© International Epidemiological Association 1999
International Journal of Epidemiology 1999;28:196–203
Printed in Great Britain
Diagnosis and treatment of cervical
intraepithelial neoplasia grade 3:
a registry-based study in the Romagna
region of Italy (1986–1993)
Monica Serafini,a,b Carlo Cordaro,a,c Emanuela Montanari,a,d Fabio Falcinia,e and Lauro Bucchia
Background Treatment of cervical intraepithelial neoplasia grade 3 (CIN3) is one of the most
unexplored issues of the monitoring of cervical cancer screening. We evaluated
(1) the frequency of major patterns of diagnosis and treatment of CIN3 (ICD-O
code 8070.2), (2) the determinants of hysterectomy as a first choice treatment,
and (3) the determinants of invasive cervical squamous carcinoma (CSC) detection
among CIN3 cases treated by hysterectomy.
Methods
A population-based, retrospective, descriptive (objective 1) and analytical (objectives 2 and 3) study was conducted by the Romagna Cancer Registry (Northern
Italy). Included were 316 CIN3 patients (median age, 38.5 years; range, 21–80)
registered between 1986 and 1993 and meeting one of the following eligibility
criteria: histological diagnosis of CIN3 on biopsy with any subsequent treatment,
histological diagnosis of CIN3 on conization, histological diagnosis of CIN3 on
hysterectomy with previous negative/benign (<CIN2) biopsy or conization.
Multivariate associations were evaluated by the multiple logistic regression.
Results
Of 316 patients, 264 (84%) were first diagnosed on biopsy, 39 (12%) on conization, and 13 (4%) on hysterectomy. Among the 264 patients diagnosed on biopsy,
the first choice treatment was local destructive therapy for 16 (6%), conization
for 155 (59%) and hysterectomy for 93 (35%). Age was the strongest uni/multivariate predictor of hysterectomy (the most frequent first choice treatment
.40 years) followed by adequacy of biopsy (inverse association) and place of
treatment (decreased probability for patients treated outside the area and in the
private sector). Among the 93 CIN3 patients undergoing hysterectomy, 23 (25%)
had a CSC diagnosed. Multivariate analysis showed that the probability of CSC
detection was related to adequacy of biopsy (inverse association), year of registration, and biopsy-to-treatment interval (inverse association).
Conclusion
Hysterectomy was a common treatment for patients with CIN3 on biopsy. Only
in a minority of hysterectomized patients was a CSC diagnosed. Difficulties and
inefficiencies in the biopsy and assessment procedure were found to be important
factors in the management and outcome of CIN3 patients.
Keywords
Cervical intraepithelial neoplasia grade 3, cancer registration, hysterectomy
Accepted
24 July 1998
a Romagna Cancer Registry, Luigi Pierantoni Hospital, 47100 Forlì, Italy.
b Center for Cancer Prevention, Ravenna, Italy.
c Pathology Department, Degli Infermi Hospital, Faenza, Italy.
d Oncology Service, Umberto I Hospital, Lugo, Italy.
e Medical Oncology Department, Luigi Pierantoni Hospital, Forlì, Italy.
Reprint requests to: Lauro Bucchi, Romagna Cancer Registry, Medical
Oncology Department, Luigi Pierantoni Hospital, 47100 Forlì, Italy.
State-of-the-art treatment of cervical intraepithelial neoplasia
grade 3 (CIN3) is based on conservative techniques.1–3 Conservative outpatient procedures such as cryotherapy, heat
coagulation, laser coagulation (also referred to as local destructive therapies) can be safely used for eradicating the lesions
with complete colposcopic visualization, no evidence of endocervical involvement, and no evidence of invasive spread. In
196
CIN3 IN ROMAGNA (ITALY)
selected subsets of cases, the modern outpatient electrosurgical
excision technique (loop excision) is the preferred approach.4 If
the criteria for local treatment are not met,1,3 the actual state of
disease should be assessed by cervical conization (also referred
to as cone biopsy) by cold knife, or electric loop, or laser.
Although it is commonly assumed that the radical protocols
suggested in the 1960s have changed over the years,4 CIN3
treatment is one of the most unexplored issues of the monitoring of cervical cancer screening. Studies of the epidemiology
of hysterectomy have only indirectly addressed the appropriateness of CIN3 treatment.5–7 Specific data have seldom
been reported. Substantial frequencies of hysterectomy (11–
48%) have been observed in areas of the Netherlands8 and
Denmark.9
The present population-based study was conducted by the
Romagna Cancer Registry (RTRo). The epidemiological rationale of the study was based on the view that cancer registries
should take on an active role in promoting evaluation studies
for interventions against cancer.10 This should be a priority
especially for those cancer control measures and those geographical areas that lack routine evaluation systems at the
population level. This is the case for cervical screening in
the greater part of Italy. The aims of the study were to assess (1)
the frequency of major patterns of CIN3 diagnosis, (2) the
frequency of major patterns of treatment, (3) the determinants
of hysterectomy for patients with CIN3 as diagnosed on biopsy,
and (4) the determinants of the detection of invasive cervical
squamous carcinoma (CSC) among CIN3 cases undergoing
hysterectomy.
197
For a limited number of cases, the clinicians responsible for the
initial diagnosis participated in the active completion of clinical
data at hospitals and medical centres located elsewhere and in
the classification of the patterns of diagnosis and treatment. The
items of clinical information collected included details on last
Pap smear, biopsy (also referred to as target biopsy), conization,
local destructive therapy, and hysterectomy. The most recent
Pap smear was included in evaluation only if performed ,7
months before biopsy. Pap smears taken at the time of biopsy
were included.
Case definition
As shown in Table 1, not eligible for the study were the patients
with (1) initial diagnosis of invasive CSC, (2) histological diagnosis
of adenocarcinoma, (3) clinical diagnosis of inoperable cervical
carcinoma, (4) cytology diagnosis of CIN3 without histological
confirmation. One further group of cases considered not to be
eligible to an analysis of the patterns of treatment comprised the
patients previously treated with subtotal hysterectomy.
Table 1 Eligibility characteristics of the total series of cases of cervical
intraepithelial neoplasia grade 3 (CIN3) and of invasive cervical
carcinoma considered for the study, and pattern of diagnosis of
eligible cases
Definition and subgroups
No.
Not eligible
Initial histological diagnosis of invasive CSCa
204
Histological diagnosis of adenocarcinoma
28
Clinical diagnosis of invasive cervical carcinoma n.o.s.b
13
Methods
CIN3 on Pap smear and no histological confirmation
12
Data collection
Previous subtotal hysterectomy and
The RTRo11,12 was established in 1986. The reporting system is
based on pathology and cytology reports, clinical reports, hospital discharges, and death certificates. Further incidence data
are actively collected at private clinics in the area and several
major hospitals and cancer centres located elsewhere in Italy.
The reported frequency of ‘death certificate only’ cases (0% for
the tumours of the uterine cervix)11 is compatible with satisfactory levels of completeness in the registration process. The
present study considered the cases registered for residents in
the districts of Forlì, Ravenna, Faenza, and Lugo (total female
population 269 865 on 31 December 1992). The study period
was 1986–1993 for Ravenna, Forlì, and Faenza, and 1991–1993
for Lugo (previously uncovered by registration). In those years,
cervical cancer screening in the area was implemented as a
routine practice in the regular healthcare system and not as an
organized, centralized, monitored procedure. Attendance was
based on self-referral to many medical centres. No target age
was identified. No standard protocol for the assessment and
treatment of screen-detected lesions was adopted. No
monitoring system was in operation.
The database of the RTRo was searched for any recorded
information on patients with in situ carcinoma and invasive
carcinoma of the uterine cervix (ICD-O topography code 180).
For each case identified, additional clinical details on diagnosis
and treatment were obtained by the local units of the registry
(i.e. the branch offices located in each district) from the hospitals and clinics (n = 26) mentioned in the notification sheets.
present histological diagnosis of invasive CSC
9
present histological diagnosis of CIN3
Subtotal
6
272
Eligible, excluded from analysis as not classifiable
for treatment
CIN3 on target biopsy and
no information on treatmentc
30
subsequent negative/benign follow-up biopsies
3
CIN3 on hysterectomy and no previous histological reportd
9
Subtotal
42
Eligible, included in analysis
Initial histological diagnosis of CIN3
on biopsye
264
on conizationf
39
on hysterectomyg
Subtotal
Total
13
316
630
a Cervical squamous carcinoma (ICD-O 8070.3, 8071.3, 8072.3).
b n.o.s. = not otherwise specified.
c Follow-up information not available to the original institution of diagnosis.
d In the absence of any previous histological report, information available was
considered to be incomplete.
e With information on any subsequent treatment. Cases subsequently
diagnosed with CSC on conization and/or hysterectomy were included.
f Cases subsequently diagnosed with CSC on hysterectomy were included.
g With information on previous negative/benign biopsy or conization (as an
evidence that available information was complete).
198
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Among potentially eligible CIN3 cases, excluded from analysis
were those not classifiable for treatment. These were characterized by (1) histological diagnosis of CIN3 on biopsy without
information on treatment, or (2) histological diagnosis of CIN3
on biopsy with negative follow-up biopsies, or (3) histological
diagnosis of CIN3 on hysterectomy without previous biopsies.
Included in the analysis were those cases meeting one of
the following criteria: (1) histological diagnosis of CIN3 (ICD-O
morphology code 8070.2) on biopsy with evidence of any subsequent treatment, (2) histological diagnosis of CIN3 on conization, and (3) histological diagnosis of CIN3 on hysterectomy
with evidence of previous negative/benign (<CIN2) biopsy or
conization. In other words, the analysis considered all cases
with a histological report of CIN3 as a result of a diagnosis/
treatment process apparently complete based on the assumption that (1) biopsy is a diagnostic procedure, (2) conization is
a combined diagnostic/therapeutic procedure, and (3) hysterectomy is a treatment procedure. In the potential sequence
biopsy-conization-hysterectomy, those cases initially diagnosed
as CIN3 and demonstrated thereafter to be microinvasive or
invasive CSC were included. Conization and hysterectomy were
assumed to be directly related to biopsy only if performed
within 18 months (i.e. 540 days) of it. Such a follow-up was
available for all cases.
Adequacy of biopsy
The histological reports collected did not have standard items
for classifying the adequacy of biopsy as defined as its ability to
inform about the actual state of disease (CIN3 versus CSC). The
following assumptions were made. All biopsies reported as
showing microinvasive or invasive CSC were considered technically adequate. For the reports of CIN3, biopsy was considered
inadequate if (1) formal statement of poor adequacy was reported, or (2) further biopsy was recommended by the pathologist, or (3) terminology about adequacy was ambiguous.
Terminology
The heterogeneous original terminology for reporting cytology
and histology diagnoses was recoded into CIN definitions according to accepted conversion criteria.1 Four Pap smears diagnosed
as atypical cells of undetermined significance were reviewed.
For reasons of comparability, smears originally reported as CIN2,
CIN3, moderate dysplasia, severe dysplasia, and high grade
squamous intraepithelial lesion were combined into one category (here designated as CIN2–3). All diagnoses intermediate
between two classifications were assigned to the most severe
one. Among histological diagnoses, CIN3 was considered a
separate entity. Microinvasive CSC was considered invasive. In
this study, both conization and local destructive therapy were
considered to be ‘conservative’ treatments.
Mantel-Haenszel χ2 test for trend (this was the case for age,
year of registration, Pap smear result, and biopsy-to-treatment
interval), the continuity-corrected χ2 test for heterogeneity
(residence, place of biopsy, place of treatment) and the Fisher
exact test (adequacy of biopsy). A P-value ,0.05 was considered significant. For each variable, the crude odds ratio (OR)
and the 95% confidence interval (CI) associated with each
category were computed according to Cornfield. Multivariate
associations were evaluated in a multiple logistic regression
model based on the backward stepwise selection. This procedure allowed the estimation of the strength of the association
between each independent variable and the dependent variable
taking into account the potential confounding effects of the
other independent variables. The covariates were removed
from the model if the likelihood ratio statistic based on the
maximum-likelihood estimates had a probability .0.10. Each
category of the predictor variables was contrasted with the
initial category. An adjusted OR with a 95% CI that did not
include 1.0 was considered significant.
As second research hypothesis, the association between the
independent variables and the frequency of invasive CSC detection versus any other outcome among patients undergoing
hysterectomy was evaluated. Differences in distribution by histological outcome were compared using the extended MantelHaenszel χ2 test for linear trend (for age, year of registration,
Pap smear result, and biopsy-to-treatment interval), the
continuity-corrected χ2 test for heterogeneity (residence, and
adequacy of biopsy), and the Fisher exact test (place of biopsy,
and place of treatment). All other methods were the same as for
the first endpoint.
Results
Case series and pattern of diagnosis
A total of 630 cases were initially identified (Table 1). Among
these, 272 (median age, 60.5 years; range, 23–91) were definitely not eligible. Of the remaining 358 potentially eligible
cases, 42 (median age, 38 years; range, 23–89) were excluded
from analysis as not classifiable for treatment. Note that 30 such
cases were accounted for by follow-up failures. Included in the
analysis were 316 cases (median age, 38.5 years; range, 21–80;
88% of the 358 eligible cases).
Based on the initial or first available histology report of CIN3
in the potential sequence biopsy-conization-hysterectomy, 264
(84%) of these were diagnosed on biopsy, 39 (12%) on conization,
and 13 (4%) on hysterectomy. No patient entered in the study
with a diagnosis of CIN3 as obtained on local loop excision. The
diagnosis process included the Pap smear in 272/316 cases
(86%).
Pattern of treatment
Data analysis
Age, district of residence, year of registration, Pap smear, place
and adequacy of biopsy, biopsy-to-treatment interval, and place
of treatment were (or were treated as) categorical variables.
As a first research hypothesis, the association between these
independent variables and the frequency of hysterectomy
versus any conservative therapy among cases with CIN3 on
biopsy as a dependent variable was evaluated. Differences in
distribution by treatment were compared using the extended
This was analysed among the 264 patients who were diagnosed
with CIN3 on biopsy (Table 2). The first choice treatment was
local destructive therapy for 16/264 cases (6%), conization for
155 cases (59%) and hysterectomy for 93 (35%). Among the 16
cases undergoing local treatment techniques, 11 were considered to be cured. The remaining five cases also underwent
conization (n = 4) or hysterectomy (n = 1).
Among the 155 cases initially treated by conization, the cold
knife was used in 122 (79%), the electric loop in eight, and
CIN3 IN ROMAGNA (ITALY)
199
Table 2 Patterns of treatment for 264 cases diagnosed with cervical intraepithelial neoplasia grade 3 (CIN3) on biopsy
Hysterectomy
Local destructive therapy
Conization
Undone
N/Ba to CIN2
CIN3
Invasive
Total
Yesb
Undone
11
–
1
–
12
N/B to CIN2
1
–
–
–
1
CIN3
3
–
–
–
3
Invasive
–
–
–
–
–
Subtotal
15
–
1
–
16
No
Undone
NAc
9
61
23
93
N/B to CIN2
14
–
–
–
14
124
CIN3
116
2
4
2
Invasive
11d
2
2
2
17
Subtotal
141
13
67
27
248
156
13
68
27
264
Total
a N/B = negative/benign.
b Heat coagulation n = 11, laser coagulation n = 4, unknown type n = 1.
c NA = not applicable according to eligibility criteria.
d Microinvasive cervical squamous carcinoma in 8/11 cases and negative cone margins in 11/11.
the laser technique in nine (unknown type n = 16). Including
the four cases previously treated by local therapy (upper section
of Table 2), a total of 159 cases underwent conization. In 15 of
these (9%) the diagnosis of CIN3 was not confirmed. In 127/
159 cases (80%) the lesion was confirmed; eight of these cases
underwent hysterectomy and two invasive CSC cases were detected. In 17/159 conizations (11%) the resected tissue showed
an invasive lesion.
Among the 93 patients undergoing hysterectomy as a first
choice treatment, radical hysterectomy accounted for 63% of
cases with the following age-specific frequencies: 23% at age
20–39, 58% at age 40–49, 75% at age 50–59, and 85% among
older patients (P = 0.0000, test for trend).
The median interval between biopsy and conization was 40
days (range, 6–514). The median interval between biopsy and
hysterectomy was 44 days (range, 5–539); the median interval
between biopsy and hysterectomy for the selected group of
patients not undergoing conization was 42 days (range, 5–539).
Determinants of hysterectomy
The decision to perform hysterectomy versus local therapy and/
or conization was analysed for 249/264 patients (15 patients
undergoing local therapy and/or conization as well as hysterectomy were excluded). There were 156/249 patients (63%)
conservatively treated and 93 (37%) treated by hysterectomy as
a unique treatment modality. Table 3 addresses the relationship
between hysterectomy and the set of available variables.
Patient’s age was the strongest univariate determinant of hysterectomy followed by the inadequacy of biopsy and the place of
treatment. Hysterectomy was the most common type of treatment above 40 years of age. Multivariate analysis confirmed
these associations.
Determinants of invasive CSC detection
Among the 93 CIN3 patients undergoing hysterectomy, 23
(25%) had an invasive CSC diagnosed (Table 4). Though greater
among women .60 years, the prevalence of CSC did not show
a significant tendency to increase linearly with age. The probability of invasive CSC detection showed a significant relationship with inadequacy of biopsy, which was confirmed in the
logistic regression model. It is worth noting, however, that most
(14/23) of invasive cancers detected among hysterectomized
women had had a CIN3 report as based on apparently adequate
biopsy. Multivariate analysis revealed also a decreased OR in the
years 1990–1991 and a weak association with the time interval
from biopsy to treatment. A decreased OR was associated with
all intervals .20 days, with a specific significance for those
varying between 41 and 60 days.
Discussion
The present study confirms that a cancer registry can participate
directly in the monitoring and quality control of cervical screening.10 Important clinical information regarding the last Pap smear,
biopsy, conization, local destructive therapy, and hysterectomy
were obtained from the database of the RTRo as well as many
hospitals and clinics involved in the registration process. As
many as 88% of eligible cases of CIN3 registered on a population basis were analysed. This suggests that the results were free
of major selection biases. The most evident limitation of the
study was the unavailability of the colposcopy diagnoses. The
colposcopy reports were traced for a minority of cases and, as
expected, were found to be incomplete and poorly standardized.
However, some major findings of the study, such as those
concerning the biopsy procedure, provided valuable though
indirect information on the role of colposcopy in the diagnosis
and treatment of CIN3.
Although the long-term outcomes remain the essential endpoint of a formal evaluation of the adequacy and appropriateness of CIN3 treatments,13 the cross-sectional data reported
here suggest that the quality of the procedures needs to be
monitored continuously.
200
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 3 Univariate and multivariate analysis of the determinants of hysterectomy among 249 cases with cervical intraepithelial neoplasia grade 3
(CIN3) on biopsy
Hysterectomya
Determinant
No (n = 156)
Univariate analysis
Yes (%) (n = 93)
ORc
(95% CI)
Multivariate analysisb
OR
(95% CI)
Patient’s age (years)
20–39
123
18 (13)
1.00
(referent)
1.00
(referent)
40–49
25
27 (52)
7.38
(3.33–16.51)
7.66
(3.47–16.89)
50–59
6
28 (82)
31.89
(10.64–100.74)
36.92
(12.34–110.46)
60+
2
20 (91)
68.33
(13.56–464.34)
75.93
(15.22–378.89)
P , 0.0001d
P , 0.0001
Residence
District A
53
36 (40)
1.00
(referent)
District B
10
4 (29)
0.59
(0.14–2.27)
District C
41
20 (33)
0.72
(0.34–1.50)
District D
52
33 (39)
0.93
(0.49–1.80)
Variable removed
P = NSe
Year of registration
1986–1987
30
17(36)
1.00
(referent)
1988–1989
35
22 (39)
1.11
(0.46–2.67)
1990–1991
40
28 (41)
1.24
(0.54–2.86)
1992–1993
51
26 (34)
0.90
(0.39–2.06)
Variable removed
P = NSd
Pap smear
Undone/N/Bf
32
22 (41)
1.00
(referent)
CIN1
75
27 (26)
0.52
(0.25–1.12)
CIN2–3
41
32 (44)
1.14
(0.52–2.47)
8
12 (60)
2.18
(0.68–7.09)
Carcinoma
Variable removed
P = NSd
Biopsy
CIN3, adequate
CIN3, inadequate
152
74 (33)
1.00
(referent)
1.00
(referent)
4
19 (83)
9.76
(2.99–35.26)
5.77
(1.40–23.83)
P , 0.0001
P = 0.0155
Place of biopsy
NHSg/Romagna
NHS/elsewhere and private sector
141
83 (37)
1.00
(referent)
15
10 (40)
1.13
(0.45–2.83)
Variable removed
P = NS
Biopsy-to-treatment interval (days)
5–20
17
12 (41)
1.00
(referent)
1.00
(referent)
21–40
74
36 (33)
0.69
(0.28–1.73)
0.53
(0.16–1.72)
41–60
42
23 (35)
0.78
(0.29–2.09)
0.65
(0.19–2.25)
61–540
23
22 (49)
1.36
(0.48–3.87)
1.78
(0.50–6.32)
P = NSd
P = 0.0738
Place of treatment
NHS/Romagna
NHS/elsewhere and private sector
104
74 (42)
1.00
(referent)
1.00
(referent)
52
19 (27)
0.51
(0.27–0.98)
0.42
(0.19–0.94)
P = 0.0417
P = 0.0340
a The Table considers those cases treated by local therapy and/or conization alone (‘No’ heading) and those undergoing hysterectomy alone (‘Yes’ heading).
b Multiple logistic regression (backward stepwise selection). The predictor variables were removed from the model if the probability of the likelihood ratio
statistic based on the maximum-likelihood estimates was . 0.10.
c Odds ratio estimate of the probability of undergoing hysterectomy.
d Test for trend.
e Not significant (P . 0.05).
f N/B = negative/benign.
g NHS = National Health Service.
CIN3 IN ROMAGNA (ITALY)
201
Table 4 Univariate and multivariate analysis of the determinants of invasive squamous carcinoma detection on hysterectomy among 93 cases
with CIN3 on biopsy
Invasive CSCa
Univariate analysis
No (n = 70)
Yes (%) (n = 23)
ORc
20–39
14
4 (22)
1.00
(referent)
40–49
24
3 (11)
0.44
(0.06–2.82)
50–59
20
8 (29)
1.40
(0.29–6.95)
60+
12
8 (40)
2.33
(0.46–12.42)
Determinant
(95% CI)
Multivariate analysisb
OR
(95% CI)
Patient’s age (years)
Variable removed
P = NSd e
Residence
District A
25
11 (31)
1.00
(referent)
District B
2
2 (50)
2.27
(0.19–27.1)
District C
17
3 (15)
0.40
(0.08–1.91)
District D
26
7 (21)
0.61
(0.18–2.07)
Variable removed
P = NS
Year of registration
1986–1987
11
6 (35)
1.00
(referent)
1.00
(referent)
1988–1989
19
3 (14)
0.29
(0.04–1.71)
0.24
(0.04–1.56)
1990–1991
25
3 (11)
0.22
(0.03–1.27)
0.16
(0.03–0.93)
1992–1993
15
11 (42)
1.34
(0.32–5.75)
2.28
(0.51–27.70)
P = NSd
P = 0.0127
Pap smear
Undone/N/Bf
15
7 (32)
1.00
(referent)
CIN1
23
4 (15)
0.37
(0.07–1.78)
CIN2–3
25
7 (22)
0.60
(0.15–2.41)
7
5 (42)
1.53
(0.28–8.38)
Carcinoma
Variable removed
P = NSd
Biopsy
CIN3, adequate
60
14 (19)
1.00
(referent)
1.00
(referent)
CIN3, inadequate
10
9 (47)
3.86
(1.17–12.9)
6.80
(1.67–27.70)
P = 0.0235
P = 0.0075
Place of biopsy
NHSg/Romagna
NHS/elsewhere and private sector
62
21 (25)
1.00
(referent)
8
2 (20)
0.74
(0.10–4.25)
Variable removed
P = NS
Biopsy-to-treatment interval (days)
5–20
7
5 (42)
1.00
(referent)
1.00
(referent)
21–40
25
11 (31)
0.62
(0.13–2.89)
0.49
(0.10–2.44)
41–60
22
1 (4)
0.06
(0.00–0.76)
0.02
(0.00–0.39)
61–540
16
6 (27)
0.52
(0.09–2.92)
0.23
(0.04–1.36)
P = NSd
P = 0.0541
Place of treatment
NHS/Romagna
55
19 (26)
1.00
(referent)
NHS/elsewhere and private sector
15
4 (21)
0.77
(0.19–2.93)
Variable removed
P = NS
The Table considers the cases reported under the ‘Yes’ heading in Table 3.
a Cervical squamous carcinoma.
b Multiple logistic regression (backward stepwise selection). The predictor variables were removed from the model if the probability of the likelihood ratio
statistic based on the maximum-likelihood estimates was . 0.10.
c Odds ratio estimate of the probability of being diagnosed with invasive CSC on hysterectomy.
d Test for trend.
e Not significant (P . 0.05).
f Negative/benign.
g National Health Service.
202
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
In this study, the very limited role of local therapies was
coupled with a considerable frequency of hysterectomy. As many
as 35% of cases reported as CIN3 on biopsy underwent hysterectomy as a first choice treatment (Table 2). The strong uni/
multivariate relation to age was expected. Hysterectomy was
the most frequent treatment .40 years (Table 3). Both analyses
showed that inadequacy of biopsy was the second most important predictor of the decision for hysterectomy. This was performed in more than 80% such cases. Both analyses showed
also a decreased probability of hysterectomy for patients treated
in public institutions located outside the area and in the private
sector, suggesting that the patient’s decision to move and the
preference for conservative therapies were related as linked
effects of an active health behaviour. This contrasts with the
belief (which is widespread in Italy) that patients who refuse
to complete the recommended protocol for assessment and
therapy at the public health care facilities in the location of
residence are at increased risk of radical treatment.
Twenty-five per cent of CIN3 patients undergoing hysterectomy were demonstrated to have an invasive disease. The
strongest independent determinant of invasive CSC detection
(Table 4) was the inadequacy of biopsy. This had a 47% predictive value for the presence of cancer. The association with
the year of registration was unexpected. Since the rate of
invasive CSC detection among patients diagnosed with CIN3 on
biopsy is inversely related to accuracy of sampling and sensitivity of histological examination and positively related to the
grade of clinical suspicion on colposcopy, the observation was
the likely effect of some substantial (though unrecognised)
variation in these factors over time. The relative frequency of
microinvasive CSC (more difficult to sample successfully) was
stable over the years (data not shown). As regards the decreased
OR associated with biopsy-to-treatment intervals .20 days, it
clearly appears that the most rapid referrals for hysterectomy
were suggested by a stronger clinic impression based on
colposcopy.
The results of the two multivariate analyses should be carefully considered. The strong association with age in the first
model was likely to reflect the role of ‘...a long tradition of dogmatic teaching and learning in gynecologic surgery’14 which is
a major cause of the unnecessary hysterectomies. However, age
was also included in both models as a correlate of those clinical
variables that were unavailable to the RTRo (such as parity
and patient’s demand)5,7 or incomplete and non-standard (such
as the diagnosis of concomitant gynaecological disorders).6,14
The aim was to obtain a reasonable estimate of the independent
role of some components of the assessment procedure in determining the choice and the outcome of hysterectomy. In fact, the
results point to problems concerning the adequacy of biopsy as
well as the ambiguity of the histological reporting as important
factors in the decision-making process. The high rate of invasive
CSC detection among total hysterectomies for CIN3 (25%) and
the twofold greater frequency among cases with inadequate
biopsy (47%) do not demonstrate that the radical treatment was
appropriate. Such data suggest only that the clinicians’ confidence in the biopsy result had to be poor. Apparently, this lends
some support to the widespread tendency of Italian gynaecologists to rely heavily upon the colposcopy findings. Unfortunately, two observations suggest indirectly that the accuracy of
the colposcopic impression was limited. First, the fact that the
majority of CIN3 patients undergoing hysterectomy had not an
invasive cancer diagnosed suggests that the colposcopic impression may have led to an overestimate of the actual state of
those diseases. Second, the fact that most patients with invasive
cancers detected among hysterectomized women had a CIN3
report as based on apparently adequate biopsies suggests that
the colposcopy findings failed to indicate the most appropriate
site for sampling. In summary, an improvement in the histological reporting terminology, a reduction in the rate of inadequate
biopsies, and an increase in the quality of the colposcopy
diagnosis (with a more accurate selection of the sampling site
and thus a greater sensitivity of biopsy for CSC) appear to be
essential prerequisites for improving the selection of patients for
conservative therapies and reducing the frequency of unnecessary hysterectomies. Otherwise, hysterectomy will probably
continue to be considered as a safe approach to the treatment of
CIN3.
One further aspect needs to be pointed out. In the large study
of Boyes et al.,13 patients treated by conization showed no
excess risk of subsequent invasive cancer compared with those
treated by hysterectomy. In fact, patients in that series
were being followed-up regularly. A less comprehensive
surveillance might have led to a less favourable outcome.15 The
implementation of fail-safe systems for ensuring that patients
undergoing conization are being closely followed-up by repeat
smears and/or colposcopy is an essential component of the
treatment protocol. In Romagna, such systems are lacking. This
may have further eroded the clinician’s confidence in the conservative treatments. Other findings confirmed that problems
with follow-up were a common correlate of cervical screening
in the area. Certain groups of cases not eligible to the study
(Table 1: patients with CIN3 on Pap smear and no histological
confirmation) or eligible but excluded from analysis (Table 1:
patients with CIN3 on biopsy and no information on treatment)
as well as some unacceptable delays between biopsy and treatment (Table 3) were compatible with serious inefficiencies in
the follow-up process.15
The diagnosis and treatment of CIN3 constitute a medical
procedure of considerable complexity. Despite ‘...the lack of even
a minimally adequate clinical research and literature...’ about
the outcomes of hysterectomy and the related quality of life,14
we do not consider the use of conservative therapies for patients
with CIN3 as a ‘question of principle’. Our data suggest that
the decision for hysterectomy may reflect difficulties and
inefficiencies in specific components of the assessment procedure. A generalized improvement in methods and the implementation of real-time surveillance and evaluation16 appear to
be the most practical approaches to this serious public health
problem.
Acknowledgements
The authors thank Rosa Vattiato, Stefania Giorgetti, Silvia
Salvatore, Carlo Milandri (Romagna Cancer Registry), Barbara
Piantini (Medical Oncology Department, Luigi Pierantoni
Hospital, Forlì), Flavia Foglietta (Oncology Service, Degli Infermi
Hospital, Faenza), Patrizia Schincaglia (Center for Cancer
Prevention, Ravenna), and Giorgio Cruciani (Oncology Service,
Umberto I Hospital, Lugo) for assistance.
CIN3 IN ROMAGNA (ITALY)
References
1 Coleman D, Day NE, Douglas G et al. European guidelines for quality
assurance in cervical cancer screening. Eur J Cancer 1993;Suppl.4:
S1–S38.
2 Pontén J, Adami H-O, Bergstrom R et al. Strategies for global control
of cervical cancer. Int J Cancer 1995;60:1–26.
3 Cannistra SA, Niloff JM. Cancer of the uterine cervix. N Engl J Med
1996;334:1030–38.
4 Ferenczy A. Management of patients with high grade squamous
intraepithelial lesions. Cancer 1995;76:1928–33.
5 Vessey MP, Villard-Mackintosh L, McPherson K, Coulter A, Yeates D.
The epidemiology of hysterectomy: findings in a large cohort study.
Br J Obstet Gynecol 1992;99:402–07.
6 Reiter RC, Gambone JC, Lench JB. Appropriateness of hysterectomies
performed for multiple preoperative indications. Obstet Gynecol
1992;80:902–05.
7 Coulter A, McPherson K. Socioeconomic variations in the use of
common surgical operations. Br Med J 1995;291:183–87.
203
9 Lynge E. Screening for cancer of the cervix uteri. World J Surg
1989;18:71–78.
10 Armstrong BK. The role of the cancer registry in cancer control. Cancer
Causes Control 1992;3:569–79.
11 Parkin DM, Whelan SL, Ferlay J et al. Cancer Incidence in Five
Continents. Vol. VII. Lyon: International Agency for Research on
Cancer, 1997.
12 Zanetti R, Crosignani P. Cancer in Italy: Incidence Data from Cancer
Registries, 1983–1987. Turin: Lega Italiana Tumori, Associazione Italiana
di Epidemiologia, 1992.
13 Boyes DA, Worth AJ, Fidler HK. The results of treatment of 4389 cases
of preclinical cervical squamous carcinoma. J Obstet Gynecol Br
Commonwealth 1970;77:769–80.
14 Haas ST. Making a decision to perform a hysterectomy. Clin Obstet
Gynecol 1992;35:865–70.
15 Chamberlain J. Reasons that some screening programmes fail to
control cervical cancer. In: Hakama M, Miller AB, Day NE (eds).
Screening for Cancer of the Uterine Cervix. Lyon: International Agency for
Research on Cancer, 1986, pp.161–68.
8 Van Ballegooijen M, Koopmanschap MA, Van Oortmarssen GJ et al.
16 Dick FJ, Murphy FA, Murphy JK et al. Effects of surveillance on the
Diagnostic and treatment procedures induced by cervical cancer
screening. Eur J Cancer 1990;26:941–45.
number of hysterectomies in the province of Saskatchewan. N Engl J
Med 1977;296:1326–28.