Performance of colposcopy in five sub-Saharan African countries

Authors


Dr R Sankaranarayanan, Screening Group, International Agency for Research on Cancer, 150 Cours Albert Thomas, 69372 Lyon Cedex 08, France. Email sankar@iarc.fr

Abstract

Objective  The performance of colposcopy provided in a screening study in five African countries was evaluated.

Design  Cross-sectional study.

Setting  Burkina Faso, Congo Brazzaville, Guinea Conakry, Mali and Niger.

Population  Women aged 25–59 years.

Methods  A total of 29 294 women participated in a cervical screening study in the five study sites, and newly trained local doctors performed colposcopy and directed biopsies as indicated. Using meta-analytical tools, four measures of colposcopy performance at different thresholds of colposcopic abnormalities were assessed. Sources of heterogeneity were also assessed.

Main outcome measures  Proportions of women receiving biopsies, adequate biopsies and women diagnosed with cervical intraepithelial neoplasia (CIN).

Results  Among 28 553 women with satisfactory colposcopy, 3101 had a colposcopic diagnosis of probable low-grade or worse lesions and 1128 probable high-grade or worse lesions. Overall, the measures that reached the set standards were proportion of biopsy taken at colposcopy threshold of probable high-grade or worse lesions (95%, 95% CI 90–100%) and proportion of adequate biopsy samples. The set standards were not met for the proportions of women diagnosed with CIN at different colposcopic abnormality thresholds. Detection of CIN2 or worse lesions increased with increasing colposcopic abnormality.

Conclusions  The performance of colposcopy in some of the African sites studied was comparable to that previously observed in other studies. With appropriate training, monitoring, continuing practice and quality assurance, adequate standards of colposcopy can be attained in sub-Saharan Africa.

Introduction

High incidence rates of cervical cancer ranging from 30 to 40 per 100 000 women are observed in sub-Saharan African countries.1,2 About 72 000 new cases and 56 000 cervical cancer deaths were estimated to occur in sub-Saharan Africa countries around 2002.2 Less than 30% of women with cervical cancer survive for 5 years after diagnosis in these countries due to advanced stage at presentation and limited access to diagnosis and treatment.3,4 In spite of the high burden of disease, there are no screening programmes in these countries. Even a sporadic screening using cytology is not feasible as there are no cytology laboratories, trained cytologists or available consumables in most sub-Saharan African countries. Many of the countries lack adequate histopathology and other diagnostic and treatment facilities for cancer in general.

As part of our efforts to evaluate the feasibility and accuracy of alternative screening methods such as visual screening with acetic acid (VIA) and Lugol’s iodine (VILI) and to catalyse dissemination of screening, early detection and prevention of cervical cancer in sub-Saharan Africa, multicentre cross-sectional studies were organised in five French-speaking countries, namely Burkina Faso, Congo Brazzaville, Guinea Conakry, Mali and Niger. The design, methods used and results from these studies have already been reported elsewhere in detail.5–7 Colposcopy was used as part of the reference standard investigations to establish the final diagnosis in these studies.5 Although colposcopy is an integral investigation in the triage, diagnosis and treatment of screen-positive women in developed countries, colposcopy was not practised routinely in Burkina Faso, Congo Brazzaville, Guinea Conakry, Mali and Niger before these cross-sectional studies. In this paper, we report on the performance of colposcopy used in our studies in these countries.

Methods

Study design

Between 2000 and 2004, the test performance of VIA and VILI was simultaneously evaluated in more than 29 000 women aged 25–64 years in five African countries (five centres) using a common protocol. None of these women had undergone any previous cervical screening, and all the participating women signed an informed consent form before screening. Details of the tests assessed have been described previously.5–7 All screened women were subsequently examined with colposcopy, irrespective of the screening test results, on the same day by 1 of 14 local doctors (including 8 gynaecologists, 4 oncologists and 2 GPs), unaware of the screening findings. The doctors were intensively trained in colposcopy and treatment methods such as cryotherapy and loop electrosurgical excision procedure before beginning the study in a 15-day training course using the International Agency for Research on Cancer (IARC) manual.8 One-day refresher sessions took place every 4–6 months during the study. Colposcopic assessment in our study followed the technique of assessing abnormal areas after application of acetic acid followed by Lugol’s iodine and assessment of Reid’s score.8 Colposcopy was categorised as unsatisfactory if the squamocolumnar junction was not visible in its entirety, and those with unsatisfactory colposcopy were excluded from analysis. Colposcopy results were reported as normal, squamous metaplasia, inflammation, probable low-grade lesion, probable high-grade lesion, preclinical invasive cancer and invasive cancer. The colposcopic assessment was then entered in the proforma, and punch biopsies were directed from colposcopically abnormal areas. As per protocol, punch biopsies were directed from women with colposcopic features of low-grade or worse lesions.

The punch biopsy specimens were processed in local pathology laboratories and reported using the cervical intraepithelial neoplasia (CIN) terminology. The histology results were reported as no neoplasia, inflammation, CIN1, CIN2, CIN3, adenocarcinoma in situ, microinvasive cancer, invasive squamous cell carcinoma, invasive adenocarcinoma, other carcinomas and inconclusive. Seven local pathologists and 11 laboratory technicians carried out histology. Histology results were assessed by the pathologists blinded with respect to the screening test results.

Measures of colposcopy performance studied

In this article, we assessed four measures of performance among women found with colposcopic abnormalities suggestive of low-grade or worse lesions: (1) proportion of women in whom a directed punch biopsy was taken, (2) proportion of women with adequate punch biopsies among women who had punch biopsies, (3) proportion of women with histologically diagnosed CIN lesions of any grade and (4) proportion of women with histologically diagnosed CIN2 or worse lesions. The last two parameters of performance correspond with positive predictive values (PPVs) of colposcopic impression for histological CIN and CIN2 or worse, respectively. These measures were also analysed for women with colposcopically probable high-grade or worse lesions. In our study, since only women with abnormal colposcopy had directed biopsies taken and since no random biopsies were taken from the colposcopically negative women, we were unable to calculate both sensitivity and specificity with no or minimal bias. However, as a proxy for but an overestimation of sensitivity, we presented the proportions of women with histologically proven CIN or worse and CIN2 or worse detected at colposcopy cutoffs of probable low-grade and probable high-grade lesions. Additionally, we explored the sources of heterogeneity by assessing the association between test performance measures and individual and study characteristics.

Quality control

All the colposcopists were monitored and periodically retrained in short sessions. The technicians and the pathologists involved in the study were trained and reoriented at the beginning of the study and retrained with some of their reports periodically reviewed by an expert pathologist from Lyon, France, during the course of the study. Internal and external quality control measures were introduced in the pathology laboratories, in which the proportion of adequate biopsy reports, the type and quality of consumables and the quality of equipment used were frequently checked. Laboratory procedures and manuals were reviewed. A sample of histopathology slides from four of the study sites (Burkina Faso, Guinea Conakry, Mali and Niger) was reviewed in a reference laboratory in Lyon, France, for assessment of inter-observer reproducibility.

Statistical analysis

Using meta-analytical methods, the performance measures were pooled, allowing for inter-setting heterogeneity using random effect models.9,10 Forest plots of the four measures of performance were presented. Sources of heterogeneity were explored by assessing the association between colposcopy performance indicators and individual and study characteristics. The influence of age, study centre, time period and severity of the colposcopic lesion on measures of performance was assessed using logistic regression.11 Age was aggregated into 10-year groups, and study period by tertiles, using date of screening. Study period was considered a proxy for accumulated experience of the test providers. Severity of colposcopic lesion was categorised into low-grade squamous intraepithelial lesions, high-grade squamous intraepithelial lesions and invasive cancer.

Results

Of the 29 294 women screened, we excluded 741 (2.5%) from the analysis (261 [0.9%] with colposcopy not performed and 480 [1.6%] with unsatisfactory colposcopy) (Table 1). Among the 28 553 women with satisfactory colposcopy, 3101 (10.9%) had probable low-grade or worse lesions and 1128 women had a diagnosis of probable high-grade or worse lesions on colposcopy (Tables 1 and 2).

Table 1.  Profile of screened women in the study centres
CentreCountryNumber screenedExcludedColposcopy satisfactoryColposcopy abnormal* (%)
Colposcopy not doneColposcopy unsatisfactoryTotal
  • *

    Colposcopically low-grade or worse lesions and number used in the final analysis.

BamakoMali777511471487627945 (12.4)
BrazzavilleCongo71861464771391017 (14.2)
ConakryGuinea9716452402859431802 (8.5)
NiameyNiger25611642582503135 (5.4)
OuagadougouBurkina Faso205619852031853202 (10.9)
Total 29 29426148074128 5533101 (10.9)
Table 2.  Number of abnormal colposcopy, biopsy taken, adequate biopsy, histology CIN or worse and histology CIN2 or worse
CentreColposcopy abnormalBiopsy takenAdequate biopsyHistology CIN or worseHistology CIN2 or worse
Among colposcopy low-grade or worse lesions
Bamako945715715504213
Brazzaville1017784784564350
Conakry802567551264136
Niamey1351291285414
Ouagadougou2021881886117
Total3101238323661447730
Among colposcopy high-grade or worse lesions
Bamako263255255185138
Brazzaville641565565402280
Conakry1191181169590
Niamey16141443
Ouagadougou8987872711
Total112810391037713522

Proportion of women who had directed biopsy

At the cutoff of low-grade or worse lesions on colposcopy, 82% (95% CI 73–91%) of women had biopsies taken (Figure 1A, Table 2). Two of the centres (Niamey and Ouagadougou) had biopsies directed in more than 90% of the women with abnormal colposcopy, and the remaining three centres had less than 80% biopsies taken from women with abnormal colposcopy results (Figure 1A). Among women with colposcopic abnormalities suggestive of high-grade or worse lesions diagnosis on colposcopy, biopsies were directed in 95% (95% CI 90–100%) (Figure 1B, Table 2). In Niamey and Brazzaville, biopsies were directed in 88% of such women, while in the other three centres (Bamako, Conakry and Ouagadougou), 95% or more had biopsies taken.

Figure 1.

Forest plots showing the proportion of the biopsy samples taken among colposcopy of (A) low-grade or worse lesions and (B) high-grade or worse lesions.

Proportion of women with adequate biopsy

Biopsy samples from almost all women in all the five study centres were scored as adequate by local pathologists (Figure 2A, B, Table 2).

Figure 2.

Forest plots showing the proportion of adequate biopsy among all biopsies taken from colposcopy of (A) low-grade or worse lesions and (B) high-grade or worse lesions.

PPV of colposcopy to detect histologically proven CIN

Among women with colposcopically diagnosed probable low-grade or worse lesions, 53% (95% CI 39–67%) had histologically diagnosed CIN of any grade (Figure 3A, Table 2). The PPV was above 70% in Bamako and Brazzaville, whereas in the other three centres, it was below 50% (Figure 3A). Among women with probable high-grade or worse lesions on colposcopy, 60% (95% CI 45–74%) had histologically diagnosed CIN of any grade (Figure 3B, Table 2); these proportions varied from 29% in Niamey to 82% in Conakry (Figure 3B).

Figure 3.

Forest plots showing the PPV of colposcopy (A) at cutoff of low grade to detect histological CIN, (B) at cutoff of high grade to detect histological CIN, (C) at cutoff of low grade to detect histological CIN2 or worse and (D) at cutoff of high grade to detect histological CIN2 or worse.

PPV of colposcopy to detect histologically proven CIN2 or worse

Among women with colposcopically diagnosed probable low-grade or worse lesions, 24% (95% CI 11–36%) had histologically diagnosed CIN2 or worse lesions (Figure 3C, Table 2); the centre-specific proportions ranged from 9% in Ouagadougou to 45% in Brazzaville (Figure 3C). Among women with probable high-grade or worse lesions on colposcopy, 44% (95% CI 24–64%) had histologically diagnosed CIN2 or worse disease (Figure 3D, Table 2). These proportions varied from 13% in Ouagadougou to 78% in Conakry (Figure 3D).

Proportion of histologically proven CIN detected by colposcopic abnormality

Overall, 92% (95% CI 88–96%) of the women histologically diagnosed with CIN were detected by an abnormal colposcopy of probable low-grade or worse lesions (Figure 4A). This proportion ranged from 73% in Niamey to 98% in Ouagadougou (>90% in four of the centres). However, abnormal colposcopy of probable high-grade or worse lesions detected on average 36% (95% CI 15–58%) of the histologically proven CIN, varying from 5% in Niamey to 65% in Brazzaville (Figure 4C).

Figure 4.

Forest plots showing the proportion of (A) histological CIN detected by colposcopy low-grade or worse lesions, (B) histological CIN detected by colposcopy high-grade or worse lesions, (C) histological CIN2 or worse detected by colposcopy low-grade or worse lesions and (D) histological CIN2 or worse detected by colposcopy high-grade or worse lesions.

Proportion of histologically proven CIN2 or worse detected by colposcopic abnormality

When probable low-grade or worse lesions were taken as colposcopic abnormality, the overall proportion of women with histological CIN2 and CIN3 lesions or worse detected by colposcopy was 95% (95% CI 93–98%), varying from 88% in Niamey to 100% in Ouagadougou (Figure 4B). At colposcopic abnormality of probable high-grade or worse lesions, the overall proportion of CIN2 and CIN3 lesions or worse detected was 59% (95% CI 46–71%), ranging from 19% in Niamey to 76% in Brazzaville (Figure 4D).

Sources of variation in the measure of colposcopy performance

All measures of colposcopy performance showed statistically significant inter-study heterogeneity (P for Cochrane’s Q test < 0.01), with the exception of the proportion of adequate biopsy taken at colposcopically high-grade or worse lesions cutoff, which was not statistically heterogeneous (P for Cochrane’s Q test = 0.735).

None of the measures of colposcopy performance at the colposcopically low-grade or worse lesions cutoff point was influenced by the three factors assessed, except the PPV of colposcopy to detect histological CIN2 or worse. Detection of CIN2 or worse lesions increased with increasing colposcopy severity, with a statistically significant increase observed when colposcopy findings were suggestive of invasive cancer.

Inter-observer agreement on histopathology

The agreement between the local pathologists and the pathologist from the reference laboratory in Lyon, France, is presented in Table 3. A moderate agreement (kappa value = 0.51) was observed when the histopathology report was grouped into five categories (inconclusive, normal, CIN1, CIN2 and CIN3 and invasive cancer). The kappa values ranged between 0.17 for Niger and 0.60 for Guinea Conakry. The agreement was substantial when the report was dichotomised into <CIN2/CIN2 or worse (kappa value = 0.71) and almost perfect when the report was dichotomised into <CIN3/CIN3 or worse (kappa value = 0.84).

Table 3.  Agreement between histology findings by African pathologists and reference pathologist from the reference laboratory in Lyon, France
Histology report by the Africa pathologistsHistology report by reference pathologistTotal
InconclusiveNormalCIN1CIN2 and CIN3Invasive cancer
  1. Agreement = 61.4%; kappa = 0.5073.

Inconclusive210003
Normal136511282
CIN11514353067
CIN2 and CIN31441224054
Invasive cancer122012540
Total5686482927246

Discussion

Colposcopy is an essential triaging investigation in the assessment, diagnosis and management of women with positive cervical screening tests. The performance of colposcopy varies according to the training and experience of the colposcopist as well as the clinical context in which it is used, ranging from primary care to referral practices. Practically, no harm occurs if colposcopy is practised competently, whereas colposcopy in incompetent hands may miss severe disease or cause harm due to false-positive conclusions. There are several pitfalls in colposcopy, which are overcome by good training, continuing practice and clinical audits.8 The standard and performance of colposcopy practice has been widely studied in developed countries.12–15 However, colposcopy has been rarely practised in most sub-Saharan African countries, with the possible exception of South Africa, and information on the performance of colposcopy providers in African settings does not exist.

In our study, all our providers were newly trained and had very little experience to begin with. The study sites were in some of the least developed healthcare systems in the world, and no organised screening and diagnostic services existed before this study. However, it is worth mentioning that this project provided an opportunity to train a core group of service providers and that the services for screening, diagnosis and treatment still exist in the study sites and have been extended to other parts of four of countries (except Burkina Faso) as the aftermath of our study. It is hoped that with continuing practice, clinical audit and continuing training programmes, the standards of colposcopy will greatly improve in these settings.

Many studies using data for women with abnormal Pap smear screening results have reported a high range of sensitivity (87–99%) but low specificity (23–87%) for colposcopy in detecting CIN.15–17 In this study, we could not obtain unbiased estimates of sensitivity and specificity since biopsies were directed for only women with abnormal colposcopy. However, abnormal colposcopy of low-grade or worse lesions detected over 90% of both CIN and CIN2 and CIN3 or worse lesions overall (and in four of the centres). The fact that these proportions dropped to 36 and 59%, respectively, when colposcopic abnormality of high-grade or worse lesions was considered stresses the need for taking biopsies from women with all colposcopy abnormalities in low-resourced settings to allow detection of as many high-grade CIN as possible.

The National Health Service Cervical Screening Programme (NHSCSP) in UK requires that excisional biopsy should be obtained from >95% of women with either colposcopically high-grade abnormalities or low-grade colposcopic changes associated with severe cytological dyskaryosis or worse.18 In our study, no excisional biopsy was performed. Nevertheless, two of the centres directed biopsies in 88% of the colposcopy high-grade or worse lesions, while four of the centres directed biopsies in 71–82% of the colposcopy low-grade or worse lesions. Failure to direct biopsy from >95% of women with colposcopic abnormalities would suggest conservatism among colposcopists. In the developing world, where there are no organised repeat screens and where there is poor compliance to follow up, this may pose a significant clinical risk to women. In one study in the UK,12 biopsy was performed in only 84% of high-grade smear referrals; however, the clinical risk to women in this setting could be minimised through adequate cytology and colposcopic follow up. Although almost all biopsy specimens in our study were reported as adequate by the local pathologists, the quality assurance pathologists scored one-fifth of the slides as inconclusive.

CIN was diagnosed in half of the women with colposcopy features of low-grade or worse lesions, and the proportion of women (60%) diagnosed with CIN among women with colposcopic features of high-grade or worse lesions was below the target of at least 65% set by NHSCSP.18 The low proportion of histologically diagnosed CIN might be due to several factors. First, this might have been due to overcalling of colposcopic lesions in order not to miss true positive disease. The colposcopists were advised to give a higher grade whenever they were doubtful. Second, as colposcopy is a subjective assessment, not all clinical appearances of CIN have distinctive colposcopic appearances, so interpretation error might have occurred.19 In a British Columbia study, only 76 and 50% of the abnormal cases on colposcopy had histologically confirmed CIN and CIN2 or worse disease; in a UK study, the corresponding figures were both 70%.12,15 These results indicate that, even by experienced colposcopists, not all CIN will have specific telltale features that enable their accurate diagnosis. Third, the colposcopist might not have sampled the most abnormal-looking area or the biopsy specimen might have been cross-cut. Some previous studies have suggested that high-grade CIN might not always be at the maximum colposcopic abnormal area.20–22 Several authors have suggested that taking multiple biopsies would improve on the accuracy of colposcopy performance.19,23,24 The best way to optimise biopsies still remains an issue. Fourth, histology, the gold standard, is itself subjective, hence false-negative and false-positive results might have occurred. In addition, this might partly reflect the suggestion that histological diagnosis be obtained for persistent low-grade abnormalities within 2 years of presentation. Definitely, a number of these women will have no cervical disease. However, this recommendation would work well in settings with well-organised cervical cancer prevention programmes that have good follow-up systems, which is far from reality in sub-Saharan countries.

The proportion of unsatisfactory colposcopy in our study fell in the lower range of that observed in other studies (range 1–23%).15,17,25 In our study, 77% of the participants were younger than 45 years, and the proportion of unsatisfactory colposcopy increased with age from less than 2% in women younger than 45 years to more than 7% in those aged 55 years and older. This explains the low proportion since unsatisfactory colposcopy is more likely to be in postmenopausal women.

The main objectives of the cross-sectional studies were to evaluate the accuracy of the screening tests and to provide affordable and accurate cervical cancer screening on a continuing basis after the study. It is therefore not surprising that the performance of colposcopy was more or less constant throughout the course of the study, reflecting the fact that the colposcopists might have been anxious not to miss lesions. In a study in New Zealand assessing nurse colposcopists in three study phases, correlation of colposcopy findings with results from histology and cytology improved with the study phase.14 It is important to note that there was moderate agreement when the histological reports from a sample of slides were compared between the local study pathologists and the pathologist from the reference laboratory in Lyon, France.

Understandably, there is a learning curve in acquiring colposcopy skills. Experienced providers are more likely to have a better performance than less experienced providers. To maintain these skills, a substantial workload of women must be seen yearly. For this reason, the NHSCSP stipulates that all practicing colposcopists must see at least 50 new abnormal cytology referrals per year and attend at least one British Society of Colposcopy and Cervical Pathology recognised colposcopy meeting every 3 years.18 Furthermore, proper training and certification of colposcopists are recommended with appropriate quality control tools. This training could be greatly aided through distance learning by telecolposcopy, where colpophotographs could be transferred through worldwide web now that many parts of the developing regions have access to internet services.

A high level of colposcopic practice depends on many other factors than those included in this study. Some of the fundamental components of high-quality practice are adequate counselling, communication with women, suitable and comfortable environment, good equipment, appropriate follow up, continual training and retraining of colposcopists and continually striving to improve. Even though restricted to few factors, this study shows that in Conakry (at colposcopically high-grade or worse lesions threshold), the standard of colposcopy was of high quality and that this can be achieved in low-resourced settings such as sub-Saharan Africa.

Although data from the early 1990s show that cervical cancer is among the top five causes of death among women aged 45–49 years and among the top ten for women aged 30–44 years in most developing countries, there are a number of other serious health needs that compete for available scarce resources.26 Cervical cancer control has been seen to be a moderately cost-effective intervention compared with other health interventions, and a very cost-effective intervention compared with other cancer control efforts.27 With innovative cheap techniques of cervical cancer screening, combined with colposcopy as part of reference investigations and use of acceptable and feasible treatment such as cryotherapy for women with high-grade CIN, cervical cancer interventions might become cheaper than interventions of other prevention/treatment conditions in the developing settings. Because of scarce resources, and for cervical cancer screening programmes to be beneficial, they need to begin on a small scale and be integrated with other reproductive health programmes and eventually expanded as resources allow.

Conclusions

Our findings provide evidence that colposcopic performance in developing countries can compare favourably with that in developed settings, provided there is adequate training, quality assurance and monitoring.12,15 Quality does not mean getting everything right but being open about the strengths and weaknesses of what health professionals do and being determined to inform and improve oneself. We believe that with appropriate and continued training, maintenance of high skills and aptitude for improvement combined with quality assurance measures, countries of sub-Saharan Africa and other regions of the developing world most hit by the burden of cervical cancer could attain high levels of colposcopy practice. As women in these regions are referred to colposcopy based on visual screening methods with a high positivity rate,5–7 it is vital to have a triaging investigation like colposcopy practised competently in public health settings.

Investigators of the IARC Multicentre Study Group on Cervical Cancer Early Detection

Bamako, Mali: Prof A Dolo, Prof S Bayo, Dr M Touré and Dr B Keita. Brazzaville, Congo: Prof C Gombe Mbalawa, Dr J Kokolo and Dr J Nsonde-Malanda. Conakry, Guinea: Prof N Keita, Prof M Koulibaly and Mr I Kabba. Lyon, France: Dr R Sankaranarayanan, Mr E Lucas, Dr B Fontanière, Dr L Frappart and Dr R Muwonge. Niamey, Niger: Dr M Nayama and Prof H Nouhou. Ouagadougou, Burkina Faso: Dr M Nacoulma and Dr B Sakande.

Disclosure of interest

We declare that none of us had any financial, personal, political, intellectual or religious conflicts of interest.

Contribution to authorship

R.M. participated in the analysis and interpretation of the data and writing of the initial draft of the manuscript. C.G.M., N.K., A.D., H.N., M.N., M.K. and S.B. participated in the conception and design of the study, and also in the monitoring, supervision, acquisition, analysis and interpretation of the data. J.N.M. participated in the acquisition, analysis and interpretation of the data. R.S. had the initial idea and was responsible for the conception and study design, monitoring, supervision, acquisition, analysis and interpretation of the data and writing of the initial draft of the manuscript; provided training of project staff.

Details of ethics approval

The study was approved by the Institutional Review Boards and Ethics Committees of IARC and the Collaborating National Institutions.

Funding

The study was funded by the Bill & Melinda Gates Foundation through the Alliance for Cervical Cancer Prevention, Seattle, WA, USA.

Acknowledgement

We thank Mrs Sue Anthony, IARC, for her assistance in preparing this manuscript.

Ancillary