Urbanization and the incidence of abnormalities of squamous and glandular epithelium of the cervix


  • Mathilde E. Boon M.D., Ph.D.,

    Corresponding author
    1. Leiden Cytology and Pathology Laboratory, Leiden, The Netherlands
    2. Stichting Bevolkingsonderzoek Baarmoederhalskanker Regio West, Organization for Cervical Screening West Netherlands (SBBW), Leiden, The Netherlands
    • Leiden Cytology and Pathology Laboratory, P.O. Box 16084, 2301 GB Leiden, The Netherlands
    Search for more papers by this author
    • Fax: 011-31-715123776

  • Hedda H. van Ravenswaay Claasen M.D., Ph.D.,

    1. Stichting Bevolkingsonderzoek Baarmoederhalskanker Regio West, Organization for Cervical Screening West Netherlands (SBBW), Leiden, The Netherlands
    Search for more papers by this author
  • Roelof P. van Westering M.D.,

    1. Stichting Bevolkingsonderzoek Baarmoederhalskanker Regio West, Organization for Cervical Screening West Netherlands (SBBW), Leiden, The Netherlands
    Search for more papers by this author
  • Lambrecht P. Kok Ph.D.

    1. Department of Biomedical Engineering, University of Groningen, Groningen, The Netherlands
    Search for more papers by this author



The large data bases of the Dutch cervical screening program can be exploited to establish the relation between urbanization and the incidence of abnormalities of the squamous and glandular epithelium, including mild or greater changes of the squamous and glandular epithelium of the cervix.


Six cytology laboratories in the context of the Dutch cervical screening program screened over 190,000 cervical smears. Urbanization (place of residence) data were derived from postal codes. All smears were coded with the Dutch national coding system, the Dutch national classification system KOPAC, in which squamous abnormalities are coded S4–S9, and glandular cell changes are coded G4–G9. From the scores per 1000 screened women, the relative risk (RR) of living in a large city compared with living in rural areas was calculated. To investigate a trend in incidence in relation to urbanization, the Schaafsma method was used.


Of the smears with positive cytology, mild squamous dysplasia (S4) had the highest incidence per 1000 screened women (4.32), and the lowest incidence was found for adenocarcinoma (in situ; G7/G9; RR, 0.07). The RR for urban women ranged from 1.73 for moderate squamous dysplasia (S5) to 7.55 for adenocarcinoma (in situ; G7/G9). For smears with positive cytology for both squamous and glandular abnormalities, the Schaafsma method indicated a significant positive trend.


The incidence of squamous and glandular abnormalities are maximal in women who live in a large city, which, in The Netherlands, is where there also is a population at high risk for human papillomavirus and bacterial vaginosis. Cancer (Cancer Cytopathol) 2003;99:4–8. © 2003 American Cancer Society.

The Dutch national screening program is well suited for an epidemiological study concerning urbanization and the incidence of abnormalities of squamous and glandular epithelium of the cervix for several reasons. First, all women in the selected cohort (age range, 30–60 years) who are on the well-maintained statutory list of inhabitants kept by local townships are invited individually for a smear test. Second, the smears are taken by the patient's own family physician in a familiar environment, a situation that makes participation in the program attractive for the women who are invited to have a smear test. Third, all participating laboratories use the same Dutch classification system (KOPAC) for refined coding of the changes in both the squamous epithelium and the glandular epithelium of the cervix. The relevant codes in this system are shown in Table 1. Fourth, The Netherlands is divided into nine large regions, and the cervical screening results from each of the nine regions (from six to nine laboratories) are collected into a central regional computer system, thus allowing analysis of large sets of data derived from women living in the same region. Fifth, the Dutch postal coding system, which is stored in the central computers for all screened women, allows the classification of habitats into villages, towns, and large city.

Table 1. Number of Patients and Scores per 1000 Screened Women
Codes of the KOPAC systemCytologyNo.aScore (per 1000 women)
  • KOPAC: Dutch national classification system.

  • a

    Total number of patients = 193,048.

Squamous abnormalities   
 S4Mild dysplasia, squamous8344.32
 S5Moderate dysplasia, squamous3171.64
 S6Severe dysplasia, squamous2681.39
 S7Carcinoma in situ, squamous750.39
 S8/S9Squamous cell carcinoma180.09
 S6/S9“Positive”, squamous3611.87
Glandular cell changes   
 G4Mild atypia, glandular2171.12
 G5Moderate atypia, glandular860.45
 G6Severe atypia, glandular500.26
 G7/G9Adenocarcinoma (in situ)130.07
 G6/G9Positive, glandular630.33

In this article, we present the analysis of 193,048 smears from women who participated in the cervical screening program during the period 1996–1999 in one of the nine regions, the West Region, with over 2 million inhabitants and 6 pathology laboratories offering cytology services to family physicians. The analysis showed that there is a numeric correlation between urbanization and the incidence of all grades of epithelial changes in both the squamous epithelium and the glandular epithelium of the cervix.


Women who Participated in the Screening Program

Cervical screening programs have existed in The Netherlands since 1975. The current national program, which started in 1996, is the fourth, and, like in the third program (1991–1995), the smears are prepared by family physicians. For non-Dutch readers, it is important to know that each woman living in The Netherlands has a family physician who she knows personally. Each family physician has 2000–3000 patients on file; if necessary, the physician refers a patient to the hospital for diagnosis and/or treatment. Smears also are prepared from women with clinical complaints; however, smears entered into the cervical cancer detection program, in principle, are from asymptomatic women only.

The canvassing system for cervical screening is as follows: The statutory files of the municipal units provide the names and addresses of the chosen cohorts of women. In these files, Dutch citizens and legal immigrants are included. The local health centers send the invitation letters either directly or through family physicians to the women. The letter explains the purpose of the program, and the woman is asked to go to her family physician to have a smear taken. Each woman between the ages of 30 years and 60 years receives an invitation once in 5 years to participate in the program. During the 4 years considered in this study (1996–1999), > 80% of the women between the ages of 30 years and 60 years received an invitation letter; the remaining 20% of women were approached in the year 2000.

Cervical Smear Coding

Since 1986, in The Netherlands, all smears have been coded with the national coding system (the KOPAC system). Consequently, over 15 million Dutch smears are coded with this system.1, 2 KOPAC is a Dutch acronym in which the P stands for plaveisel (Dutch for squamous), and the C stands for cilindercel (Dutch for glandular [endocervical] cell). In this report, we have used S (squamous) for the letter P in KOPAC and G (glandular) for the letter C in KOPAC. Using S codes, the changes in squamous epithelium are coded as S4 (mild squamous dysplasia), S5 (moderate squamous dysplasia), S6 (severe squamous dysplasia), S7 (carcinoma in situ of squamous cells), and S8/S9 (squamous cell carcinoma [SCC]). The codes S6/S9 are considered positive cytology in squamous cells. For the glandular abnormalities, G codes are parallel; that is, G4 (mild atypia of glandular cells), G5 (moderate atypia of glandular cells), G6 (severe atypia of glandular cells), and G7/G9 adenocarcinoma (in situ). The codes G6/G9 are considered positive cytology in glandular epithelium.

One smear can have both an S code and an G code in the 4–9 range. All 193,048 smears from the current study had both S codes and G codes, which were entered into the data base. Only codes 4–9 were used for the calculations in this report.

Urbanization and Trends in the Scores

The first four digits of the Dutch postal system can be used to stratify data according to urbanization. We divided women into three urbanization groups: Group 1, (semi)rural, which included villages and small townships with < 20,000 inhabitants (referred to as villages); Group 2, which included towns with between 20,000 and 250,000 inhabitants (referred to as towns); and Group 3, which included any metropolis with > 250,000 inhabitants (referred to as the large city). Towns like Delft, Leiden, and Gouda fell in Group 2, and there was only one large city for Group 3: The Hague.

For the S and G codes,4 the score per 1000 was calculated by dividing the number of women in the urbanization group by the total number of women in whom disease was detected, multiplied by 1000 (Table 1). The relative risk (RR) for large city dwellers compared with those who lived in villages was calculated as follows: the score per 1000 women for Group 3 was divided by the score per 1000 women for Group 1. Values > 1.0 indicate an increased risk, and a value of 1.0 indicated a decreased risk. The trend for the three urbanization groups was calculated using the statistical method described by Schaafsma.3 For the three different degrees of urbanization (i = 1, 2, 3) we used the three groups (Group 1, Group 2, and Group 3). For each population i, we found that, of ni tested individuals, there were si individuals identified with disease (or with a positive diagnosis, or with a certain infection, etc.). In principle, s1, s2, and s3 are the outcomes of the independent, stochastic variables S1, S2, and S3, with binomial distributions. We tested the null hypothesis H0: p1 = p2 = p3 against the alternative of increasing behavior, A: p1 < p2 < p3. With that objective, we tested s3s1 and studied the distribution of S3S1, given the condition S1 + S2 + S3 = s (as observed). This led to manageable formulas.1 In fact, (S1, S2, S3) has a multihypergeometric distribution, and the conditional variance of S3S1 can be computed explicitly to be s(ns) (n2 (n1 + n3) + 4 n1n3) n− 2 (n − 1) − 1, so that we only need to test ((s3n3s / n) − (s1n1s / n)) [s(ns) (n2 (n1 + n3) + 4 n1n3) n−2 (n − 1) − 1] − 1/2, where s = s1 + s2 + s3, and n = n1 + n2 + n3 in the standard normal distribution. This implies that the null hypothesis has to be rejected (i.e., the upward urbanization trend is significant) if the value of this quantity is greater than the value 1.645 (2.576) at the confidence level corresponding to P = 5% (P = 0.5%). Analogously, a downward urbanization trend can be tested to be statistically significant. The null hypothesis of the presence of a trend was tested at confidence levels of 0.05 and 0.005. The age distribution was similar in the three groups.


Number of Women and Scores per 1000

In total, there were 193,048 smears, of which, for instance, 834 smears were coded as S4 (mild squamous dysplasia), resulting in a score per 1000 women of 4.32. The number of diagnoses and scores per 1000 women are presented in Table 1. The highest score was for mild squamous dysplasia (S4), and the lowest score was for adenocarcinoma (in situ; G7/G9), with 13 coded smears resulting in a score of 0.07. All scores for S codes were higher compared with the scores for G codes. In addition, for both S codes and G codes, there was a ranking order, with the highest scores for S4 and G4 (the least severe changes) and the lowest scores for S8/S9 and G7/G9. Only relatively small numbers of smears (i.e., 117 of 1878 diagnoses from the total of 193,048 smears) were coded 4–9 for both squamous changes and glandular changes (S codes and G codes, respectively), as shown in Table 2.

Table 2. Cross Table of Patients with a Squamous S Score in the S4/S9 Range and a Simultaneous Glandular G Score in the G4/G9 Range
codes Glandular cell change codeSquamous abnormality codeTotal


There were 48,766 women in Group 1 (women living in villages), 105,512 women in Group 2 (towns), and 38,770 women in Group 3 (the large city). For all RR values (Group 3 vs. Group 1) calculated for smears that were assigned S codes and G codes, the values were much greater than 1.0. The RR was highest for G7/G9 (adenocarcinoma in situ), with a value of 7.55, and lowest for S5 (moderate squamous dysplasia), with a value of 1.73. Trends at the 0.05 and 0.005 confidence levels were established by the Schaafsma method3 and are tabulated in Table 2. Only for codes that were assigned to very few smears, such as S8/S9 (SCC) and G7/G9 (adenocarcinoma in situ), were the results of the Schaafsma method3 nonsignificant; that is, the positive trend was not significant. For all other codes, the results were statistically significant at the 0.05 confidence level; and, for codes S4, S6/S9, and G4, the results also were significant at the 0.005 confidence level. The age distribution was similar for the three urbanization groups.


Relatively high cancer incidence rates in urban populations have been observed for many decades.4–6 However, these differences seem to have diminished over the past few years, especially in the industrialized countries.7 In The Netherlands, life-style differences in the cities and urban areas have become less compared with the differences found 20 years ago, which may have influenced the incidence rates for cervical carcinoma, which is known to be highly dependent on life-style aspects, such as promiscuity and smoking.8

Our study (mainly concerning the lesser, preinvasive cervical lesions) indicates that, in the last decade, the women who are at high risk for developing cervical carcinoma still are the women who live in an urban setting. However, it should be noted that there are confounding factors. For instance, in the large city, the percentage of immigrant women (including women with one parent born outside The Netherlands) is more than three times as high as the percentage of immigrant women in the villages (32% vs. 8.9%). Another confounding factor was the percentage of women who accepted the invitation to have a smear test. This was as high as 78% in the villages, as low as 52% in the large city, and 68% in the towns.

We found high incidence rates in the large city, similar to our findings 20 years ago, when we analyzed the screening data from 6 regions.9 In the late 1970s, the positive rates in the six regions differed significantly and were dependent on the degree of urbanization of the region. In the late 1990s, we observed that, for all S codes (squamous) and G codes (glandular) that indicated epithelial changes in the RR of living in the large city (compared with living in villages) in the West Region of The Netherlands, the RR clearly was > 1.0. The urbanization trends, as established with the Schaafsma method (for definitions, see Material and Methods), proved to be statistically significant. This phenomenon cannot be explained by frequent concurrence of detected squamous and glandular changes in the smear (see Table 3). In this context, it is important to note that, in the KOPAC system, in principle, a smear may receive a positive code (i.e., S5 or G5) for both squamous and glandular abnormalities. It is noteworthy that none of the 18 smears that were coded S8/S9 and none of the 13 smears that were coded G7/G9 are shown in Table 2; thus, for the most severe lesions, no concurrences of S codes and G codes were noted. We assume that there must be more instances in which both the squamous epithelium and the glandular epithelium are affected10 but that, in the screening process, only the most prominent changes were recorded. The RR values for S codes were similar to the RR values for G codes, indicating that squamous cell carcinoma and adenocarcinoma of the cervix share risk factors.

Table 3. Urbanization and Cytology Findings
Codes of the KOPAC systemCytologyRRSignificance trenda
  • KOPAC: Dutch national classification system; RR: relative risk (of living in a large city compared with living in a village).

  • a

    Trend is defined in Materials and Methods.

Squamous abnormalities    
 S4Mild dysplasia, squamous2.39YesYes
 S5Moderate dyplasia, squamous1.73YesNo
 S6Severe dysplasia, squamous2.17YesNo
 S7Carcinoma in situ, squamous4.15YesNo
 S8/S9Squamous cell carcinoma2.52NoNo
 S6/S9Positive, squamous2.50YesYes
Glandular cell changes    
 G4Mild atypia, glandular3.43YesYes
 G5Moderate atypia, glandular4.53YesNo
 G6Severe atypia, glandular2.99NoNo
 G7/G9Adenocarcinoma (in situ)7.55NoNo
 G6/G9Positive, glandular3.49YesNo

The incidence of human papillomavirus (HPV) is highest in the large Dutch cities.11 It is a well-known fact that not only are SCC and its precursor lesions associated with HPV,12 but adenocarcinoma of the cervix also has been associated with HPV.13 Accordingly, screening metropolitan women can mean screening women who are at high risk for cervical carcinoma.

Another point of interest is the correlation between urbanization with KOPAC codes O3 (bacterial vaginosis [BV]) and O5 (Gardnerella), both indicating a disturbed vaginal ecology.14 Both also had an RR that far exceeded 1.0, and both had a positive urbanization trend with a threshold significance value as low as 0.5%.11 Gardnerella forms a part of the spectrum of BV.14–17 In this context, it is no surprise that, in our data base, the two KOPAC codes run parallel, and both are associated with urbanization.


All grades of squamous and glandular cell changes that were observed in cervical smears were found more frequently in women living in the large city, which, in The Netherlands, is where there also is a population at high risk for HPV and BV.


The authors thank Professor Schaafsma for useful discussions on hypothesis testing problems with the alternative restricted by a number of inequalities and Dr. Annelize F. Goedbloed for her literature studies concerning disturbances of the vaginal flora. They also thank all pathologists and cytologists of the laboratories in The Hague, Leiden, Delft, and Gouda for screening and diagnosing the smears.