© International Epidemiological Association 1999 Printed in Great Britain International Journal of Epidemiology 1999;28:327–334

Interpreting the decline in tuberculosis:

the role of secular trends in effective contact
E Vynnycky and PEM Fine

Background The dramatic decline in tuberculosis (TB) in developed countries during the past
century has been attributed to many factors, including improvements in living
and social conditions and, more recently, effective treatment. Each of these
changes should have reduced the average number of individuals ‘effectively
contacted’ (i.e. sufficiently to transmit infection) by each infectious TB case.
Method Estimates of the average number of individuals effectively contacted by each
infectious TB case, for each year since 1900 in England and Wales, are derived as
the ratio between published estimates of the annual risk of infection and estimates
of the prevalence of infectious cases, as derived using a published model of the
epidemiology of TB.
Results The results suggest that each infectious case contacted, on average, about 22
individuals in 1900 sufficiently to transmit Mycobacterium tuberculosis infection,
and that this number declined to about 10 by 1950 and to approximately one by
1990.
Conclusions Although several factors contributed to the decline in TB in developed countries
during this century, a major contributor has been the decline in the number of
effective contacts by each case over time. Similar declines have doubtless occurred
over the past century for many infections in developed countries.
Keywords Tuberculosis, epidemiology, model, decline, contact
Accepted 10 July 1998

The dramatic decline in tuberculosis (TB) in developed measured, though several studies have explored related
countries during the past century, even prior to the introduction questions. For example, one study correlated the annual risk of
of effective anti-TB drugs, has been attributed to many factors, infection to the estimated incidence of smear-positive cases in
including improvements in social conditions and nutrition,1 many different populations and concluded that an incidence of
reduced crowding2 and segregation of infectious cases either to 50 per 100 000 reflects an annual risk of infection of 1%.6,7
workhouses or sanatoria.3 Whilst the risk of tuberculous infec- Assuming an average duration of infectiousness of 2 years,
tion has declined in many developed countries, at least since the this implies that on average each smear-positive case contacts
early years of this century,4,5 the extent to which the reduced 10 individuals per year.8 This particular study did not determine
morbidity was attributable to a reduction in infection trans- whether this statistic had changed over time. Another study
mission per case, as opposed to a reduced risk of developing derived a related statistic for the Netherlands, namely the ratio
disease among infected individuals, is unknown. Reductions in between the annual risk of infection and mortality rates from all
transmission could have occurred either through changes in the forms of TB and found that it hardly declined at all during the
degree of exposure to infectious cases, for example as a result of time period considered (1921–1938).7 A more recent study, also
changing behaviour, reduced crowding or increasing segrega- carried out in the Netherlands, found that yet another measure,
tion of infectious cases to sanatoria, or else because of declines the number of individuals contacted by each (all forms) TB case,
in the susceptibility to infection of those exposed, perhaps through declined from about three individuals contacted in 1950 to
improved hygiene or nutrition. about 1.8 individuals contacted in 1980.9
The effects of these changes should be reflected in trends in In the analyses presented here, we extend these estimates
the average number of individuals ‘effectively’ contacted by using an age-structured model of the transmission dynamics of
each infectious TB case. To date, the trend in this statistic in Mycobacterium tuberculosis in England and Wales. We first estimate
any single population throughout this century has not yet been the prevalence of infectious cases since 1900, and relate these
figures to estimates of the annual risk of infection,5 in order to
assess the extent to which the decline in TB morbidity in de-
Infectious Disease Epidemiology Unit, Department of Infectious and Tropical
Diseases, London School of Hygiene & Tropical Medicine, Keppel Street, veloped countries reflects changes in the number of individuals
London WC1E 7HT, UK. infected by each infectious case.

327
328 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

Methods
General structure and assumptions in the model
Figure 1 shows the general structure of the model, which ex-
tends the work of Sutherland et al.10 and describes the dynamics
of all forms of pulmonary TB in England and Wales over the
past century. The derivation and validation of the model are
described in detail elsewhere.11 The input parameters and their
values are summarized in Table 1. Analyses presented here
relate to white ethnic males in the absence of HIV infection,11
and thus avoid the complications of gender differences, im-
migration and the HIV epidemic in the recent epidemiology of
TB in England and Wales.
Individuals are born uninfected, and face an annual risk of
infection (i(t)), dependent on calendar year (see below). Infected
individuals are divided into those infected for ,5 years who
have not experienced (primary) disease (I(a,t,s)), and those in
the ‘latent’ class (L(a,t)), who are at risk of endogenous disease
or of reinfection, followed by exogenous disease. The ‘latent’
class comprises individuals who have been infected for .5 years,
and also those who have been (re)infected for ,5 years but
have already experienced disease in the meantime.
The risks of initial infection and of reinfection are assumed
to be identical, though reinfection is less likely to lead to dis-
ease than is an initial infection due to the (partial) immunity
induced by previous infection.11 These assumptions stem from
the results from our previous analyses,11 which found that the
Figure 1 Schematic diagram of the model. Primary disease is defined as
best fit to the observed data (see below) occurred if we assumed disease within 5 years of initial infection;34 exogenous disease is here
that infection confers no protection against reinfection per se, defined as the first disease episode within 5 years of the most recent
though it did reduce the probability that the reinfection event reinfection. Endogenous disease includes disease occurring .5 years
after the most recent (re)infection event, and second or subsequent disease
led to disease. In reality, it is likely that the risks of infection
episodes occurring ,5 years after the most recent (re)infection event
Reproduced with permission from Epidemiology and Infection
Table 1 Summary of parameter values used in model

Variable Definition Assumption

i(t) Infection and reinfection rates at time t. 20% until 1880, declining by 2% pa until 1901, by 4% pa until 1950 and 13% pa
thereafter.5
v(a,t) Proportion of uninfected individuals of age a Vaccination introduced in 1954 and restricted to 13 year olds.
immunized at time t. Vaccine efficacy assumed to be 77% and vaccine coverage increasing to
approximately 80% since 1960.11
dp(a,s) Risk of developing the first primary episode at Declines with time since initial infection (Figure 2B). Cumulative risks within
time s after infection at age a. 5 years of initial infection: 4.06%, 8.98% and 13.8% for 0–10 year olds, 15 year
olds and individuals aged .20 years respectively.11
dx (a,s) Risk of developing exogenous disease at time s Declines with time since reinfection (Figure 2B). Risks within 5 years of
after reinfection at age a. reinfection: 6.89%, 7.57% and 8.25% for 0–10 year olds, 15 year olds and
individuals aged .20 years respectively.11
dn (a) Annual risk of developing endogenous disease at 9.82 × 10–8%, 0.0150%, and 0.0299% for 0–10 year olds, 15 year olds and
age a. individuals aged .20 years respectively.11
d+ (a) Proportion of total disease incidence among cases 10% for 0–10 year olds, increasing linearly to 65% for 20 year olds and aged a
assumed to be infectious. increasing linearly to 85% for 90 year olds (Figure 2C).
kL(s) Rate at which individuals who have been infected Transition occurs exactly 5 years after infection/reinfection
or reinfected for time s without developing disease i.e. kL(s) = 0 if 0 , s , 5 and ∞ for s = 5 years.
move into the ‘latent’ class.
r(a,t, ŝ) Recovery rate for cases of age a at time t and time ŝ Individuals are diseased for 2 years unless they die in the meantime
after disease onset. (see below).
m+(t, ŝ) Case-fatality of infectious pulmonary cases at Case fatality in second year after disease onset is 65% of that in first year.
time t and time ŝ since disease onset. Overall case-fatality: 50% until 1950, declining to 30% and 25% by 1953 and
1956 respectively, and constant until 1976. Identical to mortality in general
population thereafter.11
mg(a,t) Mortality rate of non-infectious and non-diseased Identical to all-cause mortality (after subtracting deaths among infectious cases,
individuals in the general population of age a at estimated in the model). Annual age-specific all-cause mortality rates obtained
time t. from Government’s Actuary’s Department since 1841. Data until 1841 obtained
by back-extrapolation.
 INTERPRETING THE DECLINE IN TUBERCULOSIS 329

Figure 2 Summary of main assumptions in the model relating to the risks of developing disease
A General relationship between the risk of developing the first primary episode (during the first year after infection) and age at
infection. An identical relationship is assumed to hold between the risk of exogenous disease and the age at reinfection and
between the risk of endogenous disease and the current age of individuals. See Table 1 for the magnitude of the disease risks.
B Risk of developing the first primary episode (or exogenous disease) in each year following initial infection (or reinfection),
relative to that experienced in the first year after infection. The relationship was derived using data from the UK MRC BCG
trial during the 1950s.35
C Proportion of respiratory disease incidence manifested as sputum-positive (i.e. infectious forms), based on Norwegian data
provided by the late Dr K Styblo (TSRU) and Dr K Bjartveit (Norwegian National Health Screening Service).
Reproduced with permission from Epidemiology and Infection

and of reinfection differ, either because the immune response of the model to various assumptions has been explored in an
resulting from a previous infection makes reinfection difficult earlier publication.11 In particular, the exclusion of the reinfec-
to establish, or because individuals who are already infected tion disease pathway considerably worsened the fit to notifica-
live in situations which predispose them to further (re)infec- tion rates, especially among the elderly.11
tion. For simplicity, the reinfection risk is assumed to be An age-specific proportion of disease is assumed to be
negligible for recently (re)infected individuals who are already ‘sputum-positive’, i.e. infectious (d+(a)). This is independent
at risk of developing their first primary episode or exogenous of the mechanism of disease onset and of whether it is a first
disease.11 or subsequent disease episode (Figure 2C). It is assumed that
The risks of developing disease are age-dependent; those of individuals are diseased for 2 years unless they die in the mean-
developing the first primary episode (dp(a,s)) and of exogenous time. The case-fatality rate is assumed to be 50% during the
disease (dx(a,s)) depend also on time since infection and reinfec- prechemotherapy era, which is similar to that found in several
tion respectively (Figures 2A and B). Neither of these risks is observational studies12 (see also reviews in6,8) and in a major
assumed to depend on calendar year—we discuss the implica- longitudinal study of the natural history of TB in the absence
tions of this assumption below. Estimates of the risks of develop- of treatment.13 Disease-attributable deaths are distributed over
ing disease have been derived by fitting model predictions to 2 years following disease onset (Table 1). Treatment first became
notifications of pulmonary TB in England and Wales from the generally available from about 1950, and lasted 6 months to
period 1953–1988, assuming that individuals either could or over 2 years during the 1950s,14 with short-course regimens
could not develop disease following reinfection.11 The sensitivity (e.g. 9 months) having been introduced in 1976 in England
330 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

and Wales.15 To incorporate the contribution of treatment to Results

the transmission dynamics of M. tuberculosis in the model, the
Figure 3 demonstrates that predicted trends in mortality of
duration of infectiousness is assumed to decrease gradually from
TB depend greatly on whether we assume that disease did or
2 years in 1950 to one year in 1976. The BCG vaccination of
did not arise following reinfection in the past. For individuals
13 year olds since 1954, assuming a vaccine efficacy of 77%11
aged ,65 years, predictions derived using the full model (i.e.
has also been incorporated (Table 1).
assuming that disease could occur as a consequence of reinfec-
Estimation of infection risk trends tion) provide a good fit to the mortality rates, which is consider-
ably better than that derived assuming that reinfection did not
The estimation of the annual risk of infection since 1900 in
lead to disease in the past. For 15–24 and 25–34 year olds, the
England and Wales has been described elsewhere.5 Briefly, the
estimates derived using the full model slightly overestimate the
annual risk of infection between 1900 and 1949 was derived
observed mortality rates, especially for the early 1900s, and
using tuberculous meningitis mortality rates among 0–4 year
the fit is best for individuals aged 35–44 and 45–54 years. The
olds, assuming (as found elsewhere4) that 1% of new infections
fit to the mortality rates during the period 1914–1918 and the
in this age group led to fatal tuberculous meningitis. The infec-
early 1940s is poor, coinciding with the two World Wars (no
tion risks thus derived led to a good fit between the expected pre-
attempt was made to incorporate the effects of war on TB trends
valence of tuberculous infection and the observed prevalence of
in the model). Model predictions suggest that if it is assumed
tuberculin sensitivity among individuals tested during the
that reinfection did not lead to disease, then the TB-specific
1949–1950 national tuberculin survey in England and Wales.16
mortality rates would have been considerably lower than those
Given the absence of appropriate data for England and Wales
observed for all age groups, and for the younger age groups
after 1950, it is assumed that the annual risk of infection
(those aged ,65 years) they would have increased until 1950
declined from 1950 at 13% per annum, as was found in the
(this is explained below). For individuals aged .65 years, the
Netherlands (which has the most reliable estimates of trends in
fit to the observed data is poor irrespective of the reinfection
infection risks over time).4
assumption, and that derived using the full model is especially
Evaluation of the reliability of model predictions poor for individuals aged .75 years.
prior to 1950 Figure 4 shows that predictions of the age-standardized TB-
specific mortality rates derived using the full model closely match
Here we examine the reliability of model predictions for the
the decline in the observed rates, though they slightly overesti-
prechemotherapy era. Given the absence of appropriate routine
mate the overall magnitude of these rates. In contrast, predictions
TB morbidity data prior to 1950, we compared model predic-
derived assuming that disease did not arise through reinfection
tions of the mortality rates (among males) from respiratory TB
underestimate the observed mortality rates, and predict that the
as an age-standardized population statistic, and in different age
overall mortality rates should have increased during this century.
groups, against those observed. The annual mortality rates from
Figure 5 summarizes estimates of the annual risk of infection
1901 until 1957 were calculated using the numbers of deaths
(Figure 5A) as derived elsewhere,5 and estimates of the overall
from respiratory TB (using the appropriate classification), ex-
prevalence of infectious pulmonary cases in England and Wales
tracted from Revisions I–V of the historic data files (available in
since 1900 (Figure 5B), as derived using the full model, which
electronic form from the Office for Population and Census Sur-
provided the best fit to the observed mortality rates (see above).
veys (OPCS), now known as the Office for National Statistics)
The prevalence of infectious cases is estimated to have declined
and population estimates calculated by OPCS in 1991.
dramatically during this century, from approximately 600 per
In previous analyses of TB trends from 1950, the exclusion of
100 000 in 1900 to about 200 per 100 000 by 1950, with the
the reinfection disease pathway from the model led to a poor
decline accelerating after 1950.
fit to notification rates especially among the elderly.11 Given
The implications for the effective contact number are shown
that the contribution of reinfection to the transmission dyn-
in Figure 5C. This suggests that the number of effective contacts
amics of M. tuberculosis is disputed,17 we also compared predic-
per case declined during this century—on average, each infec-
tions of the mortality rates against those observed, assuming
tious pulmonary case effectively contacted about 22 individuals
that disease could not result from reinfection and using the
in 1900; this declined to about 10 by 1950 and to about one by
best-fitting disease risks which resulted from this assumption.11
1990.
Estimates of the ‘effective contact number’
An effective contact between an infectious pulmonary case and Discussion and Conclusions
another individual is here defined, as by Frost,18 as that sort of These analyses present evidence that the decline in TB mor-
contact which is sufficient to lead to infection if the contacted bidity during the past century in developed countries was asso-
individual has never been infected in the past. The number of ciated with a decline in the degree of effective contact between
individuals effectively contacted by each infectious case (the individuals.
‘effective contact number’) was derived for each year during The conclusions are based in part on a model which simplifies
the period 1900–1990 as the ratio between the annual risk the epidemiology of TB, while attempting to preserve the essen-
of infection and the prevalence of infectious cases. Given the tial characteristics of the disease. Among the model’s major
absence of reliable notifications until the 1950s in England and assumptions is that of random mixing, which is likely to have
Wales, the latter were generated using the model which best been more realistic for the first half of this century than for
matched the observed magnitude and trend in the age-specific more recent years. This assumption means that our estimates of
mortality rates until 1950. the effective contact number may be particularly unreliable for
 INTERPRETING THE DECLINE IN TUBERCULOSIS 331

Figure 3 Comparison between observed age-specific mortality rates attributable to

respiratory tuberculosis in England and Wales in males since 1901, against those predicted
by the model, assuming that reinfection did and did not occur, using best-fitting disease
risks for the corresponding assumption. (Risks of disease based on the assumption that
reinfection did not affect disease risks: 2.53% and 13.6% for individuals during the first
year after infection at age 0–10 years and .20 years respectively; annual risk of developing
endogenous disease: 4.72 × 10–4% and 0.0416% for 0–10 year olds and those aged
.20 years respectively11.)

recent years. Another important assumption is that the disease the early years of this century, followed by an increase in
risks among infected individuals depend only upon age and mortality rates until about 1950.
time since (re)infection, but not otherwise on calendar time. We The finding that reinfection needs to be incorporated in the
discuss the implications of this assumption below. model for predictions to match observed trends is consistent
with the results from previous analyses of the transmission dyn-
The reliability of prevalence estimates prior to 1950 amics of M. tuberculosis in England and Wales since the 1950s.5
The full model (which assumes that disease could result from Those analyses found that if it was assumed that disease could
reinfection) appears to provide a credible description of TB not result from reinfection, then the model predicted a slower
trends in England and Wales during the prechemotherapy era, decline and lower levels in the morbidity among older indi-
with model predictions matching the magnitude and trend ob- viduals than those actually observed. In the analyses presented
served in both the age-standardized and age-specific mortality here, the importance of reinfection in the model stems from
rates (Figures 3 and 4). The alternative assumption, that reinfec- the very high annual infection risks estimated for the start of
tion could not lead to disease, predicted trends which were very this century (e.g. 12% in England and Wales in 19005), which
different from those observed, with low mortality rates during implies a very high prevalence of infection among adults alive
332 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

during this time. This suggests that, if it is assumed that indi- Although not assumed in the model, some decline in disease
viduals could not experience disease following reinfection in risks among infected individuals is not unreasonable, and
the past, then, given these high infection risks early this cen- thus a reduction in the contribution of reinfection-attributable
tury, model predictions can only match observed trends if it disease was probably not the sole factor determining tuber-
is assumed that the risks of developing endogenous disease culosis trends in England and Wales in the past. Disease risks
were much higher than those assumed and that they declined could have declined for two reasons: through a decline in inten-
dramatically over time. sity (dose) of exposure, or through an improvement in the gen-
eral health status of infected individuals, associated perhaps
with nutrition. There is some evidence suggesting that the risk
of developing primary disease given infection is higher among
individuals who are exposed to a high dose of tubercle bacilli, as
compared with those less intensely exposed.19 It is likely that
the risk of developing disease following reinfection is also dose-
dependent. Although there is no evidence that dose of initial
exposure affects the risk of endogenous reactivation many years
thereafter, such a relationship is plausible, as initial dose could
determine the number and magnitude of lesions, and hence the
probability that reactivation disease occurs. Insofar as this is
true, in populations where the overall annual risk of infection
early this century was lower than in countries such as England
and Wales, a reduction in infecting dose might have been as
important as a reduction in reinfection-attributable disease in
determining tuberculosis trends.
Figure 4 Comparison between the observed average age-standardized
mortality rates attributable to respiratory tuberculosis in England and Despite the excellent fit for younger and middle-aged indi-
Wales in males since 1901, against those predicted by the model, viduals, we note that predictions derived using the full model
assuming that reinfection did and did not occur, using the best-fitting greatly overestimate the observed mortality rates for older indi-
disease risks for the corresponding assumption. (See caption to Figure 3 viduals during the prechemotherapy era, especially for those
for values of the disease risks)
Reproduced with permission from Epidemiology and Infection

Figure 5 Estimating the effective contact number in England and Wales since 1900
A The annual risk of infection (derived using tuberculous meningitis statistics and tuberculin
sensitivity data5).
B Prevalence of infectious pulmonary cases since 1900, as derived using the full model.
C The estimated effective contact number, derived as the ratio between the annual risk of
infection and the prevalence of infectious cases
 INTERPRETING THE DECLINE IN TUBERCULOSIS 333

aged >75 years (Figure 3). There are several possible explana- number should also have declined as a result of increasing pro-
tions for this discrepancy. First, it could be true that reinfection portions of infectious individuals being removed from the popu-
was not important in determining the morbidity among the lation to sanatoria during the prechemotherapy era. During
elderly, either because older individuals had more immunity the early 1900s, there were only three TB sanatoria in England
to reinfection due to previous infection, or were less exposed to and Wales, and one TB dispensary; by the 1930s there were
new infections than were younger individuals. Second, the low more than 300 sanatoria and hospitals and about 500 TB dis-
mortality rates among the elderly could be a residue of high pensaries.29 It has also been pointed out that infectious cases
mortality rates experienced by susceptible members of these were effectively segregated even before the 1900s, as a con-
cohorts when they passed through young adult life—there has sequence of the removal of the poor to workhouses.3
been much discussion of this issue in the literature.20–23 A third Most of these factors should have influenced the effective
explanation is that some of the TB-attributable mortality at contact number for other infectious diseases in addition to TB,
older ages was not recognized. This is supported by an autopsy many of which have declined during this century. A study of
study carried out during the period 1935–1944, which found childhood bronchitis and pneumonia has suggested that the
that open tuberculous lesions were present in about 50% of decline in childhood infections was related to declines in fertil-
individuals aged .50 years who were recorded as having died ity rates, which, by reducing the size of the average household,
of causes other than TB.24 increased the average age at infection.30 It is possible that these
concurrent declines in acute infections may have themselves
Estimates of the effective contact number contributed to reductions in TB morbidity, as some studies
for tuberculosis have suggested that individuals may be more susceptible to
There is a large literature on the decline in TB and other in- tuberculous infection and/or disease after experiencing acute
fectious diseases during the past century in developed countries. viral infections.31 Ironically, the effective contact number for
Our analyses suggest that this decline was associated with a some infections may have increased over this century, as a con-
decline in the average number of individuals ‘effectively con- sequence of increasing proportions of children attending
tacted’ by each infectious case. crêches or pre-school day-care nurseries. This increased mixing
Such a decline in the effective contact number is not un- among young children is likely to be much less important for
reasonable, and can be attributed to many factors. Among the the transmission dynamics of M. tuberculosis than for acute re-
obvious influences is a decline in the degree of domestic crowd- spiratory and enteric infections, however, as very few children
ing. During the latter part of the 19th century, about 8% of the develop infectious pulmonary forms of TB.
population in England and Wales lived in accommodation with The decline in the effective contact number from 1950 esti-
more than two individuals per room, as compared with 5.5% mated here (Figure 5) is particularly interesting. Much of the
and about 1% by 1901 and 1951 respectively.25 By 1991, only decline during the 1950s is attributable to the availability of
0.25% of the population lived in accommodation with more effective antibiotics for TB after 1950, which must have had
than 1.5 individuals per room.25 The average size of each house- a marked effect on reducing exposure by shortening the dura-
hold has also decreased over time, from about 5 individuals in tion and the infectiousness of each case. It also plausible that the
1901 to about 2.5 by 1990.26 The importance of general living effective contact number after 1950 was affected by changes in
conditions in determining transmission was demonstrated by a the age distribution of infectious cases over time, as an increas-
study in 1931, which compared the prevalence of tuberculin ing proportion of infectious cases from the 1950s were middle-
sensitivity among individuals living in accommodation rated aged or elderly —it is likely that these individuals contact fewer
as ‘poor’, ‘fair’ and ‘good’ (determined on the basis of income, individuals than do younger individuals who comprised most of
locality and general hygiene) and in which there was a source the infectious sources earlier this century. It has been suggested
of infection.27 The prevalence of tuberculin sensitivity among that this particular factor contributed to the similar decline
contacts was inversely proportional to the standards of housing, which was observed in the number of individuals contacted by
and a similar comparison in households without a source of infec- each (all forms) TB case in the Netherlands between 1950 and
tion showed little association between the prevalence of tuber- 1980.9
culin sensitivity and housing standards. It is also likely that The effective contact number for a given infection influences
ventilation in homes and workplaces has improved appreciably another important epidemiological statistic, the basic reproduc-
during this century, which should have reduced the risk of tion number, defined as the number of secondary infectious
aerosol transfer of infection. cases resulting from a typical infectious case in a ‘totally sus-
Other changes must also have reduced the opportunity for ceptible’ population. For simple infections such as measles, for
exposure to infectious cases in the domestic setting. The propor- which all infected individuals develop infectious forms of the
tion of elderly individuals (important as sources of tuberculous disease, the total number of individuals effectively contacted
infection24) living either alone or in institutions has increased by a case during his/her infectious period in a ‘totally suscep-
over time, thereby reducing exposure opportunities to younger tible’ population will be identical to the basic reproduction
individuals in the population.26 Improvements in the nutri- number. For TB, the relationship between the effective contact
tional status in the population28 may have contributed to number and the basic reproduction number is complicated by
reductions in the effective contact number, by reducing the the fact that not all of those infected go on to develop infectious
probability of a contact becoming infected with M. tuberculosis disease, by the age-dependence in disease risks, and by the fact
given a certain exposure. Improvements have occurred in the that reinfection can occur. The implications of the factors for
general hygiene in the population (and in behaviour associated the basic reproduction number for TB are discussed in detail
with sneezing, coughing and spitting). The effective contact elsewhere.32
334 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

These analyses indicate that many factors have influenced the 11 Vynnycky E, Fine PEM. The natural history of tuberculosis: the im-
transmission dynamics of M. tuberculosis in the past, and provide plications of age-dependent risks of disease and the role of reinfection.
evidence that a declining effective contact number was im- Epidemiol Infect 1997;119:183–201.
portant in determining the decline in tuberculosis morbidity 12 Berg G. The prognosis of open tuberculosis. A clinical-statistical

through most of this century. It is interesting that the decline in analysis. Acta Tuberc Scand 1939;IV(Suppl.):1–207.
TB incidence was halted in several countries during the 1980s, 13 National Tuberculosis Institute, Bangalore. Tuberculosis in a rural

a change which has been attributed to several influences, in population of South India: a five-year epidemiological study. Bull
particular the HIV epidemic, changes in immigration patterns, World Health Organ 1974;51:473–88.
14 Springett VH. Ten-year results during the introduction of chemo-
the increasing problem of homelessness and underfunding of
control programmes.33 Though recent data for the US indicate therapy for tuberculosis. Tubercle 1971;52:73–87.
15 British Thoracic Association. Short course chemotherapy in pulmon-
that the recent increase was only a brief setback, comparable
perhaps to the increases observed during the two World Wars ary tuberculosis. Lancet 1980;i:1182–83.
earlier this century, the full implications in different countries 16 Medical Research Council. MRC national tuberculin survey 1949–50.

is yet to be seen. Whatever unfolds, an appreciation of the Lancet 1952;i:775–85.

factors underlying past trends should help us to interpret the 17 Stead WW. Management of health care workers after inadvertent

future epidemiology of TB. exposure to tuberculosis: a guide to the use of preventative therapy.
Ann Intern Med 1995;122:906–12.
18 Abbey H. An examination of the Reed-Frost theory of epidemics.
Acknowledgements Hum Biol 1952;24:201–33.
19 Grzybowski S, Barnett GD, Styblo K. Contacts of cases of active
We thank the British Medical Research Council for financial
support, the late Dr K Styblo (Tuberculosis Surveillance Research pulmonary tuberculosis. Bull Int Union Tuberc 1975;50:90–106.
Unit, The Hague) and Dr K Bjartveit (Tuberculosis Screening 20 Andvord KF. Hvad kan vi laere ved å følge tuberkulosens gang fra

Service, Norway) for supplying TB data from Norway, Dr J generasjon til generasjon? Norsk Magasin Laegevidenskapen 1930;91:
Watson (PHLS Communicable Disease Surveillance Centre) for 642–60.
21 Frost WH. The age selection of mortality from tuberculosis in succes-
data from the national TB surveys in England and Wales, the
UK Government Actuary’s Department for providing age and sive generations. Am J Public Health 1939;30:91–96.
22 Donovan JW. A study in New Zealand mortality. N Z Med J 1969;70:
time-specific mortality rates in England and Wales, and an
anonymous reviewer, who encouraged further discussion of the 90–103.
23 Adelstein A. Mortality from tuberculosis. Popul Trends 1977;8:20–23.
implications of infectious dose.
24 Medlar EM, Spain DM, Holliday RW. Disregarded seedbed of the
tubercle bacillus. Arch Intern Med 1948;81:501–17.
References 25 Hunt S. Housing-related disorders. In: Charlton J, Murphy M (eds).
1 McKeown T, Record RG. Reasons for the decline of mortality in England
The Health of Adult Britain, 1841–1994. Vol. 1. London: The Stationery
and Wales during the nineteenth century. Popul Stud 1962;16:94–122. Office, 1997, pp.157–70.
2 Puranen B. Tuberculosis and the decline in mortality in Sweden. In: 26 Haskey J. Population review: (6) Families and households in Great
Schofield R, Reher D, Bideau A (eds). The Decline of Mortality in Europe. Britain. Popul Trends 1996;85:7–24.
Oxford: Clarendon Press, 1990, pp.97–117. 27 Dow DJ, Lloyd WE. The incidence of tuberculosis and its relation to
3 Wilson LG. The historical decline of tuberculosis in Europe and
contagion in children under 15. Br Med J 1931;2:183–86.
America. J Hist Med Allied Sci 1990;45:366–96. 28 Charlton J, Quaife K. Trends in diet: 1841–1994. In: Charlton J,
4 Styblo K, Meijer J, Sutherland I. The transmission of tubercle bacilli:
Murphy M (eds). The Health of Adult Britain, 1841–1994. Vol. 1. London:
its trend in a human population. Bull Int Union Tuberc 1969;42:5–104. The Stationery Office, 1997, pp.93–113.
5 Vynnycky E, Fine PEM. The annual risk of infection with Mycobacterium
29 Philip R. The outlook on tuberculosis: changing orientation. Br Med J
tuberculosis in England and Wales since 1901. Int J Tuberc Lung Dis 1997; 1931;i:43–49.
1:389–96.
30 Reves R. Declining fertility in England and Wales as a major cause
6 Murray C, Styblo K, Rouillon A. Tuberculosis. Global Burden of Disease.
of the twentieth century decline in mortality. The role of changing
In: Jamison DT, Mosley WH, Meesham AJ, Bobadillo JL (eds). Disease
family size and age structure in infectious disease mortality in infancy.
Priorities in Developing Countries. New York: Oxford Clarendon Press,
Am J Epidemiol 1985;122:112–26.
1993, pp.244–59.
31 Carey SL. A unique epidemic of tuberculosis: Eskimo Point, 1963. Am
7 Styblo K. The relationship between the risk of tuberculous infection
Rev Respir Dis 1965;91:479–87.
and the risk of developing infectious tuberculosis. Bull Int Union Tuberc
32 Vynnycky E, Fine PEM. The long-term dynamics of tuberculosis and
1985;60:117–19.
8 Styblo K. Epidemiology of Tuberculosis. Selected Papers. The Hague, The other diseases with long serial intervals: implications of and for
changing reproduction numbers. Epidemiol Infect 1998;121:309–24.
Netherlands: KNCV, 1991.
33 Snider Jr DE, Roper WI. The new tuberculosis. N Engl J Med 1992;326:
9 Borgdorff MW, Nagelkerke NJD, van Soolingen D, Broekmans JF.
703–05.
Transmission of tuberculosis between people of different ages in the
34 Holm J. Development from Tuberculous Infection to Tuberculous Disease. The
Netherlands—An analysis using DNA fingerprinting. Int J Tuberc Lung
Dis 1999, (In Press). Hague, The Netherlands: TSRU Progress Report; KNCV, 1969.
10 Sutherland I, Švandová E, Radhakrishna SE. The development of 35 Sutherland I. The ten-year incidence of clinical tuberculosis following

clinical tuberculosis following infection with tubercle bacilli. Tubercle ‘conversion’ in 2550 individuals aged 14 to 19 years. The Hague, The
1982;62:255–68. Netherlands: TSRU Progress Report; KNCV, 1968.