LL'' .
Medical Hypotheses 20: 53-63, 1986
PERIOD AND COHORT TRENDS IN MORTALITY FROM CANCERS
OF THE UTERUS
IN ENGLAND AND WALES. A NEW THEORY OF CHANGING
GENE FREQUENCIES.
Phil.ip R.J. Burch,
Department of Medical Physics, University of
Leeds,
The General Infirmary, Leeds L51 3EX, England.
ABSTRACT
Analysis of the age distribution of cancers of the
uterus reveals
three clearly distinctive types: early- and
late-onset cancers of the
cervix and one form of cancer of the corpus.
England and Wales
mortality data are available for the three types
combined from 1911;
cancers of the cervix and corpus have been listed
separately from 1951.
In the highest age groups, 60-64yr and above,
death rates fell fairly
steadily from 1921-25 to 1981-82 whereas cigarette
smoking - an alleged
cause of cervical cancer - rose over much of that
period. In the
younger age groups, 55-59yr and below, the more
complicated temporal
trends include two marked cohort dips. The first
dip affected cohorts
born around 1911; this was followed by a rise up
to the 1921 central
cohort, then a steep fall to the 1936 central
cohort and, finally, a
further rise up to the 1946 cohort (at least). The
period trends are
probably caused by a decline in the impact of one
or more precipitators
of cervical cancer, such as the papilloma virus.
Cohort shifts are
attributed to autoaggressive attacks on parental
germ cells, the zygote,
or early-stage embryo that induce gene change to
affect the frequency of
women genetically predisposed to early-onset
cervical cancer. Attacks
are also triggered by one or more precipitating
agents with a temporal,
and possibly a seasonal, dependence. Implications
of this hypothesis of
cohort shifts are discussed.
INTRODUCTION
Changes with time in recorded mortality from
malignant diseases
display many enigmatic features (1.2). Of special
interest to medical
biologists are the well recognized but unexplained
cohort and period
trends in mortality from cancers of the uterus.
especially of the
uterine cervix, in England and Wales (2-6). (When
changes in the
recorded levels of a disease can be related to the
year(s) of birth
these are called 'cohort' changes; when they occur
simultaneously in
53
CZ:)
1 _~
BATCo document for Province of BritiSh Columbia 23
April 1999


different age groups, and hence different cohorts,
they are called
. peri od' changes. Temporal trends are a
combination of cohort, period
and other changes.)
Cohort trends in mortality statistics are
conveniently studied by
following persons born over a five-year period and
tracing their
age-specific death rates for the disease under
study by the customary
five-year age groups to as high an age as
possible. Period trends may
be studied by considering age-specific death-rates
by calendar year or,
say, by five-year periods. When both cohort and
period changes are
present they can be difficult or impossible to
disentangle, particularly
when age-dependent errors of death certification
are present, or when
more than one distinctive disease is included
under a broad diagnostic
category. Both complications arise in connexion
with cervical cancer and
hence the two types of trend cannot be fully
resolved.
Two conspicuous dips in cohort trends are seen in
the post-1911
England and Wales mortality data for malignant
neoplasms of the uterus,
a feature that appears to be unique to this cancer
(2). Period trends
pose no novel problem of interpretation (1) and my
main purpose here is
to propose a new hypothesis of the cohort trends.
Data are interpreted
in terms of my unified theory of growth and
age-dependent autoaggressive
disease (1, 7. 8).
AGE PATTERNS OF UTERINE CANCERS
An appreciation of the broad characteristics of
the
age-distributions of cancers of the uterine cervix
and corpus is a
necessary preliminary to the interpretation of
temporal trends. For an
unbiased assessment of the kinetics of
pathogenesis accurate statistics
would be needed for a population in which temporal
trends had shown no
appreciable variation over a period of time at
least as long as the
human lifespan (7). It is doubtful whether such
statistics are
available but for present purposes the
age-specific onset rates for
cancers of the cervix and corpus uteri, England
and Wales. 1976-80,
suffice to make an important distinction (Fig.1).
They show a bimodal
distribution for the onset of cervical cancer -
left panel - and a
unimodal distribution for cancer of the corpus -
right panel. Similar
patterns have been illustrated previously in
statistics for Denmark,
1953-62; Sweden, 1959-68; and the Gerumn
Democratic Republic, 1964-66
(I).
The bimodal distribution for cervical cancer can
be interpreted as
the sum of two 'standard' unimodal distributions
(1) with the genetics
a s well as the kinetics of the 'early-onset'
group differing from those
f or the .1 at e-onset' group. Data for the
'early-onset' group were
fitted (1) to the following stochastic equation:
dp/dt - (14kltexp(-kt 2MI-exp(-kt 2)]6 (1).
where dP/dt is the age-specific initiation-rate at
age t , assuming an
average latent period of 2.5yr between the
completion of initiation and
the diagnosis of onset; k is a kinetic constant
and 51 is the proportion
54
BATCo document for Province of BritiSh Columbia 23
April 1999


11, .
of the fediale population at risk to 'early-onset'
cervical cancer. The
age-patterns for 'late-onset' cervical cancer and
for cancer of the
corpus are also described by an equation of the
same general form as
eq.(1). but with a much lower k value (common to
both these uterine
cancers) and with S values distinctive to each
site.
CUM OF CERVIX AM CORIPM UTERI
AGf-iNCj",VCZ. ENGLANO AND IMILES. 1066-Y0
A cc---
cf-- cz-,
AGE-SKCIFIC
INCUMM.
CAMS
PER WCHU
PEN no
10-1
all
IV
A
IV 0 1 1 1 1 1 Jill I I I 1 11111
20 40 80 20 40 80
LST0,1111 AGE AT "TIATOR (.~ I ( k - 2 - S r I
Fig.1 Age-specific incidence (cases per woman per
year) versus
estimated initiation age for cancers of the
uterine cervix 0,
early- and late-onset types (left panel); of
cancers of the
corpus (A), and of the cervix and corpus combined
0 (right
panel) in England and Wales. 1966-1970 (9).
(Log-log scales.)
The interval between initiation and onset is
assumed to be 2.5
years. Curves shown in the right hand panel are
theoretical
(see text).
55
C:D
C71*1
BATC,o document for Province of BritiSh Columbia
23 April 1999


Equation (1) - for the Initiation of an
autoaggressive- disease by
seven independent forbidden clones each of which
Is initiated by two
somatic mutations (1,7) - probably describes
fairly accurately the
kinetics of the 'late-onset' form of cervical
cancer and cancers of the
corpus but. because the equation is strictly
applicable only when S Is
constant with time, it might not describe reliably
the kinetics of
early-onset cervical cancer. The important
inference to be drawn here
from the bimodal distribution for the age pattern
of cervical cancer is
that two distinctive groups of women are at risk
to the disease and that
one or taore of the genes that predispose to the
'early-onset' form
di fferfrom those that predispose to the
'late-onset' form (1). For the
.early-onset' group the initiation mode occurs at
about 47 years of age;
for the 'late-onset' group the corresponding modal
age is at about 67.
From Fig.1, right panel, it will be seen that the
contribution of
.early-onset' cases of cancer of the cervix
predominates during the
early years but at an Initiation age of about 53yr
the contributions -of
'early' and 'late' forms (combined cervix and
corpus) are equal; beyond
that age the 'late-onset' group predominates.
PERIOD AND COHORT TRENDS
Mortality statistics for cancers of the uterine
cervix in England
and Wales go back only as far as 1951 but data for
cancers of the uterus
(cervix and corpus combined) are available from
1911 onwards (9-12).
Ueaths from cancer of the cervix in England and
Wales constitute,
overall, about two thirds of those of the uterus,
but in the younger age
groups, 50-54yr and below, they form a higher
proportion, about 80 per
cent. Data for age-specific mortality from cancers
of the uterus for the
5-year age groups 25-29yr up to 65-69yr are
plotted in Fig-2 against
time, by 5-year periods -from 1911-15 to 1976-80
and for the combined
years 1981-82. Temporal trends by age group are
shown by continuous
lines; cohorts are represented by broken lines
which form an unusual and
interesting pattern.
In the age group, 65-69yr (and in the higher age
groups not
illustrated here) a fairly steady decline in rates
occurred after
1921-25 but in age groups 55-59yr and below, more
complicated changes
are seen. At 25-29yr, two very pronounced dips
were recorded around
1936-40 and 1961-65. 8y following cohorts through
to higher age groups
the influence of the first dip can be discerned up
to 65-69yr in spite
of evident dilution by 'late' group deaths. The
earliest cohort
delineated by broken lines in Fig.2 had a mean age
of 27.5yr in the
middle of the year 1933 and hence was born, on the
average, around the
beginning of the year 1906. For convenience I call
this the 'cl.906
cohort'. (Rates for the final period, 1981-82, are
for two years only
and hence the broken lines joining 1976-80 to
1981-82 rates do not
represent strict cohort trends.)
56
c::>
%,C-)
BATCo document for Province of BritiSh Columbia 23
April 1999


MAUGNANT NEOPLASMS OF UTERUS, ENGLAND & WALES.
1911-15 to 1901- 82
rRZAVS /N AGE-SPECIPIC O&ArN AdIrES
10-1
DEANS
PER
VIM
Ll
WOMAN
PER
YEAR
(kq WOO
10-3
CALENOAA PEA100
Fig. 2. Age-specific death rates for cancers of
the uterus
(corpus and cervix) in England and Wales by 5-year
periods
from 1911-15 to 1976-80 and for 1981-82 (9-12).
Temporal
trends are shown by continuous lines, cohorts by
broken lines.
Numbers of deaths are shown alongside points for
one cohort.
To illustrate cohort changes using larye numbers
of deaths I have
calculated cumulative mortality over the age
groups 20-24yr to 40-44yr
for cohorts from c1891 up to c1936. For the c1941
cohort, cumulative
mortality from 20-24yr to 35-39yr is compared with
the corresponding sum
for the c1936 cohort and for the subsequent cohort
c1946 a similar
approximation is made, with a further truncation.
The results of these
suaimations are shown in Fig.3. Error bars are
:tS.E. based simply on
the total number of deaths in the cohort and
Poisson statistics.
This figure reveals more clearly, with adequate
statistics, the
very pronounced dips corresponding to minima for
the 1:1911 and c1936
cohorts. These modulated cohort changes are
largely or wholly confined
to the early-onset form of cervical cancer.
Whether the downward trends
in mortality from the late-onset form of cervical
cancer are caused by
steady period or cohort shifts - or both - is not
certain. (Mortality
from cancer of the corpus over the period 1951 to
1970 did not change
appreciably (11).) However, the fall . in overall
rates (Fig-2) from
57
rQ
CIO
0-
BATCo document for Province of BritiSh Columbia 23
April 1999
IM-3 WS-31 WHI 194-9 10-a MG-71 IWO
I I I ' I I I ' I ' I ' I I I
III)-is IM-2s 1931-15 PAI-15 "5"i ISI-115 1911-13
1211-12


1926-30 to 1931-35, apparent in all age groups,
suggests a supervening
period trend independent of cohort effects because
the c1891 and c1896
cohort rises. as reflected in the 25-29 and
30-34yr age groups, would
lead us to expect increases, rather than the
observed decreases, in the
35-39 and 40-44yr age groups over the period
1926-30 to 1931-35.
011-
Cancer of tll~e Uterus
CavoRr AfORrALIrY. C14GLAOW0 AND WALES
ZO-,?4 ft 3]0 -
40-44y,
290 -
Z10-
170-
0191 0556 cISGI CISK CIIII c1116 cIM 0926 0931
0136 cl%l 094
CONCRr MID MAR
Fig. 3. Cumulative mortality from cancers of the
uterus
(corpus and cervix), over the age-range 20-24 to
40-44yr, by
cohort from c1891 to c1936 (see text for
definitions). Cohort
c1941 is based on cumulative mortality over the
age range
20-24 to 35-39yr; and cohort c1946 for 20-24 to
30-34yr.
INTERPRETATION OF TRENDS
If the period trends are real and independent of
cohort shifts,
they present no novel problems of interpretation
because they resemble
those associated with acute infectious diseases
(1), ischaemic heart
disease (13, 14) and oesophayeal cancer (15).
Period trends - with the
same proportional effect on all age groups for a
genetically homoyeneous
disease - can readily (only ) be explained by
corresponding trends in
the impact of a precipitator of disease, such as
an allergen or
infective microorganism (1). Hence, the steady
fall in rates (Fig.2)
affecting both early- and, conspicuously,
late-onset forms of cervical
cancer can plausibly be attributed to the
declining impact of one or
more widespread precipitators of the disease, such
as herpesvirus or
58
C:D
BATCo document for Province of BritiSh Columbia 23
April 1999


papillomvirus (16, 17), or an allergen. Cigarette
consumption by women
cannot be held responsible because it rose over a
long period while
mortality fell. (Nevertheless, the association
observed between smoking
and cervical cancer has been widely interpreted in
causal terms - see
Greenberg et at. and references in their paper
(18).) Neither can the
fall be attributed to earlier detection and
improved treatment, which
have had little impact on survival (2).
Furthermore, Alderson and
Donnan have shown that adjustments for the
increasing prevalence of
hysterectomy had little effect on mortality rates
up to 1975 (19).
The cohort changes - down, up, down, up - demand a
separate
explanation. In formal terms they correspond to
fluctuations in S. the
proportion of the cohort of women at risk, at
birth, to early-onset
cervical cancer. Where no precipitation of a
disease is involved
S defines the proportion of a population that is
genetical ly-predisposed
to a disorder; when the disease is precipitated by
an extrinsic agent
the parameter S defi nes the proportion that is
both
genetical ly-predisposed and invaded
pathogenically by the precipitator.
The cohort changes Illustrated in Figs.2 and 3
point to large
fluctuations in the frequency of one or more
predisposing genes. It has
been widely assumed, however, that changes in gene
frequency occur very
slowly and that spontaneous mutation rates in man
are typically of the
order of 10-5 per gene, per gamete, per
generation. A novel mechanism
is therefore required to explain an unusually
rapid up and down change
in gene frequency.
According to my unified theory of growth and
age-dependent
disorders all target tissues are potentially
vulnerable to
autoaggressive attack, given the appropriate
genetic predisposition. An
attack necessitates the formation through random
somatic mutation in a
growth control stem cell, of a pathogenic
forbidden clone synthesizing
.mutant mitotic control proteins' (m14CPs)
complementary to target cell
'tissue coding factors' (TCFs). There is no reason
to suppose that male
and female germ cells, the zygote, or even the
early-stage embryo, are
xempt from this generalization. Indeed, the gross
chromosomal
bnormalties of trisomy 21 and Klinefelter's
syndrome have been
:
attributed to an autoaggressive attack on parental
(mainly or wholly
female ) germ cells causing non-disjunction; the
frequencies of these
conditions per live birth, in relation to maternal
age. conform
satisfactorily to the statistics of 'autoaggressve
disorders' (20, 21).
1 have also suggested that one allele at many
MCP-TCF genes
corresponds to transcription along one strand of
DNA and that the
alternative allele corresponds to transcription
along the complementary,
anti-parallel strand (I). Initiating somatic
mutations in
autoaggressive disease might entail a spontaneous
switch in
transcription from one strand of an MCP gene to
its complementary strand
-a DNA 'strand -swi t chi ng' event (1, 22). This
theory predicts that
such genes should be quasi -pal i ndromes; many
such genes have since been
identified in the human genome. We also predict a
specific
stereochemical relation between the side chains of
at least some amino
acids and their codons; this prediction has also
been corroborated (23,
24).
59
C=)
C=)
%10
110-
BATCo document for Province of British Columbia 23
April 1999


Suppose, for simplicity, that predisposition to
early-onset
carcinoma of the cervix entails the presence of
the autosomal allelecul
in the genotype and that the 'complementary'
allele is designated cu2.
(Predisposition is likely to be polygenic but we
shall assume that the
expression of any other predisposing genes does
not change appreciably
with time.) In this model a decrease in the
frequency of women
predisposed to the neoplasm requires an
autoaggressive attack on either
a gem cel I. the zygote, or the very early-stage
embryo to convert au 1
tocu2, by inducing DNA strand-switching at the ou
locus. To explain
the temporal changes in the frequency of
predisposed cohorts
(underestimated in magnitude in F19.3 because of
the diluting effect of
contributions from late onset cervical and corpus
cases) we have to
postulate an autoaggressive attack that depends on
one or more
fluctuating environmental precipitators such as
microorganisms or
allergens. On this argument, one 'epidemic'
(positive or negative ? )
reached maximum effectiveness around 1910 and the
second around 1935
(Fig.3). That the second cohort dip is lower than
the first might be
due, at least in part, to a temporal decline in
the (independent)
precipitator of early-onset cervical cancer
itself. Because various
gross chromosomal abnormalities depend markedly on
maternal age (25) the
risk of early-onset cervical cancer should be
determined in relation to
maternal (or even paternal) age at birth. If any
such dependence should
conform to the statistics of autoaggressive
disease, corroboration of
the present thesis would be provided.
In the light of the maternal age dependence of
Klinefelter's and
Down's syndromes and its interpretation in terms
of autoaggressive
attacks on oocytes (20, 21). it is of special
interest that Videbech and
Nielsen have reported significant seasonal
variation in the birth of
males with Klinefelter's syndrome in Denmark from
1900 to 1946 and of
Down's syndrome births over the period 1900 to
1980 (26). These
observations support the concept of a seasonal
ly-dependent precipitator
of autoaggressive attacks culminating in
chromosomal non-disjunction.
Whether the date of birth of women who develop
'early onset' cancer
shows a seasonal, as well as temporal dependence,
also calls for
investigation.
CONCLUSIONS
The remarkable temporal changes in age-specific
mortality from
cancers of the uterus and, in particular, from
cervical carcincona,
challenge interpretation. Period changes without
cohort shifts are a
conspicuous and familiar feature of infectious
diseases; they present no
interpretational problem. However, for diseases
that are not generally
recognized as being either infectious or allergic,
such as ischaemic
heart disease or oesophageal cancer, but which
show period changes
implying a precipitating mechanism (13-15), our
practical problem is to
identify the actual precipitator(s). According to
the present analysis
of the cohort changes in early-onset cervical
cancer an analogous
problem arises in connexion with the
precipitator(s) of the hypothetical
autoaggressive attack on parental germ cells.
60
C:)
I-rl
110
BATCo document for Province of BritiSh Columbia 23
April 1999


If the above arqu,nents about induced genetic
change should be
established they will undermine classical ideas
about genes with a
stability that is threatened only by the rare
mutation. Recent
investigations of the structure of immunoglobulins
in particular have
served to highlight the complicated patterns of
gene rearrangement that
occur in somatic cells. Another form of genetic
lability is
hypothesized here in which a change of gene
expression (DNA
strand-switching?) is induced by an autoaggressive
attack on germ cells
and, possibly, the zygote or early-stage embryo.
If such lability
exists to an appreciable degree It presents severe
complications for
classical genetic studies based on the random
segregation of virtually
'fixed' genes. As I have argued previously (27)
autoaggressive attacks
on the early stage embryo - as well as random
mutations - will lead to
genetic discordance between monozygotic co-twins.
Elucidation of the
role of genetic factors in the aetiology of
autoaggressive disorders,
neoplastic and non-neoplastic, will call for the
identification and
characterization of the complex recognition
components of. MCP-TCF
macromolecules.
REFERENCES
1. Burch PRJ. The Biology of Cancer: A New
Approach. MTP Press
Lancaster, 1976; University Park Press, Baltimore,
1976.
2. Osmond C. Gardner Mi, Acheson ED, Adelstein AM.
Trends in Cancer
Mortality Analysed by Period of Birth and Death,
1951-80, England
and Wales. Office of Population and Censuses and
Surveys, Series
DHI No 11, London, HMSO. 1983.
3. Hill GB, Adelstein AM. Cohort mortality from
carcinoma of the
cervix. Lancet 2: 605-6, 1967.
4. Barrett JC. Age, time and cohort factors in
mortality from cancer
of the cervix. Journal of Hygiene (Cambridge) 71;
253-259, 1973.
5. Cook GA, Draper GJ. Trends in cervical cancer
and carcinoma in
situ in Great Britain. British Journal of Cancer
50. 367-375,
1984.
6. Parkin OM, Ngayen-Dinh X, Day NE. The impact of
screening on the
incidence of cervical cancer in England and Wales.
British Journal
of Obstetrics and Gynaecoloqy 92; 150-157, 1985.
7. Burch PRJ. Spontaneous auto-immunity: Equations
for age-specific
prevalence and initiation-rates. Journal of
Theoretical Biology
12 : 397-409, 1966.
8. Burch PRJ. An Inquiry Concerning Growth,
Disease and
Ageing. Edinburgh, Oliver and Boyd, 1968.
9. Registrar General. Statistical Reviews of
England and Wales for
the years 1971 to 1973. Part 1. Tables, Medical.
London. HKSO.
1973-1975.
61
C=)
BATCo document for Province of BritiSh Columbia 23
April 1999


10. Registrar General. Statistical Review of
England and Wales for the
Three Years 1968-1970. Supplement on Cancer.
Office of Population
Censuses and Surveys. London, HMSO, 1975.
11. Office of Population Censuses and Surveys.
Cancer Mortality
England and Wales 1911-1970. London, NKSO, 1975.
12. Office of Population Censuses and Surveys.
Mortality Statistics:
Cause. England and Wales 1974 to 1982. London,
HMSO. 1977-1983.
13. Burch PRJ. Ischaemic heart disease:
epidemiology. risk factors and
cause. Cardiovascular Research 14: 307-338, 1980.
14. Burch PRJ. End of static decade for coronary
disease? British
Medical Journal 289: 1142, 1984.
15. Burch PRJ. Esophageal cancer in relation to
cigarette and alcohol
consumption. Journal of Chronic Diseases 37:
793-808, 1984. .
16. Guijon FB, Paraskevas M, Brunham R. The
association of sexually
transmitted diseases with cervical intraepithelial
neoplasia: A
case-control study. American Journal of
Obstetrics, Gynecology and
Reproductive Sciences 151: 185-190. 1985.
17. Syrj5nen K, deVilliers E-M, Saarikoski S,
Castren 0, VSyrynen M.
Mgntyjlrvi R. Parkkinen S. Cervical papillomavirus
infection
progressing to invasive cancer in less than three
years. Lancet 1:
510-511, 1985.
18. Greenberg ER, Vessey M, McPherson K, Yeates D.
Cigarette smoking
and cancer of the uterine cervix. British Journal
of Cancer 51:
139-141, 1985.
19. Alderson M, Donnan S. Hysterectomy rates and
their influence upon
mortality from carcinoma of the cervix. Journal of
Epidemiology
and Community Health 32: 175-177, 1978.
20. Burch PRJ. Down's syndrome and caternal age.
Nature 221: 173-5,
1969.
21. Burch PRJ. Klinefelter's syndrome, dizyogtic
twinning and diabetes
mellitus. Nature 221: 175-7, 1969.
22. Burch PRJ, Burwell RG. Self and not-self. A
clonal induction
approach to immuinology. Quarterly Review of
Biology 40: 252-279,
1965.
23. Hendry LB, Bransome Jr ED, Hutson MS, Campbell
LK. First
approximation of a stereochemical rationale for
the genetic code
based on the topography and physiochemical
properties of "cavities "
constructed from models of DNA. Proceedings of the
National
Academy of Sciences, USA 78: 7440-7444, 1981.
62
BATCo document for Province of BritiSh Columbia 23
April 1999


24. Hendry LB, BranSOme Jr ED, Hutson MS, Campbell
LK. A newly
discovered sterochemical logic in the structure of
DNA suggests
that the genetic code is inevitable. Perspectives
in Biology and
Medicine 27: 623-651, 1984.
25. Ferguson-Smith MA, Yates JRW. Maternal age
specific rates for
chromosome aberrations and factors influencing
them: Report of a
collaborative European Study on 52965
amniocenteses. Prenatal
Diagnosis 4: 5-44, 1984.
26. Videbech P, Nielsen J. Chromosome
abnormalities and season of
birth. Human Genetics 65: 221-231, 1984.
27. Burch PRJ. Twin studies in relation to the
mechanisms of
age-dependent disease. Acta Medica Auxologica 8:
87-105. 1976.
63
BATCo document for Province of BritiSh Columbia 23
April 1999