A Practical Approach to Squamous Abnormalities on Cervical Cytology: Overview of Interpretive Criteria and Guidance for Altering Thresholds in Response to Quality Assurance Findings

Squamous intraepithelial lesions identified by the Papanicolaou (Pap) test result from the action of human papillomavirus (HPV) on the mucosal cells of the female genital tract. HPV is the most common sexually transmitted infection in the world and affects almost every person at some stage of their lives, most commonly during adolescence or early adulthood. The biology of HPV is complex [1, 2], but understanding the basics facilitates an appreciation for the significance of squamous abnormalities seen in cytology.

HPV is a family of nonenveloped, double-stranded DNA viruses of approximately 7,900 base pairs. There are many different related viruses, some of which are high risk for development of invasive carcinoma, most notably types 16 and 18, whereas others are low risk [3]. All types infect basal squamous cells, most commonly at the transformation zone where vulnerable cells are permanently exposed, making this portion of the cervix the primary target for screening and treatment. HPV can also infect squamous mucosa at sites of microtrauma where viral particles gain entry to the basal layer. Endocervical glandular mucosa may also be infected, though less commonly. After the initial infection, HPV stimulates the infected cells to begin producing new virions. As the infected cells mature and shed, the new viral particles are released and can infect other areas of the mucosa as well as the mucosa of sexual partners.

The cells that are producing HPV viral particles in this manner have a characteristic viral cytopathic effect, known as koilocytosis, that was described prior to any understanding of the role of HPV in the process. This historical accident helps explain the confusion over terminology for these lesions. In The Bethesda System (TBS) [4] and Lower Anogenital Squamous Terminology (LAST) System [5], koilocytic changes are known as low-grade squamous intraepithelial lesions (LSILs). The term “lesion” has replaced the older words “dysplasia” and “cervical intraepithelial neoplasia” since it is now known that the viral effects corresponding to LSIL are only a marker of risk for invasive cancer rather than being truly premalignant in and of themselves. Most LSILs will spontaneously regress following immune activation against the offending virus, usually within 2 years [6]. Vaccination prevents infection by priming the immune system to recognize and eliminate HPV before lesions can form [7].

The reason cervical screening exists is to identify precursor lesions for invasive cancer that can be destroyed through medical intervention, preventing morbidity and mortality. In TBS and LAST, such lesions are termed high-grade squamous intraepithelial lesions (HSILs) and are fundamentally different from the similarly named LSILs. In addition to the immediate production of virions, HPV DNA can also enter the nucleus of the infected cell and incorporate itself into the host genome. This process allows the virus to persist indefinitely, while the associated E6 and E7 gene products of HPV interfere with the normal regulatory mechanisms of the cell, leading to uncontrolled cell growth and reproduction that may produce and perpetuate mutations [2]. If untreated, malignant transformation may eventually occur. High-risk viruses are defined by their ability to cause this process, which low-risk viruses cannot do. Importantly, high-risk viruses do cause koilocytic changes as well. Most low-grade lesions are attributable to high-risk viruses because the high-risk viruses are more prevalent in addition to being more virulent, particularly type 16 [1, 8]. This explains why LSIL can serve as a marker of risk for HSIL, particularly if persistent or identified in women who are past the peak years of infection (more than 25 years of age).

LSIL and HSIL are the two most common definitive categories in TBS for the categorization of cervical cytology. However, the morphologic distinctions between LSIL, HSIL, and reactive changes are not always clear-cut. For this reason, TBS also includes the equivocal categories. Unfortunately, also, screening failures occur, and invasive squamous cell carcinoma may be seen in Pap tests, necessitating an outright malignant category as well. A detailed analysis of how to classify squamous abnormalities within TBS will constitute the bulk of this review. The basics will be outlined below, but for those first learning gynecologic cytology, reading the 3rd edition TBS book is essential [4]. The authors of TBS have waived royalties to make the book as inexpensive as possible. A recent and thorough review article is also available online at no cost [9]. The primary objective of this review will be to provide advice on how to alter established thresholds among experienced practitioners. Although performance measures identify outlier interpretive patterns that should be corrected, little literature exists that provides guidance as to how to make such corrections. This article aimed to fill that void.

Before further discussion of the categories, a brief review of normal cervical cytology is required. Cervical squamous mucosa in reproductive age women contains 3 cell types: parabasal/metaplastic, intermediate, and superficial. These cell types correspond to different degrees of maturation. Normal squamous mucosa maintains itself through the production of parabasal cells that then begin to mature as they rise to more superficial layers of the mucosa. The parabasal cells adhere to the basement membrane and divide to produce new cells. These new cells then gain cytoplasm as they mature in preparation for becoming the superficial layer that is responsible for making the mucosa resistant to friction and the passage of water. As cytoplasmic volume increases in response to estrogen, keratin and glycogen also increase. The keratin helps form the impervious barrier. Glycogen provides nutrients to Lactobacillus bacteria that in turn produce an acid environment that deters harmful organisms. In parallel with cytoplasmic expansion, the nuclei condense and become smaller.

This biological process of squamous mucosa regeneration and maturation produces the three different cell morphologies used as a baseline in gynecologic cytology. Parabasal cells have the least cytoplasm, resulting in the highest nucleus-to-cytoplasm (N/C) ratio. At the transformation zone, squamous mucosa interfaces with glandular mucosa that predominates within the endocervical canal, forming a circumferential band where parabasal-type cells are more exposed than in other areas of squamous mucosa. Squamous metaplasia, a process whereby cells within gland structures take on squamous biologic and morphologic characteristics, also occurs in the endocervix and produces a complex squamocolumnar junction. This process is ubiquitous among women, though the degree varies. Morphologically, parabasal cells, transformation zone cells, and squamous metaplastic cells cannot be readily distinguished. Often, they are referred to collectively as “metaplastic” because these are the cells from this group most likely to be sampled during Pap testing. It should be noted that among postmenopausal women with atrophy, and among women who are pregnant or exposed to artificial progestins for contraception, parabasal cells come to predominate, and this may be the only population seen in a Pap test. The cytoplasm of metaplastic cells is typically blue-green and foamy or vacuolated but may be dense. The cell shape is ovoid or round.

Intermediate cells show the flattened scale-like or plate-like abundant cytoplasm typical of squamous differentiation. The cytoplasm remains blue, however, due to relative lack of keratins. Yellow glycogen may be seen in some cells. The nuclei are about the same size or slightly smaller than those of metaplastic cells but still usually have open chromatin. Intermediate cell nuclear size is important as a comparator when evaluating the nuclear size of cells of concern. Superficial cells have abundant cytoplasmic keratin, causing them to retain the orange G stain. They also have plate-like cytoplasm. Their nuclei are smaller and darker.

Interpretation of cytomorphology is inherently subjective. Perfect inter- and intraobserver agreement cannot be achieved. However, TBS represents a significant and useful effort to create a uniform set of criteria that can be used by cytology practitioners as a basis for their interpretations, leading to greater consistency. Within TBS, there are 3 definitive abnormal categories and 3 equivocal categories for squamous lesions. The definitive categories are LSIL, HSIL, and squamous cell carcinoma. The equivocal categories are atypical squamous cells of undetermined significance (ASC-US), atypical squamous cells, cannot exclude HSIL (ASC-H), and HSIL with features suspicious for invasion. ASC-US and LSIL are the most commonly used categories, reflecting the high frequency of HPV infections. However, ASC-H and HSIL are more specific to the lesions of the greatest interest that require treatment. Analysis of these four categories will be the focus of the review.

LSIL (shown in Fig. 1a–d), as mentioned above, corresponds to koilocytosis by morphology and HPV viral cytopathic effect by biology. The morphologic criteria for LSIL are multiple but not all need to be present for the interpretation to be rendered. However, any one criterion alone is not enough. The most concrete criterion for LSIL is nuclear enlargement. At least some of the nuclei in cells interpreted as LSIL should have nuclei 3 times (or greater) the size of an adjacent normal intermediate cell nucleus. Other nuclear criteria for LSIL include irregular nuclear contours, hyperchromasia, and multinucleation. The final major criterion for LSIL is the presence of a large perinuclear halo, or clearing of the cytoplasm, that surrounds the nucleus. Cytoplasmic halos also help confirm the presence of LSIL and, along with marked nuclear enlargement, are usually the easiest features to identify during rapid screening or at low power. However, halos are relatively nonspecific, especially if not well formed with distinct cytoplasmic clearing, a crisp and distinct edge, and involvement of a sizable portion of the cytoplasm entirely encompassing the nucleus.

HSIL cells (shown in Fig. 2a–d) more closely resemble metaplastic cells than intermediate or superficial cells. They have less cytoplasm and a higher N/C ratio. The cytoplasm may be dense or finely vacuolated and usually blue-green, though orange keratinized cytoplasm may be present, typically in a minority of the cells. Halos are usually not present, though in some instances, they may be seen. The nuclear features are critical. Somewhat in contradiction of the usual expectation in cytology that more worrisome cells have larger nuclei than less worrisome cells, HSIL usually has smaller nuclei than LSIL, though still enlarged relative to normal intermediate or metaplastic cells. Remembering that active replication of HPV virions produces koilocytosis helps explain this paradox since the nuclei are in fact more active in LSIL than HSIL in many instances. In addition to enlargement, nuclear changes include hyperchromasia and nuclear contour irregularities. These changes are typically more striking and more uniformly present in HSIL than LSIL. Although hypochromatic HSIL may rarely be encountered, marked hyperchromasia is normally present and is often the most striking finding that draws initial attention to the cells during screening. Closer analysis demonstrating dramatic nuclear contour irregularities, with pronounced nuclear grooves and projections resembling “ears and noses” on a distorted face, helps confirm the interpretation of HSIL.

ASC-US and ASC-H have features suggesting the possibility of a squamous intraepithelial lesion but qualitatively or quantitatively falling short of what would be needed for a definitive interpretation. ASC-US has a specific nuclear size criterion that enlargement to at least 2.5 times the size of an intermediate cell nucleus should be present. However, the ability of the human brain to accurately distinguish the subtle difference in enlargement of 2 times or less (negative) versus 2.5 times (ASC-US) versus 3 times (LSIL) is questionable at best [10]. Therefore, despite TBS providing these numbers, the category remains somewhat subjective. ASC-H is the less commonly used category and has much higher specificity for HSIL in follow-up biopsy. This category should be used when cells concerning for HSIL are seen, but there is a lack of certainty. Usually, this is because of scarcity of the cells of interest. However, it may also occur in circumstances where there are definite koilocytes present with some cells with higher N/C ratio and more dramatic nuclear atypia also seen. Although the latter cells may be relatively numerous, confident differentiation from LSIL may be difficult due to the presence of a spectrum. In such cases, an interpretation of ASC-H and LSIL may be used. ASC-H, though less definitive, is, however, more specific for high-grade lesions and should therefore be emphasized over LSIL in the Interpretation/Result section of the report to minimize potential confusion over patient management. The combined interpretation of LSIL, cannot rule out HSIL, may be encountered in some laboratories but is discouraged by the authors of TBS. Importantly, ASC-US may also be used as the interpretation in some cases where the cells of concern have high N/C ratio. Typically, this occurs in the setting of atrophy or high-progesterone states where lack of maturation makes it difficult to distinguish basal from more superficial layers, resulting in consideration of both ASC-US and ASC-H for the same cells. Thus, ASC-US encompasses a broad range of atypical changes that encompasses concern for both LSIL and HSIL.

Invasive carcinoma is relatively rarely encountered, especially in well-screened populations. Invasive carcinoma has two major subtypes: non-keratinizing and keratinizing. Keratinizing invasive carcinoma has a very distinctive appearance with markedly atypical cells showing “tadpole” morphology with extended orange cytoplasm resembling tails. Such cells often also have very dark and pyknotic nuclei that may be angulated. Non-keratinizing invasive carcinoma closely resembles HSIL. However, the cells are typically more numerous, have more dramatic nuclear atypia and pleomorphism, may have prominent nucleoli, and, most importantly, have a background containing abundant blood and diathesis. Diathesis is granular debris derived from breakdown products of either necrotic malignant cells, reactive desmoplastic stroma, or both. In conventional Pap smears, the blood and diathesis are relatively obvious. However, liquid-based Pap test methods are designed to reduce the presence of blood and debris to improve visualization of the cells, resulting in relative scarcity of these features [11]. In liquid-based Pap tests, the diathesis is often seen “clinging” to the malignant cells and is best appreciated at high power. Abundant blood and debris may also clog the filter in ThinPreps, resulting in a “red ring” around the periphery of the circle where cells should be deposited, resulting in a deceptively hypocellular (or even unsatisfactory) Pap test. If available, correlation with patient history and the clinical impression regarding the status of the cervix may be helpful in such cases. Not uncommonly, the patient has symptoms, and a biopsy collected at the same time as the Pap test can be immediately correlated with the cytology. If a case has definite HSIL as well as some exceptionally dramatic atypia or diathesis, an interpretation of HSIL with features suspicious for invasion can be used. This category may be more useful in the context of asymptomatic or pregnant women where conveying elevated concern may help guide immediate management.

TBS establishes a relatively straightforward interpretive system for Pap tests. However, subjectivity remains substantial due to considerable morphologic overlap between reactive change, koilocytosis/LSIL, and HSIL. This subjectivity cannot be eliminated, but several methods of quality assurance (QA) have been developed to provide feedback to cytology screeners and cytopathologists. This feedback serves to indicate whether the interpretations rendered correspond well to true cervical lesions and if not, provides a signal to change interpretive thresholds in order to provide more reliable and accurate screening.

The most straightforward QA method is cytology to histology (cyto-histo) correlation [12]. By comparing cervical biopsies to prior Pap test slides and interpretations, valuable information about the accuracy of individual cytology interpretations may be gleaned. If a biopsy shows HSIL according to the LAST system (corresponding to high-grade dysplasia or cervical intraepithelial neoplasia 2–3 in previous terminological systems), review of prior Pap tests may reveal the fact that HSIL cells were not seen or misinterpreted during screening. Similarly, lookbacks following a diagnosis of invasive cancer may show screening lapses. Although less clinically significant, review of prior Pap tests preceding a biopsy with LSIL may also show that cells of interest were missed. Cyto-histo correlation of biopsies with negative findings following Pap tests interpreted as HSIL or carcinoma may also highlight screening errors resulting from over-interpretation of reactive changes or other mimics. Conversely, it could also reveal mistaken biopsy diagnoses. While performing cyto-histo correlation, it is important to remember that neither cytology nor biopsy has perfect sensitivity, that sampling error may occur, and that low-grade lesions may spontaneously regress during the interval between tests. Therefore, biopsies and Pap tests may not correlate even though no errors in interpretation or diagnosis can be identified.

Other common QA methods attempt to identify screening errors prior to the release of the original report. In the USA, rescreening is the most common method due to its codification in law. At least 10% of cases initially interpreted as negative must undergo a second screening, with positive results on the second review providing an opportunity to identify missed cytologic findings for the purpose of feedback to the primary screener. This method is also meant to reduce the error rate of the laboratory over time, though the utility of this process is questionable due to the low prevalence of significant lesions in a typical screening population [13]. An alternate method used widely outside the USA is rapid prescreening [14, 15]. This involves having two screeners review every slide, with the first screener looking quickly for 30–60 s and a second screener reviewing more definitively at ×100 magnification, with any discrepancies prompting additional review. This method has the advantage of providing QA for every slide. It also appears to provide more consistent improvement in screening performance.

QA tools may also operate more broadly to encompass large volumes of Pap tests rather than individual slides. A commonly used method is to compare cytotechnologist’s screening interpretations with the final interpretation rendered by cytopathologists [16]. It cannot be assumed that the final interpretation by the cytopathologist is more correct or accurate than the screening interpretation [17, 18]. However, in general, a high rate of discrepancies, especially if consistently tilted toward either the higher or lower end of the spectrum, may indicate skewed interpretive thresholds on the part of the cytotechnologist.

Another valuable technique is the ASC:SIL ratio. It is calculated by adding the number of ASC-US and ASC-H interpretations and dividing by the number of LSIL, HSIL, and malignant interpretations [15]. This yields a simple-to-understand number. ASC:SIL ratio should not exceed 3, with the usual target set at 1.5. In the USA, the College of American Pathologists has conducted surveys to determine ASC:SIL ratios in various laboratories around the country, with the results published to provide a benchmark including the overall mean and the 10th and 90th percentiles to provide a sense of the outer boundaries of the acceptable range of performance [19]. Most laboratory information systems allow ready calculation of this ratio for individuals and the entire group. This enables identification of practice patterns that deviate from established norms.

HPV test results may also be employed as a QA tool. The rate of positivity for high-risk HPV types in Pap tests with various interpretations can be calculated in a straightforward manner, though the analysis may be limited by the fact that not all Pap tests undergo HPV testing. However, in most laboratories that perform HPV testing, the great majority of ASC-US get tested due to the widespread use of HPV triage for ASC-US even in populations that are not receiving primary HPV screening or co-testing with both HPV and cytology. This allows for an HPV-positive rate in ASC-US that is fairly representative and reliable. Results of an individual or laboratory can also be compared to large published studies, such as those performed to evaluate HPV test platforms for regulatory approval [20, 21], and College of American Pathologists survey data are also available [22]. Speaking broadly, a rate of HPV positivity of about 30–35% for ASC-US is appropriate for typical screening populations [20‒22]. However, different rates may be appropriate for laboratories serving populations with higher or lower HPV prevalence.

ASC:SIL ratio and high-risk HPV-positive rates in ASC-US provide valuable data that can be used to evaluate performance. The information can be provided to cytopathologists to help them practice in a manner more similar to established norms. However, the effectiveness of this feedback is limited by two key factors. One is that it can be difficult to determine why ASC:SIL ratios and HPV-positive rates are different from desired targets. The subjectivity of ASC-US thresholds is high both at the lower end, where the differential is reactive change, and at the higher end, where the differential is LSIL and to a lesser degree HSIL. ASC-US rates, ASC:SIL ratio, and HPV positivity rates can therefore be altered by changing either threshold. The best approach is often unclear [23]. Longitudinal data with ongoing periodic feedback that allows for trial of various approaches with evaluation of the outcomes over time have been proposed as a means of dealing with this issue [24]. The potential for longitudinal assessment to improve interpretive consistency, however, can only be fully realized if experienced practitioners have practical guidance enabling them to consciously alter their thresholds based on principles that are easy to comprehend and apply while also being reasonable and appropriate.

Suggested methods for adjusting thresholds will constitute the remainder of this review. This information is intended as the primary purpose of this work and will hopefully provide strategies for improved interpretative reproducibility that readers will find useful. Attention will be paid predominantly to ASC-US thresholds since these decisions are the most common and the most analyzable using the QA tools described above. Ways to reduce false-negative and false-positive HSIL screening results will also be discussed, however, to aid cytology practitioners who struggle with these issues as identified through cyto-histo correlation, rescreening, or prescreening.

Efforts to reduce the ASC-US rate or improve the ASC:SIL ratio often require shifting a larger number of borderline or minimally abnormal cells into the NILM category. This means raising the threshold and allowing more Pap tests to be released with a negative result. There are several approaches that can be taken to making this shift. The easiest for most practitioners is to not overreact to single abnormal features. Remember that every adequate specimen has at least 5,000 squamous cells, and most have tens of thousands. Inevitably, upon careful screening, a few outlier cells in any Pap will show “abnormalities” that are simply part of natural variation.

It is therefore important to become comfortable ignoring isolated binucleation or nuclear enlargement in the absence of other features. Binucleation can be found in almost any Pap if specifically looked for. Large nuclei have little significance if the nuclear contour is smooth and the chromatin is pale, such that you could imagine that the total amount of chromatin present in the large nucleus is basically the same as the amount in a nearby nucleus that is smaller but proportionally darker (shown in Fig. 3c). Artifactual nuclear enlargement can also be seen in some circumstances that, if recognized, may allow the change to be ignored. In liquid-based Pap tests, it is not uncommon for cells at the periphery of the circle to show nuclear enlargement and corresponding pallor known as “edge effect” (shown in Fig. 3d). These changes resemble air-dry artifact that may cause similar changes in conventional Pap smears or Pap-stained non-gynecologic specimens with delayed or insufficient fixation. In extreme cases, the nuclei may also show degenerative changes such as cracking or a glassy appearance.

Perinuclear halos are another feature that may be misinterpreted in isolation. True koilocytic halos should be large, encompassing much of the cytoplasm, with distinct clearing and a crisp, well-defined cytoplasmic rim. Organisms, particularly Trichomonas and Candida, frequently induce small reactive halos, known as “tight halos” or “trich halos” (shown in Fig. 3a). These small, round halos encompass only a small proportion of cytoplasm in a symmetrical manner, immediately surrounding the nucleus with a lack of well-formed outer rim. Typically, they are also much more frequent than true koilocytic halos. A similar phenomenon, known as “moth-eaten” cytoplasm (shown in Fig. 3b), occurs in the infectious and reactive context as well. Moth-eaten cytoplasm refers to cytoplasmic clearings that are typically multiple, asymmetric, and often do not encompass the nucleus. Occasionally, the clearing may surround the nucleus and appear koilocyte like, though the presence of additional clearings in the same cell serves as a clue to the reactive nature of the process. Tight halos and moth-eaten cytoplasm both also occur in the setting of increased acute inflammation, another clue to their reactive origin.

Although most screeners become accustomed to recognizing and appropriately ignoring reactive pseudokoilocytosis, another under-appreciated problem is the halo-like appearance of cells resulting from folded-over peripheral cytoplasm. Not uncommonly, especially in liquid-based preparations, squamous cells edges may curl over on themselves, creating a dark rim of redundant cytoplasm at the cell periphery. At first glance, this may impart to the central portion of the cytoplasm a halo-like appearance due to its relative pallor and the presence of a peripheral rim. Careful attention to the “halo,” however, will reveal that in such instances, there is no true clearing and the central cytoplasm has the same color as adjacent cells with similar degrees of maturation, while the outer folded-over rim is approximately twice as dark.

Finally, glycogenated cytoplasm, as seen in so-called “navicular” cells due to their resemblance to a dugout canoe, may also mimic halo formation and should be discounted. The yellow tinge of glycogen in the Pap stain is the easiest clue to glycogenation rather than true clearing as the cause of pallor in the center of the cells. However, glycogen may be present in some true koilocytes, and therefore yellow color alone does not rule out LSIL.

In general, attention to the nuclear features, regardless of the cause of the halo-like appearance of the cytoplasm, will aid in separating pseudohalos from the true version, as LSIL should have distinct nuclear changes in at least a portion of the involved cells. Halos have value as an easily discerned feature during screening, but ASC-US rates can be reduced if only cells with abnormal nuclei as well as halos fall into the ASC-US category. Strictly adhering to the criteria of at least ×2.5 nuclear enlargement is a particularly effective way of avoiding overuse of ASC-US due to excessive reliance on cytoplasmic features.

Parakeratosis is another major source of interpretive problems and excessive use of the ASC-US category [25]. In LSIL or in response to inflammation or trauma, the superficial, keratinized layers of the mucosa may become thickened and take on a “spire” morphology, in which the cells form into a cone resembling a medieval castle spire roof. The advanced maturation of these cells makes the nuclei relatively small and dark, regardless of cause, and nuclear contour irregularities are also often evident, though frequently degenerative in nature (shown in Fig. 4a). Since discernment of parakeratosis due to HPV and due to nonspecific inflammation can be difficult, many cytology practitioners reflexively include such findings in the ASC-US category. However, most parakeratosis is not related to underlying koilocytosis, and this will result in excessive ASC-US rates with low HPV positivity.

“Atypical parakeratosis” that should properly reside in the ASC-US category can be discerned by looking carefully at the cell groups. The presence of significant nuclear pleomorphism and anisonucleosis helps distinguish truly worrisome parakeratosis (shown in Fig. 4b). The presence of other features such as binucleation or cytoplasmic halos also helps, though these features are less commonly discernible in this context. Nuclear size criteria may be modified such that ×2.5–3 enlargement relative to other nearby cells with a similar degree of nuclear and cytoplasmic maturation, rather than an intermediate cell, may be used as a somewhat objective measure. Lack of such enlargement is reassuring. Careful application of these suggestions can significantly reduce the fraction of parakeratotic spires interpreted as ASC-US.

Less commonly, reparative changes may result in overinterpretation as ASC-US. Most cytology practitioners easily recognize the enlarged but very pale nucleus, often containing a prominent chromocenter, which is typical of reactive change. When such nuclei are aligned and present within a sheet-like cluster of cells with elongated “stretched out” or “taffy pull” cytoplasm, or the so-called “school of fish” streaming appearance, the changes can be readily interpreted as repair and dismissed as insignificant (shown in Fig. 4c). However, some repair-like changes may cause increased concern, especially if they are more three-dimensional or haphazard in architecture. Overcalling such groups as ASC-US may be limited by looking carefully at the nuclear features. Only if marked nuclear contour irregularities, asymmetric chromatin distribution, significant anisonucleosis, frequent multinucleation, or some combination of these features is present should the cells be interpreted as atypical (shown in Fig. 4d). In the event of a known history of radiation, the threshold should be raised even higher. It is worthwhile to remember HSIL or invasive carcinoma typically have scant cytoplasm, meaning that reparative-appearing cells with abundant cytoplasm, including the marked cytomegaly typical of radiation, may be classified as NILM with reactive change.

Overinterpretation of atrophic changes is another major problem that may lead to an excess frequency of ASC-US [26, 27]. The presence of an exclusively immature population of cells makes application of the usual criteria somewhat more difficult (shown in Fig. 5a, d). The same problems may be seen in high-progesterone states such as pregnancy or progestin-based birth control methods, though usually to a lesser degree. The presence of numerous high N/C ratio cells with dark and irregular nuclei may be difficult to ignore. Excessive ASC-US rates in atrophy can be avoided if attention is paid to the baseline cytologic appearance. Many atrophic cells look concerning in isolation at high power. However, pulling back to look at the entire population of cells will usually show that similar morphology is ubiquitous. Only if the cells of concern have a distinctly different appearance from the background cells should a positive interpretation be entertained (shown in Fig. 5b, c). In principle, this is like the “two-cell population” rule employed in body cavity fluids to distinguish malignant populations from reactive mesothelial cells. One caveat is the presence of rare individual cells with marked nuclear enlargement. This occurs not uncommonly in older women and, if unaccompanied by hyperchromasia or nuclear contour irregularities, usually has no significance.

ASC-US rates may also become inflated due to knowledge of past results. Many cytology practitioners are reluctant to call a Pap negative if the woman has previous positives, particularly something high grade. It is important to remember, however, that Pap test sensitivity is low. Even if a high-grade lesion exists, the Pap result can be negative. This factor is built into the screening and treatment algorithms. It is also important to remember that women may receive curative treatment between the prior Pap and the current specimen without the knowledge of the cytology laboratory. Furthermore, women with positive results can only return to routine screening if their follow-up results are negative. Calling Pap tests ASC-US due to artificially lowered thresholds following previous positive results therefore causes more harm than benefit and should be avoided.

An even more pernicious problem is the influence of knowledge of corresponding HPV results on the psychology of cytology practitioners [28, 29]. Often, they struggle to call an HPV-positive case NILM. They seek to “agree” with the molecular testing result and stringency drops. On the other hand, worrisome findings get improperly dismissed as “reactive” if there is a known HPV negative result based on the false premise that HPV testing has 100% sensitivity [30, 31]. Treatment guidelines depend on the independence of HPV and cytology results to determine the risk level [32]. The purpose of the cytology arm of co-testing is to find the cases that HPV alone would miss [33]. When interpreting Pap tests, therefore, every effort should be made to remain maximally objective and not allow HPV results to influence the interpretation. Indeed, to help achieve this outcome, practitioners should consider raising the threshold for ASC-US in HPV-positive cases while lowering their threshold in HPV negative cases. This mindset, rather than the opposite that is too often employed, is optimal for patient care. Women with positive HPV tests have already screened positive and do not benefit from superimposed false-positive cytology. Women with negative HPV results may not be screened again for 5 years, making minimization of false-negative cytology critical to protect them.

As primary HPV screening becomes implemented, this problem of knowing and being swayed by the molecular result will become ubiquitous. Currently, many laboratories can withhold HPV results or release cytology results prior to completion of molecular testing. This will not be possible if screening only occurs as a reflex test. Studies of DNA methylation molecular testing as a more objective secondary test to triage for colposcopy after a positive high-risk HPV result have already progressed to an advanced stage in Europe [34, 35], partially due to the difficulty of remaining objective as a cytology practitioner in this situation. If cytology is to stay relevant, cytology practitioners must become comfortable interpreting HPV-positive Pap tests as NILM if strict morphologic criteria for SIL are not met.

Reduction of the ASC-US rate and ASC:SIL ratio may also be achieved by moving more borderline cases from the ASC-US category into LSIL instead. There are several strategies that may be employed to accomplish this goal. The presence of organisms should not preclude an interpretation of LSIL. Two separate processes can happen at once. As discussed above, reactive changes due to organisms may mimic koilocytosis but should not reproduce all features simultaneously. Likewise, many cytology practitioners hesitate to use LSIL if only one group of cells meets the criteria. Given the low sensitivity of cytology, finding only a single convincing group happens not infrequently, even with large corresponding lesions. Pushing such cases to LSIL is reasonable and appropriate.

Another strategy to consider may be more controversial but does have the ability to reduce the expense of screening. Young women, below age of 30 years and especially below age of 25 years, have a higher likelihood of HPV infection than older women while simultaneously having a lower risk of HSIL or invasive carcinoma. Using this knowledge, judiciously interpreting borderline cases as LSIL in younger women may be of benefit by reducing the ASC:SIL ratio and by reducing the number of women reflexed to HPV testing that has relatively low utility.

To restate the obvious, Pap tests exist to find HSIL. This is the treatable premalignant lesion. Thus, every screener should strive not to miss this finding. Unfortunately, low-grade lesions, being more common, often come to dominate the mentality of cytology practitioners. This can result in missed opportunities to identify HSIL.

If QA efforts indicate this problem, several psychological shifts may help mitigate it. First, do not stop looking for HSIL just because LSIL has already been found. HSIL often coexists with LSIL, and recognition still has significant follow-up implications. Second, take extra care when screening Pap tests with abundant inflammation. The neutrophils often serve as distractors that may make recognition of single HSIL cells relatively more difficult. There may also be an increase in associated reactive changes, requiring more vigilance to pick out significant abnormal cells. Third, look carefully at the glandular cell clusters. Sometimes HSILs in glands may manifest as “hyperchromatic crowded groups,” dense clusters of cells that usually have a rounded gland-like architecture and significant 3-dimensionality, making discernment of nuclear features difficult [36]. HSIL may also manifest as repair-like changes with prominent nucleoli [37]. If features like high N/C ratio, hyperchromasia, or nuclear contour irregularities are seen in cells around the edge of such glandular clusters, then careful searching of the background may yield individual cells or smaller clusters that can be more definitively recognized as HSIL. Finally, do not automatically dismiss cells with orange-staining cytoplasm from consideration. Keratinizing HSIL has more abundant cytoplasm and may be dismissed by the unwary as reactive despite the presence of worrisome nuclear features. Fortunately, keratinizing HSIL cells are usually a minority and more prototypical HSIL can also be found, but occasionally, these cells may be the predominant finding.

One useful method for becoming better at recognizing less obvious instances of HSIL is to practice on more obvious cases with higher cellularity. To do this, thoroughly examine cases interpreted as HSIL, looking at individual high N/C ratio cells outside the screener dots, as well as hyperchromatic crowded groups on the same slide, to gain experience recognizing HSIL features in a setting where the cells of interest likely do derive from a high-grade lesion. This exercise will help set more accurate thresholds for HSIL and aid in recognizing challenging or sparse cases as at least ASC-H or ASC-US.

QA efforts may also demonstrate that a cytology practitioner tends to overuse HSIL and ASC-H, resulting in a high rate of negative follow-up biopsies and unnecessary work-up. One frequent setting in which this problem occurs is the older patient population. The morphologic similarities between atrophy and HSIL should induce caution. Other morphologic changes related to atrophy, including so-called “transitional metaplasia” and “blue blobs,” may also raise false concern for HSIL. Transitional metaplasia takes its name from the supposed resemblance of the cells to urothelial (transitional) cells of the bladder. The primary distinguishing characteristic is a single longitudinal nuclear groove in the cells. This finding should not be mistaken for nuclear contour irregularities indicative of HSIL that would be more pronounced and erratic. Blue blobs are large but markedly degenerated atrophic cells that stain darkly, usually with a smudged appearance. Awareness of the existence of this phenomenon usually prevents overinterpretation. In general, caution is indicated in the use of ASC-H in the setting of atrophy. Cells of concern should not go into the ASC-H category just because of high N/C ratio. ASC-US is usually appropriate for cells with nuclear abnormalities that cannot be ignored but that do not dramatically stand out from the background atrophic population.

Another important consideration when attempting to pull back from overuse of ASC-H and HSIL is the size of the nuclei in the cells of concern. HSIL should have nuclear enlargement, at least in some of the cells. If all cells of concern have small nuclei, consideration should be given to two possibilities: histiocytes and endometrial cells. Histiocytes have markedly irregular nuclei and scant cytoplasm but are small cells with small nuclei lacking hyperchromasia. They also frequently have thickened nuclear membranes and partially cleared-out chromatin. Usually, histiocytes can be recognized as inflammatory with appropriate comparisons to the other cells nearby. Endometrial cells can be more problematic. Endometrial epithelium is often balled-up and degenerated, producing hyperchromatic crowded groups with high N/C ratio cells and irregular nuclei. Recognition of small cell size and when present, the finding of a stromal core in the cell groups helps distinguish endometrium. Endometrial stroma may also be problematic. Trauma to the lower uterine segment by intrauterine devices or electrocautery excision may induce the presence of very high N/C ratio cells in Pap tests that are difficult to interpret as NILM. Again, small cell size and small nuclei, as well as history, may help dismiss such cells in some cases. Decidualized stromal cells shed during pregnancy may also mimic HSIL, contributing to the often-given advice to be cautious about interpreting Pap tests as HSIL in this setting.

Finally, reparative changes should not be overinterpreted as HSIL. Typical repair with its abundant cytoplasm, pale and even chromatin, prominent nucleoli, smooth nuclear contours, and sheet-like cohesive architecture usually causes few problems. However, repair sometimes may have significant nuclear atypia and pleomorphism. In these cases, including many seen in the setting of prior chemotherapy and radiation, dramatic nuclear features may resemble malignancy according to criteria commonly used in non-gynecologic cytology specimens. The temptation to interpret such cases as HSIL or carcinoma should be resisted in a screening setting where no mass lesion is clinically evident. The presence of abundant cytoplasm and the scarcity of the cell clusters found in atypical repair should allow for an interpretation as ASC-US or NILM in most instances, especially if history is well documented.

This review aimed to help experienced cytology practitioners who seek to alter their interpretive thresholds in response to feedback from QA activities. Reconsideration of where to draw the line between normal and abnormal, and between different categories of abnormal, often proves challenging and frustrating. The effort is worthwhile, however, to help bring individuals into closer alignment with the practice of the broader cytopathology community. Increased uniformity and reproducibility of cytology results benefits patient care by facilitating the production of treatment guidelines with universal validity.

Michael J. Thrall has received royalties and authoring fees from Elsevier for the textbook Diagnostic Pathology: Cytopathology and has received honoraria for lectures on gynecologic cytology for the online Oakstone CME series.

No funding was provided for this review publication.

Michael J. Thrall is the sole contributor and is responsible for all aspects of this work.