Sources of Error

Strategies to meet the challenge of difficult cytology need to be informed by an awareness of the sources of error. The retrospective finding of recognisable abnormal or suspicious cells in many false negative slides is often attributed to inattention or to lack of knowledge. This is too simplistic. While these may at times contribute to mistakes, the sources of error and difficult diagnosis in cytology are far more complex. They include human factors, the cytological criteria that are applied to decisions, and the characteristics of the specimen.

Human factors

Limitations of vision and perception

Inherent deficiencies of the human visual system and perceptual processes contribute to errors. Lack of detail in peripheral vision, the relatively long time taken for eye movements, reaction times, and perceptual processes that select information from the visual field, all contribute to loss of information between the slide and the cytologist's conscious awareness. 1

Screening technique

Stage movement practice 2, visual search strategies, the level of vigilance to alarm signals and strategies to maintain alertness and attention are all part of screening technique, and can lead to error if not optimal. Furthermore, ineffective technique can become habitual since there is little immediate feedback from the screening task to reveal where a technique may not be working well.

Subjectivity

It is a truism that cytology is subjective. There are several aspects to this:

Interobserver variability

Each cytologist applies criteria differently 3 according
to individual experience and understanding, and applies intuition to decision making 4. Interobserver variation may be considerable, especially for non-standard appearances 5.

A cytologist may interpret the same cells differently on different occasions, attending to different features, or weighting criteria differently. This may be a major source of error. Experienced practitioners are as prone to such variability as the less experienced 6 practitioners.

Bias

Intraobserver variability occurs because human judgement is always biased in some way and different biases operate
at different times. In addition, the person making a judgement is unaware of the biases distorting their decision at
the time 7.

It is well known that powerful bias exists during slide review, and missed abnormalities are more easily found retrospectively 8, 9. However, it is less well known that during routine screening, interpretation is also biased – towards negative. For example when features resemble a benign reactive condition, it is very easy, and only natural, to jump to the wrong conclusion – ‘Only reactive’, then to seek and find evidence to support
that conclusion.

The opinions of colleagues provide a useful counter to some biases because each person brings a different perspective and set of biases to the situation. In addition, social biases also come into play whenever opinions are communicated. Communication reduces observer variation, but an agreed opinion may still be wrong 10. It should be noted that the same patterns of bias operate during diagnostic review or checking prior to reporting.

Table 1

Lack of specific knowledge

Lack of knowledge of specific abnormal criteria contributes to diagnostic difficulty and misdiagnosis11. Even experienced cytologists may have gaps in their knowledge 12.
Emphasis on lack of knowledge of specific criteria, as the key source of error, may be misplaced. Experienced cytologists often make errors despite having adequate knowledge of the relevant criteria.

Cytological criteria

Morphological overlap between diagnostic categories

There is true morphological overlap between normal and abnormal:13

• Every abnormal feature may be seen in normal cells
• Every normal feature may be seen in abnormal cells
• Every abnormal criterion may be absent or present to a greater or lesser degree in abnormal cells

Thus no single criterion can be relied upon to diagnose abnormality, and no single criterion can be used to distinguish normal from abnormal. Multiple abnormal criteria must be present to sustain an abnormal diagnosis – but multiple abnormal criteria may be present in a benign condition.

Even if abnormality has been identified, distinguishing low and high grade lesions may not be possible on morphology alone 14,15,16.

Morphological overlap gives rise to difficulties for researchers and authors when identifying, defining and testing diagnostic criteria. It also gives rise to difficulty for practising cytologists attempting to apply diagnostic criteria. Each difficult case presents a novel combination of ‘normal’ and ‘abnormal’ features.

There are further dimensions to the problem of morphological overlap and the application of cytological criteria:

Criteria are ill defined

Cytological terms are often ill defined, or refer to a spectrum of appearances. There is no strict agreement on what some terms mean. For example coarse, hyperchromatic, and irregular. Criteria used successfully by some may seem too vague to be of use to others. 17 There is often little said about the relationships between abnormal criteria, how to weight them for importance, and how criteria ought to be applied under different circumstances.
A statistical approach to criteria may give a false impression of precision and consistency within a diagnostic category by neglecting the range of appearances within and between cases, and the subjectivity in interpreting them.

Standard generalised abnormal criteria do not apply to all cases

Descriptions of cytological abnormalities in the literature are generalised. For example, criteria for High Grade Squamous Intraepithelial Lesion (HSIL) include: 14

• Nuclear enlargement
• Marked increase in the nuclear / cytoplasmic (N/C) ratio
• Hyperchromasia
• Nucleoli generally absent
• Irregular nuclear outlines

The general description might allow most cases to be diagnosed. However, many abnormal cases depart from the standard and lack one or more of the general criteria. The standard criteria can be misleading in such cases, and abnormalities may be overlooked or dismissed as normal. 11

Figure 1

Figure 2

Figure 3

Although every cytologist knows that there are exceptions, standard, generalised criteria are an important foundation for cytological thinking. This can contribute to misinterpretation. While ever we think of non-standard cases as exceptions, we are less likely to recognise them as full members of the category abnormal. Actually, they are not exceptional they are quite common.

Table 2

Contradictions in the literature

Published studies may contradict each other, showing that different criteria are useful in diagnosing a condition.
For example, alternative criteria for cervical adenocarcinoma include:

• Small nuclei and fine, even chromatin 18
• Enlarged nuclei and irregularly distributed chromatin 14

Although confusing, such contradictions may not be critical. Different authors study different sets of cases, so there are bound to be some differences. The contradictions simply reveal something about the wider range of appearances in that condition. Also contradictory criteria may be present together in a case. One author may draw attention to general appearances, another to a diagnostically significant subgroup of cells. Both perspectives provide useful information.

Cellular patterns for which adequate diagnostic criteria are not well known or agreed on

A number of specific cellular patterns have been identified as difficult to find or to interpret. These include small pale abnormal metaplastic cells and crowded sheets. They often resemble some benign entity, and discriminating criteria are not well established. These patterns are over represented in false negative smears. 11,19,20,21,22

Figure 3

Figure 4

Changes in knowledge and terminology

A challenging aspect of cytological practice is how to integrate new findings into diagnostic practice. Recent examples include transitional cell metaplasia, 23,24,25 stratified mucin-producing intraepithelial lesions, 26 and small pale dyskaryosis. 11

Revision of the Bethesda System has introduced new categories and redefined others. 27

Some cases with adequate criteria are not confirmed

Sometimes a cytological prediction of high grade abnormality is not confirmed. Apart from cytologic error, possible explanations include biopsy site selection, histotechnical factors, and histologic interpretive errors. 28,29 Further investigation or close follow-up is warranted, as subsequent investigation often reveals a significant abnormality. 30,31 In some cases however, despite unequivocal cytological abnormality, thorough investigation and follow-up never confirms a lesion. Regression of the lesion is a possibility, but such cases challenge the reliability of cytological criteria. An implication of this is that every cytological diagnosis or prediction carries an element of uncertainty within it.

Figure 5

Characteristics of the specimen

Specimen compromised

Abnormal cells may be difficult to detect, and even if detected, difficult to interpret if key criteria are compromised by degeneration, poor fixation, being obscured by inflammatory exudate 21, the presence of blood and overly thick areas of specimen.

Figure 6

Small numbers of abnormal cells

Overlooking a small number of abnormal cells has been highlighted as a reason for false negative cytology. 19 Many, if not most false negatives however, are due to misinterpretation of a large number of abnormal cells. 11,20,32 Even in cases where stand alone diagnostic cells are few, there are often many high risk pattern cells that could serve as a trigger to increased vigilance and so aid detection of the few diagnostic cells.