening and treating CIN, pathologists, researchers in cervical cancer prevention and treatment, programme directors, health educators, epidemiologists, public health officers, nurses and methodologists. A Methods Group (MG) from the MacGRADE Centre at McMaster University, a WHO collaborating centre, provided expertise in evidence synthesis and guideline development processes.

Formulating questions and determining outcomes

In February 2011, the GDG met to discuss the questions and outcomes to address in the chapter on screen-and-treat strategies to appear in the updated C4-GEP, in order to incorporate new evidence. The GDG identified 15 potential questions to guide the evidence review for screening options and treatment strategies for cervical pre-cancer. The MG surveyed the GDG anonymously online using Survey Monkey⁵ to prioritize the questions and determine which ones are clinically relevant or used in practice; 14 out of 17 members responded. Among the 15 questions, the GDG identified 7 that related to comparisons between standard screen-and-treat strategies and those that are NOT typically used in practice (e.g. an HPV test followed by cytology), and therefore these seven questions were excluded. The remaining eight questions were retained as the basis for the screening recommendations (see Box 1).

Box 1

Prioritized questions for screening options for cervical pre-cancer. Should an HPV test or VIA be used to screen? Should an HPV test or cytology followed by colposcopy (with or without biopsy) be used to screen?

During this same meeting, the GDG developed a list of outcomes that should be considered when making decisions and recommendations for the screen-and-treat strategies. These outcomes were informed by the work previously conducted for the preparation of the WHO guidelines entitled Use of cryotherapy for cervical intraepithelial neoplasia (7). Following the meeting, the MG surveyed all GDG and ERG members online using Survey Monkey to identify and rank the critical outcomes for making recommendations. Participants ranked outcomes on a scale from 1 (not at all important) to 7 (critical) in terms of importance for decision-making. Thirty of the 50 members surveyed provided responses and an average ranking was calculated for each outcome. Outcomes with an average ranking of 4 (important) or higher were included in the evidence review and considered when making the recommendations (see Box 2).

Box 2

Outcomes for screen-and-treat strategies identified as important for making recommendations (in order of importance). Mortality from cervical cancer Cervical cancer incidence

Synthesis of the evidence and preparation of evidence profiles

A screen-and-treat strategy is made up of two linked parts: the screening test(s) followed by treatment of CIN. The best studies to inform screen-and-treat recommendations are randomized controlled trials in which women are randomized to receive ‘screen-and-treat strategy A’ or ‘screen-and-treat strategy B’, and health outcomes of all of the women are measured and presented (even those who had false-negative screening results and were never treated). However, few such studies have been conducted; most studies do not link the screening strategy with the treatment and the outcomes. Instead, there are studies that measure the accuracy of a test to diagnose CIN (without reporting on treatment), and there are other studies that only measure health outcomes of screen-positive women after treatment. Hence, the outcomes for women with true-negative or false-negative screening results are not measured. This is true for the literature for cervical pre-cancer screening; few studies report health outcomes of all women who were screened, including those who were treated and those who were not. For this reason, the primary evidence used to develop these screen-and-treat recommendations could not be based on evidence from randomized controlled studies. Instead, these recommendations are based on modelling health outcomes from a series of reviews of the diagnostic accuracy of the available screening tests, and a series of reviews of the effects of the various treatments for CIN.

The MG searched the MEDLINE and EMBASE online databases up to February 2012 for screening strategies related to an HPV test compared to VIA, and VIA compared to cytology. A separate search was conducted to update a Cochrane Review that was in progress for an HPV test compared to cytology up to November 2012. Another search was conducted for colposcopy up to September 2012 (see Annex 5 for search strategies). The MG used the evidence on treatment of CIN that was concurrently being gathered for the development of WHO guidelines for treatment of cervical intraepithelial neoplasia 2–3 and glandular adenocarcinoma in situ (8). The searches were not restricted by language or study design in order not to exclude primary studies or previously published systematic reviews in this area. Reference lists of relevant studies were reviewed and the WHO GDG was contacted for additional references.⁶

At least two members of the MG independently screened titles and abstracts and the full text of relevant articles, and a third investigator resolved disagreements. The MG included observational studies for diagnostic test accuracy studies considered at low risk of bias. For example, all women in the studies had to receive both screening tests that were being compared, and all women who tested positive or negative (or a random sample of at least 10% of the women who tested negative) had to receive the ‘gold standard’ diagnostic test. Studies had to include non-pregnant women aged 18 years or older who had not been treated previously for CIN. Women could be of HIV-positive or HIV-negative status or of unknown HIV status. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist was used to develop the flow diagram for inclusion and exclusion of studies (Annex 6). A list of all studies included in the reviews of diagnostic test accuracy is provided in Annex 7.

Two members of the MG independently abstracted data about patient characteristics, setting, and diagnostic test accuracy, using a pre-tested data abstraction form. Data to assess the quality of the studies was also collected with the QUADAS tool (QUality Assessment for Diagnostic Accuracy Studies) (13). We pooled the diagnostic test accuracy data using Stata 12 data analysis and statistical software.

The MG developed a mathematical model to calculate the benefits and harms of each screen-and-treat strategy compared to other screen-and-treat strategies for women of unknown HIV status and for women of HIV-positive status. CIN2+ prevalence, natural progression data, the pooled diagnostic test accuracy results, and pooled data on treatment effects and complications were all considered in the model (see Annex 7 for references used in the model). The estimates of the expected absolute effects on health-care outcomes and a summary of the model assumptions are provided transparently in the evidence profiles for women of negative or unknown HIV status and for women of HIV-positive status (see Supplemental material, Sections A and B, available online) and for women of different ages.

Two members of the MG evaluated the quality of evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach (14, 15) and presented the evidence and its quality in the GRADE evidence profiles. Evidence for diagnostic accuracy of the screening tests is presented in evidence profiles for diagnostic test accuracy for each recommendation (see Supplemental material, Sections A and B, generally in section 2.1 for each recommendation). The evidence from the model (i.e. outcomes after a screen-and-treat strategy) is also presented in evidence profiles (see Supplemental material, Sections A and B, generally in sections 2.2 and 2.3 for each recommendation, according to age). The quality of the evidence or confidence in the effect estimates was assessed as high ⊕⊕⊕⊕, moderate ⊕⊕⊕⊖, low ⊕⊕⊖⊖, or very low ⊕⊖⊖⊖, according to the GRADE criteria. Tables to facilitate decision-making for recommendations (evidence-to-recommendations tables) were produced for each recommendation. These tables include a summary of the evidence (benefits and harms), an assessment of the quality of the evidence, relevant patient values and preferences, and any implications for use of resources and feasibility (Supplemental material, Sections A and B).

Modelling of health outcomes

A screening test with the highest diagnostic accuracy is not necessarily the test of choice in clinical practice. The decision to recommend a screening test needs to be justified by its impact on downstream patient-important health outcomes. Decision analysis is a powerful tool for evaluating a diagnostic or screening test on the basis of long-term patient-important outcomes when only intermediate outcomes – such as test sensitivity and specificity – are known. When making the decision to recommend a diagnostic or screening test, a panel should consider the health outcomes downstream from the test. For example, the health risks of interventions resulting from false-positive (FP) and false-negative (FN) findings should be compared with the health benefits associated with true-negative (TN) and true-positive (TP) findings.

To inform these recommendations, we built a mathematical model using TreeAge Pro 2012 software. In this model we calculated the proportions of TP, TN, FP and FN findings for each of the screening tests (VIA, HPV and cytology) given the pooled test-accuracy estimates and the pretest probability of having CIN in that population. We then calculated the probability of developing any of the critical outcomes for decision-making (see Box 2) based on the treatment they may receive and the pooled estimates of efficacy and potential complications of the different treatments (cryotherapy, CKC and LEEP). To calculate an overall estimate of the outcome, we added the probability of developing an outcome for each of the categories (TP, TN, FP, FN) for the same screening test and treatment option. We identified our assumptions for the models a priori. These assumptions are summarized in the Supplemental material available online, Sections A and B (below each GRADE evidence table for patient-important outcomes following different screen-and-treat strategies). We also specified a priori the sensitivity analysis we performed based on HIV status (HIV-positive compared to unknown HIV status) and different age categories.

Development of the recommendations

In early 2012 (26–28 April), the GDG, the ERG and the MG met to discuss the recommendations. One member each from the GDG and the MG chaired the meeting, which was attended by experts from around the world, representing various public health and medical disciplines. To expand the geographical representativeness of the GDG, it was decided that the ERG – a large group with members representing many countries – would participate in the development of the recommendations during that meeting. Members of the MG presented evidence profiles and evidence-to-recommendation tables, which included evidence about the benefits and harms, values and preferences, resources and feasibility.

With regard to patient values and preferences, the GDG agreed that the evidence found could be applied across all recommendations (16–18). The evidence from qualitative studies suggests that women may fear screening and may have a high level of anxiety related to colposcopy or treatment, and may feel burdened by the need for a second visit for treatment. However, once women decide to be screened they find the screening tests and immediate treatment acceptable. Evidence from the systematic reviews demonstrated that there is a preference for more frequent screening and active management among women who have screened positive for CIN1. In addition, evidence from controlled trials showed that women find treatment by cryotherapy and LEEP acceptable, and are satisfied with a screen-and-treat approach (19).

WHO has recently developed the WHO cervical cancer prevention and control costing tool (20). This tool includes two modules: one on the cost of HPV vaccination and the other on the cost of a screen-and-treat programme. The purpose of the tool is to help programme managers develop a budget for the programme. In order to develop the tool, the cost of each intervention was collected for a range of countries and the calculation tables were developed. This, in addition to the experience of the members of the ERG, was essential to the discussion of the resources needed for each of strategy.

Recommendations were made by the GDG and ERG by balancing the overall desirable and undesirable consequences of the screen-and-treat strategies, which included consideration of important outcomes, values and preferences, resources and feasibility, along with the level of certainty of that information. Members of the panel made decisions based on consensus and unanimous voting, which was not anonymous. The results of those discussions are documented in the evidence-to-recommendation tables for each recommendation, available online in Supplemental material, Sections A and B. The GDG and ERG also identified key research gaps. All the discussions and decisions took place during the April 2012 meeting and no major discord was noted.

The recommendations were assessed as ‘strong’ or ‘conditional’, in accordance with the WHO handbook for guideline development (11, 12). Strong recommendations have been worded as ‘we recommend’ and conditional recommendations as ‘we suggest’. A strong recommendation means that it was clear to the panel that the net desirable consequences of the specified strategy outweighed those of the alternative strategy. But a conditional recommendation was made when it was less clear whether the net desirable consequences of the specified strategy outweighed those of the other strategy. In this guideline, many recommendations are conditional. Table 1 provides a guide to the interpretation of the strength of the recommendations.

Table 1

Interpretation of strong and conditional recommendations.

Guideline review and approval process

The WHO screen-and-treat strategies to prevent cervical cancer underwent the following peer review process before and during development:

• The questions formulated for the development of the guidelines were circulated among the WHO Steering Group, who also discussed them with the GDG. When the GDG and the WHO Steering Group had reached agreement on the questions, these were sent to the ERG.
• The protocol for systematic reviews was circulated among the GDG. This protocol was also discussed during the ERG meeting, which was also attended by the European Guidelines Development Group in addition to the WHO Steering Group, the GDG and the MG. During that meeting the evidence that had been identified and the draft evidence profiles were discussed.
• Discussions and conference calls were regularly held with the GDG to discuss the data from the literature review, the models, the estimated parameters to include in the models, and the outcomes.
• The final draft guideline with the recommendations was circulated among the members of the GDG for review before WHO clearance.

Footnotes

5

Survey Monkey: www​.surveymonkey.com

6

Details of the methods for the systematic reviews are available at the WHO website.