Epidemiologic Aspects of Uterine Cervix Cancer

In the introduction to his classic monograph on vaginal smears half a century ago, Papanicolaou observed that “although methods for treating uterine cancer have undoubtedly been improved, the total death rate ascribed to the disease in the United States has undergone scant diminution during the past twenty-five years.”¹ The problem, in his opinion, was attributable to “the difficulties which surround early diagnosis and hence to the infrequent opportunities to treat the disease in its most susceptible stage.” Papanicolaou lived to see his pessimism disproved by his own successful efforts in the early detection of cervical cancer.

Epidemiologic aspects of any disease may not be the most appealing or interesting area of study for clinicians; however, it provides the most effective means for preserving health. In 1948, a writer in the New England Journal of Medicine proclaimed, “There is no indication for the use of vaginal smears.” It is sobering to speculate how many lives were lost because of this statement and similar sentiments delaying the adoption of widespread screening.

This chapter discusses epidemiologic aspects of cervical cancer that may not be obviously useful for clinicians in their daily practice. Principles and specific information are reviewed that may convey an understanding of general themes useful in the practice of medicine over a lifetime. An appreciation of epidemiology is essential for the practitioner so that advances like the Pap smear do not require decades before they are appreciated.

Cancer is not an officially reportable disease in many countries. Therefore, mortality statistics often have been used in estimating the incidence of cancer and associated time trends. This approach is valid only after consideration has been given to the effects of variations in survival rates and the diagnostic validity of the death certificates on the calculated rates. For instance, in rare cancers, the death certificate diagnosis is as likely to be an error as it is to be the correct, rare cancer diagnosis.²

Cervical cancer is particularly susceptible to surveillance errors because it is a disease of the developing world. Even when errors may not be an issue, such as in communities with sophisticated tracking capabilities, the reported data results may not be applicable to other populations. This is because cervical cancer is a socially determined disease to a degree not seen in other cancers. In relatively homogenous populations such as those of northern Europe, this may not be a significant limitation. However, in locales such as the United States, which is characterized by its heterogeneity, differences in socioeconomic status can be significant. In the United States, terms such as nonwhites, blacks, and African Americans do not adequately represent the populations they seek to describe. Nevertheless, investigators continue to report data on samples and populations that may have limited relevance in other areas of the country. Having acknowledged this limitation of the literature, most of the referenced sources will be from U.S. data unless otherwise stated.

Cervical cancer is the leading cancer in women throughout the developing world and the second most common female cancer worldwide. Each year approximately 500,000 women are diagnosed with cervical cancer. Mortality is staggering at 200,000 lives lost each year.³ Because it is a disease of young women in the developing world, it results in significant years of life lost and tremendous social complications for the affected societies.

Statistics on cervical cancer mortality are summarized in Table 1, where both age-specific and overall age-adjusted rates are presented. Worldwide, the incidence rate varies by more than 10-fold. Despite the aforementioned acknowledged deficiencies, some conclusions can be drawn:

TABLE 1. Time Trends in Mortality Rates from Cervical Cancer

*Average annual mortality rates per 100,000; rates for specific age groups based on average of fewer than 5 deaths per year in parentheses.
†Rates adjusted to total US population, 1950.
(Modified from Kessler II: Cervical cancer epidemiology in historical perspective. J Reprod Med 12:173, 1974; and relevant volumes of Vital Statistics of the United States for 1975, 1976, 1980, 1981, 1984, and 1985. Baltimore rates for 1974 derived from unpublished statistics of the Baltimore Department of Health, 1976.)

1. Nonwhite women are subject to a cervical cancer mortality risk that is higher than that of whites in each decade of life as well as overall.
   
2. Cervical cancer mortality rates among white and nonwhite women have been declining substantially since the late 1950s in each age group.
   
3. Mortality from cervical cancer increases continuously with age to at least age 70; half of all cervical cancer deaths in the United States occur in women older than 60 years of age.
   
4. Deaths among patients younger age 25 are primarily caused by the nonsquamous varieties of the neoplasm.
   
5. Among older nonwhite women, substantial declines in cervical cancer mortality did not occur until the late 1960s or early 1970s.
   

A distinction must be made between the improvement in outcomes seen in cervical cancer overall versus the stage-specific survivals, which have not clearly changed during the same time period. Three-year and 5-year relative survival rates among women with cervical cancer have not increased substantially since the 1950s (Table 2). As expected, survival is substantially higher for patients whose cervical cancer is localized, but for women with either localized or regional disease, there have been no substantial improvements in survival over the last three decades. Between the early 1970s and 1984, there was no consistent increase in 3- and 5-year relative survival rates for white and nonwhite women with cervical cancer when controlling for stage.⁴ Because the sizable decline in cervical cancer mortality cannot be attributed to therapeutic enhancements, the explanation must lie in the stage migration. Stage migration is the increase proportion of patients seen with stage 0 (carcinoma in situ [CIS]) versus frankly invasive cancers.

TABLE 2. Time Trends in Survival from Cervical Cancer by Stage of Disease: US Whites, Treated and Untreated

*Patients whose disease was of indeterminate stage. Their survival rates are included only under All Stages.

†Rates refer to the 68% of the patients whose disease was staged.
(Modified from Axtell LM, Cutler SJ, Meyers MH: End Results in Cancer, Report No. 4. Department of Health, Education and Welfare Publication No. [NIH] 73–272, 1973; Axtell LM, Meyers MH: Recent Trends in Survival of Cancer Patients, 1960---1971. Department of Health, Education and Welfare No. [NIH] 75–767, 1975; and 1986 Annual Cancer Statistics Review, DHHS, Publication No. [NIH] 87–2789.)

The data reveal that a significant change in the stage at diagnosis of cervical cancer has taken place since the 1950s. At that time, only one third of the cases were diagnosed while still in stage 0. Fifteen years later, the proportion had risen to approximately three quarters (Table 3). Statistics indicate that this year, approximately 60,000 women will be diagnosed with CIS compared with 6000 invasive cervical cancers.⁵ Whereas this trend toward enhanced detection of the intraepithelial disease is manifest in all ages, the greatest beneficiaries have been the younger age groups, (i.e., women who are most likely to adopt the salutary habit of periodic Pap testing). The spectacular increase in the frequency of diagnosis of CIS may be seen in the incidence statistics for the period between the early 1950s, the late 1960s, and the 1990s (Table 4). From a rate of 5.0 per 100,000 in the early 1950s, the incidence among white women increased sixfold during the latter part of the century. Among nonwhite women, whose baseline incidence is approximately twice as high as white women, the rate increased fourfold over the same time interval. The increases in in situ cancer incidence were observed at all ages younger than 60 but were more pronounced among women younger than 50 years of age. These trends were essentially the same among both white and nonwhite women. The proportion of stage 0 lesions among the latter is lower.⁶

TABLE 3. Time Trends in Stage at Diagnosis of Incident Cervical Cancer: Baltimore Women Younger Than Age 60

(Modified from Kessler II: Cervical cancer epidemiology in historical perspective. J Reprod Med 12:173, 1974.)

TABLE 4. Time Trends in Cervical Carcinoma In Situ Incidence, 1950---1969: Baltimore Women Younger Than Age 60

*Average annual incidence rates per 100,000; rates based on average of fewer than 5 cases per year in parentheses.

†Rates adjusted to total U.S. population, 1950.
(Modified from Kessler II: Cervical cancer epidemiology in historical perspective. J Reprod Med 12:173, 1974.)

Overall, stage I cancers of the cervix constitute approximately 60% to 70% of all cervical cancer diagnosis; stage II, 10% to 20%; stage III, 10% to 20%; and stage IV, 5%. Whereas socioeconomic status has an effect on the stage of disease among whites, in nonwhite women, the attained socioeconomic status is a greater cofactor in determining the proportion of cases detected in the intraepithelial or frankly invasive stages. Currently, cases of cervical cancer in United States' hospitals are overwhelmingly intraepithelial or early invasive in stage, whereas advanced metastatic patients are more common in unscreened populations. In countries where exfoliative cytologic screening is actively practiced, the clinical situation closely resembles that in the United States. Many western countries have lower rates of invasive cancer and higher proportions of intraepithelial lesions than the United States.

Women with cervical cancer are at increase risk for other cancers. These include lung, anal, vulvar, and bladder cancer. The magnitude of this elevated risk has been reported between 2- and 10-fold. These second primary cancers may share a common etiologic agent (e.g., human papillomavirus [HPV]) or may be related to the host's immune system, as in the case of smokers.^7,8 This elevated risk has not been specified specifically for adenocarcinoma of the cervix.

Evidence for the effectiveness of cytologic screening is largely indirect because a controlled clinical evaluation of the effectiveness of exfoliative cytologic screening in a defined population never has been undertaken. This is typical of the introduction of new technologies into health care practices. Only recently has there been rigorous testing of new technologies by evaluating evidenced-based outcomes. It is difficult to test established practices once they have become integrated into routine health care. However, as the developing world incorporates western practices, some investigators have seized this as an opportunity to objectively evaluate current treatment algorithms.

A group from the Cleveland Clinics, in partnership with the government of the People's Republic of China, has taken part in the initiation of cervical cytologic screening in various provinces in China (J. Belinson, personal communication, 1999). As women take part in the screening program, simultaneous colposcopy, biopsy, and HPV assay are performed on all patients. This prospective cohort investigation will generate data on the true sensitivity, specificity, and clinical utility of each of these interventions.

This lack of a complete evaluation for medical interventions is regrettable but not unusual. The reasons for this include a tendency within the biomedical community to accept new technology with great enthusiasm but without a commitment to critical evaluation. This is especially true if the modality appears to be rational, acceptable to patients and provider, and profitable.

Impressive indirect evidence on the effectiveness of cytologic screening is available from the results of the program in British Columbia. Analysis of the effects of the program between 1961 and 1971 reveals a 10-to-1 difference in incidence of invasive cervical cancer between the unscreened and the screened women (see Table 4). Critics of Pap testing argue that if the program were truly successful, the incidence of cervical cancer should have fallen among the screened women, whereas the number of cases has remained fairly constant at approximately 4.5 per 100,000. This phenomenon is easily explained in view of the definition for “screened” used by the British Columbia program (i.e., a woman who has received “one or more” Pap tests over this period of time). Thus, a woman screened once and never tested again is considered to be screened and yet has an increasing risk of cervical cancer as time passes. A better comparison would be between women never screened and those screened on a regular (e.g., annual or biannual) basis. If regular Pap testing were effective in the early detection of disease, then its initial effects in a community should be an increase in the proportion of intraepithelial neoplasms diagnosed, followed by a decrease in the death rate from invasive cervical cancer. Both of these phenomena have been observed. A longer range manifestation of successful cytologic screening, then, should be a more gradual reduction in the incidence rate of invasive cervical cancer as more of the cases are detected while still in situ (Table 5). The data also reveal declines in incidence among nonwhite women equivalent to, if not in some cases greater than, those in whites. These trends are clearly consistent with the beneficial impact of Pap testing in the community.

TABLE 5. Time Trends in Invasive Cervical Cancer Incidence, 1950---1960: Baltimore Women Younger Than Age 60

*Average annual incidence per 100,000; rates based on average of fewer than 5 cases per year in parentheses.
†Rates adjusted to total U.S. population, 1950.
(Modified from Kessler II: Cervical cancer epidemiology in historical perspective. J Reprod Med 12:173, 1974.)

Although cytologic testing may appear as a science to medical professionals outside of pathology departments, the public, and their attorneys, most tests must separate normal from abnormal results based on an arbitrary and sometimes unclear line. A lower threshold for labeling Pap smear results as abnormal would yield a greater sensitivity but lower specificity. In the 1970s and 1980s, results of approximately 2% of all Pap smears were interpreted as abnormal. By the late 1990s, results of up to 30% of Pap smears were being considered abnormal.⁹ This may result partly from a true increase in the incidence of dysplasia; however, some of this increase is from moving the bar lower; that is, more Pap smear results are being considered abnormal without a true increase in cervical dysplasia. Evidence for this can be found in the CIS incidence data from the Surveillance, Epidemiology, and End Results Program, which do not demonstrate a parallel increase in CIS during the same time that abnormal Pap smear results increased so dramatically. A diagnosis of CIS is based on histologic study, which may be more reliable and a better measure of true disease incidence compared with cytologic study.¹⁰

New variations on the Pap smear make use of the fact that the line between normal and abnormal can be set arbitrarily. When used to rescreen Pap smears initially reported as normal, a computer-assisted interpretation can detect additional abnormal findings. The clinical significance of this increase detection is doubtful because of the low absolute increase in detection. For instance, manual Pap smear interpretation in a study detected a rate of high-grade squamous intraepithelial lesion of approximately 1.5%. Rescreening the apparent normal results with the computer-assisted technology increased that by 40%, a statistically significant increase because of the numerous Pap smears in the study. However, the absolute increase was only 0.6%. In women screened annually by the manual method, a one-time increase of 0.4% is doubtful to have a clinically meaningful impact.¹¹

Any change in Pap smear practices can have a significant impact on a health care system. In the United States, approximately 75% of all women have had a Pap smear in the last 3 years.¹² This represents approximately 60 million smears each year. As stated previously, up to 20% of smears are labeled “abnormal.” Based on older data and other recent reports, a better estimate may be 2% to 5%.¹³ Regardless of which numbers are used, this translates into millions of abnormal smears each year, most of which will not be associated with a histologic normality. Unfortunately, 50% of all women diagnosed with cervical cancer each year had not had a Pap smear in more than 3 years.

Self-administered cytologic screening techniques may offer a feasible solution to the problem of periodically screening populations remote from medical centers and pathology laboratories (see later). In the United States, these techniques might be considered for the screening of presumably healthy women who, for one reason or another, refuse to appear for periodic examination. Experience shows that response rates in community Pap testing programs rarely exceed approximately 50% overall.¹⁴ In marked contrast is the experience with self-administered vaginal irrigation, in which response rates of 90% or higher are common.¹⁵ In addition, women whose smears are cytologically not normal and who do not return for diagnostic confirmation might be monitored temporarily through the mail using the self-administered test.

For convenience, screening tests may be divided into those administered by the physician (or a surrogate) and those obtained by the women herself. Scraping devices of some kind are commonly used by physicians to obtain cellular material from the vagina, cervical os, and portio. Exfoliated cells from the endocervical canal may be sampled using a brush device or saline-moistened cotton-tipped applicator.

Several methods also have been devised for the collection of specimens by the patients themselves. One involves the use of a cotton vaginal tampon encased in a sheath of nonabsorbent nylon, which is inserted by the woman herself and left in the vagina for periods of time ranging from a few minutes to overnight.¹⁶ Self-administered cytologic tests are based on vaginal irrigation techniques developed in the early 1960s.^17,18 This is a modification of Papanicolaou's original vaginal smear method that called for aspiration of exfoliated material from the posterior fornix using a pipette attached to a rubber bulb. Papanicolaou suggested in 1943 that women can easily be taught to prepare their own smears when many smears are required for study.¹ These techniques also can be used to test for HPV and organisms causing sexually transmitted diseases, including Chlamydia and human immunodeficiency virus (HIV). Other intermediate end points of malignancy may not require more than the DNA from a single cell, further enhancing the potential of vaginal lavage.

The lavage instrument is a disposable plastic pipette containing an irrigating and cell-preservative solution. It is uncapped, inserted by the woman, squeezed, released, and recapped; by simple lavage, a cellular sample is recovered in suspension. The pipette then may be either delivered or mailed to the laboratory, where all further processing is done. This consists mainly of resuspension, centrifugation, and placement of a drop of sediment onto a glass slide for microscopic examination. Advantages inherent in the lavage technique include the following:

• The sample is obtained in privacy.
  
• Additional smears from a single irrigation may be prepared as needed, without recalling the patient for HPV or other ancillary testing.
  
• Processing of the sample is inherently efficient because the pipette serves not only for collection and transportation of the specimen, but also for its centrifugation and fixation.
  
• Preparation of the slide for microscopic study may be carried out under standardized conditions in the laboratory rather than in the physician's examining room.
  
• The smeared cells are distributed as a uniform monolayer on the slide rather than in clumps.
  
• The method is more amenable to automated processing than others.
  

A significant disadvantage of the irrigation method is that cell samplings from the vaginal pool necessarily contain relatively fewer abnormal cells than would a direct scraping of the cervical lesion. However, some lesions cannot be visualized, and even cervical scrapes sometimes must be taken blindly. Another claimed disadvantage is that irrigation smears take longer and are more expensive to process and to interpret than scrape smears. This probably is not the case if the time and expense required for the physician to obtain the scrape before processing are included. Another criticism has been directed at the diminished accuracy or validity attributed to the technique and the “high” false-negative rate presumably associated with it. Comparable data on the validity of vaginal irrigation and cervical scrapes in detecting cervical neoplasia are relatively scanty but range from a high of 100%¹⁹ to a low of 61%.²⁰ Reported false-negative rates tend to be lower for the cervical scrape, although not invariably. Conversely, some cervical cancers that have yielded negative results on cervical scrape have been detected through vaginal irrigation.^21,22

The significantly higher acceptance rate for vaginal irrigation, compared with Pap testing among participants in screening programs, largely compensates for the inferior validity of the pipette. In the laboratory (i.e., once the cytologic specimen has been obtained), the two techniques differ primarily in the rate at which they yield false-negative results. Ignoring the complementary (in this case, less important) problem of false-positive tests, the test sensitivity of the two techniques is assumed to be 90% and 60%, respectively. These figures probably favor the scrape unduly and minimize the true validity of irrigation. In any event, of every 1000 women whose specimens reach the laboratory, 900 would be effectively screened by the scrape and only 600 by irrigation. The absolute advantage of the scrape here is 30%; its relative advantage is 50%.

Assuming the given rates of test validity, what would happen to the effectiveness of the screening programs in light of the different acceptance levels for the two methods in the field, namely 50% versus 90% (Table 6)? Because of the reluctance of women to appear for Pap testing, only 450 of 1000 women sought would be screened effectively. In contrast, irrigation would effectively screen 540 of the target population. The absolute advantage of irrigation in this case is 9%, with a 20% relative advantage.

TABLE 6. Exfoliative Cytology by Scrape and Irrigation Estimate of Effectiveness in the Community

*Advantage of irrigation: absolute: 90/1000 = 9%; relative: 90/450 = 20%.
†Estimated as complement of false-negative rate.
(Adapted from Kessler II: Mortality from cervical cancer: Can it be prevented now? Compr Ther 2:38, 1976.)

These figures suggest that cervical scraping remains the best method for detecting early cervical neoplasia among women who can be induced to present themselves for periodic cytologic screening. Conversely, self-administered methods might be preferred for other women. Irrespective of the method chosen for community-wide cytologic surveys, several operational questions must be answered before the fieldwork begins: Who should be screened? How frequently?

In theory, all women in a given community are at risk of developing cervical cancer, and, therefore, all should be screened. In practice, mass screening is difficult, expensive, and never complete. Thus, the notion of concentrating screening efforts on selected population subgroups is inherently appealing. The question becomes one of establishing workable criteria for the high-risk group to be screened. Selection on the basis of high-risk characteristics may be possible in communities where these groups can be readily identified. In some instances, however, it may be considered ethically or politically undesirable to deal exclusively with groups so classified. In these circumstances, if not in general, selection on the basis of age alone appears to be the simplest approach. For instance, screening only women between ages 30 and 45 years may be most efficient because CIS peaks during this age interval; therefore, this should encompass the most detectable preinvasive cases. Second, over 95% of clinically invasive cervical cancer becomes manifest at age 30 or thereafter; thus, few of these cases will be missed by ignoring women younger than 30. Third, some evidence shows that CIS among women younger than 30 progresses to invasion slowly and over intervals of many years; therefore, the preinvasive cases occurring in such women may be safely detected at age 30 or older. Fourth, a significant proportion (perhaps one half) of all invasive cervical cancers becomes manifest before age 45. Finally, if most invasive lesions arise from earlier, preinvasive lesions, then screening women aged 30 to 45 will eventually eliminate the progression of this disease in most women, without requiring the more complex task of screening all women.

Based on different assumptions in different populations, similar arguments can be made for other age ranges. Countries throughout the world have done exactly that when determining their national health care policy regarding who to screen and how often. There is no universally accepted age to initiate screening, no agreed-on age to stop screening, and no routine frequency. In the United States, cervical cancer screening programs often are part of routine obstetric well-woman care. Although many high-risk patients are undoubtedly screened thereby, the impact on morbidity in the total community often is marginal. There may be a serious down side to universal screening starting at 18 years of age because of the consequence of false-positive results. False-positive results in a young women may lead to unnecessary cervical surgery. When the patient becomes older and is therefore at higher risk for cervical intraepithelial neoplasia, additional cervical surgeries may be required. After multiple procedures, the patient may become reluctant for another surgery before childbearing, fearing the potential obstetric and fertility complications.

Publicity may be a vital part of screening programs for cervical cancer. In countries where population registers are maintained by law, the women to be screened can be readily identified and reminders sent. Conversely, in such countries as the United States where screening targets usually are not identifiable in advance, large-scale public relations efforts may make a substantial difference in response rates. Unfortunately, there also are certain inadvertent effects to such publicity. First, when exhortations on cancer screening are broadcast repeatedly, anxieties tend to be roused, and physicians may be swamped with healthy but nervous patients demanding reassurance. This is routinely demonstrated when a breast cancer study is publicized in the media.²³ Second, publicity campaigns draw the attention of many nontargeted individuals who also may insist on being screened, thus overburdening the system further. The increased cost of screening the additional women involved is obvious.

Several motivational studies have been undertaken on populations undergoing cervical cancer screening in the United States and elsewhere. Physician attitudes toward cytologic examination were a significant factor in deciding whether residents sought cytologic testing. Most investigators agree that nonrespondents in screening surveys tend to come from the lowest socioeconomic classes and from the older age groups.

Another unresolved problem is the frequency of cytologic rescreening of the target populations. To a large extent, this is a function of the rate at which preinvasive lesions progress to invasion (see later) and the financial resources available to the community for such programs. The latter is not easy to estimate, requiring judgments concerning direct and indirect costs of screening as well as lifetime medical expenses and work productivity saved through prevention or early detection of disease.

Some investigators agree that an initial screening should be followed within a relatively short interval (e.g., 1 year) by a second test and that subsequent screenings should be scheduled at longer intervals (e.g., every 3 to 5 years). The rationale is to minimize the false-negative rate of the women who are initiating screening soon after their entry into the program, with subsequent examinations offered at less frequent intervals compatible with the community's expected disease incidence and available resources for screening.

Once a woman is diagnosed with a cervical abnormality, any discussion of frequency regarding the general asymptomatic population is irrelevant. Women with any cytologic abnormality should be rescreened as often as necessary. Rescreening intervals should consider the overall probability of detecting a clinically significant lesion over a specified period of time. It is difficult to derive a rational interval because of several reasons. First, each screening episode may not be considered as having independent probability of detection. Second, the progression rate and time to recurrence is not predictable for the individual patient.

Rescreening is particularly difficult to accomplish in countries like the United States where residential mobility is great, insurance companies change often, many are uninsured, and where population registries for tracking are nonexistent. By contrast, the circumstances are more favorable in the Scandinavian countries where lists of female residents by age are readily available and where residential changes are recorded by law.

In Canada and Europe, the expenses of cervical cancer detection programs usually have been borne by government, with no charges being levied on the women being screened. This is clearly advantageous where the goal is to encourage participation rather than to maximize statistical ratios of “yield” to cost. Conversely, several screening surveys in the United States are supported by patient fees to private physicians or cytologic laboratories. Many high-risk patients are certain to be excluded with such a system. In this context, mention should be made of a cost-benefit analysis carried out on cervical cancer patients screened and treated in Rochester, Minnesota.²⁴ An average of 3.23 additional years of life were attributed to the screening. If the taxes paid on income earned during these additional years are included, cost is low.

Cost is not only financial. Cost of a program also must consider emotional or psychological costs of screening results. Both false-positive and false-negative results can have a “cost” in terms of quality of life. As a result of screening, other health care opportunities may be missed or become unavailable because of limited resources.

In 1842, Rigoni-Stern published an analysis of cancer mortality based on deaths in Verona from 1760 to 1839.²⁵ On the then-valid assumption that mortality approximates incidence, he calculated age-specific proportionate mortality rates and noted that among unmarried women and religious sisters, breast cancer was unusually frequent and uterine cancer unusually infrequent when compared with married or widowed women. Noting that uterine cancer increased in frequency among women 30 to 40 years of age and declined after 60, he speculated that the susceptibility of the uterus to cancer might vary with “the natural exercise of its functions.”

Several investigators during the last two decades have confirmed Rigoni-Stern's observations concerning the rarity of uterine cervix cancer among nuns. Gagnon examined the medical files from many convents of French-Canadian nuns, covering an annual average of 13,000 women over a 20-year period, and found 14 cases of corpus cancer but none of the cervix.²⁶ He then reviewed the pathologic archives of several hospitals and identified three cervical cancers compared with 19 corpus cancers among nuns. Finally, he made an ad hoc survey of four religious orders, with an average annual population of 3280 women whose medical histories over the previous 25 years were documented, and found a total of 222 cancers, 7 of which involved the uterine corpus and none, the cervix.²⁷ Gagnon was impressed not only by the small absolute number of cervix uteri carcinomas among the nuns in these studies but also by their infrequency relative to corpus cancer, a reversal of the pattern usually observed. He attributed this to the rarity of chronic cervicitis, which he regarded as a major carcinogenic factor in virginal women.

In another investigation, Towne diagnosed cervical cancer in 3 nuns among 3083 seen in her private medical clinic and 3 others over a 20-year period in several large religious communities of about 10,000 women.²⁸ Although she concluded that cervical malignancy can develop in women without regard to virginity or parity, the unusually low rate of cervical cancer, both absolute and relative, to that of corpus cancer is apparent.

A reduced mortality from cancer of the uterus among Jewish women had been consistently observed previously. This is based largely on biased and older studies.²⁹ Incidence data on cervical cancer among Jewish women half a century ago revealed an incidence of cervical cancer among the Jews of New York City of 5.0 of 100,000.³⁰ However, even early investigations demonstrated considerable variation in rates between Israeli Jews of different ethnic or geographic origins. The presence of immunohistochemical staining for HPV types 16 and 18 in tissue of Israeli Jewish women with cervical and vulvar neoplasia, however, is no longer different from that of other populations.³¹ The consequence of this already may be present. The prevalence of abnormal cytologic findings among the Israeli women is now the same as that of the non-Jewish group. Moreover, it was found that there was an increase of 29.2% in the prevalence of cervical premalignant lesions among the Israeli women. These authors state that this population can no longer be considered as being at low risk for this disease. In a geographic-based study in northern Israel, Jewish women were the highest risk group for cervical neoplasms.³²

Previously offered theories for the formerly low rate of cervical cancer in Jewish women have included traditional sexual abstinence during and shortly after menses, circumcision of the male, and unspecified genetic factors. Scandinavian countries do not practice circumcision but still have a cervical cancer rate as low as that of Israel. Moslems also practice circumcision, as well as menstrual abstinence, and have similar rates. One other population group has been reported to experience a reduced risk of cervical cancer. Cross and coworkers also found the incidence of this neoplasm to be lower among Amish women than among their rural neighbors in Holmes County, Ohio.³³ They suggest that generations of endogamous mating and monogamous sexual behaviors in the Amish have limited the spread of an agent critical in the etiology of cervical cancer. This prophetic statement heralded the discovery of HPV.

American blacks constitute a high-risk group for cervical cancer. Their age-adjusted mortality rate for this neoplasm between 1959 and 1961 in the United States was 2.53 times higher than that of whites. The mortality among nonwhites was higher than among whites at every age and was unaffected by urban or rural residence.³⁰ Although population-based surveys of this disease in the orient are lacking, cervical cancer is said to be the most common malignant tumor among Hindu women in India, with rates approximately double those among Moslems. The same appears to hold for women in Korea and China.

Prostitutes, independent of ethnic group, experience an extraordinary high risk of cervical cancer. Nearly 9% of a group of prostitutes in a London prison were diagnosed as having uterine CIS during the course of a cytologic survey in 1966.³⁴ Two studies of prison populations in the United States and one in Canada reveal a fourfold to sixfold increase in cervical cancer among female inmates compared with women attending planned parenthood clinics, industrial employees, and prevalence statistics from published population studies.^35,36,37

The HIV-infected patient is at higher risk for cervical neoplasms because of the immunocompromise associated with HIV disease and because the socioeconomic status risk factors for HIV and cervical neoplasms are similar. Again, notice that the term HIV encompasses a large and heterogeneous population. Early studies on the association of cervical neoplasms and HIV included women with advanced immunosuppression, that is, patients with acquired immunodeficiency syndrome or those falling into the Centers for Disease Control category 3, or “C” patients.³⁸ This is because early in the HIV epidemic, women with asymptomatic disease were not detected. Therefore, early statistics on HIV and cervical neoplasms do not reflect the current situation for most of the HIV-infected asymptomatic, non-immunocompromised women. Stratification by semiquantitative measure of immune competence such as viral load or CD4 cell counts may provide a more accurate description of a particular HIV population and allow application of prior studies to current populations. With these limitations acknowledged, some descriptive data follow.

Dysplasia rates generally are twofold higher in HIV women than in noninfected women.³⁹ HPV infection is present in approximately 50% of HIV-positive women. HIV-positive prostitutes are found to be HPV positive in 90% of women studied. In a prospective cohort, 20% of HIV-positive women developed cervical intraepithelial neoplasia over a 30-month period (8.3/100 women per year). Most (90%) were low-grade lesions.⁴⁰ Initially, cervical smears were reported as having a lower sensitivity in HIV-positive women because of frequent concurrent cervicitis.⁴¹ However, subsequent studies demonstrate that if there is any decrease in sensitivity, the positive predictive value of Pap smears is equal in HIV-positive and HIV-negative women.³⁹ Recurrence rates and time-to-event data suggest a median of 1 to 2 years for cervical intraepithelial neoplasia after standard treatment. These earlier studies indicate that the natural history of cervical neoplasms was different in HIV-positive women. This may be true; however, recent data indicate less differences between HIV-positive and HIV-negative women.⁴² Again, this may show that as the epidemic matures, more women with asymptomatic HIV are being included in the reports.

The spread of HPV largely explains the increase in the incidence of cervical intraepithelial neoplasm over the last five decades. HPV is a common human pathogen found in up to 90% of individuals. There are at least 70 different subtypes, each with a particular tropism. The virus is clearly transmitted through sexual contact; however, nonsexual transmission also is possible.⁴³ In a study of households with HPV-positive, sexually active young adults, the adolescent household siblings were found to convert to HPV positive without sexual contact. Vertical transmission from mother to fetus during delivery also has been documented.⁴⁴ Disease occurs in few of the infected individuals. Risk factors for colonization to proceed to disease include immunosuppression, smoking, and persistent infection. Immunocompromised conditions closely associated with HPV disease include HIV infection and organ transplant status. Smoking may predispose to HPV disease through several mechanisms.⁴⁵ Most HPV infections are transient but recur according to periodic genital assays.⁴⁶ Absence of persistent type specific HPV infection over a 6-month period suggests that significant HPV disease will not develop.

The two most significant risk factors for the development of cervical cancer are HPV infection and access to Pap smear screening. HIV has emerged as the next most important factor. HIV is a host factor that promotes HPV progression to disease. Certain diets also may contribute to HPV disease.⁴⁷ Similarly, tobacco use may function as a marker for HPV infection. However, it also may represent a cofactor that promotes the progression of HPV infection to disease. Many other demographic characteristics are associated with these two dominant risk factors. Because of the well-established link between HPV and Pap smear access with the development of cervical cancer, other risk factors should be interpreted as surrogates for HPV and Pap smear status. The long history of epidemiologic studies that linked socioeconomic status characteristics with cervical cancer contributed enormously to the eventual discovery of the causative agent, HPV. It should be viewed as an epidemiologic success story with some major errors, including the previously assumed role of herpes simplex virus (HSV) as the cause of cervix cancer. Other previously considered risk factors for cervical cancer include early age of first intercourse, multiple sexual partners, history of sexually transmitted diseases, oral contraceptive use, intercourse with the partner of a cervical cancer patient, ethnic background, occupation, tobacco use, and diet. Most of these factors should be viewed as risk factors for HPV infection or progression instead of causes of cervical cancer.

Inferential data on the natural history of cervical dysplasia have been derived from laboratory animal studies. In one experiment, a group of mice underwent cervical applications of methylcholanthrene, a carcinogen, until local inflammation (but not dysplasia) developed.⁴⁸ A second group received double this dose until basal cell hyperplasia ensued. Within a year, most of the animals in the first group had dysplasia; all but one in the second group had invasive carcinoma. The authors believed that this experiment demonstrated that there is a critical point in time before which a dysplastic lesion will not progress and after which carcinoma ensues despite withdrawal of the carcinogen. In another similar experiment, mice that did not develop cancer showed ultimate regression of the cervical lesion to normal.⁴⁹ Whether dysplasia progresses to carcinoma in mice was shown to result from a combination of factors, including carcinogen dose and frequency of application, mouse strain, and individual host susceptibility.⁵⁰ These experiments suggest that only the early, nonproliferative forms of carcinogen-induced dysplasia in mice are reversible.

The natural history of dysplasia in women has been studied in various groups, usually hospital patients, over varying periods of time (Table 7). Cytologic smears or biopsies have been the basis for diagnosis; the inferential problems associated with each are discussed elsewhere in this chapter. No precise picture of the clinical course of dysplasia emerges from these investigations, although its variability (i.e., some progressing, many stable, others regressing) is obvious. As might be expected, the more severe dysplasia seems more likely to progress; however, this may reflect changes in the pathologists' diagnostic certainty. Currently, a cytology consultation requires that a medical history be supplied. The cytologist is aware of the prior cervical abnormalities when reviewing the slide at hand. This suspicion bias may affect the interpretation and diagnosis.

TABLE 7. Studies on the Biologic Behavior of Cervical Dysplasia

*Ob/gyn, obstetric---gynecology; Gyn, gynecology.
*Progressed to carcinoma in situ or invasive cancer, unless otherwise noted.
†Progressed to “severe” dysplasia, carcinoma in situ or invasive cancer.
‡Additional patients lost to follow-up or excluded for various reasons in parenthesis.
§Includes 32.5% that progressed within 1 year and 3.8% that progressed after 1 yr.
||Includes 17.1% in which the dysplasia was incidentally noted at hysterectomy.
ÿIncludes 19.6% that progressed within 1 year and 3.1% that progressed after 1 yr

A recent prospective trial may provide an estimate of the natural history of dysplasia. After follow-up of between 6 and 24 months, women initially diagnosed with mild or moderate dysplasia were found to revert to normal approximately 23% of the time. Follow-up smears remained unchanged in 50% and progressed to severe dysplasia in the remaining 25%.⁵¹ These results are consistent with another study that reviewed several natural history reports.⁵² In this summary article, 7% of atypical smears progressed to a more severe lesion, and 21% of mild abnormalities progressed to a severe lesion. Invasive cancer followed 0.25% of initially atypical smears, 0.15% of mild abnormalities, and 1.44% of severely abnormal smears. Follow-up smears became normal in 68% of initially atypical smears, 47% of mild abnormalities, and 35% of severely abnormal smears, all at 24 months.

Squamous carcinoma of the cervix usually evolves from in situ lesions that, in turn, traverse a dysplastic stage. A critical and unanswered question concerns the proportion of preneoplastic and preinvasive lesions, which in the normal course of events can be expected to transform and invade.

Cervical intraepithelial lesions, including CIS, are classic intermediate end points of malignancy. Intermediate end points of malignancy are better for determining the effects of different intervention on cancers because the time frame is shorter, and it may be possible to follow an intermediate end point such as cervical dysplasia whereas it is not possible to follow cervical cancer. There may be better candidates available soon, such as the HPV E6 or E7 protein expression, but the histologic diagnosis of dysplasia currently is a suitable marker.

Several investigators have prospectively followed cases of cervical CIS to ascertain the proportion that eventually becomes invasive. A wide variation in this proportion has been observed (Table 8). This is largely attributable to three factors: mode of diagnosis, length of observation, and the molecular heterogeneity of the single clinical entity known as dysplasia. In addition, in the process of diagnosing CIS, an adequate histologic specimen is required. Obtaining the biopsy specimen is destructive and therefore may alter the natural history of the CIS. Thus, it is likely that a localized CIS may be removed entirely ( i.e., cured) by a fortuitously located biopsy. Single-point biopsies may be effective in removing CIS.⁵³ This may reduce the proportion of cases that are observed to become invasive. In some studies, more destructive procedures were used. These included conization, cervical amputation, and even hysterectomy. The lack of progression in these series is therefore not surprising. Although the biopsy may cure a lesion, it also may miss it partially or entirely. To the extent that this occurs, the apparent risk of invasion increases. Thus, cytologic smears taken to measure the progress of the lesion after biopsy might be, if only temporarily, falsely negative.

TABLE 8. Studies on the Biologic Behavior of Cervical Carcinoma In Situ

Gyn, gynecology; Ob/gyn, obstetric---gynecology.
*Patients excluded from analysis for various reasons in parenthesis.
†Median age.
‡28.0% if 2 “questionably invasive” cases are included.
§1.7% if 3 microinvasive cases are included.

The variability of time periods during which CIS has been followed for evidence of invasion is evident from Table 8. Considering the presumably protracted course of most cases of CIS, some of the reported variation in their invasiveness probably stems from variations in length of observation.

In addition to prospective follow-up of patients with CIS, some information on natural history also has been acquired from retrospective reviews of biopsy specimens taken from patients known to have subsequently developed invasive cervical cancer.⁵³ For instance, 1023 cases of clinical carcinoma of the cervix seen over a 16-year period were reviewed. In 24 of these, a cervical biopsy had been taken 1 or more years before the diagnosis of invasive cancer. These biopsies were reviewed, and additional sections were studied by recutting the original paraffin blocks. For 17 of the patients, changes characteristic of CIS were noted from 1 to 20 years previously. In four other patients, there were no abnormalities, and in the remaining three, changes characteristic of “basal cell hyperactivity” were recorded. The authors conclude that a many cases of carcinoma of the cervix are preceded by CIS for a variable length of time.

The relationship between in situ and invasive cervical cancer has been examined indirectly by comparing the estimated incidence of each condition under Stewart's assumption that all infiltrative cervical cancer must come from in situ cancer.⁵⁴ Applying this method to data for British Columbia from 1955 to 1959, Boyes and colleagues estimate that 61.5% of in situ squamous carcinomas eventually become invasive.⁵⁵ The validity of this approach is unclear, however, because the calculation was based on a ratio of invasive to in situ cancer incidence rates, which recently have been changing in opposite directions. A related methodologic fallacy involves estimating the duration of in situ cancer from the differences between the mean ages of women with invasive and in situ lesions. Knox points out that usually one of the conditions (CIS) has been determined on a population basis and the other (cervical cancer) on a selected basis; or, the mean age in one case represents onset and in the other case, discovery.⁵⁶ Therefore, a lesion's persistence over several years does not necessarily indicate a definite progression to invasive cancer. The age distribution and prevalence of in situ and invasive cervical cancer also have been regarded as consistent, with a biologic relationship between the two. Thus, Gore and Hertig suggest that age curves tend to be parallel, with the CIS curve occurring 10 to 12 years earlier than that for invasive carcinoma.⁵⁷ Ancillary evidence has been offered by Dunn and Martin, who conclude from their statistical analysis of San Diego cytologic screening data that 10% to 15% of cervical carcinomas become invasive early in their development.⁵⁸ In addition, Figge and coworkers screened 39,343 women and found that 34, after having a previously negative smear, developed invasive cancer from 8 months to 3 years later.⁵⁹

Observer variability is a serious problem both to the clinician and to the researcher seeking to understand the natural history of cervical lesions. The subjective nature of microscopic diagnosis has been strikingly demonstrated by several investigators. Siegler submitted slide sections from 20 cervical lesions to 25 pathologists.⁶⁰ The number of slides diagnosed as positive for carcinoma or CIS ranged from 13 by one pathologist to 0 by three others. No more than 4 of the 25 pathologists even agreed on the total number of positive diagnoses. McGarrity submitted 71 slides previously diagnosed as intraepithelial neoplasms to three other pathologists.⁶¹ They agreed with the original diagnoses in 65% of the cases; 28% were considered benign, and 7%, invasive cancer. Holmquist and colleagues requested seven pathologists to independently classify 118 histologic slides prepared from cone biopsies of the uterine cervix.⁶² The results were analyzed by calculating the percentages of agreement for paired comparisons of all possible pairs of pathologists. The average agreement was 54%, with values ranging from 46% to 64%. Most of the disagreement occurred in slides classified by some as CIS, by others as microinvasion, and by others as atypical squamous hyperplasia. Histologic sections of 22 cases diagnosed as stage I invasive cervical cancer were reexamined by five pathologists. For only 10 (45.5%) of the cases was the diagnosis confirmed. Six (27.3%) were reclassified as preinvasive, and three (13.6%) were considered either not cancer or uncertain. This diagnostic variability suggests that the apparent progress of a histologic (or cytologic) lesion may reflect changes in the pathologists' degree of certainty about the diagnosis. Knox observes that it is unwise to assume that progression in a smear sequence depicts progression in the histologic lesion instead of the resolution of uncertainty.⁵⁶

The natural history of cervical CIS has not been fully elucidated, at least partly because of the numerous artifacts and other problems already discussed. Variations in the clinical behavior of cervical lesions led Lawson to suggest that cervical cancer probably has two forms: one that evolves rapidly, gives little opportunity for early diagnosis, and is highly lethal; and the other, which is slowly progressive and amenable to treatment.⁶³ Wespi had previously argued that only a few invasive cervical cancers traversed a significant in situ stage.⁶⁴ Maliphant suggests that the biologic behavior of cervical lesions might be determined mostly by factors related more to host resistance than to intrinsic properties of the tumors.⁶⁵ The impact of HIV disease on cervical neoplasms appears to have confirmed that host defenses are important in the natural history of CIS.³⁸

Ashley is impressed by apparent differences in the age distributions of the squamous and glandular types of cervical carcinoma, the latter occurring primarily among the elderly and the former more frequently in younger women.⁶⁶ He stresses the different shapes of the mortality and morbidity curves of cervical cancer by age and concludes that there may be different types of cervical cancer at different ages.

The observed relationships of cervical cancer to coitus, venereal diseases, and prostitution led several investigators to suggest the possibility of a venereally transmitted etiology in this disease. The hypothesis remained conjectural until Rawls and colleagues isolated genital herpesvirus (HSV-2) from patients in male venereal disease clinics.⁶ Thus began a decade of misled research that is an example of the importance of properly designed epidemiologic studies to generate hypotheses, not to determine causality. Associations were reported without careful consideration of a possible undetected covariable. The discovery of the HPV had to wait for years after HSV was being recovered from patients with cervical cancer until the necessary laboratory techniques were developed. The HSV theory was based on a series of seroepidemiologic studies designed to measure antibodies to HSV-2 in women with and without cervical neoplasms. Despite inconsistencies such as the failure to isolate any HSV virus from husbands of 22 women with cervical cancer, the theory continued to be popular.⁶ Whereas each evidential item is inadequate to confirm causality per se, the net weight of the evidence—biologic, virologic, immunologic, endocrinologic, cytopathologic, epidemiologic, and clinical—was sufficient to lead many to erroneously conclude that HSV was the putative human carcinogen.

However, based on this epidemiologic model, an investigation of the role of the male in cervical carcinogenesis was undertaken. Two large groups of women underwent long-term prospective evaluation. The first group was all the other wives of men who, at some other time, were married to women who developed cervical cancer. The second, or control, group consisted of otherwise comparable women married to men without such histories. A flowchart of the study protocol is depicted in Figure 1. A total of 1983 other wives and 868 control wives have been fully traced.⁶⁷ Cervical cancer or CIS was detected in 45 (2.3%) of the other wives and in 10 (1.2%) of the control wives. A total of 14% of the other wives had either cervical cancer or a cervical cytologic specimen that was abnormal. The corresponding statistic for control wives was 8%. These differences in the prevalence of cervical cancer and of nonnormal cervical cytologic findings are statistically significant. The 45 other wives, together with their probands, constitute what may be termed “marital clusters” of cervical cancer. The first wives in these clusters were married at a mean age of 19.1 years; the second wives were married at 33.5 years. On the average, cervical cancer was diagnosed 16.3 years after the first marriages and 6.3 years after the second.

If women marrying men whose previous wives had cervical cancer experienced a significantly increased risk of this disease themselves, two alternative inferences might be made. First, their increased risk might be attributed to an extrinsic characteristic of the spouses, such as a venereal carcinogen or procarcinogen. Second, the clustering might result from an intrinsic factor, such as the tendency for certain men to marry women who are destined to develop cervical cancer. The second of these possibilities cannot be summarily dismissed because cervical neoplasia is associated with certain social and sexual characteristics, which could influence particular men to repeatedly choose such women for their marital or coital partners.

However, direct evidence favoring an etiologic role of HSV-2 in human cervical carcinoma is not convincing. Instead, the transfection assays with HPV confirm its role as the etiologic agent of cervical cancer.

Whatever pathogenic factors are yet to be discovered, in human cervical cancer, early detection through screening remains the best method for prevention of mortality and morbidity. Thus, exfoliative cytologic testing combined with colposcopy promises to detect cervical lesions early enough to prevent death and serious morbidity. An unresolved problem is the failure of many women to avail themselves of Pap testing on a regular basis. The Norwegian Cancer Society discontinued offering mass cytologic screening because although approximately 75% of the targeted population were being screened by their own physician, the remaining 25% constituted a core of women who refused to accept cytologic screening from any provider. Thus, the program proved ineffective and was dropped.

Several years ago, the demographic characteristics of women who volunteered for Pap testing in a program paid for by a philanthropic organization were reported. Television, radio, and print media had been used to induce the local population to volunteer for screening. When compared with all women in the community who were of the same race, the respondents to the screening program were of considerably higher socioeconomic class. Nearly 60% of them had received college-level education compared with 26% for women in the county as a whole. Over 40% of the women screened had undergone a Pap test within the previous year, 80% could name their own gynecologist or family physician, and only 6% had never had a Pap test. The findings suggest that a well-intentioned community outreach program may not reach the intended target. A prime objective must be the development of mechanisms to reach the core of susceptible women who continue to elude early cervical cancer detection.