059-116 24/01/05 11:05 Page 59

Chapter 2 Screening tests

Cervical cytology cytological findings into five categories as to what other non-neoplastic condi- or classes (Table 14) (Papanicolaou, tions were combined in Class II. Such 1954). At the time the classification was ambiguity in the Papanicolaou classifi- Cytological testing involves collection developed, there was only limited cation resulted in its non-uniform use by of exfoliated cells from the cervix and understanding of the relationship different cytologists. Modifications of the microscopic examination of these cells between cervical cancer precursor Papanicolaou classifications are still after staining. The concept of utilizing lesions and invasive cancers. More- used in some countries. In the exfoliative cytology to identify women over, invasive cervical cancer was com- Netherlands, a modified Papanicolaou with invasive cervical cancer was intro- mon and cervical cytology was initially system (CISOE-A) is used for classifica- duced by Papanicolaou and Babes in viewed as a way of detecting early- tion. This redefined and subdivided the the 1920s (Papanicolaou, 1928; stage, easily treated cancers. Therefore, Papanicolaou classes in order to make Papanicolaou & Traut, 1941). Subse- the Papanicolaou classification system the terminology correlate with histo- quently, Papanicolaou refined the tech- focused on how closely the exfoliated pathological terminology (Hanselaar, nique and demonstrated that conven- cells resembled those from an invasive 2002). tional cytology could also be used to cancer. Although the Papanicolaou identify precancerous lesions of the classification was modified many times World Health Organization terminology cervix (Papanicolaou, 1954). The shift over the years, the problems inherent in In the 1950s, some cytologists began in emphasis from using cytology as a this classification remain. For example, to promote a more scientifically accu- way to identify cases of invasive cervi- although is clear how Class I and Class rate terminology that would allow cyto- cal cancer to using it to identify women V translate into known histological enti- logical diagnoses to translate directly with high-grade precursor lesions who ties, Classes II, III or IV correlate less into histological diagnoses. This termi- are at risk for subsequently developing clearly with standard histopathological nology (Table 15) was later adopted by invasive cervical cancer was highly sig- lesions. For example, should a carci- the World Health Organization (WHO) nificant, as it meant that cervical cytol- noma in situ be classified as Class IV (Riotton et al., 1973). The WHO ogy could be used to actually prevent and all grades of dysplasia as Class III, terminology allows more precise the development of cervical cancer or does mild dysplasia correspond to correlation between cytological and rather than simply identify cases at an Class II? There was also no consensus histopathological findings, but is early stage. In the 1960s, cervical cytol- ogy began to be widely used in many Table 14. The original Papanicolaou classification developed countries as a technique for cervical cancer prevention. Although Class Description the method was introduced over a half century ago, cytology-based screening I Absence of atypical or abnormal cells programmes continue to be the main- II Atypical cytology, but no evidence for malignancy stay of cervical cancer prevention. III Cytology suggestive of, but not conclusive for, malignancy IV Cytology strongly suggestive of malignancy Cytological terminology V Cytology conclusive for malignancy Papanicolaou classes The terminology developed by From Papanicolaou, 1954 Papanicolaou separated cervical

59 059-116 24/01/05 11:05 Page 60

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 15. Comparison of different terminologies used for cytologic The Bethesda System terminology reporting By the late 1980s, advances in our understanding of the role of human Papanicolaou World Health CIN Bethesda System papillomavirus (HPV) in the pathogen- class system Organization esis of cervical cancer needed to be incorporated into cytological terminol- Class I Within normal limits ogy. Moreover, it was recognized that Class II clinicians were often confused by the Benign cellular changes non-standard terminologies used to ASC report cytological results and that this had a potential adverse impact on Mild dysplasia CIN1 Low-grade SIL clinical care.Therefore, in 1988, the US Class III Moderate dysplasia CIN2 National Institutes of Health held a Severe dysplasia CIN3 High-grade SIL conference in Bethesda, Maryland, to develop a new terminology that would Class IV Carcinoma in situ CIN3 ensure better standardization and accommodate current concepts of the Class V Microinvasive carcinoma Invasive Invasive carcinoma pathogenesis of cervical disease, Invasive carcinoma carcinoma so that cytological findings could be transmitted to clinicians as accurately Abbreviations: CIN, Cervical intraepithelial neoplasia; ASC, Atypical squamous cells; SIL, Squamous intraepithelial lesions and concisely as possible. The From Papanicolau (1954), Riotton et al. (1973), Richart (1968, 1973), Solomon et al. terminology that resulted is known as the (2002) Bethesda System. In 1991 the Bethesda System was slightly modified on the basis of experience obtained during the difficult to use since it includes a num- Cervical intraepithelial neoplasia (CIN) first three years of its use and it was ber of different entities. These are mild terminology further modified in 2001 to take into dysplasia, moderate dysplasia, severe As a result of advances in understand- account the results of new research and dysplasia, epidermoid carcinoma in ing of the pathogenesis of cervical over a decade of experience with the ter- situ, epidermoid carcinoma in situ with cancer, the cervical intraepithelial neo- minology (Luff, 1992; Solomon et al., minimal stromal invasion, invasive epi- plasia (CIN) terminology was 2002). dermoid microcarcinoma and invasive introduced in the late 1960s (Richart, The Bethesda system is viewed epidermoid carcinoma. Studies have 1968, 1973). The CIN concept empha- with caution in the United Kingdom, shown high rates of intra-observer and sized that dysplasia and carcinoma in which retains its own British Society for inter-observer variation with cervical situ represent different stages of the Clinical Cytology (BSCC) ‘dyskaryosis’ cytology in general (Yobs et al., 1987; same biological process, rather than terminology (British Society for Clinical Klinkhamer et al., 1988; Selvaggi, separate entities. It had a major impact Cytology, 1997). This can largely be 1999; Stoler & Schiffman, 2001). on how precancerous lesions were mapped to the Bethesda system for Classification systems that utilize more treated, since all types of cervical comparison of data in a research set- diagnostic categories have inherently cancer precursor were considered to ting, except for the borderline category, higher rates of variability than do classi- form a biological and clinical contin- which may include koilocytes. Due fication systems with fewer diagnostic uum. In the CIN terminology, mild largely to the robust nature of the categories (Yobs et al., 1987; Selvaggi, dysplasia is classified as CIN 1, ‘severe dyskaryosis’ category, fear of 1999; Stoler & Schiffman, 2001; Kundel moderate dysplasia as CIN 2 and increasing the overtreatment inherent & Polansky, 2003). Other limitations of severe dysplasia and carcinoma in situ in cervical screening and the difficulty the WHO terminology are that it does are grouped together and classified as of achieving inter- and intra-observer not adequately deal with non-neoplas- CIN 3 (Table 15). The CIN terminology agreement on ‘low-grade’ reports, the tic conditions nor with specimen ade- is still widely used in many countries United Kingdom continues to use this quacy. Despite its limitations, many for reporting both histological and terminology. cytologists around the world continue to cytological diagnoses. There are three distinct parts to utilize the WHO terminology. each Bethesda System report: a state-

60 059-116 28/01/05 12:00 Page 61

Screening tests

ment of the specimen adequacy, a A ‘satisfactory for evaluation’ spec- 98% for preparations with over 5000 general categorization and a descrip- imen must be appropriately labelled. To cells. tive diagnosis (Table 16). These cate- ensure proper identification, the There is much controversy over the gories assist clinicians by providing woman’s name or identifying number importance of identifying a transforma- answers to three basic questions: (1) should be written on, or affixed to, the tion zone component (e.g., squamous Do I need to repeat the cervical cytol- slide before it is sent to the cytology metaplastic cells) or endocervical cells ogy? (2) Was the cervical cytology laboratory. Cytology laboratories in a cervical cytology preparation. normal? (3) If the specimen was not should not accept unlabelled slides Because the majority of high-grade completely normal, what specifically and should return them to the submit- precursor lesions arise within the was wrong? ting clinician. It is also critical that transformation zone, it was widely Because cervical cytology is con- the smear-taker provide pertinent clini- believed until recently that specimens sidered a screening, rather than diag- cal information to the laboratory that lacking a transformation zone compo- nostic, test, the 2001 Bethesda will evaluate the specimen, including nent (TZC) or endocervical cells (EC) System reports cytological findings as the woman’s age, date of last men- should be considered somewhat less an ‘interpretation’ or ‘result’ rather than strual period, previous history of than ‘satisfactory for evaluation’. This as a ‘diagnosis’. This stresses the fact abnormal cervical cytology specimens view is supported by several studies that cytological findings usually need or treatment for cervical disease, and that have shown the prevalence of SIL to be interpreted in the light of clinical the source of the specimen (e.g., to be higher among cytology speci- findings, and that the test is designed vaginal or cervix). The minimal cellular mens that contain TZC/EC than to reflect the underlying disease state requirements for a specimen to be among those that do not (Vooijs et al., but does not always do so. considered ‘satisfactory for evaluation’ 1985; Mitchell & Medley, 1992; In this Handbook, the Bethesda in the 2001 Bethesda System Szarewski et al., 1993; Mintzer et al., System (SIL) terminology is used for vary depending on whether the speci- 1999). However, other studies have cytological interpretation of screening men is a conventional cytology speci- failed to confirm this association and, tests unless otherwise reported. men or a liquid-based cytology speci- perhaps more importantly, several ret- men. For classification of conventional rospective longitudinal cohort studies Specimen adequacy cytology specimens as ‘satisfactory for have found that women lacking The 2001 Bethesda System requires evaluation’, an estimated 8000 to TZC/EC are no more likely on follow- that every cervical cytology specimen 12 000 well visualized squamous cells up to be diagnosed with squamous be assessed with respect to its ade- need to be present. For liquid-based lesions than are women whose speci- quacy (Solomon et al., 2002). cytology specimens, an estimated mens contain TZC/EC (Mitchell & Specimens are classified into one of 5000 cells need to be present (Figure Medley, 1991; Mitchell, 2001). One ret- two categories: ‘satisfactory for evalua- 29). Although the selection of these rospective case–control study of true tion’ or ‘unsatisfactory for evaluation’. cut-offs is fairly arbitrary, the limit of This represents a departure from the 5000 cells for a liquid-based cytology 1991 Bethesda System, which also specimen to be classified as ‘satisfac- included a third category for specimen tory for evaluation’ is based on a cell- adequacy that was called ‘satisfactory counting study in which referent for evaluation but limited by…..’ or samples were diluted to produce SBLB. This ‘satisfactory but limited preparations with defined numbers of by…’ category was most frequently squamous cells (Studeman et al., used when a specimen lacked either 2003). A clear demarcation in endocervical cells or squamous meta- sensitivity was observed using the plasia from the transformation zone but SurePath™ procedure (see below) was in all other aspects ‘satisfactory’. between specimens with less than With the 2001 Bethesda System, 5000 squamous cells and those with cytology specimens previously classi- 5000 cells or more: the sensitivity for a fied as ‘SBLB’ are classified as ‘satis- reference diagnosis case of low-grade Figure 29 Liquid-based cytology: factory for evaluation’ and a quality squamous intraepithelial lesion (LSIL) superficial and intermediate squa- indicator comment is made indicating increased from 73% for specimens mous cells and a cluster of columnar what limiting features are present. with less than 5000 squamous cells to endocervical cells (obj. 5x)

61 059-116 24/01/05 11:05 Page 62

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

positive and false negative cervical squamous metaplastic cells. Speci- glandular cell abnormalities. This cate- cytology specimens from women with mens in which inflammation, blood or gory is used whenever there are CIN 3 found no difference in true posi- poor preservation cause 50–75% of epithelial cell abnormalities, except for tive rates between cases with or with- the epithelial cells to be obscured benign reactive or reparative changes. out TZC/EC (O’Sullivan et al., 1998). A should be classified as ‘satisfactory for The 2001 Bethesda System intro- prospective study of women with nor- evaluation’, but a quality indicator com- duced a new general categorization, mal cytology at entry found that ment made indicating that there are ‘other’. This category is used whenever although specimens containing EC at partially obscuring factors. there are no morphological abnormali- the subsequent test were at signifi- A specimen is classified as ‘unsat- ties in the cells per se, but there are cantly higher risk of both low- and isfactory for evaluation’ when either the findings indicative that the woman is at high-grade squamous intraepithelial minimal number of epithelial cells some increased risk. An example is lesions than those without EC, the required for interpretation is not pre- when benign-appearing endometrial presence or absence of EC at entry sent or blood, inflammation or poor cells are identified in a woman 40 had no significant effect (Mitchell, preservation obscures more than 75% years of age or older. 2001). In another compelling study on of the epithelial cell component (Figure the lack of importance of EC, all nega- 30). Cases which the laboratory can- Squamous cell abnormalities tive cervical cytology specimens not process, such as those received Atypical squamous cells (ASC): obtained in the Netherlands between unlabelled, are also classified as Epithelial cell abnormalities are subdi- 1990 and 1991 were matched with ‘unsatisfactory for evaluation’ and no vided into four categories (Table 16). results of subsequent cytological and interpretation is rendered. ‘Atypical squamous cells’ (ASC) is histological examinations (Bos et al., used when cytological findings are 2001). There was no significant differ- General categorization considered suggestive but not diag- ence in the number of women subse- The ‘general categorization’ is included nostic of a squamous intraepithelial quently diagnosed with CIN between as an optional component of the lesion (SIL) (Figure 31). The term ASC women whose initial cytology speci- Bethesda System to allow clinicians to was retained in the 2001 Bethesda mens contained EC and those that did readily determine whether any degree System because of the wide recogni- not. Moreover, the proportions of of abnormality is present. With the tion that these cells imply a significant women diagnosed with cervical cancer 2001 Bethesda System, all cytology risk for an underlying high-grade cervi- were the same in both groups. It is also specimens are classified into one of cal intraepithelial lesion (SIL). In vari- important to recognize that EC are less three general categories.These include ous studies, the prevalence of CIN 2 or frequently found in cervical cytology ‘negative for intraepithelial lesion or 3 in women with ASC has varied specimens from women using oral malignancy’, ‘epithelial cell abnormali- between 10% and 20% (Wright et al., contraceptives, who are pregnant or ties’ and ‘other’. These categories are 2002a). The ASC category roughly who are postmenopausal (Davey et al., mutually exclusive and specimens correlates with the ‘borderline 2002). It has therefore been argued should be categorized according to the dyskaryosis’ category used in the that specimens lacking EC or a TZC most significant findings. should not be considered unsatisfac- ‘Negative for intraepithelial lesion tory and may not need to be repeated or malignancy’ includes all specimens (Davey et al., 2002; Birdsong, 2001; in which no intraepithelial lesion or Bos et al., 2001). The 2001 Bethesda malignancy is identified. This includes System recommends that reports cases with common infections such as should state whether or not EC or a Trichomonas vaginalis, fungal organ- TZC are present. Specimens lacking isms such as Candida species, endocervical cells or squamous meta- Actinomyces or herpes simplex virus, plastic cells should be classified as a shift in bacterial flora consistent with ‘satisfactory for evaluation’ and the bacterial vaginosis, reparative/reactive quality indicator comment should indi- changes, changes associated with cate that these components are not intrauterine devices, radiation reac- Figure 30 Unsatisfactory smear present. The numeric criterion for stat- tions or atrophic changes. because of inflammation. Cell cluster ing that such a component is present is The category ‘epithelial cell abnor- difficult to analyse. Repeat after local 10 well preserved endocervical or malities’ includes both squamous and treatment (obj. 10x)

62 059-116 27/01/05 15:27 Page 63

Screening tests

Table 16. The 2001 Bethesda system minor cytological abnormalities. The term ASC should be used only when Specimen adequacy the cytological findings are suggestive, but not diagnostic, of SIL. Currently, Satisfactory for evaluation (note presence/absence of endocervical transformation approximately 4–5% of all cervical zone component) cytology specimens are classified as Unsatisfactory for evaluation (specify reason) - Specimen rejected/not processed (specify reason) ASC in the USA (Jones & Davey, 2000). - Specimen processed and examined, but unsatisfactory for evaluation of epithelial The ‘atypical squamous cell’ cate- abnormality because of (specify reason) gory is formally subdivided into two subcategories: ‘atypical squamous General categorization (optional) cells – of undetermined significance’ Negative for intraepithelial lesion or malignancy (ASCUS or ASC-US) and ‘atypical Epithelial cell abnormality squamous cells – cannot exclude a Other high-grade SIL’ (ASC-H). This subdivi- Interpretation/result sion was felt to be important because Negative for intraepithelial lesion or malignancy Organisms Trichomonas vaginalis Fungal organisms morphologically consistent with Candida species Shift in flora suggestive of bacterial vaginosis Bacteria morphologically consistent with Actinomyces species Cellular changes consistent with herpes simplex virus Other non-neoplastic findings (Optional to report; list not comprehensive) Reactive cellular changes associated with inflammation (includes typical repair), radiation, intrauterine contraceptive device Glandular cells status posthysterectomy Atrophy

Epithelial cell abnormalities Squamous cell Atypical squamous cell (ASC) of undetermined significance (ASCUS) Figure 31 Parakeratotic cell (arrow), cannot exclude HSIL (ASC-H) with an eosinophilic cytoplasm denser Low-grade squamous intraepithelial lesion (LSIL) than normal superficial cells and a rel- High-grade squamous intraepithelial lesion (HSIL) (can use modifiers to separate into CIN 2 and CIN 3) atively regular but enlarged nucleus: Squamous-cell carcinoma ASCUS (rule out LSIL) (obj. 10x) Glandular cell Atypical glandular cells (AGC) (specify endocervical, endometrial or not otherwise specified) Atypical glandular cells, favour neoplastic (specify endocervical or not women with ASC-H (Figure 32) are at otherwise specified) considerably higher risk for having CIN Endocervical adenocarcinoma in situ (AIS) 2 or 3 and of being high-risk HPV Adenocarcinoma DNA-positive than are women with Other (List not comprehensive) ASCUS (Genest et al., 1998; Sherman Endometrial cells in a woman ≥ 40 years of age et al., 1999, 2001; Selvaggi, 2003). Information from the US National From Solomon et al. (2002) Cancer Institute ASCUS–LSIL Triage Study (ALTS) clinical trial indicates that the risk that a woman with ASC-H has United Kingdom. However, neither the also clearly separates ASC from reac- CIN 2 or 3 is over twice that of a WHO terminology nor the CIN termi- tive/reparative changes and an interpre- woman with ASCUS (Table 17) nology incorporates a category similar tation of ASC should not be made (Sherman et al., 2001). Moreover, the to ASC. The 2001 Bethesda System whenever a cytopathologist identifies prevalence of high-risk HPV DNA-

63 059-116 28/01/05 16:46 Page 64

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 17. Prevalence of high-risk HPV DNA and CIN 2 and CIN 3 in women may vary in size and, in many cases of with ASCUS and ASC-H in the ASCUS-LSIL triage study (ALTS)* LSIL that are characterized by marked HPV cytopathic effects, are only twice Cytology No. % high-risk HPV % biopsy- % biopsy- the size of a normal intermediate-cell result DNA-positive confirmed CIN 2+ confirmed CIN 3+ nucleus. The nuclei are usually hyper- chromatic, and multinucleation is com- ASCUS 764 63.2% 11.6% 4.7% mon. The chromatin is finely granular ASC-H 116 85.6% 40.5% 24.1% and uniformly distributed. The cells typ- HSIL 213 98.7% 59.2% 37.6% ically occur as individual cells or as sheets of cells with well defined cell * Study provides the results for liquid-based cytology specimens that were tested for borders. high-risk types of HPV using Hybrid Capture 2 High-grade squamous intraepithe- From Sherman et al. (2001) lial lesion: Because the Bethesda System combines moderate and severe dysplasia together with carci- to as ‘koilocytosis’, a term derived from noma in situ in the HSIL category, the Greek koilos, meaning hollow. there is wide variation in the cytological The classical studies of Reagan appearance of HSIL. When applying and others identified the key cytologi- the 2001 Bethesda System, many cal features of CIN 1 (Table 18) cytopathologists utilize the option of (Reagan & Hamomic, 1956). The cells subdividing HSIL into CIN 2 and CIN 3 are of the superficial or intermediate- lesions. As the severity of the lesion cell type. They are classically increases, the degree of differentiation described as having nuclei 4–6 times and the amount of cytoplasm the size of a normal intermediate-cell decreases, the nuclear:cytoplasmic nucleus (Figure 33). However, nuclei ratio increases, and the degree of

Figure 32 Inflammatory smear with Table 18. Cytological features of squamous intraepithelial lesions parabasal squamous cells with enlarged nuclei: ASC-H (ellipse) (obj. Bethesda system LSIL HSIL 20x) CIN terminology CIN 1 CIN 2 CIN 3

positivity among women with ASC-H is WHO terminology Mild Moderate Severe Carcinoma in situ almost as high as that of women with a dysplasia dysplasia dysplasia high-grade squamous intraepithelial lesion (HSIL) cytological result. Cell type Superficial Parabasal Basal Basal, spindle, Therefore the recommended manage- or intermediate pleomorphic ment of women with ASCUS and ASC- Cell arrangement Singly or sheets Singly or Singly or Singly or sheets H differs (Wright et al., 2002a). sheets sheets or syncytia Low-grade squamous intraepithe- lial lesion: The LSIL category in the Number abnormal + ++ +++ ++++ Bethesda System includes both HPV Koilocytosis +++ + +/– +/– effects and CIN 1 (i.e. mild dysplasia). Most cytologists consider Nuclear size +++ ++ + + the cytopathic effects of HPV, including Hyperchromasia + ++ +++ ++++ multinucleation, perinuclear halos and nuclear atypia with irregular Nuclear:cytoplasmic + ++ +++ ++++ nuclear outlines and hyperchromasia, ratio to overlap the cytological features of CIN 1. These features are referred From Reagan & Hamomic, 1956

64 059-116 24/01/05 11:05 Page 65

Screening tests

demonstrate minimal differentiation, of carcinoma in situ is composed of the nuclear:cytoplasmic ratio is greatly pleomorphic, highly atypical cells, increased. In severe dysplasia, there many of which have thick keratinized are usually considerably greater num- cytoplasm. These cells are often spin- bers of neoplastic cells that are typi- dled or tadpole-shaped and have cally found individually. Carcinoma in extremely dense nuclear chromatin situ can be of the small-cell type, of the (Figure 36). large-cell non-keratinizing type or of Invasive squamous-cell carcinoma: the large-cell keratinizing (pleomor- Cytologically, squamous-cell carcino- phic) type. Although separation of car- mas of the cervix are subdivided into cinoma in situ into these three different keratinizing and non-keratinizing types. cytological types has little clinical sig- Non-keratinizing carcinomas (Figure nificance, all three have quite different 37) typically have large numbers of cytological appearances. Small-cell malignant cells that form loose cell Figure 33 LSIL: typical eosinophilic lesions consist of small basal-type sheets and syncytial arrangements. and basophilic koilocytes associated cells similar to those seen in severe The cells have enlarged nuclei with with some parakeratosis and binucle- dysplasia but which demonstrate even coarsely clumped chromatin, promi- ated cells (obj. 20x) less cytoplasm and higher nu- nent macronucleoli and focal chro- clear:cytoplasmic ratios (Figure 35). matin clearing. A key cytological fea- Because of their small size, these cells ture is the presence of a ‘dirty’ back- nuclear atypia also increases. HSIL of can easily be overlooked during rou- ground containing blood and necrotic the moderate dysplasia type typically tine screening and such cases account material. This is often referred to as a contains cells similar to those seen in for a disproportionate percentage of tumour diathesis. This characteristic LSIL, as well as atypical immature false negative cytological results. The background is usually less prominent cells of the parabasal type (Figure 34). cells of large-cell non-keratinizing in liquid-based cytology specimens. The nuclei of these cells are more lesions typically form syncytial-like cell Cervical smears from women with hyperchromatic and irregular than typi- sheets in which individual cell mem- keratinizing carcinomas contain malig- cally seen in LSIL. In severe dysplasia, branes are difficult to identify. These nant cells of a variety of shapes and the overall size of the cells is reduced cells have enlarged, hyperchromatic sizes (Figure 38). Some of the cells are compared to mild and moderate nuclei and minimal amounts of cyto- pleomorphic or tadpole-shaped with dysplasia, but because the cells plasm. The keratinizing large-cell type nuclei that are irregular in shape and

Figure 34 Parabasal cells arranged in Figure 35 HSIL (severe dysplasia): Figure 36 HSIL (severe dysplasia): a pile with nuclear enlargement, irreg- inflammatory smear containing many basal cells with enlarged nuclei and ular nuclear outlines and coarse chro- parabasal cells with enlarged nuclei irregular or very dense and opaque matin. HSIL (moderate dysplasia) (obj. with irregular chromatin (black arrow). chromatin (arrow), accompanied by an 20x) Some cells with a mildly eosinophilic atypical mature cell (obj. 40x) cytoplasm (ellipse) (obj. 20x)

65 059-116 24/01/05 11:05 Page 66

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

squamous abnormalities and most cytologists tend to be less comfortable recognizing and diagnosing them. In addition, the criteria used to differenti- ate reactive endocervical changes from neoplasia are less well estab- lished than those used for squamous lesions. Cytologists even have difficulty in differentiating atypical endocervical cells from cases of CIN 2 or CIN 3 that have extended into endocervical crypts. This accounts for the high Figure 37 Invasive squamous cell Figure 39 Smear from the transfor- prevalence of squamous abnormalities carcinoma mation zone and endocervix (approximately 30%) detected in One cluster of pleomorphic and poorly Sheets of atypical glandular cells women referred to colposcopy for AGC differentiated malignant cells and one (AGC) with enlarged nuclei with simi- (Eddy et al., 1997; Veljovich et al., isolated cell of abnormal shape lar chromatin pattern in all cells (A and 1998; Ronnett et al., 1999; Jones & (arrow). Inflammation, blood and B: obj. 20x) Davey, 2000; Krane et al., 2004). necrosis in the background (obj. 20x) The cytological features of atypical glandular cells vary depending on the degree of the underlying histopatho- logical abnormality and whether or not the cells are endocervical or endome- trial in origin. Atypical glandular cells of endocervical origin frequently form dense two- or three-dimensional aggregates that have minor degrees of nuclear overlapping. In some cases, the chromatin is somewhat granular and nuclear feathering can be seen at the periphery of the cellular aggre- gates (Figure 39). In cases interpreted as atypical glandular cells – favour Figure 38 Invasive squamous cell Figure 40 Endocervical adenocar- neoplasia, there is more marked cyto- carcinoma cinoma in situ (AIS) logical abnormality and typically a Pleomorphic malignant cells, isolated Atypical columnar endocervical cells, greater number of abnormal cells. or in clusters, sometimes keratinized with enlarged, elongated and hyper- Adenocarcinoma in situ: In cases or necrotic with bizarre cell shapes chromatic nuclei. Typical feathering of adenocarcinoma in situ, there are (arrow). Inflammation, blood and and palisading. (obj. 20x) usually a larger number of atypical necrosis in the background (obj. 10x) glandular cells that form crowded cel- lular clusters (Figure 40). The sheets are usually three-dimensional. The quite hyperchromatic. Unlike non-kera- glandular cells (AGC), atypical glandu- cells within these sheets occasionally tinizing squamous-cell carcinoma, lar cells – favour neoplasia, adenocarci- form rosettes and have extensive keratinizing squamous-cell carcinomas noma in situ and adenocarcinoma. feathering of the cells at the periphery. usually do not have a ‘dirty’ background Whenever possible, atypical glandular Individual endocervical cells are highly or evidence of tumour diathesis. cells are categorized as to whether they atypical with enlarged round, oval or are endocervical or endometrial in origin. elongated nuclei that vary in size from Glandular cell abnormalities Atypical glandular cells (AGC): cell to cell. In most cases, the chro- Glandular cell abnormalities are cate- Glandular cytological abnormalities matin is coarsely clumped and multiple gorized into four categories: atypical are considerably less common than mitoses are seen.

66 059-116 24/01/05 11:05 Page 67

Screening tests

Adenocarcinoma: Invasive adeno- Conventional cervical cytology actively menstruating, blood and cellu- carcinomas should be subclassified The importance of proper specimen lar debris from the endometrium tend into the endocervical or endometrial collection cannot be overemphasized. to obscure the cells on the smear, par- type whenever possible. The cytologi- Although no formal studies have ticularly during the first few days. cal diagnosis of invasive adenocarci- demonstrated that educating clinicians Similarly, a cytology specimen should noma is relatively straightforward. on the optimal technique of obtaining not be obtained when an abnormal Adenocarcinoma cells from either an cervical cytology samples improves vaginal or cervical discharge is endocervical or an endometrial pri- specimen quality, there is considerable observed. Women with a discharge mary type have enlarged nuclei, high anecdotal evidence that this is impor- should be evaluated for cervicitis and nuclear:cytoplasmic ratios, coarsely tant (Krieger et al., 1998). One half to vaginitis using appropriate tests and clumped chromatin and prominent two thirds of false negative cervical be treated before the cytology speci- nucleoli (Figure 41). They can occur cytology results are attributable to men is taken, otherwise the specimen singly or in clusters. either poor patient conditions at the may be compromised by the inflamma- time the cervical specimen is collected tory exudates or mildly reactive cells Other terminologies or the manner in which it is collected may be misinterpreted as a significant Although the 2001 Bethesda System (Morell et al., 1982; Gay et al., 1985; cytological abnormality. classification is applied in many coun- Vooijs et al., 1985; Agency for Health There is controversy as to the ideal tries, other classification systems are Care Policy and Research, 1999). timing of post-partum smears. Smears also widely used. As mentioned Therefore it is important that clinicians obtained less than eight weeks post- previously, many countries prefer to and nurses obtaining specimens be partum are often difficult to interpret subclassify high-grade intraepithelial adequately trained in specimen collec- because of marked inflammation and lesions into at least two categories. tion and that they avoid situations that reparative changes, so a high rate of This is the approach used in the United may reduce the performance of the mild cytological abnormalities may be Kingdom, where squamous intra- test (McGoogan et al., 1998). This is diagnosed. Another factor that can epithelial abnormalities are divided into especially important in low-resource adversely affect the interpretation of five categories (borderline changes; settings, where women may undergo cervical cytology specimens is severe mild, moderate, severe dyskaryosis screening only once or twice in their atrophy. and severe dyskaryosis or possibly lifetime. Although one should strive to col- invasive cancer) (British Society for lect specimens under ideal conditions, Clinical Cytology, 1997). Preparing the woman failure to comply with suggested Whenever possible, appointments for a screening intervals presents a greater cervical cytology examination should risk to women. For previously non- be scheduled approximately two compliant women, particularly those at weeks after the first day of the last risk for cervical neoplasia, a smear menstrual period. Patients should be obtained under less than ideal condi- instructed to avoid sexual intercourse tions is preferable to no smear at all. and douching for 24 to 48 hours before having the cytology specimen col- Equipment lected. In addition, women should not To collect a conventional cervical cyto- use any intravaginal products or medi- logical specimen, the equipment cine for several days before the smear required is a speculum, a light source, is taken. Women using an intravaginal a collection device, a glass slide and estrogen product should discontinue fixative. Since most cervical cancer its use several days before the exami- precursors and invasive cancers occur nation. in the transformation zone, the use of Circumstances that may interfere specially designed devices that sample Figure 41 Histologically proven inva- with the interpretation of a cervical this area is recommended. The most sive adenocarcinoma cytology test include active menstrua- common is a wooden or plastic spatula More or less cohesive mallignant tion, significant cervical or vulvovaginal that conforms to the curvature of the columnar cells next to a less atypical infections and a timing less than eight portio. It is critical that the endocervical cell group (obj. 40x) weeks post-partum. When a woman is canal be sampled in order to obtain

67 059-116 24/01/05 11:05 Page 68

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

reasonable sensitivity (Martin-Hirsch that a test requisition is accurately and the cervix with a large cotton-tipped et al., 1999) and many spatula-type legibly filled out before collecting the applicator can produce a better-quality devices have extended tips designed specimen. The information most com- smear (Kotaska & Matisic, 2003), but to collect cells from this area. Either a monly requested by laboratories vigorous cleansing may remove many moistened cotton swab or a brush-type includes: of the most easily exfoliated cells. endocervical sampler device (e.g., Saline should not be used to help clear cytobrush) can be used to collect a • Woman's name and indication if debris from the surface of the cervix. It second sample directly from the endo- there has been a name change in is also preferable not to apply 3–5% cervical canal after the portio has been the last five years. Some laborato- acetic acid to the cervix before taking sampled (Koonings et al., 1992; ries also use unique patient identi- the cytology specimen, as this can Kohlberger et al., 1999). Recently fier numbers reduce the cellularity of the smear and developed collection devices that sam- produce poor staining (Griffiths et al., ple the endocervix and exocervix • Date of birth or age 1989; Cronje et al., 1997). simultaneously do not provide a signif- Before the specimen is collected, icantly lower false negative rate than • Menstrual status (date of last men- the cervix should be carefully the combination of spatula and a coni- strual period, whether the woman inspected with the naked eye for cal cervical brush (Szarewski et al., is pregnant, post-partum, on hor- grossly visible masses or ulcerations 1993). mone replacement therapy, or has that may indicate an invasive cervical There is no consensus as to had a hysterectomy) cancer. If a grossly visible lesion is whether a single-slide technique, with identified, the woman should be both samples of the ectocervix and • Previous history of abnormal cervi- referred for further confirmation. In endocervix placed on the same slide, cal cytology, or treatment for CIN or many cases, the lesion can be directly or a technique in which the two sam- cancer sampled and the cellular sample ples are put on two separate slides is obtained can be submitted separately preferable. Comparative studies of the • Whether the clinician considers the for cytological assessment. The proce- two techniques have reported similar woman to be at high risk for devel- dure for collecting cells from the cervix results (Saitas et al., 1995; oping CIN or cancer. Possible risk varies depending on the type of device Quackenbush, 1999). The single-slide factors include smoking, infection used and the number of slides to be approach has the advantage of reduc- with HIV, lack of previous screening prepared. If a spatula and conical cer- ing screening time and laboratory and multiple partners. vical brush are utilized, the first step is workload and it decreases the storage to place the spatula firmly against the space required for archiving slides. • Specimen source – vaginal or cer- ectocervix with the long projection When a single-slide technique is uti- vical extending into the endocervical canal. lized, there also is no consensus on The spatula is then rotated several whether the specimens from the ecto- Good visualization of the cervix is times 360° around the portio and cervix and endocervix should be mixed important for obtaining an adequate removed. It is important to ensure that together on the slide or kept separate specimen. Cervical cytology speci- the entire squamocolumnar junction is as in the V (vagina) C (ectocervix) E mens are generally collected with the sampled, since this is the site where (endocervix) technique. woman in the dorsolithotomy position. most CIN lesions develop. In most A sterilized or single-use bivalve women, the spatula will come into con- Collecting the sample speculum of appropriate size is tact with the squamocolumnar junction A conventional cytology specimen is inserted into the vagina in such a man- if the pointed end is placed in the os, typically obtained using a spatula and ner as to allow complete visualization but in young women with a large conical cervical brush. The slide must of the cervical os and as much of the ectopy, the spatula may need to be first be labelled with the woman's transformation zone as possible. The moved laterally to sample a peripher- name or number. Laboratories should cervix should not be contaminated with ally positioned squamocolumnar junc- have a written protocol specifying what lubricant or water-soluble gel that may tion. When rotating the spatula, it is is considered adequate labelling and obscure the smear. Therefore the easy to miss part of the cervix; this can should not accept inadequately smear must be obtained before any be alleviated by directly visualizing the labelled specimens. The person col- bimanual examination. Gentle removal cervix while sampling. Transfer is best lecting the specimen should ensure of excess mucus and discharge from performed by using the spatula to thinly

68 059-116 27/01/05 15:31 Page 69

Screening tests

spread the cells onto the glass slide. It container of alcohol or allowed to fix for and specificity of cervical cytology are is important to ensure that as much cel- 20 to 30 minutes in the alcohol, lower than previously thought (Fahey lular material as possible is transferred removed and allowed to air-dry. et al., 1995; McCrory et al., 1999; from both sides of the spatula. Various different spray fixatives are Nanda et al., 2000). [The Working The endocervical canal is then available. Only spray fixatives specifi- Group considered the estimates of sampled, using a conical cervical cally designed for cytological speci- cytology test performance obtained brush, which is placed in the endocer- mens should be used and the manu- through these meta-analyses to be of vical canal so that the last few bristles facturer's instructions for a given prod- concern, given current cytology prac- remain visible and then gently rotated uct must be followed. The fixative tices. In particular, it felt that it is very 90° to 180° once. One such rotation should be liberally applied such that unlikely that specificities as low as will adequately sample the endocervi- the slide appears moist over its entire 60–70% would be observed in a mod- cal canal and generally does not pro- surface. In order to prevent ern cytological screening practice.] duce bleeding. Material from both disruption of the cellular layer on the Table 19 presents the sensitivities and sides of the spatula should be spread slide, the container of spray fixative specificities of conventional cervical onto the slide. should generally be held 15–25 cm cytology observed in a number of If collection devices that simultane- from the slide during application. recent large cervical cancer screening ously sample both the endocervix and studies. Even within the confines of the ectocervix are used, the manufac- Performance of conventional cytology research studies, a wide range of per- turers' directions should be followed for Despite the proven effectiveness of formance has been reported. each type of device. cervical cytological screening in reduc- Cell fixation must be performed ing the incidence of cervical cancer, Liquid-based cervical cytology within a few seconds of specimen col- over the last decade the accuracy of Liquid-based cytology (LBC) was intro- lection in order to prevent air-drying, cervical cytology has been questioned. duced in the mid-1990s as a way to which obscures cellular detail and hin- Two factors need to be considered improve the performance of the test. ders interpretation (Somrak et al., when assessing the accuracy of any Rather than having the clinician pre- 1990). Immersing the slide in alcohol screening or diagnostic test. One is pare the cytological specimen at the or spraying it with a specially formu- whether the test is specific in detecting bedside by spreading the exfoliated lated spray fixative can prevent air-dry- a given condition; the other is the sen- cells onto a glass slide, the cells are ing. With immersion fixation, the slide sitivity of the test for detecting the con- transferred to a liquid preservative is either immersed in alcohol and dition. Several large meta-analyses solution that is transported to the labo- transferred to the laboratory in the have indicated that both the sensitivity ratory, where the slide is prepared.

Table 19. Performance of conventional cytology in various large research studies

Author Country Ages Study Sensitivity (%) Specificity (%) Histological size cut-off Cuzick et al. (1999a) United Kingdom 34+ 2988 86 98 CIN 2+ Hutchinson et al. (1999) Costa Rica 18+ 8636 55 98 CIN 2+ Ratnam et al. (2000) Canada 18–69 2098 56 62 CIN 2+ Denny et al. (2000a)* South Africa 35–65 2944 70 85 CIN 2+ Denny et al. (2002)* South Africa 35–65 2754 40 96 CIN 1+ Cuzick et al. (2003) United Kingdom 30–60 11 085 77 96 CIN 1+ Petry et al. (2003) Germany 30+ 8466 44 98 CIN 2+ Salmerón et al. (2003) Mexico 15–85 7868 59 98 CIN 1+ Sankaranarayan et al. (2004b) India 25–65 10 591 65 92 CIN 2+

Cytological cut-off for referral for all studies is ASC or greater except for those studies marked by asterisk, where a cut-off of LSIL or greater was used. Sensitivity and specificity are estimated cross-sectionally (see Chapter 4)

69 059-116 24/01/05 11:05 Page 70

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

A number of different LBC tech- tory cytology specimens, the availabil- prepared second, this design would niques are in use worldwide. These ity of residual cellular material for sub- seem to lead inherently to bias against include ThinPrep®, SurePath™, sequent molecular testing or for mak- LBC. Therefore it has been argued that Cytoscreen™, Cyteasy®, Labonord ing additional glass slides, and possi- split-sample studies do not demon- Easy Prep, Cytoslide, SpinThin and bly increased detection of HSIL. strate the full benefit that could be PapSpin. The first two of these are obtained when LBC is utilized in rou- approved for use in the USA by the Performance of liquid-based cytology tine clinical practice. Studies utilizing Food and Drug Administration (FDA) methods historical controls avoid the need to and are the most widely used methods Numerous studies have evaluated the prepare several cytology specimens worldwide. They are therefore the best comparative performance of the two from a single woman, but introduce characterized in terms of performance. most commonly used LBC methods other potential biases, including the With the ThinPrep method, clumps of (ThinPrep and SurePath) and conven- comparability of the populations being cells and mucus are broken up by tional cytology with respect to test pos- compared. mechanical agitation and then the itivity, their sensitivity and specificity for Other significant limitations found liquid preservative solution is filtered identification of CIN, the time required in many of the studies evaluating LBC through a membrane filter with a pore for evaluation of the specimens, and include failure to compare test perfor- size specifically designed to trap specimen adequacy. Although there is mance with a reference standard of epithelial cells while allowing contami- reasonable agreement that LBC ‘blinded’ colposcopy/biopsy and a nating red blood cells and inflamma- improves specimen adequacy and study population of women followed up tory cells to pass through. The epithe- reduces screening time compared to for a prior cytological abnormality lial cells collected on the membrane conventional cytology, there is consid- rather than women undergoing routine filter are then transferred onto a glass erable controversy surrounding the rel- screening. A review of new cervical slide and stained. This produces a ative sensitivity and specificity of the cytology methods conducted in 2001 relatively thin, monolayer-type prepa- two approaches, largely due to a lack for the US Preventive Services Task ration. The ThinPrep-2000 processor of well designed comparative studies. Force and the Agency for Healthcare allows one specimen to be processed Most comparative studies have uti- Research and Quality found that out of at a time, whereas the newer lized one of two types of study design: 962 potentially relevant studies, not ThinPrep-3000 processor is more fully split-sample studies and historical con- one met their predefined inclusion cri- automated and allows up to 80 sam- trol studies. Split-sample studies col- teria (Hartmann et al., 2001). This was ples to be processed at a time. In con- lect cells from the cervix using a single commonly due to lack of an adequate trast, with the SurePath method, collection device and a conventional reference standard, but most studies clumps of cells and mucus are broken cervical cytology specimen is prepared were excluded for more than one rea- up by aspiration through a syringe. The first. Residual cells remaining on the son. At the time the review was con- cell suspension is then layered on top device are then transferred to a liquid- ducted, only one study, from Costa of a density gradient and the red blood based cytology preservative. Therefore Rica, had applied a definitive clinical cells and inflammatory cells are sepa- each woman acts as her own control reference standard to a random sam- rated from the epithelial cells by den- and detection rates in conventional ple of women with normal screening sity gradient centrifugation. The result- and LBC specimens are compared. test results and allowed the relative ing cell pellet containing predominantly The other widely used study design, sensitivity and specificity for LBC and epithelial cells is then inserted into a known as ‘direct to vial’, compares the conventional cervical cytology to be robotic workstation, where it is resus- performance of LBC collected in the calculated (Hutchinson et al., 1999). pended and transferred to a glass routine manner (direct transfer to the The Costa Rica study was a split-sam- microscope slide. The SurePath preservative solution) during a given ple study rather than a direct-to-vial method allows up to 48 samples to be time period with historic control data study. The other studies that were processed at a time. obtained using conventional cytology. reviewed used various types of clinical LBC is purported to have a number Both study designs have significant reference standard, including a of advantages over conventional cervi- limitations. With split-sample studies, it combination of histological follow-up cal cytology.These include a more rep- is difficult to ensure that the two cytol- and conventional cytology follow-up resentative transfer of cells from the ogy specimens are comparable. Since with incomplete data, a consensus collection device to the glass slide, a the conventional cytology slide is pre- expert panel diagnosis of the index reduction in the number of unsatisfac- pared first and the LBC specimen is specimens, histological follow-up or

70 059-116 24/01/05 11:05 Page 71

Screening tests

consensus expert panel diagnosis in findings, it is quite possible that the specificities of histologically confirmed cases of missing follow-up, and histo- interpretation of cervical biopsy speci- lesions. They have come to somewhat logical follow-up of HSIL combined mens will be biased. Large, random- conflicting conclusions (Table 22). It is with a balanced follow-up diagnosis of ized controlled clinical trials comparing important to note that the comparative all other available follow-up data. The the performance of LBC and conven- utility of LBC relative to conventional most common reference standard tional cytology need to be conducted cervical cytology will vary from one used in studies of LBC performance by laboratories in which the techniques setting to another. The National Health has been the expert panel review of are well established. Although the Service of the United Kingdom selected cytology specimens. results of no such studies are yet avail- recently agreed to introduce LBC Unfortunately, with expert panel able, one large randomized trial is cur- throughout the country, in view of the review, the screening test findings are rently under way in the Netherlands reduction of inadequate specimens not related to the true disease status of (M.A. Arbyn, personal communication). from 9% with conventional cervical the cervix, making determination of Several systematic, evidence- cytology to 1–2% with LBC (National false negatives and false positives, and based reviews of the published litera- Institute for Clinical Excellence, 2003). hence sensitivity and specificity, ture on LBC have been published Table 20 presents data from a impossible. Cervical biopsy diagnoses (Nanda et al., 2000; Payne et al., 2000; number of ‘direct-to-vial’ studies. obtained as part of routine follow-up of Bernstein et al., 2001; Hartmann et al., Although there is considerable women with abnormal cervical cytol- 2001; Sulik et al., 2001; Abulafia et al., variation between the studies in the ogy results are another commonly 2003; Klinkhamer et al., 2003; Arbyn et prevalence of HSIL identified using used reference standard in studies of al., 2004a). These reviews are based either conventional cytology or LBC, on LBC. However, unless the pathologist on test positivity ratios or detection average the use of LBC increased the is blinded to the original cytological rates, i.e., relative sensitivities and rate of detection of HSIL in these stud-

Table 20. Comparison of identification of SIL using conventional cytology with LBC in representative "direct-to- vial" studies

Reference LBC Population Conventional Liquid-based cytology Increase test No. LSIL HSIL No. LSIL HSIL in HSIL

Bolick & Hellman (1998) TP Screening 39 408 0.8% 0.3% 10 694 2.3% 0.8% 173% Dupree et al. (1998) TP Screening 22 323 0.9% 0.2% 19 351 1.4% 0.3% 50% Papillo et al. (1998) TP Screening 18 569 0.9% 0.5% 8541 1.6% 0.7% 55% Carpenter & Davey (1999) TP High-risk 5000 4.4% 1.9% 2727 6.9% 2.4% 26% Diaz-Rosario & Kabawat TP Screening 74 756 1.6% 0.26% 56 339 2.7% 0.52% 102% (1999) Guidos & Selvaggi (1999) TP Screening 5423 1.0% 0.3% 9583 3.6% 1.0% 233% Vassilakos et al. (1999) SP Screening 88 569 1.6% 0.4% 111 358 2.5% 0.7% 79% Hatch (2000) TP High-risk 16 260 2.9% 1.5% 7934 6.1% 3.2% 116% Tench (2000) SP Screening 10 367 0.6% 0.5% 2231 1.0% 0.7% 46% Weintraub & Morabia (2000) TP Screening 126 619 0.5% 0.1% 39 455 1.8% 0.5% 400% Obwegeser & Brack (2001) TP Screening 1002 3.7% 1.8% 997 4.7% 1.6% – 11% Baker (2002) TP Screening 4872 2.8% 0.7% 3286 4.1% 1.0% 43% Cheung et al. (2003) TP Screening 191 581 1.0% 0.25% 190 667 1.7% 0.24% – 4% Moss et al. (2003) TP Screening 67 856 2.3% 1.4% 34 128 2.6% 1.7% 21% SP Screening 43 280 2.3% 1.4% 47 642 2.3% 1.2% –14% Colgan et al. (2004) SP Screening 445 225 1.4% 0.40% 445 011 1.8% 0.35% –

Abbreviations: TP, ThinPrep; SP, SurePath

71 059-116 24/01/05 11:05 Page 72

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 21. Comparison of specimen adequacy in conventional cytology with LBC in "direct-to-vial" studies

Reference LBC Population Conventional Liquid-based cytology test No. Limited Unsatisf. No. “Limited’ Unsatisf. (%) (%) (%) (%)

Bolick and Hellman (1998) TP Screening 39 408 17.8 1.0 10 694 11.6 0.3 Dupree et al. (1998) TP Screening 22 323 2.0 19 351 3.8 Diaz-Rosario and Kabawat TP Screening 74 756 22.0 0.2 56 339 18.7 0.7 (1999) Carpenter and Davey (1999) TP High-risk 5000 19.4 0.6 2727 10.5 0.3 Guidos and Selvaggi (1999) TP Screening 5423 21.4 1.2 9583 0.7 0.5 Vassilakos et al. (1999) SP Screening 88 569 4.7 1.5 111 358 1.2 0.2 Tench (2000) SP Screening 10 367 31.0 2.9 2231 15.8 0.4 Weintraub and Morabia (2000) TP Screening 130 050 27.8 0.3 39 790 8.1 0.2 Obwegeser and Brack (2001) TP Screening 1002 2.5 0 997 5.5 1.4 Baker (2002) TP Screening 4872 18.2 0.7 3286 9.1 0.8 Cheung et al. (2003) TP Screening 191 581 2.6 0.48 190 667 0.5 0.32 Moss et al. (2003) TP Screening 74 584 9.7 34 813 2.0 SP Screening 47 632 9.1 21 456 0.9

ies. The wide variety of study popula- ited by obscuring factors’ such as blood, tion time. LBC seems to be associated tions makes comparisons difficult. This inflammation or poor preservation than with shorter interpretation times than is because estimates of performance does conventional cytology. In addition, required for conventional cytology are influenced by outlying results of a in many studies both methods have specimens. Payne et al. (2000), in their few studies. In a comprehensive formal reduced the number of specimens clas- systematic review for the United meta-analysis of all published ‘direct-to- sified as ‘unsatisfactory for evaluation’. In Kingdom National Health Service, pro- vial’ studies that adjusted for outlying a recent pilot study in the United vided estimates of three minutes for results, Arbyn et al. (2004a) found a Kingdom (Moss et al., 2003), the use of LBC compared with 4–6 minutes for pooled ratio for detection rate of HSIL in ThinPrep reduced the ‘inadequate’ rate conventional cytology. This is not sur- ThinPrep specimens versus conven- from 9.7% to 2.0%. The use of SurePath prising given that the total surface area tional cytology of 1.72 (95% CI reduced the ‘inadequate’ rate from 9.1% that needs to be screened is consider- 1.42–2.08) and for SurePath specimens to 0.9%. For all study sites combined, ably less for both ThinPrep and versus conventional cytology of 1.47 there was an 82.7% reduction (rate ratio SurePath than for conventional cytol- (95% CI 1.14–1.89). It is important to 0.173, 95% CI 0.17–0.19). The reduc- ogy specimens. The need for continu- bear in mind the limitations to interpre- tion was significant in each of three age ous adjustment to focus is also tation of these studies, as described groups: 20–34, 35–49 and 50–64 years. reduced using LBC, since the cells above, and that the actual number of In a meta-analysis of the comparative tend to be in the same plane of focus. additional cases classified as HSIL performance of LBC, Arbyn et al. With conventional cytology specimens, using LBC is quite small—only about (2004a) estimated the ratio of the inade- the screener needs to continually three cases per 1000 women screened. quacy rate versus conventional cytology adjust the focus to evaluate clusters of of ThinPrep in ‘direct-to-vial’ studies to cells. Payne et al. (2000) reported, Specimen adequacy be 0.70 (95% CI 0.39–1.27) and of however, that cytologists in Edinburgh The effect of LBC on specimen ade- SurePath to be 0.13 (95% CI found screening monolayers to require quacy rates has been evaluated in a 0.07–0.26). more intense concentration than number of the ‘direct-to-vial’ studies screening conventional cytology speci- (Table 21). Both the ThinPrep and Specimen interpretation time mens, making it more tiring. In part, SurePath methods appear to produce A few studies have evaluated the this reflects the fact that occasionally fewer specimens classified as either ‘lim- impact of LBC on specimen interpreta- only one or two HSIL cells are present

72 059-116 24/01/05 11:05 Page 73

Screening tests

Table 22. Systematic reviews of comparative performance of LBC and conventional cytology

Author Subgroup Indicator Key conclusions

Nanda et al. (2000) ThinPrep Histologically confirmed lesion Higher sensitivity of LBC, but only three studies were evaluated

Payne et al. (2000) Test positivity or histologically Some evidence that LBC offers an confirmed lesion improvement in sensitivity

Bernstein et al. (2001) ThinPrep Test positivity ThinPrep is as good as, or superior to, conventional cytology for diagnosing CIN

Hartmann et al. (2001) All studies Histologically confirmed lesion Current evidence is inadequate to gauge whether LBC is "better" than conventional cytology

Sulik et al. (2001) Histologically confirmed lesion LBC demonstrated higher sensitivity (90%; 95% CI 77–96%) than conventional cytology (79%; 95% CI: 59–91%) for CIN 2 or more severe

Abulafia et al. (2003) ThinPrep only Test positivity ThinPrep tends to be more sensitive than conventional smears in detecting CIN

Klinkhamer et al. (2003) All studies Histologically confirmed lesion Indications that SurePath has lower sensitivity than conventional for ASC or greater No definitive statement can be made for detection of LSIL or higher or HSIL or higher for SurePath because of conflicting results Indications that ThinPrep has higher sensitivity than conventional for ASC or greater Likely that ThinPrep has higher sensitivity than conventional for LSIL or higher Likely that ThinPrep has higher positivity rate and greater absolute sensitivity than conventional for HSIL

Arbyn et al. (2004a) Split-sample studies Test positivity More LSIL in LBC than in conventional cytology Positivity rates for ASC and HSIL not statistically different Direct-to-vial studies More LSIL detected by LBC 80% (95% CI 52–112%) ThinPrep; 54% (95% CI 25–90%) SurePath More HSIL detected by LBC 72% (95% CI 42–108%) ThinPrep; 47% (95% CI 14–89%) SurePath Positivity rates for ASC were the same There was no reduction in positive predictive value for CIN 2 and CIN 3 of LBC versus conventional cytology

73 059-116 24/01/05 11:05 Page 74

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

in an LBC specimen, necessitating process of maintaining minimum stan- Training careful scrutiny of every individual cell. dards and continually striving for excel- Training of both the cytotechnicians lence. Quality assurance should be a who perform the initial screening in the Availability of residual cellular mater- coordinated effort that is designed to laboratory and the pathologists who ial for molecular testing control, detect and prevent the occur- provide the final interpretation is critical One of the major benefits of LBC in rence of errors and hopefully to to obtaining optimal performance of a many settings is the availability of improve patient care. In general, there cervical cytology programme. Cyto- residual cellular material for molecular are three stages to the process of pathologists should either receive for- testing for agents such as Chlamydia quality control (Bozzo, 1991): mal training in an established acade- trachomatis or HPV. In the USA, the mic programme or be trained in an 2001 Consensus Guidelines for the • Setting standards for what one established national centre for cervical Management of Women with Cytologic wishes to control and defining the cytology for at least six months (Miller Abnormalities considered HPV DNA benchmarks; et al., 2000). This training should typi- testing of residual LBC fluid to be the • Developing a mechanism for cally include not only the interpretation preferred approach to managing assessing what one wishes to con- of cervical cytology specimens, but women with ASCUS cytological results trol; also cervical histopathology. A cyto- (Wright et al., 2002a), on the grounds • Defining the response to be taken pathologist who will run a laboratory is that such reflex HPV DNA testing when deficiencies are identified. generally selected for leadership offers the advantage that women do potential and ability to organize, run not need to return to the office or clinic For cervical cytology screening, and manage a successful cytopathol- for an additional clinical examination. quality assurance programmes can ogy laboratory, and will require training In addition, the 40–60% of women who include a number of types of activity in laboratory management skills. are high-risk HPV DNA-negative will and should take into account country- In most developed countries, be spared a colposcopic examination and location-specific needs. What may cytotechnicians undergo 1–2 years of and can be rapidly assured that they be considered acceptable or even formal didactic training in order to do not have a significant cervical mandatory in one setting may serve develop a high level of competence in lesion. A comprehensive study of simply to limit the availability of screen- evaluating all types of cytological spec- triage methods for women with ing in other settings. It is critical, imens, including gynaecological cytol- ASCUS indicated that reflex HPV DNA however, that any cervical cytology ogy. However, in some countries con- testing provides the same or greater laboratory or programme have an sideration is now being given to inten- life expectancy benefits and is more established quality assurance sive six-month training programmes cost-effective than either a programme programme. In general, it is preferable focusing only on gynaecological cytol- of repeat cytology or immediate col- for cytology services to be centralized ogy. In addition, cytotechnicians should poscopy (Kim et al., 2002). as much as possible, to facilitate periodically participate in competence- quality assurance. The use of comput- based education programmes. Unfor- Quality assurance and quality erized data collection systems that can tunately, cytotechnology training pro- control issues integrate cytological findings, histo- grammes are not available in many An advantage of cervical cytology over logical findings and follow-up informa- developing countries and extended for- screening methods such as visual tion is highly desirable (Miller et al., mal training programmes are not an screening is that even though quality 2000). option. In these settings, cytotechni- assurance and quality control pro- cians are often ‘bench-trained’, being grammes can be developed for both, Preanalytic quality control tutored by a person with some level of the availability of archival glass slides Preanalytic quality control measures training in interpreting gynaecological facilities such programmes. Various include the records that laboratories cytology specimens. Training in this definitions for quality control and qual- should maintain relating to specimen manner should be avoided unless the ity assurance are used by laboratories. receipt, preparation of specimens, laboratory where it occurs processes In general, quality control can be staining of specimens and upkeep of at least 15 000 specimens annually thought of in terms of the actual equipment and microscopes, as well and training should last for at least six assessments that are done to ensure as records of personnel and their train- months (Miller et al., 2000). Although high quality and quality assurance can ing and education. there is little evidence that cytotechni- be thought of in terms of the entire cians who are trained in such a ‘hands-

74 059-116 24/01/05 11:05 Page 75

Screening tests

on’ fashion perform less well than It is also important that a laboratory as negative by each cytotechnician those who receive formal training, the process a minimum number of speci- must be reassessed by either a pathol- variability in training inherent in this mens per year in order to maintain an ogist or a supervising cytotechnologist approach is a cause for concern. adequate level of competence (Krieger before the result is reported (Federal Whenever possible, cytotechnicians et al., 1998). In reviews of US laborato- Register, 1992). This regulation is con- should receive formal, structured, ries by the College of American troversial for a variety of reasons. One competence-based training in inter- Pathologists, screening error rates is the level of discrepancy that is con- preting cervical cytology specimens. were found to be greatest in laborato- sidered significant. It has been argued The International Federation of ries processing less than 5000 speci- that negative specimens classified as Cytology has an international qualifica- mens per year and having no dedi- atypical (e.g., ASC) upon review tion for cytotechnicians, which can be cated screening cytotechnicians should not be considered errors, used to ensure that competence has (College of American Pathologists, because of the inherent subjectivity of been obtained. 1997). In the United Kingdom, labora- this diagnosis (Krieger et al., 1998). tories are now required to process Another problem with performing 10% Workload limits – maximum and min- at least 15 000 specimens per year. rescreening of negative cases is that imum Evaluation of a minimum annual num- significant lesions are quite uncommon It is now widely accepted that, because ber of specimens is also to be consid- in the reviews. Given an underlying of the repetitive nature of screening ered desirable in low-resource set- rate of SIL of only 2–3% in the cytology specimens, there should be tings. The Peruvian Society of Cyto- screened population and assuming workload limits on the number of spec- pathology does not certify laboratories that even a poorly performing cytotech- imens that a cytotechnician can screen that process under 5000 specimens nician will be able to identify 75% of in any given period. In the USA, federal annually (Salvetto & Sandiford, 2004). specimens containing SIL, large num- regulations require that anyone per- A recent World Health Organization bers of specimens must be rescreened forming primary screening of cervical consensus panel recommended that to determine which cytotechnicians or cytology specimens should evaluate no each laboratory should process at laboratories are performing poorly.This more than 100 cytology specimens per least 20 000 specimens yearly in order lack of statistical power greatly ham- 24-hour period and in not less than to maintain acceptable skills (Miller et pers its use as a quality control mea- eight hours (Federal Register, 1992). In al., 2000). sure (Hutchinson, 1996). addition, every laboratory must estab- The technique of rapid rescreening lish individual workload limits for each Review of abnormal cases of all negative specimens has been the cytotechnician, based on their experi- It is generally accepted that a patholo- subject of a number of studies and ence and skill. This must be gist should review all specimens appears to present an attractive alter- reassessed every six months using lab- deemed by the screening cytotechni- native to the 10% rescreening oratory-defined performance stan- cian to have any degree of cytological approach. Using this technique all, or dards. In many European countries, abnormality (American Society of most, of the specimens classified as this workload limit is considered too Cytopathology, 2001). Identification of negative by a laboratory undergo a high and other limits are used. In the discordant cases provides an element second, more rapid evaluation by a dif- United Kingdom, for example, time lim- of quality control for the screening ferent screener. This is the approach to its rather than slide limits are used. process and allows identification of rescreening adopted in the United Cytotechnicians are restricted to specific cytotechnicians and specific Kingdom by the National Health screening for only four hours per day, areas of cytology requiring additional Service (NHSCSP, 2000). Another regardless of whether they are screen- education. It is important for quality approach is referred to as ‘prescreen- ing conventional or LBC specimens. monitoring that all reviews be docu- ing’, in which all specimens undergo a Since LBC specimens can be screened mented. rapid review before the intensive more rapidly than conventional cytol- screening. In a recent meta-analysis of ogy specimens, this means that greater Rescreening of negative cases published data, Arbyn et al. (2003) numbers of LBC specimens can be Some form of rescreening of speci- found that the pooled estimated sensi- screened by each cytotechnician. A mens initially considered negative is tivity of rapid prescreening for HSIL or recent consensus panel recommended important for quality control. In the more severe lesions was 86% and that that a daily workload limit of 60 cases USA, federal regulations stipulate that the technique showed diagnostic prop- was preferable (Miller et al., 2000). at least 10% of all samples interpreted erties that support its use as a quality

75 059-116 24/01/05 11:05 Page 76

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

control measure. The same group pre- programme divided by the total num- low-resource settings was promoted viously demonstrated that rapid pre- ber of LSIL or greater specimens iden- (Stjernswärd, 1987). Over the last ten screening was superior to 10% ran- tified through regular screening and years, the use of dilute (3–5%) acetic dom rescreening in identifying cases the rapid rescreen process combined acid applied to the cervix before that were missed (Arbyn & Schenck, (Krieger et al., 1998). inspection (visual inspection with 2000). acetic acid, VIA) has been investi- Proficiency testing gated. More recently, the application of Cytology–histology correlations and Proficiency testing programmes pro- Lugol’s solution has been used and is clinical follow-up vide laboratories, cytopathologists and referred to as visual inspection with If a laboratory has access to histologi- individual cytotechnicians with sets of Lugol’s iodine (VILI). cal specimens obtained at the time of stained cytology specimens on which Visual tests are inherently subjec- colposcopy for an abnormal cytological the interpretation has been agreed to tive. Published studies of the test per- finding, it should compare all premalig- according to a set procedure. The slide formance characteristics vary with nant and malignant cytological results sets are then evaluated by the person regard to important methodological with the histopathological observa- being tested and their interpretation is aspects that result in biases and other tions. This allows the laboratory to compared with that of the panel or with difficulties in generalizing the findings refine its cytological criteria. If histolog- their peers (Coleman & Evans, 1999; to other populations. For example, ical specimens are not available, the NHSCSP, 2000). This allows the per- studies may use different definitions of laboratory may attempt to obtain refer- formance of both whole laboratories test positivity. Differences in training of ral and follow-up information. It is and individual cytotechnicians or personnel and in the light sources important that the laboratory obtain fol- cytopathologists to be compared used also generate variability in test low-up information on women with against others in an unbiased manner. performance characteristics across dif- HSIL to ensure that they have not been Periodic retesting should be conducted ferent study settings. Varying abilities lost to follow-up. every 6–12 months (Miller et al., 2000). of colposcopists to detect lesions and Individuals who perform poorly on pro- of pathologists to interpret histology Measuring the performance of the ficiency testing should receive addi- accurately also affect the assessment laboratory tional training to improve their skills of test performance. Laboratories need to carefully and and any who continue to perform Colposcopy with directed biopsy is continuously monitor their perfor- poorly after retraining should be reas- the usual reference standard by which mance as a whole, as well as that of signed to non-screening tasks. the performance of visual tests is individual cytotechnicians. Information Recent evidence suggests that assessed, but biases may impair the that can be useful for a given labora- performance on proficiency testing validity of the assessment. Verification tory includes the percentages of spec- provides some evidence of the real- bias arises if colposcopy is not applied imens classified as having a given world performance of cytotechnicians equally to all women because of the result (e.g., ASC, LSIL, HSIL, etc.), the (Keenlyside et al., 1999). A recent study design. Blinding between those rate of unsatisfactory specimens, the report from Peru and Nicaragua has performing the visual test under evalu- ASC:LSIL ratio, the laboratory turn- shown that proficiency testing can be ation and those performing the refer- around time, etc. One of the most implemented successfully in develop- ence colposcopy is crucial to avoid important measures is screening sen- ing countries (Salvetto & Sandiford, information or expectation bias. Test sitivity (Krieger et al., 1998; NHSCSP, 2004). performance for detection of CIN 2 or 2000; American Society of Cyto- worse lesions and potential biases of pathology, 2001). However, it is very all studies reviewed are summarized difficult to determine the sensitivity of Visual inspection below. screening in a real-world laboratory setting. One approach that has been The use of visual inspection methods Unaided visual inspection (VI) proposed to estimate screening sensi- to screen for cervical neoplasia began Visual inspection (VI) (also called tivity in a laboratory is to calculate the with the use of Schiller’s test in the ‘down-staging’ or ‘unaided visual ‘false negative proportion’, which is 1930s (Schiller, 1933). In the 1980s, inspection’) consists of a clinical exam- essentially the number of false nega- the idea of looking at the cervix with ination of the cervix using only a tive LSIL or greater specimens identi- the naked eye for early detection of speculum and a light source. Test pos- fied through a 100% rapid rescreen disease (known as ‘down-staging’) in itivity is defined by the presence or

76 059-116 24/01/05 11:05 Page 77

Screening tests

absence of specific characteristics, usu- of intracellular proteins, resulting in One study used a gold standard for ally with low and high thresholds of pos- noticeable opacity and a decrease in enhanced disease ascertainment that itivity (Table 23). Only one of the six the usual reddish hue imparted by the was based on directed biopsy of any published studies reporting test charac- subepithelial vasculature. This effect, abnormal area(s), four-quadrant biop- teristics of VI (Table 24) did not suffer called acetowhitening, is not specific to sies and endocervical curettage (ECC) from obvious verification bias (see cervical neoplasia and may also occur in all women (Belinson et al., 2001). The Glossary and Chapter 4) (Basu et al., in immature squamous metaplasia and other study that had enhanced design 2002); it found sensitivity to be low in inflamed, regenerating cervical features was based on 55 000 women (< 50%) irrespective of the threshold epithelium. The degree of opacity due enrolled at 11 sites in six West African used to define test positivity. The other to the acetowhite reaction varies countries and India (Sankaranarayanan five studies used cytology as the refer- according to the thickness of the neo- et al., 2004a). Each site followed a com- ence standard. In all studies, the high- plastic change present in the epithe- mon testing protocol that included VIA, threshold definition of test positivity lium and thus according to the grade of VILI and colposcopy with directed (corresponding to a 5–10% positivity intraepithelial neoplasia. biopsy, as required, performed by sepa- rate) was associated with rather low The most common features rate individuals. Although similar train- sensitivity (30–60%). Gains in sensitivity observed using VIA are summarized in ing methods and test result definitions using the low-threshold definition of test Table 25. VIA results are reported were used, there was substantial varia- positivity led to concomitant decreases using negative and positive categories. tion in the reported positivity rates in specificity.Thus it is clear that VI lacks VIA-positive cervices are illustrated in (7–27%), sensitivity (56–94%) and sufficient sensitivity for use as a primary Figure 42. specificity (74–94%). screening test. In 17 published studies, test positiv- Numerous studies have shown VIA ity rates ranged from 3% to 53% (Table to have sensitivity similar to that of cer- Visual inspection using acetic 26). Seven studies were designed to vical cytology for identifying women acid (VIA) minimize verification bias. In two other with HSIL, but much lower specificity VIA involves naked-eye inspection of the studies (Denny et al., 2000a, 2002), (Table 27). Only two studies compared cervix one minute after application of a only women negative by cytology, VIA, the accuracy of VIA and HPV DNA 3–5% solution of acetic acid using a cot- HPV DNA testing and cervicography testing (Table 28); these showed the ton swab or a spray. Test positivity is were not subjected to colposcopy, two tests to have similar accuracy. The based on the appearance of acetowhite reducing susceptibility to bias. Seven of reproducibility of VIA has been docu- areas in the transformation zone, close these nine studies (Londhe et al., 1997; mented to be equivalent to that of his- to the squamocolumnar junction or the University of Zimbabwe/JHPIEGO tology, cytology and colposcopy os.The cervix is examined using a bright Cervical Cancer Project, 1999; Denny (Sellors et al., 2002). In the multicentre light source such as a torch or halogen et al., 2000a; Belinson et al., 2001; study in Africa and India, the agree- focus lamp. VIA is also known as direct Denny et al., 2002; Cronjé et al., 2003; ment between master trainers and visual inspection (DVI), acetic acid test Sankaranarayanan et al., 2004a), local providers using 36 cervical pho- (AAT) and cervicoscopy. accounting for more than 95% of the tographs was fair (raw agreement, Dilute acetic acid causes what is total sample size, reported sensitivities 64.5%; kappa, 0.38) (Sankaranara- thought to be a reversible coagulation of approximately 75%. yanan et al., 2004a).

Visual inspection using acetic Table 23. Test definition for visual inspection acid with low-level magnifica- tion (VIAM) Test definition Characteristics VIAM is VIA with low-level magnifica- tion (2–4 x), using a hand-held device Normal Normal-looking cervix, nabothian cysts to inspect the cervix one minute after Positive (low threshold) Cervicitis, erosion, polyp, wart, unhealthy cervix, application of acetic acid. Table 29 pre- reddish-looking cervix sents test results from four studies comparing VIA and VIAM. None of Positive (high threshold) Low-threshold features plus bleeding on touch, bleeding these studies documented any signifi- erosion, hypertrophied elongated cervix, growth, ulcer cant difference in test performance characteristics between VIAM and VIA.

77 059-116 24/01/05 11:05 Page 78

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 24. Studies of visual inspectiona Study Sample Population (age, Provider Reference Positivity Sensitivity SpecificityComments size recruitment, diagnosis rate (%) (%) (%) location) Singh et al.b (1992) 44 970 Cytology/ (H) 11 63 89 Verification Opportunistic histology (L) 69 bias India

Bhargava et al.b (1993) 3608 Midwife Cytology/ (H) 5 25 96 Verification Opportunistic histology (L) 65 92 37 bias India

Sujathon et al. (1995) 3602 30+ Cyto- Cytology (L) 63 89 50 Not Opportunistic and pathologist designed for referred accuracy India estimation

Nene et al.c (1996) 2135 35–60 Health Cytology/ (H) 6 60 94 Verification Community-based worker histology (L) 57 90 43 bias

Wesley et al. (1997) 2843 30+ Health Cytology/ (H) 6 29 94 Verification Opportunistic worker histology (L) 45 66 55 bias India

Basu et al. (2002) 6399 Community-based Health Colposcopy/ (H) 7 32 93 India worker histology (L) 25 49 76

a Some test characteristics of the table are not exactly those reported in corresponding publications. Estimates have been computed when they were not provided or have been corrected to achieve comparability between studies. This correction was performed to take into account differences in study design or analysis, due to various factors: different threshold of test positivity, different disease defini- tion, only a subset of the population used for estimation of characteristics, improper computation method, etc. b Detection of any lesions c Detection of cancer Test positivity was defined at high-threshold (H) and at low-threshold (L), the sensitivity was estimated with the threshold CIN2–3 unless otherwise specified The study with no verification bias is highlighted

A correlation study, with a sample size but fell into disuse as cytological Lugol’s iodine stains glycogen of 2080 previously screened women testing became available (Schiller, 1933; stored in cervical epithelial cells. Mature and a positivity rate of approximately Wright, 2003). Several decades later, squamous epithelium stores more 5% for VIAM, reported poor associa- research on visual inspection methods glycogen than either columnar epithe- tions between VIAM and HPV test pos- led to the observation that nurses and lium or immature squamous metaplas- itivity and between VIAM and cytology midwives recognized non-staining areas tic epithelium. The application of iodine test results (Rodriguez et al., 2004). on the cervix after application of Lugol’s solution to the cervix thus results in solution more readily than acetowhite black or dark brown staining of mature Visual inspection using Lugol’s areas (Sankaranarayanan & Wesley, squamous epithelium. Columnar epithe- iodine (VILI) 2003), which led to renewed interest in lium does not stain and retains its red- The use of Lugol’s solution to aid this technique (referred to in the past as dish hue. Areas of immature metaplasia inspection of the cervix with the naked ‘Lugol’s iodine test’ and ‘Schiller’s iodine stain a very light brownish hue, if at all. eye was described in 1933 by Schiller, test’). Neoplastic squamous epithelium

78 059-116 25/01/05 10:37 Page 79

Screening tests

Table 25. Test definition of visual inspection with acetic acid

Denomination Possible thresholds Characteristics

ABC

Normal Normal looking cervix: no white lesion, smooth, uniform, featureless Atypical cervix: ectopion, polyp, cervicitis, inflammation, Nabothian cysts negative

Indeterminate Severe inflammation or cervicitis so that cervix cannot be adequately negative assessed for acetowhite lesion negative

Ill-definite lesion Pale white lesion (acetowhite lesion), poorly circumscribed and faintly acetowhite Focal, small punctuated areas of acetowhitening usually involving the transformation zone positive

Definite lesion Dense white lesion with sharp border; one border abutting the squamo- positive columnar junction

Suspicious cancer positive Cervical ulcer or growth cauliflower-like growth or ulcer Fungation mass

contains little or no glycogen and does and areas partially denuded of squa- et al., 2004a) involved 54 981 women not stain with Lugol’s iodine, taking a mous epithelium do not stain with iodine aged 25–65 years. The reference stan- bright mustard or saffron yellow colour. and remain colourless in a surrounding dard was colposcopy-directed biopsy. Atrophic epithelium stains partially with black or dark brown background. Results VIA and VILI were performed indepen- Lugol’s iodine, which makes interpreta- of VILI are categorized in Table 30. dently by blinded individuals in order to tion difficult in postmenopausal women. Images of VILI-positive cervices are minimize information (expectation) A condylomatous lesion may not stain or shown in Figure 43. bias. In this setting, VILI was more only partially stain with Lugol’s iodine. The single published report of VILI sensitive than VIA and equally specific Areas of leukoplakia (hyperkeratosis) test characteristics (Sankaranarayanan (Figure 44). The reproducibility of

Figure 42 Example of VIA-positive lesions

79 059-116 24/01/05 11:05 Page 80

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening Unbiased esti- test accuracy test accuracy test accuracy mation not Not designed for test accurary test accuracy estimation Minimal verifica- tion bias Only 74% of the Not designed for Not designed for enrolled patient underwent colposcopy estimation from the possible data published estimation estimation estimation Not designed for Not designed for Comments 84 75 49 83 64 98 68 97 91 (%) Specificity (%) 50 88 78 67 65 77 95 29 87 Sensitivity 18 25 53 3 18 40 3 36 10 rate (%) Positivity diagnosis Colposcopy/histology VIA+ Histology for VIA+ or Pap+ for VIA+ or Pap+ for Colposcopy/histology Colposcopy Colposcopy/histology VIA+ or Pap+, for Colposcopy/histology Colposcopy/histology Colposcopy/histology VIA- and 20% of VIA+ or Pap+ for VIA+, HPV+ Pap+, for VIA+ or Pap+ for Colposcopy/histology or cervicography+ or cervicography+ Reference a Nurse Clinician Clinician Smear- Nurse Nurse taker Midwife Nurse Cytotech. Provider 15–45 17–83 Opportunistic Mean age: 34 20–83 Opportunistic Opportunistic 20+ 25–55 22–70 Opportunistic India India South Africa Opportunistic Zimbabwe India Opportunistic South Africa practice Family South Africa location) USA Italy Opportunistic Opportunistic (age, recruitment, 2690 2036 372 6298 2426 2935 2148 1268 2944 Sample Population size . et al et al. . (1996) . (1993) . (1992) . (1997) . (2000) . (2001) et al et al et al et al et al et al (1998) JHPIEGO (1999) (1999) Sankaranarayanan of Zimbabwe/ University Sankaranarayanan Denny Cronjé 26.Table of visual inspection with acetic acid Studies Study Slawson Cecchini Megevand Londhe

80 059-116 24/01/05 11:05 Page 81

Screening tests bias bias Verification Minimal verification Comments :74–94] b 74 49 82 59 78 [R 86 79 91 (%) Specificity :56–94] b (%) 71 79 [R 87 92 94 77 70 37 Sensitivity :7–27] b 28 [R 53 42 48 31 16 25 10 rate (%) Positivity diagnosis Colposcopy/histology Colposcopy/histology + ECC Colposcopy/histology histology 4 quadrant Colposcopy/histology Histology Colposcopy/histology VIA+ HPV-, Pap+, for VIA+ or Pap+ for or cervicography+ Histology Reference Gynaecol. Gynaecol. Nurse Nurse Midwife Provider 30–60 35–45 India, Africa (11 studies) location) Mean age: 37 35–65 19–45 21–65 25–65 Nurse Opportunistic China 25–65 (age, recruitment, Referred Symptomatic India Opportunistic Mexico Opportunistic South Africa Opportunistic South Africa Opportunistic Philippine Opportunistic Pakistan Opportunistic size 1997 Sample Population 402 2754 376 1093 3316 54 981 . . et al et al (2001) . (2003) (2002) (2003) 501 . (2003) . (2001) et al et al. et al et al. et al. et al are not those reportedSome test characteristics of the table in the corresponding publications. not provid- were they been computed when Estimates have within the studies reported the range R stands for a Belinson (2002) Outcome threshold CIN 2–3 bias are highlighted The studies with no verification Singh 26 (contd) Table Study b studies. between comparability been corrected to achieve in study design or analy- ed or have into account differences to take performed This correction was factors: due to various sis, estimation of characteris- used for disease definition, only a subset of the population different threshold of test positivity, different improper computation method, etc. tics, Denny Cronjé Ngelangel Rodriguez-Reyes Tayyeb Sankaranarayanan (2004a)

81 059-116 24/01/05 11:06 Page 82

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 27. Comparison of VIA and cytology accuracy in published studiesa

Study Sample Positivity rate (%) Sensitivity (%) Specificity (%) size VIA Cytology VIA Cytology VIA Cytology

Slawson et al. (1992) 2690 3 6 29 87 97 95 Cecchini et al. (1993) 2036 25 4 88 63 75 96 Megevand et al. (1996) 2426 3 13 65 100 98 88 Londhe et al. (1997) 372 53 6 78 22 49 95 Sankaranarayanan et al. 2935 10 10 87 86 91 91 (1998) University of Zimbabwe/ 2148 40 13 77 44 64 91 JHPIEGO (1999)# Sankaranarayanan et al. 1268 36 16 95 62 68 87 (1999) Denny et al. (2000) 2944 18 15 67 80 83 87 Cronjé et al. (2001) # 6298 18 2 50 19 84 99 Singh et al. (2001) 402 42 42 87 81 82 79 Denny et al. (2002) # 2754 25 70 57 79 96 Ngelangel et al. (2003) 3316 (VIA) 3195 10 2 37 14 91 98 (Cytology) Tayyeb et al. (2003) 501 31 16 94 47 78 89 Cronje et al. (2003) # 1093 53 9 79 53 49 95 Sankaranarayanan et al.b 22 663 17 9 72 65 84 92 (2004b) [Rc: 38–81] [Rc: 86–99]

a Some test characteristics of the table are not the ones reported in corresponding publications. Estimates have been computed when they were not provided or have been corrected to achieve comparability between studies. This correction was performed to take into account differences in study design or analysis, due to various factors: different threshold of test positivity, different disease definition, only a subset of the population used for estimation of characteristics, improper computation method, etc. b Subset of five Indian studies from Sankaranarayanan et al.(2004a) c R stands for the range within the studies reported Cytology threshold: ASCUS+, unless otherwise indicated (# , LSIL+) Outcome threshold: CIN 2–3

Figure 43 Example of VILI-positive lesions

82 059-116 24/01/05 11:06 Page 83

Screening tests

Table 28. Comparison of VIA and HPV testing accuracy in published studies

Study Sample size VIA HPV testing Positivity Sensitivity Specificity Positivity Sensitivity Specificity rate (%) (%) (%) rate (%) (%) (%)

Denny et al. (2000) 2944 18 67 83 16 73 86 Sankaranarayanan et al. 18 085 11 65 89 7 65 94 (2004c)a [Rb: 54–79] [Rb: 89–90] [Rb: 6–9] [Rb: 45–81] [Rb: 92–95]

Outcome threshold: CIN 2–3 aSubset of five Indian studies from Sankaranarayanan et al. (2004a) bR stands for the range within the studies reported

Table 29. Comparison of accuracy of visual inspection with acetic acid, with or without magnification (VIA and VIAM), in published studies

Study Sample size VIA VIAM Positivity Sensitivity Specificity Device Positivity Sensitivity Specificity rate (%) (%) (%) rate (%) (%) (%) Denny et al. (2000) 2944 18 67 83 x 2.5 18 67 83 (hand-held device) Denny et al. (2002) 2754 25 70 79 x 4.5 27 74 77 Aviscope Ngelangel et al. (2003) 3316 (VIA) 10 37 91 Speculo- 11 34 90 3447 (VIAM) scope (6 x 16 magnification) Sankaranarayanan et al. 16 900 14 (2004d) (3 studies) [Ra: 11–19] [Ra: 56–71] [Ra: 82–90] x 4 magnifi- 14 64 87 cation [Ra: 11–18] [Ra: 61–71] [Ra: 83–90] Hand-lens (2 studies) and Aviscope (1 study)

Outcome threshold: CIN 2–3 a R stands for the range within the studies reported

VILI appears to be greater than that of part, the subjective nature of visual process measurements (e.g., test pos- VIA. inspection. Definitions of test result itivity rates, histological confirmation categories should be standardized to rates) and on-site supervision are Quality control for visual improve reproducibility. Due to the sub- critical to support high-quality visual inspection tests jective nature of visual inspection inspection-based screening services. The substantial variability in test per- methods, it is difficult to maintain the Although reliable methods of correlat- formance characteristics of visual quality of assessment among trained ing daily competence with proficiency inspection tests reflects, at least in staff. Adequate training, routine testing results have not been

83 059-116 27/01/05 15:38 Page 84

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 30. Categorization of visual inspection with Lugol’s iodine (VILI) test results

VILI-negative Patterns include the following:

• A normal pattern of dark brown or black staining of the squamous epithelium and no change in colour of the columnar epithelium, or

• patchy, indistinct, ill-defined, colourless or partially brown areas, or

• pale areas of no or partial iodine uptake present on polyps, or

• a leopard-skin appearance (associated with T. vaginalis infection), or

• pepper-like, non-iodine uptake areas seen in the squamous epithelium, far away from the squamocolumnar junction, or satellite, thin, yellow, non-iodine uptake areas with angular, or digitating margins, resembling geographical areas seen far away from the squamocolumnar junction

VILI-positive Presence of a dense, thick, bright, mustard-yellow or saffron-yellow iodine non-uptake area seen in the transformation zone close to or abutting the squamocolumnar junction or, in the absence of a visible squamocolumnar junction, close to the os, or when the entire cervix turns bright yellow

VILI-positive, Frank, nodular, irregular, ulceroproliferative growth visible on the cervix, which turns densely yellow invasive cancer on application of iodine

Adapted from Sankaranarayanan & Wesley (2003)

established for cytological screening programmes (Vooijs et al., 1998), visual inspection-based screening pro- grammes may utilize periodic assess- ments of practitioners’ skills using collections of VIA and VILI cervical photographs. There is no consensus on the num- ber of visual inspections that should be performed correctly by an individual before he/she is deemed competent, nor on the minimum daily rate that is required to maintain skills. The periodic computation of test positivity rates for visual inspections performed by each individual may be useful for monitoring visual inspection. However, because there is no permanent record when Figure 44 Plot of sensitivity and specificity for VIA and VILI for each of the studies. visual assessments are made by VIA The size of the bubble reflects the precision of the estimates. The bigger the or VILI, unless a photographic image is bubble, the lower the variance of the sensitivity and specificity and the higher the precision taken for subsequent review by a due to a larger study sample size. The bubbles with thick borders supervisor (Sellors et al., 2002; Wright, represent the pooled estimates. (Adapted from Sankaranarayanan et al., 2004a) 2 2 2003), these positivity rates are not Note: bubble size = k/(dse + dsp ), where k is a constant, d is the difference between the lower and upper confidence limits, se is sensitivity and sp is specificity. verifiable by audit. Visual inspection

84 059-116 24/01/05 11:06 Page 85

Screening tests

methods coupled with immediate wives, paramedical workers and trained of a trained colposcopist and access to cryosurgical treatment for test-positive non-medical personnel, after a short a competent histopathologist able to cases (‘see and treat’) have also been period of training (1–3 weeks). In addi- perform assessment of removed suggested for cervical screening in tion, results are available without delay, biopsy material. The findings from low-resource settings (Denny et al., allowing immediate referral for confirma- video colposcopy seem to agree with 2002, Gaffikin et al., 2003). Because tory testing. However, if immediate treat- those obtained with traditional optical biopsies to exclude invasive carcinoma ment is performed, high rates of colposcopy (Ferris et al., 2000b). are not performed before ablative overtreatment may result, given the rela- treatment, the ‘see and treat’ tively low test specificity of VIA and VILI. Colposcopic findings approaches carry the risk of under- Terminology to describe the morpho- treatment of invasive carcinoma and logical findings in a standard fashion reduced opportunities to diagnose Colposcopy has evolved over the years (Dexeus et potentially curable invasive disease. It al., 1977; Jordan, 1985; Sellors & has been argued that, outside of Colposcopy is a procedure that allows Sankaranarayanan, 2003; Walker et research settings, it may not be feasi- illuminated stereoscopic and magni- al., 2003) (Table 31). Many of the qual- ble to monitor the safety or confirm the fied (typically 6–40 x) viewing of the itative descriptions have been quanti- effectiveness of ‘see and treat’ pro- cervix and the vagina. For colposcopy, fied as to the degree of abnormality grammes (Suba et al., 2004). the woman is placed in the lithotomy and have been combined into a scor- position, with the cervix exposed with a ing system (Table 32) that is used by Advantages and potential hazards bivalve speculum in place, and various many colposcopists to grade abnormal of visual inspection methods solutions (normal saline, 3–5% dilute squamous epithelial areas (Reid & Concerns about personal modesty and acetic acid and Lugol’s iodine) are Scalzi, 1985; Reid, 1993). The uncom- discomfort caused by the vaginal applied to the cervical epithelium in mon glandular epithelial lesions tend to speculum are common to all screening sequence. A green filter is rarely used be more difficult to diagnose and techniques. Local irritation of tissues except when the subepithelial vascular appear as strikingly dense acetowhite and allergic reactions to iodine or vine- pattern is examined (Jordan, 1985; or milky-white areas compared to the gar are rare. In both VIA and VILI, Sellors & Sankaranarayanan, 2003). surrounding villi of columnar epithe- staining of the epithelium is temporary, The aim is to examine the transforma- lium. Microinvasive and frankly inva- although iodine staining lasts longer tion zone, an area bounded laterally by sive squamous cancers are densely (up to 45 minutes) than acetowhitening the original squamocolumnar junction, acetowhite with markedly atypical (Sankaranarayanan et al., 2004a). in which metaplastic squamous epithe- blood vessels (bizarre, irregular Application of visual inspection meth- lium develops, the medial or internal branching and gross fluctuations in ods should probably be restricted to border being defined by the new calibre and course). The surface con- women under the age of 50 years. With squamocolumnar junction. This latter figuration gradually changes from increasing age, the squamocolumnar junction defines the upper limit of the small protuberances, excrescences or junction migrates inward from the readily squamous metaplastic process, which microconvolutions in microinvasive visible portion of the ectocervix towards in certain conditions may become cancer to frankly invasive cancer with the endocervical canal, so lesions prob- abnormal. When such abnormal areas strikingly raised edges, irregular sur- ably become more difficult to identify with develop within this zone, they are face contour, and strikingly bizarre visual methods in older women. In addi- graded according to morphological fea- blood vessels that bleed sponta- tion, the accuracy of visual inspection tures, namely, acetowhiteness, mar- neously or on touch. may be highly dependent on the under- gins, blood vessels and iodine uptake. lying prevalence of sexually transmitted Hinselmann (1925) first described Histological confirmation diseases, which may increase the level colposcopy.The modern colposcope is a Biopsies are obtained under colpo- of inflammation and render visual binocular microscope with a variable- scopic visualization from the locations inspection difficult to assess. intensity light source providing a stereo- with the most severe changes, in order Visual inspection-based tests are scopic view of the cervix, with a field of to histologically confirm the degree of simple, safe and well accepted. They view and depth of focus that vary severity of the neoplastic process. require a very low level of infrastructure inversely with the magnification selected. Since it is essential to rule out the and can be performed by a wide range of Provision of a high-quality colpo- presence of cancer, it is standard personnel, such as doctors, nurses, mid- scopic service requires the availability practice in some settings to obtain a

85 059-116 24/01/05 11:06 Page 86

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 31. International terminology for colposcopy from the International result of the long-standing tradition Federation for Cervical Pathology and Colposcopy rooted in German medical teaching from the time of Hinselmann in I Normal colposcopic findings Hamburg (Jordan, 1985; Dexeus et al., Original squamous epithelium 2002). When colposcopy has been Columnar epithelium Transformation zone evaluated for primary screening, it has been usually accompanied by simulta- II Abnormal colposcopic findings neous cytology (Dexeus et al., 2002). Flat acetowhite epithelium The rationale behind this combined a Dense acetowhite epithelium testing approach is that it decreases Fine mosaica false negative and false positive rates Fine punctuation Coarse punctuationa associated with cytology alone and Iodine partial positivity also reduces the need for call-back for Iodine negativitya repeat cytology, the colposcope being Atypical vesselsa used as a guide to collection of the cytology specimens (Van Niekerk et III Colposcopic features suggestive of invasive cancer al., 1998). Within Germany at least, IV Unsatisfactory colposcopy there is some reluctance to support the Squamocolumnar junction not visible continued use of colposcopy as a Severe inflammation, severe atrophy, trauma screening tool to assist in the taking of Cervix not visible cytological specimens, since there is V Miscellaneous findings no evidence that the quality of smears Condylomata is improved (Hilgarth & Menton, 1996). Keratosis Furthermore, constraints limiting the Erosion application of colposcopy to universal Inflammation screening include its high cost relative Atrophy Deciduosis to cytology, the availability and acces- Polyps sibility of adequately trained colpo- scopists, and the lower ability of col- From Walker et al. (2003) poscopy to detect endocervical lesions a Major changes (Van Niekerk et al., 1998; Belinson et al., 2001). Since colposcopy was introduced histological sample from the endocer- The colposcope can also be used in the 1960s to the English-speaking vical canal if the new squamocolumnar to assess the remainder of the lower world, it has been selectively applied junction (and thus the entire transfor- genital tract (vagina, vulva and peri- for diagnosis in women who are mation zone) cannot be examined. anal skin), especially if no cervical referred because of an abnormal cyto- Debate continues as to whether histo- lesion is found in a woman with abnor- logical test. Current indications for col- logical sampling of the endocervical mal cytology. Women who are HIV- poscopy are listed in Table 33. canal should be performed routinely in positive tend to have multifocal disease all women undergoing colposcopic involving the vagina, vulva and peri- Biases and caveats in the examination or only in circumstances anal areas, and therefore these assessment of colposcopy such as when the new squamocolum- regions need to be examined Most assessments of the sensitivity nar junction cannot be seen or the col- (Abercrombie & Korn, 1998). and specificity of colposcopy and poscopic examination is deemed to be directed biopsy are susceptible to bias. unsatisfactory. It is also suggested to Primary screening and diagnosis The colposcopic impression confounds be used when the colposcopic exami- Colposcopy continues to be used the reference standard of diagnosis nation is satisfactory but a cytological routinely as part of a standard gynae- (histology) since it dictates where the test indicates a higher grade of lesion cological examination by many clini- histological specimen is obtained from, (Spirtos et al., 1987; Fine et al., 1998; cians in some European and Latin leading to an inflated estimate of the Pretorius et al., 2004). American countries, probably as a accuracy of colposcopy. In contrast to

86 059-116 24/01/05 11:06 Page 87

Screening tests

Table 32. Combined colposcopic index, commonly used to score and document abnormal areas seen on colpo- scopic examination

Colposcopic sign Zero points One point Two points

Colour Less intense acetowhitening (not Intermediate, shiny, Dull, oyster-white completely opaque); indistinct, grey-white shade semi-transparent acetowhitening. Acetowhitening beyond the margin of the transformation zone Snow-white colour with with intense surface shine

Lesion margin and surface configuration Feathery, indistinct, or finely scalloped Regularly-shaped Rolled, peeled edges edges lesion with sharp, Internal margins Angular, irregularly shaped, geographic straight edges separating lesions margins with differing scores, Satellite lesions with margins well the more central one removed from the new squamocolumnar with the higher junction score tending to be Lesion with a condylomatous or nearest the new micropapillary contour squamocolumnar junction

Blood vessels Fine punctuation or mosaic pattern Absent vessels (after Coarse punctuation application of acetic or mosaic pattern acid

Iodine staining Positive iodine staining (mahogany- Partial iodine uptake Negative for uptake brown colour) giving a veriegated giving a mustard Negative iodine staining in an area that pattern yellow appearance scores 3 or less on the first 3 criteria in area that is significant (4 or more points) by the other 3 criteria

From Reid & Scalzi (1985); Reid (1993) A score of 0–2 is compatible with CIN 1; 3–5 with CIN 1 or 2; and 6–8 with CIN 2 or 3.

studies in which colposcopy is used for when it is used as a screening tool. If China, Belinson et al. (2001) observed primary screening (with or without cytol- possible, all women evaluated with a that increased use of technology alone ogy), studies assessing colposcopy as a test under assessment should have the does not guarantee that detection diagnostic procedure are conducted on reference standard applied to avoid ver- improves. Important factors are women referred with abnormal screen- ification bias and where this is not possi- whether the quality of light used opti- ing cytology and having, therefore, a ble, statistical correction should be mizes perception, the adequacy of higher probability and possibly a more made. When colposcopic findings are training of the personnel, and the severe spectrum of cervical pathology. compared with the pathological diagno- attributes of the population studied, Since women with more pronounced sis, the colposcopist and the pathologist such as prevalence of cervical inflam- findings and disease may be selected by should be blind to corresponding infor- mation. The definition of abnormality screening, the performance of col- mation from the other test. and certainty thresholds used by col- poscopy in a diagnostic capacity may In relation to a large multidiscipli- poscopists in a study is important, exceed its accuracy and reproducibility nary study of precancerous lesions in since these determine the replicability

87 059-116 24/01/05 11:06 Page 88

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 33. Indications for colposcopy detection of CIN 2 or worse lesions were 62.4% (234/375; 95% CI • Positive screening test result suggesting an increased probability of cervical 57.3–67.3%) and 93.7% (7612/8122; neoplasia, e.g., cytologya, visual inspection with acetic acid (VIA) and/or Lugol’s 95% CI 93.2–94.2%), respectively iodine (VILI) (Pretorius et al., 2004). Among the women with satisfac- • Suspicious-looking cervix (where cancer cannot be excluded); regardless of the screening test result tory colposcopy in the same study, directed biopsy detected 57.1% of • Presence of clinically apparent leukoplakia since a hyperkeratotic area may obscure high-grade lesions and cancers, while a lesion and preclude adequate cytological sampling of the underlying area four-quadrant biopsy and ECC detected 37.4% and 5.5%, respec- • Presence of external genital warts; regardless of the screening test results (in tively. Among women referred for a some systems) (Howard et al., 2002; Li et al., 2003) cytological abnormality, directed biop- • Women at increased risk of cervical neoplasiab sies were 4.8 times more likely to show a high-grade lesion or cancer than a Abnormal cytology including ASCUS (with positive oncogenic HPV test), LSIL, HSIL four-quadrant biopsies (26.5% versus b Those who are HIV-positive; those with external genital warts 5.5%). The yields of CIN 2 or higher from four-quadrant biopsies for women referred because of HSIL, LSIL or ASCUS with a positive HPV test were of findings and the test cut-off for what in the previous meta-analysis, were 17.6%, 3.6% and 1.7%, respectively. are the minimal criteria for abnormality. similar. One of 20 women in whom CIN 2 or A recent study of the diagnostic worse was detected only by ECC had Studies of diagnostic colposcopy accuracy of colposcopy in China cancer despite satisfactory col- Two meta-analyses have been per- (Belinson et al., 2001) included poscopy. formed on the accuracy of diagnostic methodological features intended to A cohort study of 255 colposcopi- colposcopy applied to women referred reduce selection bias and to assess cally negative women with abnormal with abnormal cytology. Mitchell et al. the degree to which colposcopically cytology and 726 controls with normal (1998a) performed a systematic review directed biopsy is confounded with the cytology were followed for five years to of 86 articles published between 1960 colposcopic impression and the refer- assess the probability of false-negative and 1996, nine of which met the inclu- ence standard. In this study, vaginal colposcopy (Milne et al., 1999). sion criteria and eight were eligible for and cervical specimens from 8497 Subsequent neoplasia was found in meta-analysis. At the cut-off level of women (aged 27 to 56 years) were 19% versus 3% of controls (p < normal versus abnormal on col- screened for 13 oncogenic types of 0.0001). poscopy, the average weighted sensi- HPV (Hybrid Capture 2 assay) and by tivity, specificity and area under the liquid-based cytology (AutoCyte, Studies of screening colposcopy receiver operating characteristic TriPath, Burlington, NC) (Pretorius et In a cross-sectional study, 1997 (ROC) curve of histological CIN 2 or al., 2004). Colposcopy was performed unscreened Chinese women (aged more were 96%, 48% and 80%, on 3063 women who had an abnor- 35–45 years) first were assessed by respectively. At the cut-off level of nor- mality on screening and a directed VIA performed by a gynaecologist and mal and LSIL versus HSIL and cancer biopsy was obtained from any abnor- then a second gynaecologist (blinded on colposcopy, the corresponding mality. If colposcopy showed no lesion to the VIA results) performed col- results were 85%, 69%, and 82%. This in a quadrant of the transformation poscopy with directed biopsies being suggests that, independent of preva- zone, a biopsy was obtained in the taken from abnormal areas (Belinson lence and compared with low-grade original squamocolumnar junction in et al., 2001). All women also had a lesions, high-grade lesions and cancer that quadrant. An ECC was then per- biopsy taken from each of the four are diagnosed with higher sensitivity. formed after biopsies had been quadrants (and all had an ECC) in Olaniyan (2002) reviewed publications obtained. Based on all of the women order to estimate the performance of from 1966 to 2000 and the results of who had colposcopy (including 11 with colposcopy in a screening setting. his meta-analysis, based on eight unsatisfactory colposcopy), the sensi- Sensitivity and specificity of col- studies, seven of which were included tivity and specificity of colposcopy for poscopy and directed biopsy for high-

88 059-116 24/01/05 11:06 Page 89

Screening tests

grade CIN or cancer were 81% (95% degrees of difference within certain entire transformation zone by loop CI 72–89%) and 77% (95% CI morphological parameters (Table 32). electrosurgical excision procedure 75–78%) compared with the combined These included reference to the colour (LEEP) and confirmed an association histological findings from the directed, of the cervical epithelial, blood vessel between lesion area and histological four-quadrant and ECC specimens. structure and the surface configuration grade. A study that estimated lesion A similar study in Germany of the epithelium of the transformation size from cervigrams concluded that enrolled 4761 women 18–70 years of zone, as well as the degree of iodine lesion size affects the sensitivity of age who were screened by conven- staining. However, few major studies cytology (Barton et al., 1989). tional cytology (obtained under colpo- have studied the incorporation of this Colposcopically inapparent high-grade scopic vision), HPV testing of cervico- scoring system within a colposcopic lesions, remaining after a directed vaginal samples by PCR and probing management regime. One retrospec- biopsy was taken, were evenly distrib- for 13 high-risk types and colposcopy tive study of 134 women with biopsy- uted among the four quadrants at 2, 4, 8, when they visited one of ten gynaecol- proven lesions using the modified Reid and 10 o’clock (Pretorius et al., 2004). ogists for standard care (Schneider et index score showed that it gave more While most studies of colposcopi- al., 2000). Biopsy and EEC were per- accurate prediction of low-grade ver- cally directed biopsy have shown less formed where appropriate and if col- sus high-grade disease than when the than perfect sensitivity for detecting the poscopy was normal, biopsies at 6 and 1976 International Nomenclature for presence of a higher-grade lesion 12 o’clock and ECC were obtained. Colposcopic Classification was found on a subsequent LEEP speci- The sensitivity and specificity of employed (Carriero et al., 1991). men (Howe & Vincenti, 1991; Barker et screening colposcopy for detecting at Prospective research on the pre- al., 2001), the rate of underestimation least CIN 2, by histological confirma- dictive validity of visual signs in 425 among HIV-positive women may be tion, were 13.3% (95% CI 7.0–20.5) women with abnormal cytology substantially higher (Del Priore et al., and 99.3% (95% CI 99.0–99.6), referred to a Canadian colposcopy 1996). respectively. clinic has shown that among three Five studies of the simultaneous morphological characteristics routinely Reproducibility of colposcopy use of colposcopy and cytology to evaluated within the abnormal transfor- Observer agreement studies of visual detect cervical cancer, performed mation zone (borders, degree of ace- methods have been conducted using more than 30 years ago, showed that towhitening, abnormal blood vessels), cervical photographs taken after the the combined sensitivity of the two performance based on acetowhitening application of dilute acetic acid. methods for cervical cancer varied was as good as all three signs com- Between three expert colposcopists, from 95.0% to 99.4% (Dexeus et al., bined (Shaw et al., 2003). A prospec- intra-observer and inter-observer 1977). A recent case series from a tive study of 2112 women referred to agreement was poor to good when German university using colposcopy the Cook County Hospital Dysplasia assessing border characteristics (range and cytology for primary screening Clinic in Chicago did not use standard- of inter-observer kappa, 0.13–0.41; of showed that the sensitivity of col- ized grading criteria, but did show an intra-observer kappa, 0.26–0.58) and poscopy for detecting at least CIN 2 association between histology and col- the colour of acetowhitening (range of was 90.8% (148/163) based on poscopic impression (p < 0.001), inter-observer kappa, 0.21–0.47; of directed biopsy (Hilgarth & Menton, although agreement was poor (kappa, intra-observer kappa, 0.34–0.75). 1996). A similar study in the USA, 0.20) (Massad & Collins, 2003). There was excellent agreement as to based on 196 women who were The size of a lesion (categorized as the site of the lesion from which a screened opportunistically in a gynae- the number of quadrants with positive biopsy should have been obtained cologist’s practice, gave estimated histology) affects the sensitivity of col- (raw agreement, 95.3%, 143/150) sensitivities of screening cytology, col- poscopy for detecting at least CIN 2 (Sellors et al., 1990). Ferris et al. poscopy and their combination of 48%, when the lesion grade on referral cytol- (2000b) studied the inter-observer 76% and 91% (Davison & Marty, ogy or histology is controlled (Pretorius agreement within pairs of colpo- 1994). The estimated specificities were et al., 2001). Colposcopy had a sensi- scopists using optical and video colpo- 100%, 96% and 96%, respectively. tivity of 65% (95% CI 47–79%) if the scopes and found that colposcopic lesion involved only one quadrant of impression agreement with histo- Validity of visual signs the cervical surface and 100% if more pathology (kappa, 0.60; 95% CI Reid and Scalzi (1985) published a surface was involved (Belinson et al., 0.53–0.68), biopsy intent agreement scoring system which quantified the 2001). Shafi et al. (1991) excised the (79.9%) and biopsy site selection

89 059-116 24/01/05 11:06 Page 90

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

agreement by quadrant (A, 78.3%; B, (Sellors et al., 1990; Etherington et al., obtained by punch biopsy. 81.3%; C, 85.3%; D, 82.7%) were not 1997; Milne et al., 1999; Harper et al., Psychological morbidity should be significantly different (p ≥ 0.3), despite 2000; Li et al., 2003). Harper et al. appreciated and counselling consid- the use of different colposcopes. (2000) showed that a telescopy net- ered (Howard et al., 2002a). Studies Similar agreement was obtained when work that allows transmission and using measures of anxiety such as the telecolposcopy (using a video colpo- sharing of static colposcopic images State-Trait Anxiety Inventory have con- scope) was viewed by an expert colpo- for consultation and teaching purposes sistently shown that anxiety scores scopist at a remote location and com- on a regular basis was technically fea- before colposcopy are markedly ele- pared with the video colposcopy per- sible, acceptable to women and health vated to levels seen in patients await- formed by an expert on-site.The kappa care providers living in remote areas, ing surgery, and fall immediately after values for colposcopic impression and and gave good inter-observer agree- colposcopy is completed. The fears histopathology agreement varied ment between the on-site colpo- that women have before colposcopy between 0.16 and 0.31 (p values not scopists and the off-site review colpo- relate to cancer, fertility, danger to part- given) and for biopsy intent, kappa was scopists as to degree of abnormality ner, social stigma and pain or embar- 0.32 (p = 0.002) (Ferris et al. 2002). (kappa = 0.68; 95% CI 0.54–0.82). rassment during the procedure. Other Assuming that colposcopists use Ferris et al. (2003) showed that net- women may have a significant level of the same definitions, reproducibility of work telecolposcopy using high-speed anxiety about the examination colposcopic assessment depends in telecommunications lines and com- because of a possible history of sexual part on colposcopists using similar puter telecolposcopy using modems abuse. Educational booklets and coun- ‘thresholds of certainty’ for categoriz- and telephone lines to transmit static selling are effective in reducing anxiety ing findings as to normal versus abnor- images was superior to cervicography (Ferris et al., 2003). Colposcopy ser- mality and grade. as measured by the number of con- vice providers need to be sensitive and firmed CIN lesions detected and time- responsive to women’s needs in order Quality control liness of results. On-site colposcopy to provide an acceptable service and Like other medical services, col- had the highest sensitivity to detect to optimize adherence to appoint- poscopy services can be audited and CIN because of the stereoscopic ments. compared with national standards, vision, the ability to manipulate the such as those established for the cervix and view the acetowhite reac- English National Health Service, for tion as it occurs, and the ability to Cervicography process and outcome (Ferris et al., resolve vascular and epithelial fea- 2002). Indicators recommended for tures. Compared with telecolposcopy, Cervigrams are replicate photographs periodic audit include waiting time for the ability to assess whether a colpo- of the cervix taken after application of colposcopy by grade of referral smear; scopic examination is satisfactory 5% acetic acid, using a camera with a adequacy of communication between appears to be better with on-site col- fixed focal length and internal light primary and secondary level; fre- poscopy (Sellors et al., 1990). source. The images are projected onto quency of procedures; agreement Documentation of colposcopic a screen at a fixed distance to simulate between colposcopic diagnosis, refer- images and data using the latest digi- magnification and are interpreted by a ral cytology and histology; treatment tal photographic and information sys- trained evaluator. method by histological diagnosis; effi- tems allows not only recording and It is now possible to achieve equal cacy of treatment (e.g., whether histo- comparison of colposcopic findings visual resolution with digital cameras, logical evidence of CIN is present in with subsequent examinations, but producing images that can be immedi- over 90% of women undergoing ‘see also the retention of data for audit, ately downloaded and transmitted for and treat’); and follow-up rates at one post-treatment follow-up and compari- expert review, the images being evalu- year (Luesley, 1996). son of data between units. ated with computer-generated magnifi- Cervical imaging using colpopho- cation as needed. Future efforts related tography, video colpography, and Potential side-effects of to cervicography will depend on digital telecolposcopy has been studied. All colposcopy techniques capable of generating methods give a true representation of A routine colposcopic examination images as good as those using high- what is seen at colposcopy and have involves some discomfort due to the quality film, with the advantages of ‘tele- been recommended for teaching and insertion of the vaginal speculum and medicine’-based screening and central- audit, as well as for patient care more when a tissue specimen is ized image analysis (Wright, 2003).

90 059-116 24/01/05 11:06 Page 91

Screening tests

Cervigrams are interpreted using second evaluator yielded a kappa sta- graphy appeared somewhat less accu- the categories presented in Table 34. tistic of 0.5, indicating only moderate rate than cytology, primarily because With these criteria, initial studies agreement beyond that expected by of inferior specificity. This was a result showed poor reproducibility because chance (Schneider et al., 2002). of the overcalling of acetowhite epithe- of differing distinction of the very subtle The deficiencies and inconclusive lial changes. acetowhite lesions that represent results of several small-scale studies The percentage of technically inad- either immature squamous metaplasia of cervicography were summarized by equate cervigrams varies widely by or HPV changes. Nuovo et al. (1997). In later large-scale study; satisfactory results depend on In a large inter-observer study evaluations, summarized in Table 35, the experience of the evaluator (De among 3637 women, a comparison of cervicography proved insufficiently Sutter et al., 1998). Adjudicated cervi- dichotomous results (positive versus accurate to serve as a stand-alone gram reviews and histological re-con- not) assigned by the initial versus the screening test. In summary, cervico- firmation of CIN 2, CIN 3 or cancer did improve performance over a single Table 34. Cervigram classificationa interpretation, but suggested the upper limit of sensitivity (Schneider et al., 2002). The sensitivity and specificity Not referred for colposcopy: depended on the cut-point of positivity Negative: No definite lesion is visible and targeted disease end-point, but no choice of cut-point generated excellent Atypical 1 (A1): A lesion inside the transformation zone is visible; based overall accuracy of detection of CIN 3 on the lesion’s site and morphology, the lesion is and cancer. presently considered to be of doubtful significance A major limitation of cervicography (and possibly, by extension, other sta- Atypical 2 (A2): A lesion outside the transformation zone is visible; based on the lesion’s site and morphology, the lesion is tic visual techniques) is the poor sensi- presently considered to be of doubtful significance tivity among older women, whose transformation zones are often beyond Technically defective: The cervigram slide is not adequate for evaluation the field of vision (Schneider et al., 1999). Women aged 40–60 years would be expected to represent a size- Referred for colposcopy: able proportion of women being Positive (all categories below): A lesion is visible and colposcopy is recommended screened in the low-resource settings because of the lesion’s site and morphology, or no definite lesion is visible, but the where a non-cytological technique appearance warrants colposcopy to exclude significant disease. such as cervicography might be partic- ularly helpful (Wright, 2003). However, Positive 0 (P0): Probably normal variant; appearance warrants in this group, the targeted precancer- colposcopy to exclude significant disease ous lesions can be small enough to be easily missed. In one evaluation, the Positive 1A (P1A): A lesion extending into the canal, the visible portion of which is presently considered to be of doubtful sensitivity of cervicography for detec- significance tion of CIN 2, CIN 3 and cancer was only 30.0% among 2196 post- Positive 1B (P1B): Compatible with a low-grade lesion menopausal women, compared with 54.7% among 6264 pre-menopausal Positive 2 (P2): Compatible with a high grade lesion women, using a positive cut-point, and findings were similar using an atypical Positive 3 (P3): Compatible with cancer cut-point (see Table 34 for definitions) a As of 1 January 1995, National Testing Laboratories worldwide revised the atypical (Schneider et al., 1999). category. Previously, atypical 1 referred to trivial lesions outside the transformation The relative accuracy of direct zone and atypical 2 referred to trivial lesions inside the transformation zone. visual inspection compared with dis- tant, expert review of a static visual Modified from Schneider et al. (1999) image is not clear. There appears to be a trade-off between colposcopic

91 059-116 24/01/05 11:06 Page 92

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 35. Selected screening studies of cervicography

Study Population Disease target Cervicography Sensitivity Specificity PPV (%) NPV (%) threshold (%) (%)

Coibion et al. Belgium, n = 4015 > CIN 1, n = 123 Old Atyp.a 86 99 76 99.0 (1994)

Schneider et al. Germany, n = 967 > CIN 2, n = 38 Atyp. or Pos.a 45 91 17 97.6 (1996)

Baldauf et al. France, n = 1351, > CIN 1, n = 168 Positivea 51 96 44 97.1 (1997) mixed screening/ referral population

De Sutter et al. Belgium, n = 5192 > CIN 2, n = 33 Positivea 55 97 11 99.7 (1998)

Schneider et al. Costa Rica, > CIN 2, n =136 Atyp. or Pos. 63 85 6 99.3 (1999) n = 8460b

Positive 49 95 14 99.1

Denny et al. (2000) South Africa, > CIN 2, n = 79 Positive 58 91 58 93.4 n = 2611

Costa et al. (2000) Italy, n = 992 > CIN 2, n = 90c Atyp. or Pos. 76 91 51 97.4

Cronjé et al. (2001) South Africa, > CIN 1, n = 342 Positive 42 79 32 84.8 n = 1747d

Cronjé et al. (2003) South Africa, > CIN 2, n = 90 Positive (not P0) 49 88 26 95.0 n = 1093

Ferreccio et al. Costa Rica, > CIN 3, n = 110 Atyp. or Pos. 62 85 5 99.4 (2003) n = 8457 including follow-up

a Evaluation not performed according to National Testing Laboratory criteria. b Population-based screening of a high-risk province, where attempts were made to vary cervicography cut-point and disease end- point to explore performance. c Women with negative colposcopy presumed to be disease-negative. d Analysis of subgroup of large group of screened women. Subgroup included those with biopsied acetowhite lesions, as well as 1/5 of women with seemingly normal cervix. Predictive values not adjusted for sampling.

Sensitivity and specificity are estimated cross-sectionally (see Chapter 4)

expertise and the loss in visual dis- visual inspection by nurses, due to ogists and nurses with varied training crimination inherent in examining an increased specificity (Rodriguez et (Ferris & Litaker, 2004). image compared to real-time exami- al., 2004). However, in a statistically Since it has been concluded, on nation. In one cross-sectional screen- more powerful study, distant review by the basis of accumulated data, that ing study with limited statistical power experts of digitized, static colposcopy cervicography is inadequate as a due to small numbers of precancer- images was significantly less sensi- stand-alone screening technique, ous outcomes, cervicography was tive (but more specific) than col- research has shifted to evaluation apparently more accurate than direct poscopy performed by local gynaecol- of combining cervicography with

92 059-116 24/01/05 11:06 Page 93

Screening tests

cytology or HPV for screening, and to Techniques to detect the presence mucosal sites other than genital, and a possible role for cervicography in of HPV in cervical cell specimens have thus is not suitable, in principle, as a the triage of women with equivocal evolved considerably in the last 25 screening tool. cytology. These topics are considered years, from (i) simple scoring of koilo- Early clinical studies used non- below in the section on combined cytes (a type of cytopathic effect taken amplified DNA hybridization methods techniques. to indicate the presence of HPV in the (without signal amplification) to gauge host epithelial cells) in cervical smears the screening utility of HPV testing to (Komorowski & Clowry, 1976) to (ii) identify and manage cervical lesions. HPV DNA testing immunocytochemical staining (Syrjänen Such methods are no longer used, & Pyrhonen, 1982); non-amplified however, because of their insufficient Research on the use of HPV DNA nucleic acid hybridization methods, sensitivity and specificity for epidemio- assays as a potential cervical cancer such as (iii) dot blot (Parkkinen et al., logical and clinical studies (Franco, screening tool began in the late 1980s, 1986), (iv) Southern blot (Okagaki et 1992; Schiffman & Schatzkin, 1994). as a reflection of the emerging evi- al., 1983) and (v) filter in-situ hybridiza- Only the commercially available HC dence that these viruses played a tion (Schneider et al., 1985); signal- assay and a few PCR protocols have causal role in the genesis of cervical amplified, immunoassay-based nucleic been the focus of investigations con- neoplasia (zur Hausen, 1976; acid hybridization techniques such as ducted in the last 10 years. Although a Deligeorgi-Politi et al., 1986). Although (vi) the Hybrid Capture™ (HC) assay number of biotechnology companies much of that research began with a (Farthing et al., 1994); and (vii) a vari- are currently developing HPV DNA focus on viral detection as an end in ety of type-specific (Dallas et al., 1989) diagnostic systems for clinical use, few itself (reviewed by Schiffman, 1992), and general or consensus-primer are yet available commercially or have attention soon turned to the potential (Gregoire et al., 1989; Manos et al., reached the stage of large-scale clini- clinical utility of HPV testing for identi- 1989; Snijders et al., 1990; Roda cal studies. For this reason, this fying cervical cancer precursors Husman et al., 1995; Kleter et al., overview focuses primarily on the HC (Lörincz et al., 1990; Wilbur & Stoler, 1998; Gravitt et al., 2000) polymerase assay and on the more popular PCR 1991; Lörincz, 1992). The basic chain reaction (PCR) techniques. In protocols that have been used in assumption was that standardized addition, adaptations of the solution- screening studies. molecular testing of exfoliated cervical based, non-amplified hybridization and cells for the putative causal agent of PCR protocols have been used to Hybrid Capture™ assay cervical cancer could have acceptable detect HPV DNA in histological sec- Most clinical investigations of HPV diagnostic performance, while being tions or smears, to allow confirmation testing have used first- or second-gen- more reproducible and more easily of the presence of the virus in particu- eration Hybrid Capture™ (HC) sys- adapted for automated, high-volume lar target cells. Such in situ techniques tems (Digene, Inc., Gaithersburg, MD), testing in clinical practice than conven- (Gupta et al., 1985; Nuovo et al., 1991) the only HPV test currently approved tional cytological testing. Concerns in have been useful in molecular pathol- by the US FDA. The HC system is a the USA about the quality of smears ogy studies, but have found little inter- nucleic acid hybridization assay with processed in cytopathology laborato- est as potential screening tools for cer- signal amplification for the qualitative ries added pressure to study the vical cancer and its precursors. detection of DNA of high-risk, cancer- potential use of HPV testing as an Serological assays to detect antibodies associated HPV types in cervical spec- adjunct to cytology (Reid et al., 1991; to HPV capsid or functional protein imens. It cannot determine the specific Reid & Lörincz, 1991), despite some antigens have also received attention HPV type present, since detection is opposing views (Nuovo & Nuovo, as investigational tools in epidemiolog- performed with a combined probe mix. 1991; Beral & Day, 1992). More ical and clinical studies (Jochmus- The first HC assay (HC1) was a tube- recently, cytology has been character- Kudielka et al., 1989; Galloway, 1992). based detection system and probed for ized not only as a sufficient screening However, as with in situ assays, they only nine of the high-risk HPV types: test, but also as a likely necessary have not been considered as candi- 16, 18, 31, 33, 35, 45, 51, 52 and 56. component of future screening pro- date methods for screening cervical The second-generation HC system grammes based on HPV or visual test- cancer precursors. Serology detects (HC2) has improved reagents and is ing, due to the low relative specificity of humoral immune response to HPV based on a microplate assay lay-out non-cytological methods (Suba & antigens, which may reflect cumulative that targets 13 high-risk HPV types: Raab, 2004). exposure to HPV infection acquired in 16, 18, 31, 33, 35, 39, 45, 51, 52, 56,

93 059-116 24/01/05 11:06 Page 94

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

58, 59 and 68. A probe set for a few with polyclonal IgG antibodies that are station named Rapid Capture non-oncogenic HPV types (6, 11, 42, specific for RNA–DNA hybrids, regard- System™ (Digene) is available, which 43, 44) has been available for both the less of sequence homology. Any such performs specimen transfer, all pipet- HC1 and HC2 assays but its utility has hybrids will then be captured by the ting operations, incubations, shakings not been sufficiently investigated in solid-phase-bound antibodies, hence and washings. However, the denatura- clinical or epidemiological studies. It is the name 'hybrid capture'. After wash- tion of specimens in the sample device often designated as probe A, whereas ing steps to remove unbound mole- tubes still has to be performed by probes for high-risk HPV types are cules, a solution of a conjugate hand. This automatic station increases referred to as probe B. reagent consisting of the same anti- the accuracy of the test and allows a HC2 is an entire system that can RNA–DNA hybrid antibody covalently single user to test 352 specimens be used with a dedicated cervical sam- linked with the enzyme alkaline phos- within four hours. pler kit containing a special cervical phatase is added to the wells. Since it is based on signal, rather conical brush and a vial with specimen Conjugate antibody molecules will than target amplification (as in the transport medium (STM). The brush is then bind to any solid-phase-bound case of PCR protocols), HC2 is less designed for optimal collection of cells hybrids. After further washing to prone to cross-specimen contamina- from both the ectocervix and endo- remove unbound molecules, a solution tion, thus obviating the need for special cervix. The brush is shaped as a cone containing a chemiluminescent dioxe- laboratory facilities to avoid cross-con- (Christmas-tree-like) that fits the cervi- tane substrate is added to the wells. tamination (Coutlee et al., 1997). In cal canal and samples the endo- and Cleavage of the substrate by the alka- practice, only the high-risk probe mix ecto-cervix. This brush is inserted gen- line phosphatase releases a lumines- (probe B) is used for cervical lesion tly into the cervical canal and fully cent reaction product into the solution. screening, which reduces the time and rotated three times. It is then retrieved The intensity of the light emitted is pro- cost to perform the test. At the stan- without touching the vaginal wall and portional to the amount of HPV DNA dard FDA-approved cut-off of 1 pg/ml inserted into the collection tube con- originally present in the specimen and (RLU ≥ 1.0) and even at higher taining STM. The tip is broken and the is measured in a luminometer provided discriminant levels, there is cross-reac- tube is closed. According to the manu- with the system. The reaction signal of tivity between certain HPV types not facturer, specimens in STM can be each specimen is expressed on a present as targets in the probe B set held at room temperature for up to two scale (relative light units or RLU) rela- (e.g., 53, 66, 67, 73) and the RNA weeks and can be stored for an addi- tive to the average reactivity measured probes used in that set (Peyton et al., tional week at 4°C. If not tested in the in triplicate wells with a positive control 1998; Vernon et al., 2000; Howard et first three weeks after collection, they containing 1.0 pg of HPV16 DNA per al., 2002b, 2004). Cross-reactivity with can be stored at –20°C for up to three ml. Specimens yielding RLUs greater non-cancer-causing types would have months. than or equal to 1.0 are considered an adverse impact on test specificity in HC2 is a solution hybridization positive; some studies have assessed settings with high prevalence of the assay that uses long synthetic RNA the validity of this cut-point using ROC low-risk types. On the other hand, probes that are complementary to the curve analysis (Schiffman et al., 2000). cross-reactivity with other high-risk DNA sequence of the 13 high-risk HPV In most clinical settings, the manufac- types not represented in the probe B types (or to the probe A types) listed turer (Digene) certifies the laboratory set may be beneficial for test sensitiv- above. The initial reaction step dena- that intends to perform HC2 testing, ity (Castle et al., 2003). tures the exfoliated cells in STM, thus thus ensuring quality control. releasing host and any existing HPV Because the RLU signal is propor- Polymerase chain reaction DNA molecules to the solution. HPV tional to the amount of HPV DNA pre- PCR is based on the repetitive replica- DNA molecules then bind (i.e., sent in the specimen, the HC2 assay tion of a target sequence of DNA hybridize) with the respective RNA has occasionally been used to infer flanked at each end by a pair of spe- probe, resulting in the formation of viral load, on a semi-quantitative basis cific oligonucleotide primers, which ini- DNA–RNA hybrids reflecting the com- (Clavel et al., 1998; Sun et al., 2001; tiate the polymerase-catalysed chain position of HPV types present in the Cuzick et al., 2003). The assay is easy reaction. Because of the exponential mixture. This hybridization step occurs to perform in clinical practice and increase in the amount of target DNA in solution inside the wells of a amenable to automation, which makes sequence after a few reaction cycles of specially treated 96-well plastic it attractive for high-volume screening denaturation, annealing and extension, microtitration plate previously coated use. To this end, a robotic assay work- PCR has very high levels of molecular

94 059-116 24/01/05 11:06 Page 95

Screening tests

sensitivity and permits the detection of Several procedures are well estab- detect essentially all types of HPV that less than 10 copies of HPV DNA in a lished to minimize the potential for con- infect the mucosal areas of the lower mixture. Therefore, PCR has a lower tamination, the most important of genital and upper aerodigestive tracts. threshold of molecular detection for which is the separation of pre-amplifi- Two of these, the MY09/11 (Manos et HPV DNA than the HC assay. PCR is cation and post-amplification areas. al., 1989) and the GP5/6 (Snijders et based on target amplification with Judicious analysis of sequence al., 1990; Van den Brule et al., 1990) type-specific or consensus or general homology among different genes of systems have evolved into variants primers. The latter are able to amplify distinct HPV types using software that with better primer composition and sequences from several different HPV aligns DNA sequences will reveal internal oligonucleotide probing, such types because they target conserved countless segments that could serve as the PGMY09/11 (Gravitt et al., DNA regions in the HPV genome. The as candidates for PCR primer design. 1998, 2000) and the GP5+/6+ (Roda amplified DNA products can be In fact, many type-specific and con- Husman et al., 1995; Jacobs et al., revealed by ethidium bromide staining sensus HPV testing PCR protocols 1995, 1997; Van den Brule et al., 2002) following agarose or acrylamide gel have been published in the last 15 protocols. Over the years, the original electrophoresis, which permits pre- years. However, because of the radioactively labelled hybridization sumptive verification of the expected requirements for validation, repro- probes have gradually been aban- molecular weight of the amplified tar- ducibility, and general acceptability, rel- doned in favour of biotinylated probes get, thus confirming positivity. atively few have become established to and enzyme immunoassay formats. Verification can also be done by meth- the point of being widely used in clini- The third protocol is designated SPF10 ods that further probe the post-amplifi- cal and epidemiological studies. LiPA, for line probe assay based on the cation products for their sequence Primer systems targeting sequences in SPF10 primer set (Kleter et al., 1998; homology with the target. Dot blot, the L1, E1, E6 and E7 genes have Quint et al., 2001). Although these Southern blot or line strip hybridization been most commonly used. Because three consensus protocols amplify tar- are used to this end and generally of their well conserved sequences, L1 gets within the L1 gene of HPV, they result in improved molecular sensitivity and E1 have been targeted by the con- do so for segments of considerably dif- and specificity as compared with elec- sensus primer protocols. E6 and E7, ferent sizes: 450 base pairs (bp) for trophoresis and staining (Gravitt & on the other hand, have more MY09/11, 140 bp for GP5/6, and 65 bp Manos, 1992; Gravitt et al., 1998). sequence variation among HPV geno- for SPF10 LiPA. The size of the ampli- Finally, use of restriction enzymes to types, making them less suitable as fied product is not a trivial matter. analyse the fragment length signatures targets for amplification of a broad Although discrimination of sequence in combination with probe hybridization spectrum of HPV types (Gravitt & homology is better for longer gene (Bernard et al., 1994) and direct DNA Manos, 1992). segments and thus would in theory sequencing provide the highest possi- The most widely used PCR proto- permit improved HPV type resolution, ble resolution to distinguish the HPV cols are of the consensus or general shorter fragments tend to yield better types present in a biological specimen. primer (degenerate or non-degener- sensitivity with severely degraded The very high sensitivity of PCR is ate) type, i.e., they can potentially specimens, such as paraffin-embed- its very limiting factor in terms of clini- amplify sequences of multiple HPV ded, archival tumour tissue. Damage is cal applicability. Molecular threshold types with one primer set in one reac- often pronounced in DNA extracted does not correlate directly with clinical tion pass. The size of the amplified from such archival specimens, result- sensitivity and specificity (Snijders et product is the same irrespective of the ing in DNA fragments of less than 200 al., 2003). Because millions of copies HPV type present in the starting mix- bp. In these circumstances, a protocol of the DNA target can be produced ture, and thus electrophoresis cannot targeting a short fragment, such as from a single molecule, there is a high reveal the actual type present in the GP5+/6+ or SPF10 LiPA, tends to yield probability of contamination of other sample. The post-amplification hybridi- fewer false negative results (Gravitt & specimens and control samples with zation or sequencing techniques Manos, 1992). HPV sequences in airborne droplets described above must be used to iden- The newly developed Roche proto- and aerosolized reaction mixtures. In tify the HPV type or types originally type Microwell plate assay (Roche fact, cross-contamination was a major present. Three consensus primer sys- MWP) employs an oligonucleotide set problem in some early applications of tems (and their technical variations) which amplifies a short fragment of the PCR in HPV testing. Extreme care is based on L1 sequence detection have L1 gene of high-risk HPV types (170 needed in PCR testing laboratories. become well established. They can bp, compared with 450 bp with

95 059-116 24/01/05 11:06 Page 96

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

PGMY09/11). This amplicon is immobi- tized glass slide. Target DNA is sub- ing and for cytology in the same stud- lized using a pool of capture molecules jected to a standard PCR in the pres- ies. Data on the performance of HPV bound to the wells of a microtitre well ence of fluoresceinated nucleotides testing in triage studies are presented plate and visualized by colorimetric (labelled with Cy5 or Cy3) employing later in this chapter. For all studies, detection. The new test was developed primers for both the beta-globin specificity estimates are based on to employ the TaqGold DNA poly- (PC03/04) and for the L1 region (mod- women free of histologically demon- merase, which minimizes the amount ified GP5/6 primers) of several HPV strable squamous lesions. of non-specific amplification and types. Randomly labelled PCR prod- The studies vary considerably in increases the sensitivity of the test. ucts are then hybridized to specific terms of investigational design, choice Since it amplifies a shorter fragment, it oligonucleotides on the chip, which is of population and methods, which, as is considered to be more sensitive than afterwards scanned by laser fluores- expected, leads to enormous variabil- PGMY09/11 PCR and also suitable for cence. In the case of multiple infec- ity in the results observed. Most stud- less well preserved specimens; it has tions, multiple hybridization signals can ies assessed HPV test performance on been reported that these primers be seen. Because signal detection in the basis of prevalent lesions using detect about 13% more HPV in cervi- microarrays is subject to variation, simple cross-sectional designs or ret- cal smears than the PGMY primers additional levels of control would be rospective case series, whereas some (Iftner & Villa, 2003). However, desirable. These should include quality assessed both prevalent and seem- because these primers were designed control of the efficiency of the PCR ingly short-term incident lesions based for high-risk types only (HPV 16, 18, reaction and the hybridization condi- on cross-sectional investigations with 31, 33, 35, 39, 45, 51, 52, 56, 58, 59 tions, include a measurement of the extended follow-up (ASCUS-LSIL and 68), this test is not truly generic, homogeneity of the probes on the Triage Study (ALTS) Group, 2003a, b). but rather comparable to the HC2 test. chips and allow some sort of quantifi- Lesion definition varied across studies In addition, the use of PCR assays cation. In addition, the read-out and included either CIN 1 or CIN 2/3 or aiming at maximum sensitivity for requires expensive equipment for sig- worse lesions, diagnosed by histology detection of HPV in a screening setting nal detection and would need to be on specimens obtained by colposcopy- irrespective of concomitant disease performed with the help of special soft- guided biopsy. Sometimes the SIL ter- may be inappropriate with regard to ware that allows threshold settings. minology was used for these histologi- clinical usefulness. cal diagnoses. In some studies, the The reproducibility and agreement Sensitivity and specificity of colposcopic result was used if no of HPV testing results among the three HPV assays biopsy was taken. Some studies used most popular PCR protocols, as well Dozens of studies have provided data direct community recruitment, but usu- as between them and the FDA- on the diagnostic performance of HPV ally the study population was clinic- approved HC2 assay, for overall HPV DNA testing methods. However, only based. None of the studies was based detection have been extensively stud- some of them provided direct compar- on long-term follow-up for more relevant ied. While agreement at the overall isons with cytological testing in detect- end-points, such as incidence of CIN 2 positivity level may be considered ade- ing high-grade precancerous lesions or 3 or cancer or mortality from invasive quate in clinical settings, concordance and clearly specified the type of popu- cervical cancer (see Chapter 4). at the level of type detection leaves lation, i.e., whether it was a primary For many of the studies, the pur- much to be desired (Qu et al., 1997; screening or secondary triage study. pose was to compare HPV testing with Kleter et al., 1998, 1999; Peyton et al., The vast majority of studies either did other screening technologies for cervi- 1998; Swan et al., 1999; Gravitt et al., not provide data on cytology or pre- cal cancer (primarily cytology). None of 2000; Castle et al., 2002c; Van Doorn sented data on mixed series of sub- the investigations was a randomized et al., 2002; Castle et al., 2003). jects and could not be unequivocally controlled trial; all were based on Biomedlab Co. (Republic of Korea) designated as screening or triage set- concomitant testing for HPV and cytol- has developed an HPV oligonucleotide tings. Table 36 summarizes the main ogy alone or with additional tests. Such microarray-based system for detection features of selected published studies investigations are known as split-sam- of HPV types that currently allows that provided data on the comparison ple studies because the cervical detection of 22 HPV types, by immobi- of HPV testing with cytology in primary specimen, collected in single or multi- lizing HPV type-specific oligonu- screening for cervical cancer and its ple exfoliative procedures using a cleotide probes and a control (beta- precursors; it also gives estimates of swab, a cytobrush or other collection globin probe) on an aldehyde-deriva- sensitivity and specificity for HPV test- device, is split into several sub-

96 059-116 24/01/05 11:06 Page 97

Screening tests b =424) N b b =2281) = 1703) =5651) N N =2861) N Bias-controlled HPV indices based on Clinician-collected HPV indices excluded, LSIL in histology HC2 ( Bias-adjusted and HC2, all ages HC2, ages 18–30 specificity includes cytology Conventional cervical samples ThinPrep cytology LSIL in histology HC2, ages 31–40 HC2, age > 40 HPV tested for bias-adjusted All ages, based on HC2 ( HPV indices excluded, based on HC1 ( CIN 1 Comments paired set All ages, ThinPrep cyto- with with conventional logy ( paired set All ages, cytology ( N in primary screening for a 95 93 91 99 96 96 98 97 99 94 94 Pap 87 86 61 95 91 83 96 82 94 80 89 HPV 85 88 90 94 Specificity (%) 68 88 44 79 27 61 46 78 20 78 Pap 87 100 100 80 95 68 84 75 88 89 HPV 93 88 95 73 81 93 Sensitivity (%) . (2000), all et al (2000) free of cytological Women Multiple screening practices, Unscreened population, Subset of sample in Kuhn of HIV sero-status) pective free of cytological Women cross-sectional plus 8 months free of cytological Women abnormalities at enrolment Multiple screening practices, Completed recruitment (irres- recruitmentcommunity abnormalities at enrolment HPV Population-based, et al. Unscreened population, sample of 10% random testing follow-up for referred women Pap-/HPV- colposcopy referral colposcopy recruitment,community all colposcopy underwent women Womack study in colposcopy underwent women cross-sectional plus 15 months testing, HPV positivity follow-up alone not a criterion for referral immediate colposcopy positivity not a criterion for abnormalities at enrolment, features Study Type-specific HC2 HC1, HC2 HC1, HC2 HC1, HC2, PCR GP5+/6+ HC2 HPV test PCR (16, 18, 31, 33) HC2 HC1, HC2 PCR MYO9/11 HC2 20–45 34+ 35–65 18–69 35–65 18–70 25–55 (years) 18+ 35–45 Age 15–76 7932 2009 2988 1356 2098 2944 4761 2073 size 8554 Study

. (2000), (2000), (2001), 1997 (1995), (1999), . (2000), et al . (2000), (2000), . (2001), testing data on the comparison of HPV testing with cytological Selected studies that provided et al. et al. et al. et al et al. et al. et al et al et al.

South Africa South Africa Study, country Canada Germany UK Costa Rica China France Schneider Ratnam Kuhn Cuzick UK Cuzick Belinson Schiffman Blumenthal Wright (2001), Zimbabwe Clavel lesions:cervical cancer and its precursor and estimates of screening performance characteristics 36. Table

97 059-116 24/01/05 11:06 Page 98

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening =1550) N b b b b 30, detection of ≥ =4121) CIN 2, HPV tests: CIN 2, HPV tests: N ≥ HC2, CIN 2/3+, bias- Kolkata CIN 2/3+, clinician- Mumbai > CIN 3, HPV tests: detection of All ages, adjusted HC2, CIN 3+, bias- Trivandrum HC2/PCR, bias adjusted Age < 30, detection of Age Age > 30, paired set Age > 30, paired set collected samples with conventional cytology ( with ThinPrep cytology ( HC2/PCR, bias- adjusted adjusted ≥ HC2/PCR, bias- adjusted** Comments 99 87 98 99 90 99 98 89 96 96 95 Pap 95 92 92 94 73/79 95 95 71/78 83/87 HPV 90 88 Specificity (%) 37 37 59 70 57 40 72 Pap 46 36 58 84 98 46 93 69 91/88 97 HPV 81 100 100 74/70 63/57 Sensitivity (%) ., 2002). Unless otherwise indicated, results are based on conventional et al attending opportunisticWomen 41% ThinPrep cytology, ment, Primary screening network of attending primaryWomen sites screening in 3 different planning clinic recruit- Family clinics and gynaecological prac- compared with clinician-collected cervical samples. Those positive cytology were HPV or by for and biopsy colposcopy for referred screening. Self-collected vaginal sample of Pap-/HPV- random colposcopy for referred women tices: patients attending routine of all Germany. representative Bias controlled cervical to be screening stratified in India. All subjects investi- and, when colposcopy gated by a biopsy. underwent necessary, bias in the verification Averted design. features Study HC2 HC2 MYO9/11 HC2 and PCR HPV test HC2, PCR 15–85 25–65 30+ (years) 18–50 Age 18 085 8466 Study size 4075 . (2003), 7868 et al. (2003), et al et al. (2004), India LSIL threshold (majority of studies) or LSIL persistent ASCUS (Kulasingam bias:Verification of disease status among all participants bias-controlled denotes verification of estimates based on and bias-adjusted denotes correction Mexico et al. Germany Sankaranarayanan cytology. as disease outcome. CIN 2/3 lesions or worse are for Unless otherwise stated, estimates shown women. sample of test-negative of disease status in a random the verification Salmeron Study, country a b Kulasingam (2002), UK 36 (contd) Table Petry

98 059-116 24/01/05 11:06 Page 99

Screening tests

samples for testing. Studies varied in specificity (their estimates should be protocol in which blind biopsies are terms of timing of collection, collection considered relative). These studies collected, it is still possible that a frac- method, or whether or not visual meth- relied on the fact that with two or more tion of the existing lesions will remain ods for cervical inspection were used tests, there were always combinations undetected, because of either their as adjunct screening techniques. of either cytology-negative or HPV- location or size. Therefore, in any One advantage of HPV DNA test- negative women with verified disease cross-sectional survey of screening ing is that it is suitable for self-sam- status available for analysis. However, efficacy, the ethically acceptable gold pling, as in many of the studies shown the biasing effects of the unequal veri- standard for cervical lesions (col- in Table 36. Self-sampling is likely to fication of disease status can be strong poscopy-guided biopsies) is an imper- improve compliance, and is particularly and may lead to estimates of screen- fect one because of inadequate sam- appealing in populations with social or ing efficacy that cannot be generalized pling of the entire cervical tissue that is religious limitations on the acceptabil- for cost considerations and other pub- at risk for squamous-cell malignancy. ity of vaginal examinations. lic health uses (Franco, 2000). Such Only a more aggressive diagnostic A few large randomized controlled verification bias was averted (by apply- approach such as a detailed histologi- trials of HPV testing in primary cervical ing the gold standard of disease verifi- cal examination of serial sections from cancer screening are currently in cation to all women) or corrected (by cone biopsies or from specimens col- progress. Of note are the UK "HPV in extrapolating the screening results lected LEEP or by large loop excision Addition to Routine Testing" (HART) from a random fraction of women with of the transformation zone (LLETZ) investigation (Cuzick et al., 2003), the negative screening tests to those with- would approach the definition of an Dutch POBASCAM trial (Bulkmans et out colposcopic verification) in a few acceptable gold standard of disease, al., 2004), the UK "A Randomized Trial studies, as indicated in Table 36. but such an approach even in a sam- in Screening to Improve Cytology" An important assumption in dealing ple of test-negative women would be (ARTISTIC) (H. Kitchener, personal with the issue of verification bias is the unethical as well as impractical. communication), the Osmanabad trial expectation that the gold standard of Even if tissue sampling could be in India (R. Sankaranarayanan, per- colposcopy-guided biopsy provides done optimally with respect to lesion sonal communication), the Italian trial perfect ascertainment of disease. site and time of development, one (G. Ronco, personal communication), Studies that either avoided or cor- needs to consider also the misclassifi- the Canadian Cervical Cancer rected for the putative bias assumed cation of lesion outcome status that Screening Trial (CCCaST) (E. Franco, that a colposcopy-guided biopsy accu- exists even with histopathological personal communication) and a trial in rately reveals the existence of cervical ascertainment. Studies that involve Finland (Nieminen et al., 2003). lesional tissue, which was then used to multiple expert pathologists indicate The majority of the estimates in ascertain the distribution of diseased that the reproducibility in grading Table 36 must be interpreted with cau- and non-diseased women, allowing histopathology specimens is not high, tion because of selection biases and the computation of adjusted estimates even with large specimens, such as other issues that affect computation of of screening validity. While the LEEP-obtained tissue samples. screening performance indices. For approach is correct for its intended Therefore, a study that is simply based instance, the sensitivity and specificity purpose, i.e., to obtain an improved on lesion ascertainment by a single estimates of most studies shown in estimate of the distribution of disease expert pathologist will be more prone Table 36 are relative, not absolute, conditional on test results, it should be to lesion misclassification than one because they are not based on interval recognized that a simple colposcopy or employing a panel of readers that cancer incidence and are subject to even a colposcopy-guided biopsy can- reaches a consensus diagnosis in verification bias (see Chapter 4). The not guarantee that a lesion will be every case. latter occurs whenever the probability detected. In many test-negative Furthermore, as the design of of disease verification via the gold women, the colposcopist cannot visu- screening efficacy studies evolves standard is dependent on the screen- alize lesional tissue and may decide from the traditional single-opportunity ing test result. In general, such studies that the colposcopic impression of no sampling, cross-sectional layout to used a design in which only women disease alone serves as definitive long-term, repeated sampling investi- with one or more positive screening diagnosis. However, a lesion could be gations over many years, disease case tests were referred for colposcopy and hidden in the endocervical canal and definition becomes a more dynamic biopsy, which prevented the unbiased not visible. Although this pitfall could be process, requiring the juxtapositioning estimation of absolute sensitivity and minimized by adopting a colposcopy of screening and diagnostic test results

99 059-116 24/01/05 11:06 Page 100

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

obtained from multiple samples sive cancer that is diagnosed between cytological test. Only minor discomfort collected over time. This process the screening tests. and very minimal risk are associated involves combining the results from dif- Another issue that affects screening with obtaining exfoliated cervical cell ferent diagnostic approaches, which performance of HPV tests is the type of samples. On the other hand, little is may be differentially triggered by the specimen. In theory, clinician-collected known about the psychological and severity of the lesion grade presumed cervical specimens are ideal in terms of emotional impact of communicating by the test (HPV or cytology), e.g., col- sampling exfoliated cells from the target positive HPV test results to women. As poscopy with simple biopsy for equivo- tissue. Therefore, clinical correlations knowledge about HPV has become cal or low-grade lesions, LEEP for between lesion severity and presence more widespread, there has been a high-grade lesions, etc. Natural history of HPV should be optimal with such gradual shift in how the medical and investigations of HPV and cervical specimens. However, in public-health public-health communities consider neoplasia are examples of studies that practice, convenience for the patient cervical cancer prevention; the per- have to grapple with this added com- and cost-saving considerations have spective has moved from an oncologi- plexity by having to differentiate led some to propose self-sampling as a cal one to a model in which a sexually between prevalent and incident viable alternative to collection by a clini- transmitted infection is the target lesions, progression and regression, cian; the assumption is that the loss in (Franco, 2003). Implementation of test- and relating them to screening test screening accuracy would not be sub- ing for HPV in primary screening for performance. Calculation of sensitivity stantial to the point of offsetting the ben- cervical cancer would lead to a large and specificity in such studies involves efits of simplifying specimen collection proportion of women having to be told the combination of diagnostic informa- (Sellors et al., 2000). Self-sampling of that they harbour a sexually transmit- tion over multiple samples, which genital specimens remains an attractive ted viral infection that can ultimately greatly reduces the chance that any option in developing countries and in cause cancer. There is a dearth of lesions are missed through the pitfalls remote regions where health-care research on the merits and conse- described above for a cross-sectional providers cannot be available at point- quences of conveying this information. study relying on colposcopy-guided of-care settings. However, issues of The vast majority of such women will biopsies alone. On the other hand, the validity, acceptability and training pre- not be required to change their lifestyle repeated sampling layout of these sent obstacles to wider application of or to be referred for a more aggressive investigations obviates the need for self-sampling. diagnostic procedure on the basis of invasive diagnostic procedures among this information, since their infection women testing consistently negative for Costs and potential hazards will be found to be transient. Therefore, both HPV and cytology over many vis- The most important obstacles to more it is debatable whether conveying this its. The longitudinal nature of the inves- widespread acceptance of HPV testing information would bring any real bene- tigation ends up providing the test and in cervical cancer screening are its fit to a screening participant. Practically diagnostic data that approaches the high unit cost and the fact that the nothing is known on the potential neg- true distribution of disease dynamics, technology is not in the public domain, ative impact, including social and legal conditional on study duration. as it is for cervical cytology. The cost- implications, of imparting this informa- Therefore, correction for verification effectiveness of HPV testing is heavily tion. Also the dynamics of between-gen- bias is not a critical issue in these lon- dependent on assumptions related to der transmission of HPV infection are gitudinal studies with intensive follow- the intrinsic cost of the test, the infra- poorly understood. Such information is up of test-negative women and structure available in the setting where important in screening contexts, e.g. for repeated histological sampling of test- the screening will be implemented, the health providers to convey meaningful positive cases. However, such studies length of interval between screening information on risk to couples. do have to contend with the issue of visits, and the existing expenditures Another concern with the use of distinguishing between prevalent and incurred by quality assurance imposed HPV testing in cancer screening is the incident lesions to properly assign the by local legislation. potential for a breakdown in quality- distribution of disease for the purposes There are no additional physical control safeguards if too many of gauging screening test efficacy. hazards associated with the applica- commercial test suppliers enter the As described in Chapter 4, the tion of HPV testing technology for the market without a certain level of ideal estimation of sensitivity assumes purpose of cervical cancer screening, regulatory control of performance stan- follow-up and clinical surveillance, via as the specimen used in the test is the dards by health-care or government a cancer registry or otherwise, for inva- same as that collected for a traditional agencies. At present there are only a

100 059-116 24/01/05 11:06 Page 101

Screening tests

few commercial suppliers of HPV diag- so as to allow more slides to be fields on a given slide; a motorized nostic systems. The two major ones screened without increasing the num- computer-controlled microscope stage can afford to keep up strict quality-con- ber of staff. then takes the cytotechnician directly trol standards in reagent batch produc- The PAPNET system, that is no to these specific microscope fields. tion and performance characteristics longer commercially available, included TriPath Imaging has published results by passing on the costs to the con- neural network software and traditional using a similar type of device, referred sumers or their health insurers, private imaging technology. It selected 128 of to as the Focal Point location-guided or public. However, increased competi- the most suspect fields in conventional screening device, that is not yet FDA- tion resulting in diminishing market cervical cytological specimens and approved (Wilbur et al., 2002). This shares and reductions in the cost of presented these images on a video device is based on the earlier AutoPap testing might lead test manufacturers review screen. The cytotechnician then device. to relax their standards of quality. Such interpreted the images on the screen Several studies have reported the a scenario could prove disastrous in and decided whether to carry out man- test accuracy of automation-assisted many respects, since there are theo- ual screening. Another system that screening (Kok & Boon, 1996; Wilbur retically many more variables that can was introduced in the 1990s was et al., 1996; Koss et al., 1997; affect the performance of HPV testing AutoPap. This computerized scanning Michelow et al., 1997; Bartels et al., than there are for cytology-based device was originally designed for 1998; Doorneward et al., 1999; Halford screening. It is imperative, therefore, algorithmic classification of conven- et al., 1999; PRISMATIC Project that early performance and proficiency tional cervical cytology specimens, but Management Team, 1999; Bergeron et standards be agreed upon by public was later approved for liquid-based al., 2000a; Duggan, 2000; Kok et al., health agencies involved with quality cytology specimens. The device was 2000). They show generally a better assurance of cervical cancer screening. initially approved in the USA by the test sensitivity with at least the same FDA as a method to be used for qual- specificity as conventional screening. ity control. In the quality control mode, Most studies were retrospective (qual- Other emerging techniques only those specimens classified as ity control) and/or involved rather small normal (‘within normal limits’) were numbers of smears. One larger This section describes three new reviewed through the device. prospective study, conducted by the developments in screening methods: Subsequently AutoPap was approved PRISMATIC Project Management (1) computer-assisted cytological inter- by the FDA for primary screening Team (1999), including 21 700 smears, pretation of cervical smears, (2) use of (Dunton, 2000). In the primary screen- also showed equal sensitivity but bet- physical real-time devices and (3) ing mode, all slides are processed ter specificity for automated screening, detection of molecular surrogate mark- through the device and then, on the as well as higher productivity. Results ers of cancer progression. basis of an ‘abnormality index’ of only two randomized prospective assigned to the slide by the algorithmic public health trials in a primary screen- Computer-assisted reading of processing feature, each slide is either ing setting have been reported. One of cervical smears filed without manual review by a these studies found clearly higher The aim of automation-assisted cytotechnician (up to 25% of all speci- detection rates of in situ and invasive screening is to increase the sensitivity mens) or reviewed manually in the nor- carcinoma (Kok & Boon, 1996). of cytological testing by finding, for mal manner. However, the second study, integrated instance, small abnormal squamous Both of these devices have served in the Finnish organized screening pro- and glandular cells, known to be very as prototypes for newer devices that gramme and involving several cytolog- difficult to detect in conventional are being developed to help automate ical laboratories, did not clearly confirm screening; it should also increase the evaluation of cervical cytology this result (Nieminen et al., 2003) specificity by selecting only lesions specimens. Recently, the ThinPrep (Table 37), showing sensitivity and corresponding to objective repro- Imaging System (Cytyc, Boxborough, specificity nearly equal to those of tra- ducible criteria. Automated screening MA, USA) has received FDA approval ditional cytological screening (Table should also increase productivity by for use in primary screening of liquid- 38). excluding normal slides or part of the based cytology specimens in the USA. The few randomized prospective slides from manual screening by This system uses image analysis and studies and other performance studies selecting most atypical images from a algorithmic processing to identify a have shown that automation-assisted slide to be checked by the cytologist, fixed number of the worst microscopic screening may be feasible as a part of

101 059-116 24/01/05 11:06 Page 102

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 37. Comparison of histologically verified cervical lesions between the PAPNET® arm and the convention- al screening arm: number (N) and proportion (per 1000) of screenees, odds ratios (OR), with 95% confidence intervals (CI) (logistic regression)

Histological PAPNET arm Conventional arm OR Significance diagnosis Total 65 527 Total 25 767 N per 1000 N per 1000

Invasive cancer 44 0.67 8 0.31 2.16 p < 0.05 In situ carcinoma 79 1.20 18 0.68 1.76 p < 0.05 CIN 3 124 1.89 44 1.70 1.11 NS

NS, not significant From Kok & Boon (1996)

Histological PAPNET arm Conventional arm OR 95% CI diagnosis Total 36 225 Total 72 461 N per 1000 N per 1000

Invasive cancer 3 0.08 4 0.06 1.50 0.30–6.80 CIN 3 51 1.4 100 1.4 1.02 0.72–1.42 CIN 2 51 1.4 104 1.4 0.98 0.70–1.36 CIN 1 40 1.1 96 1.3 0.83 0.57–1.20 Normal and other 36 080 996 72 157 996 1.00 Reference

From Nieminen et al. (2003)

Table 38. Specificity of the PAPNET and conventional Pap-smear test with cut-off levels of ASCUS+ and LSIL+ for invasive cancer and for an outcome of CIN 2+ or invasive cancer in primary screening setting

Negative histology Negative Pap smear Specificity % Cytological threshold: ASCUS+ Outcome: invasive cancer PAPNET 36 222 33 447 92.3 Conventional 72 453 67 241 92.8 Outcome: CIN2+ PAPNET 36 171 33 447 92.5 Conventional 72 353 67 240 92.9

Cytological threshold: LSIL+ Outcome: invasive cancer PAPNET 36 222 35 972 99.3 Conventional 72 453 71 890 99.2

Outcome: CIN2+ PAPNET 36 171 35 970 99.4 Conventional 72 353 71 887 99.4

From Nieminen et al. (2003)

102 059-116 24/01/05 11:06 Page 103

Screening tests

routine primary screening and with the nals and computer software integrates Molecular surrogate markers devices tested it seems to perform at the information and provides a diagno- Because both HPV DNA testing and least as well as conventional screening sis of CIN or the absence of CIN cytological screening yield consider- in an organized well functioning pro- (Mould & Singer, 1997; Singer et al., able numbers of women to be referred gramme. Automation-assisted screen- 2003). In a study of 651 women in ten for colposcopy and biopsy, who are ing may improve the results of a sub- international centres, the relative subsequently found not to harbour optimal screening organization, but sensitivity for histologically confirmed high-grade disease, it seems worth- may have no advantage over a well CIN 2 or worse lesions as diagnosed while to look for markers which, at one organized, high-quality screening pro- by TruScreen was 70%; the corre- test occasion, might identify women gramme other than possibly handling sponding sensitivity for cytology was susceptible to progression with a high more samples with same quality. 69% and for a combination of predictive value. A new generation of automated TruScreen and cytology was 93% Biomolecular pathways leading devices for use with liquid-based cytol- (Singer et al., 2003). from HPV infection to the development ogy is now being launched, the perfor- Fluorescence spectroscopy is of cervical dysplasia and cancer are mance of which has not yet been eval- based on the measurement of auto- becoming well understood (zur uated in randomized trials fluorescence from tissue molecules Hausen, 2000, 2002). Continued such as FAD, NADPH and collagen expression of the viral early onco- Physical real-time devices that emit light after excitation with low- genes E6 and E7 appears to be an Advantages of physical real-time power laser light of certain wave- essential factor in the neoplastic trans- devices would be to allow non-invasive lengths (Burke et al., 1999; Ferris et formation and maintenance of immor- on-spot diagnosis, an ability to acquire al., 2001a; Follen Mitchell et al., 1999). talized cell lines, apparently by inacti- an objective machine-generated result, Multi-modal spectroscopy integrates vation of the tumour-suppressor pro- applicability in primary health-care set- several types of spectroscopy such as teins p53 and pRb, respectively (zur tings, the non-requirement for highly intrinsic fluorescence, diffuse Hausen, 1994, 2000) (see Chapter 1). trained colposcopists, and high accept- reflectance and light scattering. These Certain key DNA, RNA or protein ability by women (Soler & Blumenthal, integrated techniques allow the exami- markers arising during the neoplastic 2000; Basen-Engquist et al., 2003; nation of both biochemical characteris- transformation process might be Wright, 2003). tics and morphological features measured to predict the progressive Normal and neoplastic cervical (nuclear size, blood perfusion, cellular character of disease in screening, epithelia have different physical and changes), so as to optimize the distinc- diagnosis and prognosis. However, it is biochemical properties, yielding dis- tion between normal and abnormal tis- not clear that one common molecular tinct patterns in conductance of electri- sue (Nordstrom et al., 2001; carcinogenetic pathway is involved and cal pulses and reflectance of light Georgakoudi et al., 2002). Spectro- it is therefore possible that no single waves (Mahadevan et al., 1993; scopic instruments are continually molecular marker will ever on its own Ramanujam et al., 1994; Richards- being improved (Drzek et al., 2003). allow distinction between progressive Kortum et al., 1994; Wright et al., The performance of fluorescence and non-progressive disease. 2002c). These differences have been spectroscopic devices has been found Potential markers of progression applied in fluorescent spectroscopic promising in several small trials, include messenger RNA for the E6 or devices which capture electro-physical usually conducted by the manufacturer E7 proteins, HPV DNA sequences signals from the stimulated cervix and and most often on selected groups of integrated into the human genome, analyse the patterns using algorithms women. In a series of 111 women, over-expression of cell cycle regulator to discriminate between normal and accuracy to detect CIN 2 or worse proteins or proliferation protein mark- neoplastic tissue (Burke et al., 1999; lesions was higher for multimodal ers, and determination of certain Follen Mitchell et al., 1999). The spectroscopy than for cytology (area genetic or immunological profiles TruScreen (formerly Polarprobe, under the ROC curve (AUC): 95% for (Sotlar et al., 1998; Arias-Pulido et al., Polartechnics Limited, Sydney, Austra- spectroscopy and 78% for cytology) 2002; Bibbo et al., 2002; Kadish et al., lia) is a portable device that measures (Ferris et al., 2001a). Larger multi- 2002; von Knebel Doeberitz, 2002; the response of the cervical surface to centre trials are needed to confirm Altiok, 2003; Sherman, 2003; low-voltage electric stimuli and light these preliminary results. Solomon, 2003; Wang & Hildesheim, waves of four different wavelengths. 2003) (see Table 39). The sensor captures the emitted sig-

103 059-116 27/01/05 15:45 Page 104

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 39. Potential markers for HPV-induced cervical intraepithelial neoplasia or cancer

Marker Change in expression Family Rationale

Protein markers p16INK4A Increased Cyclin-dependent kinase (CDK) E7-mediated degradation of the Rb gene inhibitor gene yields enhanced transcription of the gene coding for p16.

p53 Decreased Anti-tumour regulating protein, Viral E6 protein from oncogenic HPV involved in apoptosis types binds p53, facilitating its degradation through the ubiquitation pathway.

Ki-67 Increased Cell proliferation marker Abnormal cell proliferation beyond basal cell layers.

PCNA Increased Cell proliferation marker Abnormal cell proliferation beyond (proliferating cell nuclear basal cell layers. antigen)

Cyclin E Increased Protein associating with CDK2 Cyclin E associated with CDK2 drives cells from G1 to S phase through phosphorylation of pRb and other targets. Mcm5, Cdc6 c-myc Increased Proliferation markers Abnormal cell proliferation beyond basal cell layers.

Telomerase Increased Nucleoprotein consisting of hTR Controls length of telomeres and plays (RNA) and hTERT (enzyme) a role in cell immortalization RNA markers E7 or E6 mRNA Presence Viral mRNA, trancripts of E6 or Presence of mRNA for E6 or E7 indi- E7 gene cates active expression of of oncogenes. Presence of E6 or E7 mRNA in the absence of L1 HPV DNA might indicate integration of viral DNA. DNA markers Decrease in the ratio Viral DNA sequences Viral integration often occurs at the E2 E2/E6 and E2/E7 viral integrated in host genome gene of the HPV genome. Disruption of DNA the E2 gene yields a more intensive transcription of the oncogenes E6 and E7. In the episomal state E2 and E6 DNA are present in equal amounts, while in the integrated form, less or no intact E2 is present.

Markers of genetic host p53 with arginine at position 72 of p53 – polymorphism of p53 gene should have a higher affinity for the E6 oncogene. Arg/Arg homozygotes should therefore have a higher risk of cervical neoplasia than Arg/Pro heterozygotes or Pro/Pro homozygotesa

a Increased risk for CIN or cervical cancer among Arg/Arg homozygotes is an inconsistent finding in the literature (see Chapter 1).

104 059-116 28/01/05 12:05 Page 105

Screening tests

p16 The potential use of p16 immuno- normally confined to the basal or Cyclin-dependent kinases (CDKs), staining in cytological smears, and the suprabasal epithelial cell layers. cyclins and CDK inhibitors are key correlation with cytological and histo- Expression of Ki-67 allows distinction molecules that control the cell cycle logical results, has been examined of atrophic cells (negative for Ki-67) and coordinate DNA synthesis, chro- recently (Table 41). p16 immunostain- from neoplastic cells (positive for Ki- mosome separation and cell division ing has been found to facilitate the 67) in menopausal women (Ejersbo et (Morgan, 1997). Viral oncoproteins retrieval of dysplastic cervical cells on al., 1999; Mittal et al., 1999; Bulten et interfere directly or indirectly with sev- a slide (Bibbo et al., 2002, 2003; al., 2000). Expression beyond the inner eral of these CDKs (Cho et al., 2002). Sahebali et al., 2004). third of the cervical epithelium is p16INK4A inhibits the CDK4/6 interac- Strong nuclear and cytoplasmic observed in case of CIN and cancer tion with cyclin D1, preventing progres- p16 staining in conventional or liquid- (Bulten et al., 1996; Keating et al., sion through the G1/S checkpoint of based cervical smears gave a sensitiv- 2001). Several authors have found a the cell cycle (Keating et al., 2001). ity for detection of HSIL or worse of significant correlation between the Accumulation of p16INK4a mRNA and about 90–100%; the specificity varied presence or intensity of Ki-67 and the protein has been reported in response between 36% and 100% (Klaes et al., severity of cytological abnormality in to inactivation of the retinoblastoma 2001; Saqi et al., 2002; Murphy et al., cytological preparations (Dunton et al., gene product (pRb) through binding 2003; Nassar et al., 2003). Sporadic 1997; Sahebali et al., 2003). Dunton et with viral E7 (Xiong et al., 1993; immunoreactivity in normal squamous al. (1997) found a sensitivity of 89% for Serrano, 1997). However, this over- metaplastic, inflammatory cells and Ki-67 immunostaining in a set of expressed p16 is inert since pRb-func- more systematic staining of endome- selected abnormal smears for detec- tion is neutralized by E7 (Medema et trial and tubal metaplastic cells and of tion of histologically confirmed CIN 2+ al., 1995; Khleif et al., 1996). Over- bacteria has been reported (Bibbo et lesions, whereas the specificity was expression of p16 protein is consid- al., 2002, 2003; Saqi et al., 2002; 65%. ered to be a marker for progression Riethdorf et al., 2002). from HPV infection to cervical cancer These accuracy measures have Other proliferation or cell cycle regu- (von Knebel-Doeberitz, 2002). been computed from very small and lating markers Immuno-detection of p16 using highly selective series and cannot be Several other proteins are over- monoclonal antibodies in histological considered representative for real expressed in proliferating cells and cer- material was described by Klaes et al. screening or clinical situations, but the tain cell progression regulators have (2001). Other studies of p16 results are promising. been proposed as potential markers for immunoreactivity in histological mater- Possible advantages of immuno- cervical neoplasia, such as proliferating ial with different levels of abnormality staining of protein markers include the cell nuclear antigen (PCNA) (Demeter and HPV infection status (Table 40) higher reproducibility of microscopic et al., 1994; Mittal et al., 1993), Mcm5 have used various types of primary interpretation, quicker detection of and Cdc6 (Williams et al., 1998) and and secondary antibody and chro- stained lesions and appropriateness cyclin E (Altiok, 2003). mogen. The sensitivity of p16-immuno- for automated detection. A disadvan- Proliferation markers are physio- histochemical detection of CIN 2 or tage is the presence of background logically present in basal or para-basal worse lesions in histological prepara- staining and positive staining of epithelial cells, and are an objective tions varied between 70% and 100%, endometrial or tubal cells, requiring the indicator of neoplasia when observed while the specificity ranged from 34% determination of criteria to define posi- beyond the lower cell layers. In cervical to 100% (Keating et al., 2001; Bibbo et tivity, balancing the sensitivity against smears lacking architectural informa- al., 2003; Murphy et al., 2003; Negri et specificity. tion, the presence of proliferation al., 2003; Zielinski et al., 2003; Wang et Biochemical detection of p16 protein markers is less informative and can al., 2004b). in lysates of cervical swab samples using easily yield false positive results. Klaes et al. (2002) showed improved a sandwich ELISA assay is a potentially inter-observer concordance in histologi- simple approach for resource-poor set- Telomerase cal interpretation of p16-immunostained tings (Herkert et al., 2004). Telomeres are repeated arrays of six material (group kappa = 0.94; 95% CI nucleotides (TTAGGG) at the chromo- 0.84–0.99) compared with haema- Ki-67 some ends that protect chromosomes toxylin–eosin stained material (group Expression of the Ki-67 protein occurs against degradation and aberrant kappa = 0.71; 95% CI 0.65–0.78). in proliferating cells and its presence is fusion or recombination (Collins &

105 059-116 24/01/05 11:06 Page 106

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 40. Overview of p16 immunoreactivity in histological material by severity of lesion and by HPV status Reference Detection system: primary and p16 Lesion, HPV status N % p16+ antibody and chromogen positivity

Sano et al. -Mouse monoclolnal antibody (JC8) Diffuse HPV 16 60 100.0% (1998) -Biotinylated horse anti-mouse antibody Other hrHPV+ 28 96.4% -Chromogen: DAB HPV 6 or 11 34 0.0%

Keating et al. -Clone G175-405 (Pharmigen, Diffuse Normal 24 0.0% (2001) -Biotinylated goat anti-mouse antibody LSIL 24 37.5% -Chromogen: DAB HSIL 37 70.3% hrHPV+ and CIN 40 70.0%

Klaes et al. -Clone E6H4 (MTM Lab., Heidelberg) Diffuse Normal, hrHPV– 32 0.0% (2001) -Biotinylated horse anti-mouse antibody Normal, hrHPV+ 10 0.0% -Chromogen: aminoethylcarbazole Inflammation, hrHPV– 30 0.0% with hydrogen peroxide in acetate buffer Inflammation, hrHPV+ 18 0.0% Reserve cell hyperplasia, 13 0.0% hrHPV- Reserve cell hyperplasia, 8 12.5% hrHPV+ CIN1, hrHPV– 32 46.9% CIN1, hrHPV+ 15 86.7% CIN2, hrHPV– 14 100.0% CIN2, hrHPV+ 18 100.0% CIN3, hrHPV– 9 100.0% CIN3, hrHPV+ 51 100.0% Invasive cancer, hrHPV– 5 100.0% Invasive cancer, hrHPV+ 55 96.4% Bibbo et al. -Clone E6H4 (MTM Lab., Heidelberg) Focal Normal 3 0.0% (2002) -Mouse non-avidin-biotin Envision+ or diffuse CIN 1 19 73.7% polymer (Dako) CIN 2 11 90.9% -Chromogen: DAB CIN 3 14 100.0%

Bibbo et al. -Clone E6H4 (MTM Lab, Heidelberg) Focal Chronic cervicitis 5 0.0% (2003) -Mouse non-avidin-biotin Envision+ or diffuse Squamous metaplasia 2 0.0% polymer (Dako) CIN 1 5 40.0% -Chromogen: DAB CIN 2 4 100% CIN 3 11 100% Murphy et al. -Clone G175-405 (Pharmingen, >10% posi- Normal 21 0.0% (2003) San Diego) tive staining -Biotinylated universal antibody, avidin– cGIN 5 100.0% biotin complex (Vector Laboratories, CIN 1 38 92.1% Burlingame) CIN 2 33 72.7% -Chromogen: DAB CIN 3 46 91.3% Squamous-cell carcinoma 8 100.0% Adenocarcinoma 2 100.0%

Negri et al. -Clone E6H4 (MTM Lab., Heidelberg) Diffuse Reactive cells 15 0.0% (2003) -Avidin–biotin kit (Lab. Vision Corp., Endocervical glandular 4 0.0% Fremont) atypia -Chromogen: aminoethylcarbazole Adenocarcinoma in situ 8 100.0% Adenocarcinoma 18 94.4%

Zielinksi et al. -Clone E6H4 (MTM Lab., Heidelberg) Diffuse Adenocarcinoma, hrHPV– 5 20.0% (2003) & strong -Biotinylated rabbit anti-mouse antibody Adenocarcinoma, hrHPV+ 20 95.0% -Chromogen: DAB or aminoethylcarbazole Adenocarcinoma endom., 15 0.0% hrHPV- Abbreviations: DAB, 3,3’-diaminobenzidine; hr, high risk; cGIN, cervical glandular intraepithelial neoplasia

106 059-116 27/01/05 15:41 Page 107

Screening tests

Table 41. Overview of p16 immunoreactivity in cervical smears material by severity of cytological abnormality

Reference Detection system: primary Preparation Cytological lesion/histological N % p16+ antibody and chromogen lesion in corresponding biopsy

Klaes et al. (2001) -Clone E6H4 (MTM Lab., Conventional Pap I/II 36 0.0% Heidelberg smears Pap IIID+ (LSIL+) -Biotinylated horse anti-mouse antibody 7 100.0% -Chromogen: aminoethylcarbazole Bibbo et al. (2001) -Clone E6H4 (MTM Lab., ThinPrep Within normal limits 2 0.0% Heidelberg) LBC LSIL -Mouse non-avidin-biotin HSIL 26 96.2% Envision+ polymer (Dako) Chromogen: DAB Saqi et al. (2002) -p16 antibody (Neomarkers, SurePath Within normal limits 25 4.0% Fremont) LBC AGUS 5 60.0% -Envision + system (Dako) LSIL 30 80.0% HSIL 10 90.0% Squamous-cell carcinoma 1 100% Adenocarcinoma 2 100.0% Bibbo et al. (2003) -Clone E6H4 (MTM Lab., ThinPrep Chronic cervicitis 5 0.0% Heidelberg) LBC Squamous metaplasia 2 0.0% - Mouse non-avidin-biotin CIN 1 5 40.0% Envision+ polymer (Dako) CIN 2 6 83.3% - Chromogen: DAB CIN 3 12 100.0%

Murphy et al. -Clone G175-405 ThinPrep Normal 12 0.0% (2003) Pharmingen, San Diego LBC cGIN 1 100.0% -Biotinylated universal CIN 1–3 20 100.0% antibody, avidin-biotin complex (Vector Labora- tories, Burlingame) -Chromogen: DAB

Nassar et al. -Monoclonal antibody Surepath Not neoplastic 10 50.0% (2003) (Neomarkers) LBC Benign cell changes 9 11.1% Mouse non-avidin-biotin ASCUS 14 14.3% Envision+ polymer (Dako) LSIL 4 50.0% HSIL 1 100.0%

Negri et al. (2003) -Clone E6H4 (MTM Lab, ThinPrep AGUS 10 100.0% Heidelberg) -Avidin-biotin kit (Lab Vision Corp., Fremont) -Chromogen: aminoethyl- carbazole

Nieh et al. (2003) -Clone E6H4 (MTM Lab., ASCUS Reactive 21 0.0% Heidelberg) Pap smears CIN 1 24 8.3% -Mouse non-avidin-biotin CIN 2/3 17 94.1% Envision+ polymer (Dako) Squamous carcinoma 2 100.0% -Chromogen: DAB Adenocarcinoma in situ 2 100.0%

Abbreviations: DAB, 3,3’-diaminobenzidine; LBC, liquid-based cytology; cGIN, cervical glandular intraepithelial neoplasia; AGUS, atypi- cal glandular cells of undetermined significance

107 059-116 24/01/05 11:06 Page 108

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Mitchell, 2002). They become progres- (PreTect HPV-Proofer, NorChip AS, DNA ratio assessed with real-time sively shorter as cells multiply, result- Klokkarstua, Norway) which detects PCR is another potential progression ing in chromosomal instability and E6 mRNA from HPV16 and E7 mRNA marker. However, other authors have senescence when a critical short from HPV types 18, 31, 33 and 45. reported exclusively episomal HPV length is reached (Counter et al., The presence of E6 and E7 mRNA DNA in tumours (Cullen et al., 1991; 1992). The enzyme telomerase is a and absence of viral L1 DNA (negative Pirami et al., 1997). ribo-nucleoprotein composed of an test result on consensus PCR) indicate RNA part (hRT) and a catalytic part integration of viral DNA in the human Micro-array technology (hTERT), which controls telomere genome, yielding enhanced transcrip- It is believed that profiles of multiple length and is believed to play a role in tion of the E6–E7 sequence. Molden et host–virus interaction factors will immortalization of cells (Mathon & al. (2004) found that rates of HPV- reveal possibilities for accurate individ- Lloyd, 2001; Blasco, 2002). Its activity Proofer positivity and presence of HPV ualized risk assessment and prognosis is increased in CIN and cancer. The DNA (measured with GP5+/6+ con- prediction. By the use of DNA microar- intensity of telomerase activity is sensus PCR and type-specific PCR) ray technology or DNA chips, expres- reported to be correlated with the increased with the severity of cytologi- sion of many genes can be analysed at severity of the abnormality in biopsies cal or histological cervical abnormality. once using only a small amount of and in cervical scrapings, but reliable Nevertheless, lower proportions of sample (Hughes & Shoemaker, 2001). detection of hTR, hTERT and telom- mRNA-positive results were observed The first step consists in extraction of erase activity is still limited by analyti- in normal cases, ASCUS and LSIL mRNA from a tissue sample. Using cal deficiencies (Oh et al., 2001; (see Table 42). reverse transcriptase, complementary Jarboe et al., 2002; Fu et al., 2003). DNA (cDNA) is synthesized, which is Viral DNA integration markers labelled with a fluorescent molecule. Detection of viral oncogene tran- Testing for HPV integration appears to This labelled cDNA is subsequently scripts increase the predictive value that an divided over a slide or membrane Viral mRNA can be detected using HPV-positive sample is derived from where hundreds or thousands of (nested) real-time PCR or nucleic acid tissue containing progressive CIN or known target DNA sequences are sequence-based amplification assay cervical cancer (Klaes et al., 1999). fixed. Hybridization of the labelled (NASBA) (Smits et al., 1995; Sotlar et Viral integration often occurs at the E2 cDNA with target DNA is detected as a al., 1998; Deiman et al., 2002). gene of the HPV genome. Disruption coloured light signal at a particular Presence of viral mRNA transcripts of the E2 gene is believed to result in locus on the array, which indicates coding for the E6 and E7 proteins from more intensive transcription of the expression of a particular gene. high-risk HPV might be a more specific oncogenes E6 and E7. In the episomal Post-translational changes also predictor of progressive infection than state, E2 and E6 DNA are present in play an important role in pathogenesis, simple presence of HPV DNA equal amounts, while in the integrated and can be studied using protein arrays (Nakagawa et al., 2000; Cuschieri et form, less intact E2 is present (zur or proteomics techniques (Wulfkuhle et al., 2004). A commercial kit exists Hausen, 2002). A decrease in E2/E6 al., 2003; Lee et al., 2004).

Table 42. Positivity rate for HPV DNA and HPV mRNA in a series of 4136 women presenting at an outpatient gynaecological service, Oslo, Norway

Normal ASCUS LSIL CIN 2 CIN 3 Squamous cancer N 3950 57 20 5 12 1

HPV mRNA+ 95 12 6 2 9 1 2.4% 21.1% 30.0% 40.0% 75.0% 100.0% HPV DNA+ 368 27 15 2 10 1 9.3% 47.4% 75.0% 40.0% 83.3% 100.0% p-value <0.0001 0.08 0.009 1.00 0.62 1.00

From Molden et al. (2004)

108 059-116 24/01/05 11:06 Page 109

Screening tests

General comments to interpret and manage the increased sitivity and specificity that are not Research on molecular markers has so complexity of results. The results so far totally comparable. Most of the designs far been largely restricted to correlation are promising but far from complete. were cross-sectional without correction studies documenting the presence or for verification bias (see Chapter 4). absence or the intensity of the consid- Screening with more than one Therefore, the results and conclusions ered marker in cytological or histologi- technique depend on the number of tests cal material from selected women. The main classes of available screen- applied. If two tests only are consid- These test accuracy measures can be ing techniques are cytology, visual ered, the cross-sectional sensitivity assessed for detection of CIN 2+ but inspection and HPV DNA testing. It is estimate for the test combination is are not representative for real screen- possible to consider combinations of bound to be 100%. The same applies, ing, triage or follow-up settings. two techniques within a class (e.g., albeit not as a logical consequence, if Potential advantages from the use conventional and liquid-based cytol- too few women with dual negative tests of molecular markers in future clinical ogy), but most interest has focused on are subjected to a commonly accepted practice include: triage of women with combining two techniques of different reference standard such as colpo- minor cytological abnormalities classes in the hope of gaining benefit scopic examination with guided biopsy. (ASCUS and LSIL) with higher speci- from complementarity.Thus, researchers Colposcopy itself is not sufficiently ficity than HPV DNA detection; have examined cytology plus visual sensitive to rule out missed disease improvement of the accuracy of histol- techniques, cytology plus HPV DNA and therefore its own errors must be ogy as gold standard for screening test testing and, to a limited degree, HPV recognized when considering results assessment, by more accurate and DNA testing plus visual techniques. that depend upon it. In general, the reproducible classification of histologi- Because of the practical limitations of sensitivity estimate of any combination cal squamous and glandular cervical resources, there has been only occa- of two tests is smaller if more tests are lesions and clearer distinction between sional interest in combining more than used for detection of disease and cervical and endometrial glandular two techniques (Reid et al., 1991). those who tested negative on every lesions; selection of best treatment Whenever two screening tech- test are not subjected to the reference procedures; prognosis prediction; and niques are combined, with abnormal standard (colposcopy). Only one study last but not least, more accurate pri- results from either test taken to consti- (Sherman et al., 2003b) has been mary screening for cervical progres- tute an overall positive result, the sen- based on interval cancer incidence, sive cancer precursors. sitivity will be higher than that of either the ideal to estimate the true sensitivity test alone (Franco & Ferenczy, 1999). (see Chapter 4). In the absence of data However, the key question is whether on the expected incidence, the risk in Combinations of different the increase in sensitivity is sufficiently screen positives versus that in screen modalities greater than random to merit consider- negatives was used as an indicator of ation. Increased sensitivity will typically sensitivity. As the previous sections have demon- lead to an offsetting decrease in speci- strated, no single currently available ficity and the trade-off must be exam- Cytology plus HPV testing screening test for cervical cancer pro- ined to determine the overall effect of The residual cytology specimen from vides an optimal trade-off between the combination on screening accu- liquid-based cytology or a co-collected sensitivity and specificity. Because var- racy. Various statistical methods for specimen can be tested for oncogenic ious screening techniques are avail- evaluating the added value of adding a HPV types. There is much evidence able, applying them in combinations second test have been suggested, but that screening of women with both might be advantageous. Although none has been fully accepted. The cytology and HPV DNA tests combinations of tests necessarily best statistical methods generate increases sensitivity for detection of require extra resources, the added roughly equivalent conclusions prevalent CIN 3 or cancer sufficiently testing accuracy might increase the (Macaskill et al., 2002; Ferreccio et al., to permit longer screening intervals detection of treatable disease and 2003), although the interpretations than with cytology alone. After consid- allow lengthened screening intervals. depend on varying regional standards eration of the accumulated evidence Research is in progress to find the of acceptable safety and cost. regarding increased sensitivity, combinations that are most comple- The studies on combination of dif- decreased specificity and the possibil- mentary, to determine how these tests ferent modalities have been run with ity of lengthened screening intervals should be combined, and to clarify how designs that provide estimates of sen- using the combination, the US FDA

109 059-116 24/01/05 11:06 Page 110

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

approved HC2 for HPV DNA testing as prevalent or incipient CIN 3 and cancer Real-time cervical scanning based an adjunct to cytological testing for (Sherman et al., 2003b). However, optico-electrical devices might improve women aged 30 years and older. The because persistent infection with onco- the sensitivity of cytology (Singer et al., supporting evidence has been sum- genic types of HPV is the necessary 2003), but the influence on specificity marized by Franco (2003), with addi- cause of virtually all cases of cervical of adding such new techniques is not tional recent support (Cuzick et al., cancer, a negative test for oncogenic yet clear. 2003; Ferreccio et al., 2003). However, HPV has unusually high negative pre- in two other recent studies (de dictive value, or reassurance, in the HPV plus visual inspection Cremoux et al., 2003; Coste et al., context of negative or even mildly In many developing countries, 2003), HC2 performed worse than in abnormal cytology. The uncommon approaches that do not rely on an other published studies and conven- combination of a negative HPV test extensive infrastructure of highly tional cytology performed considerably and an HSIL cytological result merits trained personnel must be considered. better than is usually reported. Several further evaluation because of its rarity Because first-rate cytological screen- of the supportive studies are summa- and the possibility that one of the two ing programmes are difficult to create rized in Table 43. The increase in sen- results is in error. and maintain, there is interest in estab- sitivity from adding HPV testing was lishing programmes that rely on more generally greater than the decrease in Cytology plus visual techniques easily performed and standardized specificity and, in some studies, cytol- The interest in combining cytology with techniques. It may be feasible to com- ogy actually added little to the perfor- some kind of visual inspection is nat- bine an HPV test for primary screening mance of HPV testing. ural, since cytology and colposcopy with triage modalities other than cytol- In practice, the introduction of HPV have comprised the strategy responsi- ogy, such as direct visual assessment testing into routine screening in combi- ble for a half century of successful by non-physician health-care pro- nation with cytology produces multiple screening for cervical cancer. viders. The use of HPV testing plus risk strata ranging from very low to Population screening using cytology visual inspection is best considered as very high absolute risk (positive predic- and colposcopy concurrently has been a sequential strategy, to maximize sen- tive value) of prevalent or incipient CIN explored (Olatunbosun et al., 1991), sitivity with acceptable specificity (see 3 or cancer. A woman with HSIL cytol- but is much too expensive and below). ogy (especially with a positive onco- demanding of limited expertise for use genic HPV test) has a high risk of in most regions. The search for an eas- Combining two techniques from underlying CIN 3. In contrast, negative ier alternative to colposcopy has led to the same class cytology (whether conventional or liq- studies of cytology and cervicography The combination of two cytological uid-based) with a negative HPV test is and of cytology and direct visual techniques can be seen as a logical associated with a risk of CIN 3 within inspection. Two representative studies extension of re-screening of slides or two years that approaches zero to examine these combinations are of computer-assisted screening, as (Ferreccio et al., 2003). These summarized in Table 44. Cervicogra- discussed in the first section of this extremes of risk stratification are easily phy is no longer available commer- chapter. The combination of conven- managed. However, strategies to man- cially, but it is worth noting that studies tional and liquid-based cytology is not age the very large numbers of women of cervicography as an adjunct have particularly complementary (Ferreccio who are HPV-positive and cytology- suggested that a visual technique et al., 2003) and holds little promise negative need to be developed and might complement conventional cytol- because of the expense of conducting evaluated. Repeating viral and cytolog- ogy.The two kinds of technique tend to the two tests for each woman ical tests between six and twelve detect different groups of women with screened. Similarly, there is probably months has been proposed (Wright et CIN 2 or 3 without an obviously unac- no reason to combine two visual al., 2004) as an interim measure until ceptable loss in specificity (Ferreccio techniques (e.g., cervicography and more data are available to develop et al., 2003), resulting in overall direct inspection) because of the corre- truly evidence-based guidelines. increased accuracy.The cost-effective- lated nature of the results and limita- The negative predictive value of ness of combining cytology with some tions of the techniques (Shastri et al., adding an HPV test to cytology is its kind of visual inspection (Shastri et al., 2004). HPV testing with multiple tech- major utility. HPV infection is so com- 2004) should be compared with alter- niques could reveal some testing mon that a positive test conveys only a native strategies and evaluated more errors, but not enough to justify the moderate positive predictive value for fully and formally on a regional basis. high cost.

110 059-116 24/01/05 11:06 Page 111

Screening tests 98.6 99.1 99.5 98.4 99.5 96.9 99.7 99.7 100 98.6 99.8 99.6 99.9 99.2 99.3 99.9 99.8 99.9 99.8 98.3 99.9 100 99.1 NPV % 8 9 7 9 9 9 9 7 16 10 15 12 11 36 18 10 14 13 16 39 12 12 34 15 23 17 15 13 PPV % 96 97 91 94 92 98 88 98 94 76 95 93 91 85 86 78 88 88 88 99 88 94 85 85 92 95 86 93 Spec. % 77 34 68 63 40 59 68 46 86 100 92 98 100 72 76 94 85 91 91 70 96 95 86 84 93 97 64 97 Sens. % ASCUS ASCUS ASCUS ASCUS ASCUS ASCUS Borderline ASCUS 10,1 pg/ml Either + Either + 2 pg/ml Either + Either + Either + Either + Mild 1 pg/ml ASCUS +/– 1 pg/ml Mild/2 pg 1 pg/ml 1 pg/ml 1 pg/ml Either + 1 pg/ml Either + Cut-points Cytol. Cytol. Cytol. Cytol. Cytol. Cytol. Cytol. LBC LBC+HC2 HC1, HC2 HC2 Cytol.+HC Cytol.+HC2 HC2 HC2 Cytol.+HC2 Cytol.+HC2 Cytol.+HC2 HC2 HC2 LBC PCR Cytol.+PCR LBC+PCR Cytol.+HC2 HC2 Test CIN3 CIN3 CIN3 CIN2/3 CIN2/3 CIN3 CIN2/3 CIN2/3 CIN threshold 171 110 93 30 56 37 90 86 No. of CIN 20810 8551 7732 2098 1356 7908 10358 1997 No. of women No No No 8 No 5 5 No Colposcopy negatives for P P P+I P P P P+I P identification 4 2 4 2 2 tests No. of Case . . et al et al et al. et al. (2003) Study (2000) et al . (2001) Salmeron 3 et al. Wright (2000) Belinson 5 (2003) Ratnam Petry (2003) Sherman 2 et al . (2003a) Ferreccio et al . (2003) Cuzick 43.Table screening and HPV DNA testing for Combined use of cytology detected at screening;P, I, interval cancers; cytology; Cytol, conventional LBC liquid-based cytology; chain reaction; PCR, polymerase capture; hybrid HC, value; predictive positive value predictive negative PPV, NPV,

111 059-116 24/01/05 11:06 Page 112

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 44. Combined use of cytology and visual-based methods for screening: two representative studies

Study No. of No. of CIN Test Cut-points Sensit. (%) Specif. (%) PPV (%) NPV (%) cases CIN

Ferreccio 8551 110 CIN3 Cytol. ASCUS 63 94 12 99.5 et al. (2003) LBC ASCUS 86 88 9 99.8 Cervicog. A 62 85 5 99.4 Cytol. + Cervicog. LSIL or P 75 91 10 99.6 LBC + Cervicog. ASCUS or P 93 84 7 99.9

Shastri et 4039 57 CIN2/3 Cytol. LSIL 57 99 38 99.4 al. (2004) VIA P 60 88 7 99.3 VILI P 75 84 7 99.6 Cytol. + VIA LSIL or P 83 87 9 99.7 Cytol. + VILI LSIL or P 89 83 7 99.8

Abbreviations: Cytol. = conventional cytology; LBC = liquid-based cytology; A = equivocal cervigram; P = positive cervigram; PPV, positive predictive value; NPV, negative predictive value; VIA, visual inspection with acetic acid; VILI, visual inspection with Lugol’s iodine

Sequential tests (triage) Triage is of most value when the in approximately 5% of screening cyto- The classical scheme of secondary screening test lacks specificity and/or logical tests. Similarly, in the United cancer prevention consists of three the diagnostic procedure is expensive Kingdom, approximately 4% of all steps: sensitive screening of asympto- or a limited resource. An efficient triage smears show borderline or mildly matic individuals to identify those at test should reduce overtreatment, dyskaryotic changes. The threshold of risk of disease, specific diagnosis of patient anxiety and inconvenience, as test positivity at equivocal cytology the disease state and treatment of well as overall management costs, substantially increases sensitivity for those with cancer or a cancer precur- usually by reducing the number of identifying histological CIN 2/3 (Kinney sor (see Figure 45). Triage is an addi- diagnostic procedures performed—all et al., 1998), but at the cost of tional step interposed between screen- without excessively sacrificing sensitiv- repeated cytology or referring millions ing and diagnosis to further stratify ity for detection of disease. However, of women for colposcopy and biopsy. individuals with positive primary the sequential use of imperfect tests Many of these women are not infected screening results according to their tends invariably to reduce sensitivity with oncogenic HPV types, and some risk for the disease state. In other somewhat. 90% do not have prevalent CIN 2 or words, a second test is performed only CIN 3 and are not destined to develop if the first test is neither completely nor- Triage of cytology it in the immediate future. In this setting mal nor definitely indicative of need for In cervical cancer screening, the pri- of lower specificity, a triage test that treatment. In this respect, triage is con- mary test has traditionally been a pro- further stratifies women according to ceptually related to the consideration gramme of repeated cytological tests, cancer risk is appealing. of residual error in stepwise prediction which generally succeeds because of A multicentre, randomized clinical models. the typically long natural history of trial was conducted by the US National The utility of a triage test in the HPV persistence leading to cervical Cancer Institute to compare different context of a cervical cancer screening carcinogenesis. A single cytological strategies for managing the 2–3 million programme will depend not only on the test is not sufficiently sensitive to serve women with ASCUS and 1.25 million performance characteristics of the test as an adequate screening test. In an women with LSIL cytological results in itself in relation to the primary test, but effort to maximize sensitivity and neg- the USA each year (Schiffman & also on the target screening popula- ative predictive value, a test finding of Adrianza, 2000). About 40–50% of tion, the prevalence of disease, the ASCUS or above is used as the women with ASCUS are HPV-positive cost of follow-up, the available threshold for referral for additional fol- (Manos et al., 1999; Solomon et al., resources (logistic and monetary) and low-up in the USA. ASCUS is a com- 2001), the actual proportion depending patient compliance (Solomon, 2003). mon cytological interpretation, applied on the patient population and the

112 059-116 24/01/05 11:06 Page 113

Screening tests

cytomorphological threshold utilized. HART study in women over the age of for all women with AGC or ‘atypical Importantly, virtually all of the occult 30 years has confirmed the validity of endocervical cells’ (Wright et al., CIN 2 or 3 associated with ASCUS is this approach (Cuzick et al., 2003). 2002c). One study has suggested the found in the HPV-positive fraction. With conventional cytology smears, possible utility of HPV triage following Therefore, in the context of ASCUS there is no residual sample available an AGC result (Ronnett et al., 1999). cytology, triage by HPV testing can for HPV testing, but in the USA, an Finally, colposcopic referral is recom- save approximately 50% of women additional specimen is now often co- mended for another relatively uncom- from unnecessary colposcopy without collected to be used for triage if an mon equivocal interpretation, ASC-H, compromising sensitivity. ASCUS interpretation is obtained (oth- because of the high risk of underlying This result has been supported in a erwise the co-collected specimen is CIN 2 or 3 (Wright et al., 2002c). recent meta-analysis of 15 comparable discarded). When oncogenic HPV is International variation in cytological studies of HPV DNA testing as an detected in conjunction with an terminology, compounded by the use alternative to repeat cytology in ASCUS cytological interpretation, the of different morphological criteria for women who had equivocal results on a tendency at present is to report both similarly termed diagnoses, might previous cytological test (Arbyn et al., findings, rather than to upgrade the imply that results could not be general- 2004b). The pooled sensitivity and cytological interpretation to SIL (Levi et ized between countries (Scott et al., specificity to detect histologically con- al., 2003). 2002). However, use of an atlas of firmed CIN 2 or worse were 84.4% The ASCUS-LSIL Triage Study cytology images, with known HPV sta- (95% CI 77.6–91.1%) and 72.9% (95% (ALTS) and other studies have shown tus and disease outcome, should allow CI 62.5–83.3%), respectively, for over- that cytologically identified LSIL, when the performance of HPV triage to be all HPV testing. Restriction to the nine interpreted stringently (as in, for exam- transferred between classification sys- studies (Table 45) where the HC2 ple, France, Sweden and the USA) is tems and screening programmes with- assay was used yielded a pooled sen- so highly associated with HPV that an out the need for costly repetition of tri- sitivity of 94.8% (95% CI 92.7–96.9%) HPV triage test (as a sequential test) is als (Solomon, 2003). and a pooled specificity of 67.3% (95% not useful, due to low specificity In regions where expert colpo- CI 58.2–76.4%). (ASCUS-LSIL Triage Study (ALTS) scopic services are limited or expen- Consensus management guide- Group, 2000, 2003a; Arbyn et al., sive, the possibility of triage with lines for follow-up of an ASCUS cytol- 2002; Scott et al., 2002). In a meta- another visual technique is attractive. ogy result, based on the accumulated analysis of studies based on HC2 for However, several evaluations of triage evidence, were developed for the USA detection of CIN 3 (Arbyn et al., 2002), by cervicography or visual inspection under the sponsorship of the American the estimates of relative sensitivity, after cytological testing (Costa et al., Society for Colposcopy and Cervical specificity, positive predictive value and 2000; Denny et al., 2000b; Mould et al., Pathology (ASCCP). Acceptable negative predictive value were 95.7% 2000; Blumenthal et al., 2001; Ferris et options following an ASCUS cytologi- (95% CI 91.2–100), 32.9% (95% CI al., 2001b) have indicated lower accu- cal interpretation include repeat cytol- 17.8–48.0), 32.4% (95% CI 13.4–51.3) racy than HPV DNA testing, with inad- ogy, immediate colposcopy or HPV and 98.8% (95% CI 97.1–100), equate sensitivity. testing (Wright et al., 2002c; American respectively. Cuzick et al. (2003) sug- College of Obstetricians and Gyneco- gested that HPV testing might be use- HPV first, then triage by cytology or logists, 2003). ful for triage of mild dyskaryosis based visual inspection Triage by HPV DNA testing of on high negative predictive value, The combination of HPV as an adjunct women with ASCUS is now very com- despite the high HPV prevalence asso- to cytology may be an interim strategy mon in the USA, where, if initial liquid- ciated with this cytological interpreta- in an evolution that ultimately leads to based cytology is used, ‘reflex’ HPV tion. primary screening by HPV with triage testing (see Glossary) is considered Atypical glandular endocervical by cytology. In fact, HPV testing fol- the preferred triage approach, as it cells (AGC), the glandular counterpart lowed by cytology is a rational obviates the need for a repeat visit of ASC, is a much less common cyto- approach for older women, given the (Wright et al., 2002c). HPV testing has logical interpretation with a higher risk higher sensitivity of HPV testing and also been recommended to be intro- for underlying precancerous lesions or the greater specificity of cytology duced in the United Kingdom for bor- cancer than ASCUS. Current ASCCP (Sasieni & Cuzick, 2002). derline cytological cases on a pilot guidelines recommend colposcopic The performance of cytology as a basis (Cuzick et al., 1999a, b) and the evaluation with endocervical sampling triage test might be very different from

113 059-116 24/01/05 11:06 Page 114

IARC Handbooks of Cancer Prevention Volume 10: Cervix cancer screening

Table 45. Triage of ASCUS cytology by HPV DNA testing (Hybrid Capture 2) for detection of histologically con- firmed CIN 2+

Study Sensitivity Specificity PPV NPV Test positivity Prevalence of CIN 2+

Manos et al. (1999) 0.892 0.641 0.151 0.988 0.395 0.067 Bergeron et al. (2000b) 0.833 0.616 0.208 0.968 0.432 0.108 Fait et al. (2000) 0.857 0.971 0.906 0.954 0.235 0.248 Lin et al. (2000) 1.000 0.745 0.692 1.000 0.527 0.365 Shlay et al. (2000) 0.933 0.739 0.230 0.993 0.313 0.077 Morin et al. (2001) 0.895 0.742 0.162 0.992 0.292 0.053 Rebello et al. (2001) 0.857 0.759 0.581 0.932 0.413 0.280 Solomon et al. (2001) 0.959 0.484 0.196 0.989 0.568 0.116 Zielinski et al. (2001) 0.917 0.687 0.149 0.993 0.347 0.056

PPV, positive predictive value; NPV, negative predictive value Modified from Arbyn et al. (2004)

its characteristics as a screening test The use of this approach depends could appear in several of the cate- (Solomon, 2003). If used as a triage upon the development of a widely gories depicted in Figure 45. Thus col- test, there would be a dramatic reduc- applicable inexpensive and rapid HPV poscopy is used in some settings, par- tion in the number of tests overall and test. For such a combined approach, ticularly in Europe, as an adjunctive a marked increase in the yield of posi- revised (more sensitive) criteria might screening test, but it can also be cate- tive results, altering the ratio of nega- be needed for visual assessment, to gorized as a triage modality. tive to abnormal specimens. It is prevent a serious loss of overall sensi- Although colposcopically directed unclear how this would affect the sen- tivity (Denny et al., 2000b). A disadvan- biopsy has been used as the gold sitivity and specificity of cytological tage would be the loss of the limited standard for diagnosis, recent findings testing. ability of cervical cytology to detect suggest that it misses about a quarter non-cervical neoplasia, particularly or more of prevalent CIN 3 (ASCUS- HPV then visual inspection endometrial cancer, in a strategy LSIL Triage Study (ALTS) Group, It might be possible to combine an restricted to HPV and visual inspection. 2003b). This implies that women with inexpensive, rapid HPV test of the an apparently negative diagnosis on kinds now under development with Triage following visual inspection colposcopy remain at increased can- simplified visual inspection to produce Some investigators have considered cer risk, possibly requiring more than screening strategies with good charac- two-stage cervical cancer screening in resumption of routine screening (Viikki teristics. Evidence supporting this pos- which visual inspection would be fol- et al., 2000). Moreover, the three-stage sibility comes from a very few studies lowed by a second test (Denny et al., strategy of screening, requiring a where HPV testing and cervicography 2000b). However, in a simple applica- return for histological diagnosis and a were both analysed (Ferreccio et al., tion requiring both results to be posi- third visit for treatment, leads in many 2003; Jeronimo et al., 2003). tive before treatment, the overall sensi- regions to unacceptable loss to follow- Screening, triage and even treatment tivity would be limited by the numbers up. Consequently, there is reason to services could be combined in the of cases missed by either test, regard- explore other ways of combining the same visit and thereby reduce loss to less of the order in which they are first test, triage, diagnostic test and follow-up in areas remote from health applied. management. Women diagnosed with clinics. Simple visual assessment cate- less than CIN 2 by colposcopy are at gories (e.g., normal versus lesion Follow-up of positive test approximately 10% risk of CIN 2 or treatable by cryotherapy versus lesion results CIN 3 within two years; this risk is sim- requiring a gynaecologist) could be The conventional confirmatory test fol- ilar regardless of whether the colpo- calibrated to maximize reliability and lowing an abnormal primary screening scopically directed biopsy result was sensitivity if screening was done only result has been colposcopically ‘negative’ or ‘CIN 1’ (Cox et al., 2003). once or twice in a woman’s lifetime. directed biopsy. Certain procedures There is insufficient evidence regard-

114 059-116 24/01/05 11:06 Page 115

Screening tests

SCREENING Triage DIAGNOSIS Diagnosis TREATMENT Post- and risk treatment clarification monitoring Cytology Cryotherapy LBC Colposcopically *** Cytology Smear directed Molecular Laser therapy **** **** Repeat biopsy and markers of **** HPV DNA HPV DNA cytology histological risk ? LEEP *** **** **** diagnosis **** Colposcopy VIA HPV DNA Cold-knife Cervico- **** conization graphy Colposcopy Colposcopy

Figure 45 Sequence of cervical cancer screening and prevention The classical steps of cancer screening and prevention are screening, diagnosis and treatment, in bold capitals. Triage and Diagnosis and Risk Clarification are steps to clarify the risk of respective subpopulations (adapted from Solomon, 2003).

ing the optimal management of women efficiently identify women with occult intraepithelial (Sherman et al., 2002). diagnosed with less than CIN 2 by col- CIN 3 and permit the majority of women Very early detection leads to a greater poscopically directed biopsy. In ALTS, to safely return to routine screening. likelihood of overtreatment of lesions, various re-triage strategies combining More sensitive screening and particularly CIN 2, that might otherwise follow-up cytology and HPV testing triage strategies that translate into regress. Identifying markers of risk of were compared (Guido et al., 2003). A increased detection of ‘early’ and often progression to cancer is a priority in single HPV test at 12 months gave the very small CIN 3 lesions may lead to order to reduce unnecessary treat- best trade-off of sensitivity and referral earlier treatment, but the impact on ment and attendant complications and percentage. As an alternative, semi- cancer outcomes has not been estab- costs associated with treating all cases annual cytological sampling appeared lished. Many small high-grade lesions of CIN 2 or 3. Novel approaches were to be useful. Further studies are needed might regress, and others could be considered in the previous section of to find assays or strategies that more detected later, when larger but still this chapter.

115