Parasitology (1988), 97, 193-212 193

With 6figuresin the text

A general model for the African trypanosomiases

D. J. ROGERS
Department of Zoology, University of Oxford, South Parks Road, Oxford 0X1 3PS

(Accepted 13 August 1987)

SUMMARY
A general mathematical model of a vector-borne disease involving two vertebrate host
species and one insect vector species is described. The model is easily extended to other
situations involving more than two hosts and one vector species. The model, which was
developed from the single-host model for malaria described by Aron & May (1982), is applied
to the African trypanosomiases and allows for incubation and immune periods in the two
host species and for variable efficiency of transmission of different trypanosome species from
the vertebrates to the vectors and vice versa. Equations are derived for equilibrium disease
prevalence in each of the species involved. Model predictions are examined by 3-dimensional
phase-plane analysis, which is presented as a simple extension of the 2-dimensional phase-
plane analysis of the malaria model. Parameter values appropriate for the African trypano-
somiases are derived from the literature, and a typical West African village situation is
considered, with 300 humans, 50 domestic animals and an average population of 5000 tsetse
flies. The model predicts equilibrium prevalences of Trypanosoma vivax, T. congolense and
T. brucei of 47-0, 45-8 and 287% respectively in the animal hosts, 24-2, 3-4 and 0-15% in
the tsetse vectors, and a 7-0% infection of humans with human-infective T. brucei. The
contribution to the basic rate of reproduction of the human-infective T. brucei is only 0-11
from the human hosts and 2-54 from the animal hosts, indicating that in the situation
modelled human sleeping sickness cannot be maintained in the human hosts alone. The
animal reservoir is therefore crucial in determining not only the continued occurrence of the
disease in humans, but its prevalence in these hosts as well. The effect of changing average fly
density on equilibrium disease prevalences is examined, together with the effect of seasonal
changes inflynumbers on disease incidence. In a seasonal situation changes inflymortality
rates affect both future population size and infection rate. Peak disease incidence lags behind
peakflynumbers, and that in the less favoured host lags behind that in the more favoured
host. Near the threshold fly density for disease transmission disease incidence is more
changeable than at higherflydensities and may even exceed equilibrium prevalence at the
same averageflydensity (because most hosts are susceptible at the time thatflynumbers
begin their annual increase). The implications of the model for disease control are discussed.
Identifying the precise role of the animal reservoir may suggest that treating such animals
will achieve a greater reduction of human sleeping sickness than direct treatment of the
humans alone. Statistically significant results of control campaigns may also be more easily
shown by monitoring the non-human reservoirs. The model provides a means by which a
correct perspective view can be obtained of the complex epidemiology and epizootiology of
the African trypanosomiases.

INTRODUCTION
The African trypanosomiases affect livestock production in an area of 10 million km2
(FA0-WH0-0IE, 1982; Murray & Gray, 1984). Locally, human populations may be
devastated by serious epidemics of sleeping sickness (Duggan, 1970; WHO, 1986).
The trypanosomiases involve several parasite species transmitted amongst many
different types of vertebrate hosts by 20-30 species of tsetse flies. The very complexity
of this transmission has discouraged the development of mathematical models of the
194 D. J. ROGERS
sort that have been available for many decades for other vector-borne diseases (e.g.
malaria; Ross (1911); Macdonald (1957); Bailey (1982)), yet field evidence suggests
that the epizootiology and epidemiology of the trypanosomiases may not be as complex
as they first appear. For example, over a wide range of field conditions there is a simple
linear relationship between the daily probability of susceptible cattle contracting
trypanosomiasis and tsetse fly challenge; the latter is the product of fly infection rate
and 'apparent density', a crude but acceptable relative measure of tsetse population
size (Rogers, 1985). Also, the prevalence of infection in a group of uninfected cattle
increases with time in a way which is satisfactorily described by a Poisson model with
a fixed mean daily probability of infection (Cawdery, 1958). In the case of human
trypanosomiasis local reduction in the amount of contact between humans and flies
reduces proportionately the number of cases reported annually (Morris, 1946; Rogers,
1985) although such relationships are unlikely to apply over extensive geographical
areas because of the focal nature of the disease.
Models of human and animal diseases tend to tread a delicate balance between
complexity and simplicity. At the one extreme a large number of usually simple
equations can be combined together to simulate the system under study. It is rarely, if
ever, possible to determine the equilibrium conditions or the stability properties of such
simulation models without running them for many generations. The set of equations
required tends to be unique to each case so that a simulation model developed for one
disease is rarely applicable to another. At the other extreme a very much smaller set
of equations attempts to describe the time-course of changes in the proportions of
susceptible, infected and immune host animals, and these equations can often be solved
for their equilibrium conditions analytically. At the risk of ignoring possibly important
details in the transmission process, analytical models provide much greater insight into
equilibrium and stability conditions and the threshold phenomena which are important
in determining disease control strategies (Anderson, 1982). Furthermore, analytical
models for any particular disease can be readily modified for other diseases of the same
general type (e.g. directly transmitted, vector borne, etc).
Whilst the simulation approach has already been applied to African trypanosomiasis
(Habtemariam, Ruppanner, Riemann & Theis (1983a-c) modelled a single-parasite,
single-vector, single-host species situation) none of the modern developments of
analytical disease modelling (Anderson, 1982) has made any impact in this area until
very recently (Milligan & Baker 1988). The present paper describes an analytical model
for the African trypanosomiases that incorporates two vertebrate host species and one
tsetse vector species; the equations can easily be adapted for further host and/or vector
species. The model can be used to identify research priorities (i.e. crucial areas of
current ignorance) and suggests several novel approaches to the theory and practice
of sleeping sickness control.

The basic vector-host model

The starting point for the trypanosomiasis model described here was the basic Ross-
Macdonald model for malaria (Ross, 1911, 1916; Macdonald, 1952, 1957, 1973) as
described by Aron & May (1982). A novelty of Aron & May's approach is the use of a
graphical technique (phase plane analysis) to examine potential equilibrium points for
each component of the system in turn, and for all components combined. The basic
model initially assumes that the populations of both the vectors and humans are
constant and that infections are acquired by the human population from the vector at a
 A model for the African trypanosomiases 195

rate dependent on the rate of man-biting by each individual vector (a bites/unit time),
the proportion of vectors infected (y), the proportion of infected bites that give rise to
infection (6) and the ratio of vector numbers (M) to human numbers (N). Such
infections are lost at a fixed rate (r, the per capita recovery rate), so that the average
duration of infection in an infected host is 1/r days. The vector population acquires
infection at a rate dependent upon the prevalence of the disease in the vertebrate
population (a;) and the rate of man-biting by the vector (a); it loses infections through
the natural mortality of the flies (rate u, so that the average survival time of the vectors
is \ju days).
Thus, each group of infected individuals has a single gain term to describe the
acquisition of the disease and a single loss term (recovery in humans, mortality in flies).
Initially it is assumed that on recovery humans return directly to the susceptible
category, and that flies lost through mortality are replaced by newly emerged
individuals.
These processes are described by the following pair of equations:

-j- = -—„ -—rx for the infected vertebrates,

and
-j- = ax( 1 — y) — uy for the infected flies. „
(it \L)
The equilibrium conditions of these equations are examined by setting each in turn to
zero, from which we obtain
¥ or
y =—:— —. for the vertebrates, ,_.
" abm (i — x) (3)
where
M

and ax . ,,
3y =-,—; r for the vectors. ...
(u + ax) (4)
These equations describe the relationship between the proportion of infected vectors
(y) and the proportion of infected vertebrates (x) when either vertebrate prevalence
(dxjdt = 0, equation (3)) or vector prevalence (dyjdt = 0, equation (4)) is at its
equilibrium.
Two examples of the relationship between vector and vertebrate prevalences for
these equilibrium conditions are shown in Fig. 1, redrawn from Aron & May (1982). In
Fig. 1 a the graph of dx/dt = 0 indicates that for a fixed vector infection rate, the
vertebrate infection rate increases if it is to the left of the line dxjdt — 0 and decreases if
it is to the right. In the case of the equilibrium line for disease prevalence in the vector
(dyjdt = 0), for a fixed human prevalence the infection rate in flies increases if it is below
the line dyjdt = 0 and decreases if it is above it. The point of intersection of the two lines
is a point of equilibrium of both prevalences and is therefore the equilibrium point for
the disease system. In the case of Fig. 1 a, and starting at low prevalences of both x and
y, clearly the vector infection rate increases, which in turn leads to an increase in
vertebrate infections and so on until the joint equilibrium is reached. Fig. lfe shows the
situation in which the dyjdt = 0 line falls below the line dxjdt = 0 and the disease cannot
maintain itself (vector prevalence falls and in turn lowers vertebrate prevalence
eventually to disease extinction). The disease will therefore persist in this system (i.e.
196 D. J. ROGERS
10 (a)

Equilibrium
point
0-5

"8
I'c
o
10
e
o
Q.
O
Q.
dx/dt = 0

0-5

Human proportion infected, x

Fig. 1. Phase-plane relationships of equilibrium disease prevalences in humans and vectors
in a single-host, single-vector species model (re-drawn from Aron & May (1982)). (a) The
smaller arrows indicate changing prevalences in each species towards the equilibrium value,
whilst the larger arrow indicates movement towards the joint equilibrium point, (b) Situ-
ation in which the disease cannot be maintained (equilibrium point at zero).

Fig. 1 a) only if the initial part of the dy/dt = 0 graph (at low values of x) has a higher
slope than the initial part of the dxjdt = 0 graph, i.e. if
a r
u abm'
or
a2bm
> 1.
ur (5)
The left-hand side of formula (5) is usually called the basic reproductive rate of the
disease (Ro) and is the number of new infections that eventually arise from a single-
current infection during the period of its infectiousness (i.e. before recovery, treatment
or death). Diseases only increase at their basic reproductive rates when they are close to
extinction (i.e. when all hosts are susceptible) and the higher the value of the basic
reproductive rate the more difficult it will be to eradicate the disease. By setting the
 A model for the African trypanosomiases 197
left-hand side of formula (5) equal to unity the threshold for disease transmission can be
calculated. This is the point at which only one susceptible host is contacted, and
eventually becomes infected, during the infectious period of a single infected individual.
In the case of vector-borne diseases the threshold is defined in terms of the ratio of
vectors to hosts (m) and is found from
_tir
m
~lM>- (6)
The squared term in the denominator is a common feature of models for vector-borne
diseases and arises because infections must first be picked up by susceptible vectors
whilst biting at a rate a and later transmitted whilst biting at the same rate. (Recent
developments in the field of modelling malaria epidemiology have been covered by Aron
& May (1982) and Bailey (1982).)

The model applied to the African trypanosomiases

The basic equations described above have been modified to take into account the
following features of the African trypanosomiases.
(1) More than one vertebrate species is usually involved; this paper describes a two-
host system, e.g. man and domestic animals (in the case of human sleeping sickness in
West Africa) or domestic and wild animals (for cattle trypanosomiasis).
(2) The vertebrates must first incubate the infections for a period of time before
transmission can take place. If the incubation rate is i the average incubation latent
period will be 1/i days.
(3) Once the vertebrates recover or are cured of the disease, they are immune to re-
infection for a further period. This immunity is lost at a constant rate (v, average period
of immunity = l/v days) and hosts then re-enter the susceptible category.
(4) Not all infected blood meals eventually produce mature infections in the vectors;
c, the average proportion of such meals which give rise to mature infections is smaller
for T. congolense than it is for T. vivax (Rogers, 1980). This proportion is assumed to be
constant throughout the life of thefly.In the case of T. brucei, however, the parameter
c is relatively high for the first blood meal, if taken by the flies within a short time of
their emergence from puparia, and effectively zero thereafter (Wijers, 1960; Watson,
1963; Gingrich, Ward, Macken & Schoenbechler, 1982a), except perhaps in starved
older flies (Gingrich, Ward, Macken & Esser, 19826).
(5) After taking an infected blood meal vectors must survive for a further period
before they can transmit the disease; losses occur through natural mortality during this
period, reducing the prevalence in the vector.
These modifications lead to the following equations for T. vivax and T. congolense-
type infections:
~^- = a1b1mly(l-xl-w1-zl)-r1xl,

for the first vertebrate host species and

~^- = a2b2m2y(l-x2-w2-z2)-r2x2,
for the second vertebrate host species, where w1 and w2 are the proportions of hosts
currently incubating infections and zx and z2 are the proportions of hosts which are
immune. The circumflex accents above the variables indicate evaluation a vertebrate
incubation period before the present, since the numbers of newly infected individuals
198 D. J. ROGERS

added at the present time depend upon conditions then. Changes in the proportions of
infected vectors are described by

where / is the proportion of flies incubating infections (i.e. not yet capable of
transmission) and the circumflex accent indicates evaluation at a time of one-vector
incubation period before the present. The model therefore assumes that hosts cannot be
infected if they are already incubating, infected or immune; little is currently known
about the effects of super-infection on the durations of incubation, infected or immune
periods of the trypanosomiases so that this cannot yet be allowed for in the model.
Mixed infections in vertebrates other than man are quite common. In this case the
equations should be written for each trypanosome species separately, without making
any allowance for hosts already infected with a different species (but see Willett (1972)
for examples of non-random associations of trypanosome species). In the case of the
vectors, equation (9), the terms in the first parenthesis allow for tsetse to pick up
infections from the two host types (at a daily biting rate of «! from species 1 and a2 from
species 2), the terms in the second parenthesis describe the proportion of susceptible
flies (fewer uninfected vectors are available if many are already incubating or infected),
and the exponential term allows for the loss of flies during cyclical development of their
infections to maturity (taking time T days). The model therefore assumes that flies once
infected remain so for life.
The relationships between the equilibrium disease prevalences in the vector and
vertebrate populations are obtained by setting each of the equations (7)-(9) equal to
zero and solving for y (at equilibrium, with no change in each variable, xx = xlt y = y,
w = w, etc. so that the circumflex accents are lost from each equation). Re-arranging
equations (7)-(9) is made easier by first substituting into them the equilibrium values
for incubating and immune hosts (i.e. w and z) and incubating flies (/), which are
defined as follows,
* rx
W=
T' (10)
* _rx
z
" • (11)
and
*
*
I" c(a,0x + a2x2) "I
I f*irt 'V -I /i iv \ I
(-,-MTI
[u + c (a^j + a2x2) J (12)
In each case the equilibrium values are the input rate into the category concerned,
divided by the output rate. Substitution then gives the following equilibrium relation-
ships between vector and vertebrate prevalences

y— Li for vertebrate species 1, (13)

LJ for vertebrate species 2, (14)

]
 A model for the African trypanosomiases 199

and f c(a,x +a x2) 1 _,.,, ,, ,

y = —, ) 1x—, 2 2/ . e^T r
for the vector. .t _
* [« + c(a1a;1 + o2x2)J (15)
These three equations can be solved simultaneously for xlt x2 and y. For later
reference the relationships between equilibrium prevalences in the two vertebrate
species (assuming prevalence in the vector is also at equilibrium) are as follows:
or I f ii I
X
2 = —\ i AH ~F~\ ^"" a i f ° r vertebrate species 1, (16)
and
x, = — .,,,. *-r\ r~ a 2 for vertebrate species 2,
1 2 (17)
where ax\c(B(\-x2f2)-r2x2) J *
A = a1blm1e~uT,

B = aib2m2e~uT,

and

Equilibrium equations for T. brucei

In the case of T. brucei infections equation (9) applies only to newly emerged flies for
the period t days during which they are susceptible to infection whilst taking their first
blood meal (t is less than the average duration of the feeding cycle, d days). Assuming u'
is the daily emergence rate of teneral flies from puparia (related to the adult female
population a puparial period ago) the total of unfed tenerals less than t days old (and
therefore susceptible to infection) is given by

— M — e-(ai+a2+u)t\
a1 + a2 + u ' (18)
where the first term is the total teneral population (expressed as a proportion of the
total fly population), which is reduced at the rates (at + a2) through feeding and u
through mortality, and the second term is the proportion of these flies which are less
than t days old. This function (equation (18)) replaces the second term in parenthesis in
formula (9) (which represents the susceptible flies) and results in changes in the
equilibrium disease prevalences in both vector and vertebrate populations; equations
(15), (16) and (17) are replaced by the following (where it is assumed that at equilibrium
vector birth and death rates are equal, i.e. u' = u):
y = (alxl + a2x2) c'E e'uT for the vector species, (19)
where
E =

and c' is the proportion of first infected blood meals that give rise to infection in newly
emerged flies when they feed within t days of emergence,

vertebrate species 1, .„„

(ZU)
and
x, = — IT>ru? TT—«2 for vertebrate species 2.
2 v (21)
al [c'BEil-x^) J
200 B. J. ROGERS

0-8
Fig. 2. Phase-plane relationships for the two-host single-vector species model described in
the text. The inclined, almost flat plane records equilibrium prevalence in the vector species
with changing prevalences in the two vertebrate species. Changes in equilibrium prevalence
in each of the vertebrate species is shown by one of the two vertical planes. The disease
equilibrium point is at the intersection of the three equilibrium planes. Projections of pairs of
equilibrium lines onto the various faces of thefigure(views A, B and C) are described by pairs
of equilibrium equations given in the text (see text for further details). The model is that of
equations (13)—(17) with the following arbitrary parameter values N1 = N2 = 300, V = 5000,
pl = 0-3, p2 = 0-7, u = 0025, d = 4, 1/i, = 1 / i 2 = 0, l/^ = 35 days, l/r2 = 20 days, ljv1 =
l/w2 = 20 days, T = 25 days, bt = 062, b2 = 0-3, c = 0-1.

Equilibrium prevalences, reproductive rates and transmission thresholds

The phase-plane relationships for equations (13)—(17) are shown in Fig. 2. The sides
and base of the block shown in Fig. 2 illustrate the predictions of pairs of equations for
equilibrium prevalences in the two vertebrate species and in the vectors (view A
equations (13) and (15), assuming x2 = 0; view B equations (14) and (15), assuming
x1 = 0; and view C equations (16) and (17), which assume that prevalence in the vector
is at equilibrium). The inclined plane within the block (the dyjdt — 0 surface drawn with
dashed lines in Fig. 2) shows the range of equilibrium disease prevalence in the vector
for any combination of prevalences in the two host species (notice that vector
prevalence is rather less variable than vertebrate prevalence). A projection of view C
onto this plane (shown by the thick lines on the plane itself) fixes the (unique)
equilibrium point for the disease system (where prevalence in each of the three species
involved is at equilibrium).
The basic rate of reproduction of the disease, Ro, and thresholds for transmission are
complicated in the case of the trypanosomiases by the presence of more than one
vertebrate species, and are best calculated by considering the fate of one infected fly
which during the remainder of its life (lasting l/u days) will bite individuals of host
 A model for the African trypanosomiases 201

Table 1. Basic reproductive rate of the trypanosomiases (Ro) and thresholds for
transmission predicted by the trypanosomiasis model described in the text
(See text and Table 2 for parameter and variable definitions.)
Ro Threshold
u
1
/VA a£h
H )

where h = NJN2
m
2 — mlN1INi
For T. vivax and T. congolense D = 1
For T. brucei c' replaces c and D = —^-, ; •—r—-

species 1 and 2 at rates of ax and a2 respectively. All of these hosts are assumed to be
uninfected. The total number of vertebrates successfully infected by this fly will
therefore be a^Ju of species 1 and a2b2/u of species 2. During the infection of these
hosts (lasting l/rx and l/r 2 days respectively) they will be bitten by uninfected flies,
eventually giving rise to the following numbers of newly infected flies (allowing for the
death of some of them during incubation):
- ^ . — . a1m1 . ce~uT from species 1
u r^
and
—— . — . a2m2 . ce~uT from species 2. ._„.
The sum of these two figures, the total number of newly infected flies, is the basic rate
of reproduction of the disease in the fly population (and therefore in the total vertebrate
population, since at very low prevalences a doubling of fly infection rates will double
vertebrate infection rates), and the condition for disease persistence is therefore,
ce~uT ia.fb1ml a22b2m2\ .
(23)
Inequality (23) is the two-species equivalent of formula (5) for malaria, allowing for
the additional features of trypanosomiasis already mentioned. The threshold for
transmission is obtained by setting the left-hand side of formula (23) equal to unity and
re-arranging as shown in Table 1, where the formulae for the case of T. brucei are also
listed. The additive nature of the terms in parentheses of inequality (23) indicates that
the disease may persist in species 1, even when this species alone is incapable of
maintaining the disease, because of the presence of species 2 which acts as a reservoir of
infection for species 1. This is shown graphically in Fig. 3. In this case the transmission
threshold for species 1 (i.e. the ratio of vectors to total numbers of species 1) is
determined by species 2 (see later).

Performance of the model

Parameter values used in the model
Parameter values applicable to the African trypanosomiases are based on previous
analyses and published field work (Table 2). The figures given for the proportion of
202 D. J. ROGERS

Prevalence in species 1,x,

Fig. 3. In a two-host vector-borne disease model the presence of a second, infected (i.e.
reservoir) host species can elevate the equilibrium prevalence line of the vector so that the
disease persists in the first species where otherwise it could not do so. Without the reservoir
species the equilibrium point for the first species is at zero, as in Fig. 16.

T a b l e 2. Variables and parameters of the two species trypanosomiasis model described

in the text
(The values chosen are those appropriate for a West African village situation, although the
transmission parameters c and b are estimated from East African field data and c' from
laboratory experiments (see text for more details).)
(a) General
N1 N u m b e r of animals, species 1 (e.g. humans) 300
N2 Number of animals, species 2 (e.g. domestic hosts)
3) 50
Number of tsetse 5000
Proportion of tsetse blood meals from species 1 0-3
Proportion of tsetse blood meals from species 2 0-7
u Daily mortality rate of flies 0030
d Duration of feeding cycle in flies 4 days
2 = p2/d; ml = V/N1, m2 = V/N2

(b) Disease specific T. vivax T. congolense T. brucei

l/i1 Incubation period in species 1 — — 12 days
l/t2 Incubation period in species 2 12 15 12 days
l/rl Duration of infection in species 1 — — 70 days
l/r2 Duration of infection in species 2 100 100 50 days
ljvl Duration of immunity in species 1 — — 50 days
ljv2 Duration of immunity in species 2 100 100 50 days
T Incubation period in tsetse 10 20 25 days
6j Probability of infected fly bite — — 0-62
producing infection in species 1
b2 Probability of infected fly bite 0-29 0-46 0-62
producing infection in species 2
c Probability of any infected blood meal 0177 0025
eventually giving a m a t u r e infection
in a fly
c' Probability of first feed only (taken during 0065
first day) eventually giving a mature
infection in a fly
 A model for the African trypanosomiases 203
blood meals taken by flies from humans or from other vertebrates are similar to those
found in human sleeping sickness foci in West Africa (Sachs, Mehlitz & Staak, 1980;
Gouteux, Laveissiere & Boreham, 1982 a). Fly mortality rate is an average figure based
on analysis of fly population age structure (Rogers, Randolph & Kuzoe, 1984) and
biting rates were calculated assuming a 4-day feeding interval (Rogers, 1977).
Figures for the durations of infection and immunity in Table 2 were difficult to
establish because of the great variability in the expression of the different diseases in
different host types (Murray, Grootenhuis, Akol, Emery, Shapiro, Moloo, Dar, Bovell &
Paris, 1981; Murray, Morrison & Whitelaw, 1982; Mehlitz, 1982), or even in the same
host type under different conditions of stress (Hoare, 1972). From the point of view of
disease epidemiology or epizootiology sterile immunity may be no different from
premunity (if premune animals contain too few trypanosomes to infect the vectors),
although in each case resistance is mostly or only to homologous strains. Thus the
figures for these parameters are crude average, or intermediate, values from the wide
ranges given in Hoare's book and elsewhere (Hoare, 1972).
The parameters for transmission from the vector to the vertebrate (b) are the product
of the likelihood that infected flies emit trypanosomes during any particular blood meal
(Otieno & Darji, 1979) and the probability that such trypanosomes eventually establish
themselves in vertebrates (data from Wilson, Dar & Paris (1972); for more details see
Rogers (1980)). The parameters for transmission of T. vivax and T. congolense infections
from the vertebrate to the vector (c) are derived from an analysis of data on fly and
vertebrate infection rates collected in the Serengeti National Park (Rogers & Boreham,
1973; Geigy, Mwambu & Kauffmann, 1971). Analysis of the fly infection rate data
provides an estimate of the rate at which flies are developing mature infections, whilst
the wildlife sample indicates what proportion of fly blood meals actually contain
trypanosomes; a comparison of these figures shows what proportion of blood meals
infected with each trypanosome species will eventually give rise to mature infections in
the vectors. There was a slight error in the original analysis of these data (Rogers,
1980); Table 2 gives the correct figures. In the case of T. brucei infections the parameter
c' is estimated directly from laboratory experiments (Wijers, 1960; Watson, 1963), and
it is assumed that flies can be infected only on the first day of emergence (i.e. t = 1 in
equation (18)). To some extent changes in t and c' tend to compensate for each other,
since a larger value of t is associated with a smaller average value of c'.
Finally, the human, animal and fly population sizes given in Table 2 were taken
as those 'typical' of a village situation in West Africa (the figures chosen refer to
Degbezere in Ivory Coast, where human sleeping sickness is endemic; Baldry, (1980)).
In the model no attempts were made to allow for a possible increase in the feeding
rate of infected flies (Jenni, Molyneux, Livesey & Galun, 1980; Roberts, 1981: but see
also Moloo, 1983; Moloo & Dar, 1985), nor for different mortality rates of male and
female tsetse; these refinements were considered premature at this initial stage of
modelling.
Equilibrium prevalences, basic rates of reproduction and transmission thresholds
Table 3 lists the predicted equilibrium prevalences, basic rates of reproduction and
thresholds for transmission of T. vivax, T. congolense and T. brucei using the figures in
Table 2. When species 1 is not involved (i.e. in the cases of T. vivax and T. congolense in
the present example) basic rates of reproduction and transmission thresholds were
calculated from the formulae in Table 1 setting the transmission parameter b1 = 0.
Prevalences are quite dissimilar in the vectors and vertebrates, and reflect the pattern
204 D. J. ROGERS

Table 3. Predictions from the trypanosomiasis model, using the parameters and variables
of Table 2
(Basic reproductive rates and thresholds for transmission were calculated using the formulae
in Table I. For T. vivax and T. congolense infections 6, = 0. The last column shows what
would happen if the alternative species was incapable of supporting the disease (i.e. b2 = 0 for
species 1 or b1 = 0 for species 2).)
T. brucei

T. vivax T. congolense Species 1 + 2 Species 1 or 2

Equilibrium trypanosomiasis — — 7-0% —
prevalence in species 1 (man)
Equilibrium trypanosomiasis 47-0% 45-8% 28-7% —
prevalence in species 2 (animals)
Equilibrium trypanosomiasis 24-2% 3-4% 0-61% —
prevalence in tsetse
Basic reproductive rate 388-2 64-4 2-65 Species 1 = 011
Species 2 = 2-54
Threshold for transmission — — 6-29 Species 1 = 1530
(flies/host), species 1
Threshold for transmission 0-26 1-55 37-7 Species 2 = 393
(Mies/host), species 2

very often seen in field data (i.e. very low levels of T. brucei infections in flies and
high infection rates in animals; more equal infection rates with T. vivax). The basic
reproductive rate of T. vivax is more that 150 times greater than that of T. brucei in the
same host type, whilst the transmission threshold is correspondingly lower. In all cases
T. congolense occupies an intermediate position. Such predictions coincide with field
experience where T. vivax occurs more or less wherever flies are found (Table 3 suggests
that this disease persists when there is only one fly for every three to four animals),
whilst T. brucei has a more restricted distribution (about 40 flies per animal are required
for this disease to be maintained. Table 3). The final column in Table 3 shows the
separate contributions of the two host species to the overall basic rate of reproduction
of T. brucei. Note that the contribution of species 1 (i.e. humans) falls well below unity,
indicating that this disease cannot be maintained in the human hosts alone. Species 2
acts as a reservoir of the human disease. The last two figures in this column are the
thresholds for transmission in each host in the absence of the disease in the other host.
Human hosts make very little difference to the threshold for transmission in the animal
hosts, but when such animals are uninfected more than 150 flies per human are required
to maintain human sleeping sickness, a more than 24-fold increase over the ratio
required when both species are infected.
The effects of vector density on equilibrium prevalence levels of a human-infective
T. brucei are shown in Fig. 4 (all other parameters and variables as in Table 2). Below a
density of about 2000 flies the disease cannot be maintained. Above this value the
prevalence in the more favoured host (e.g. pigs in a West African village situation)
increases more rapidly than in the less favoured hosts (humans). The upper limit to
prevalence is set mainly by the rate of recovery from infections (parameter r) and the
durations of the incubation and immune periods (parameters i and v). The more
favoured host obviously reaches this upper limit (as fly numbers increase) before the
less favoured host. For the former host type, therefore, there may not be a pronounced
relationship between prevalence and vector densitj' over the same range of vector
 A model for the African trypanosomiases 205

_ 40 Animals
r

30

20

2 10

Flies

5000 10000 15OO0

I I
Tsetse density
Fig. 4. The effect of changing tsetse population size on equilibrium prevalences of Trypano-
soma brucei, predicted by the trypanosomiasis model. The threshold for transmission is
approximately 2000flies.The large dot is situated at the equilibrium prevalence point for the
human hosts when 7% of blood meals are taken from them. The arrow points to prevalences
which would be reached if flies took 70% of their meals from humans whilst retaining the
high infection levels acquired from the previously favoured animal hosts. The horizontal
lines beneath the abscissa indicate the ranges of fly densities used in the seasonal incidence
model shown in Fig. 6.

numbers where the less favoured host shows a gradual increase in prevalence. For
example, in the range of vector densities of 5000-15000 in Fig. 4, there is a 39%
increase in prevalence in the more favoured host, but a 214% increase in prevalence of
the human host. At lower vector densities, however, prevalence in the former host type
also decreases rapidly as vector numbers decline (e.g. 2000-4000 flies in Fig. 4).
The effect of suddenly diminishing the abundance of the more favoured host is also
indicated in Fig. 4. This was an attempt to mimic the sudden abandonment of pig-
rearing that occasionally occurs in some villages in West Africa (either because of
disease or through a sudden shift of cultural practices). In such cases the domestic
animal population, when present, supports a large fly population (Rogers et al. 1984,
Fig. 4), but at the same time constitutes a 'protective screen' providing most fly blood
meals and thus protecting humans from large numbers of fly bites and therefore, it is
thought, from infections (Gouteux Laveissierre & Boreham, 19826). The model sug-
gests, however, that this is an over-simplification because in an equilibrium situation
(and using the data of Table 2) humans are equally infected whether they provide a
small or large percentage of fly blood meals (e.g. 7% or 70%). The reason for this is that
when there is a small number of domestic animals present each one supports a large
number of flies; animal and therefore fly infection rates are high, and the few flies that
feed on man cause a 2-6% infection rate in humans. When the domestic animals are
removed so that the flies rely mostly on human hosts, each human is supporting fewer
flies than did each animal so that in the long-term (i.e. at equilibrium) human infection
rate instead of rising actually falls a little, to 2-0%. In the short-term, however,
immediately after the removal of the animal population, flies are both heavily infected
and take a large proportion of their meals from humans. Human infection rates would
then temporarily rise towards 36% (the equilibrium set by this high fly infection rate
and feeding pattern) unless there is a rapid turnover of the fly population (quickly
reducing infection rate) or a rapid diminution of fly numbers. A 'mini-epidemic' of
206 D. J. ROGERS
human sleeping sickness could therefore accompany the sudden removal of domestic
animals from a village.
Effects of changing the values of single parameters
The sensitivity of T. brucei equilibrium prevalences to changes in the values of the
various parameters was examined by halving or doubling each value in turn, whilst
holding all others constant (the starting values are those listed in Table 2 and fly
density was assumed to be 5000; additional maximum limits of 10 were put on b1
and b2, and of 09995 on p1 and p2). The results are shown in Fig. 5 as the log of the
proportional change in prevalence (i.e. log (prevalence after/prevalence before change))
in each species involved (the log ratio was chosen so that equal proportional increases or
decreases give histograms of equal height in Fig. 5). The thin horizontal lines in Fig. 5
mark the points at which prevalences are twice or one half their original values so that
when the histograms exceed these limits there is a disproportional change in disease
prevalence as that particular parameter value is halved or doubled. In general,
changing parameter values has a rather more marked effect on changes in disease
prevalence in the less-favoured of the two vertebrate species (species 1 in Fig. 5) because
this species is farther from the level of maximum disease prevalence (see Fig. 4).
Doubling fly mortality rate, feeding cycle duration or vector incubation periods and
halving the duration of infection in the more favoured vertebrate species, or the
transmission parameter c', consistently cause a disproportional decrease in prevalence
levels (not unexpectedly since a number of these terms appear as powers or exponents in
the various equations). Other terms connected with feeding rates (pl and p2 in Fig. 5)
also frequently have a major effect on disease prevalence. The remaining terms in Fig. 5
are not unimportant; rather they determine the equilibrium disease prevalences around
which the changes represented in the figure occur.
Predicted seasonal incidences of trypanosomiasis
Finally, the equations were used to explore the changing seasonal incidences of
T. brucei infections when fly numbers were made to vary in an annual cycle of
abundance around their mean level. These cycles were induced in a model population
by varying seasonally fly mortality rate over a range similar to that encountered in the
field (Rogers et al. 1984); changing mortality affects not only the size of the fly
population but also its infection rate. A minimal amount of density dependence was
included in the fly population model to prevent the population drifting away from its
initial level (Varley, Gradwell & Hassell, 1973); t h e survival function chosen was that of
Maynard-Smith & Slatkin (1973). The model, which was based on equations (7)—(9)
(modified according to equation (18)), using the various values for T. brucei in Table 2
and initial conditions determined by appropriate equilibrium equations, allowed for
incubation periods in the two vertebrate species so that the numbers of individuals
added in any time interval to the infectious categories depended upon the conditions of
the various populations an incubation period before. In the case of the tsetse population
the number of newly emerged flies which were susceptible to infection and could later be
added to the infected category was determined by the adult fly population one puparial
and one incubation period previously, assuming an equal sex ratio and 50% losses in
the puparial stages.
The model was written (using a time step of two days which gave results indis-
tinguishable from daily time-steps) so that the effects of seasonal variation in fly
numbers around different average levels could be investigated. Two examples are given
 A model for the African trypanosomiases 207

-L J_ J_
d i. /•j v, v2 f bt t>2 c
06
Prevalence increases
0-4 - Species 1

0-2 - r -
]
-0-2

-0-4
-0-6 -
1 -0-8 -
(0
'prevalence beifore'

Prevalenc e d ecreases
-10

-1-2
—oo ^
-1-4 V 1 1 1 1 1 1
Pi Pi u d i, Species 2
0-4

I 0-2
Increase
prevalence a1ter cf

CO

-0-2 Decrease

-0-4

-0-6

Fig. 5. The effect of having ( • ) or doubling (II) the values of the various parameters of the
trypanosomiasis model on equilibrium disease prevalences in each of the species involved.
Results are expressed as the logarithm of the ratio of prevalence after the change to
prevalence before the change (which shows equal geometric changes equally). Parameter
descriptions and starting values are given in Table 2. The thin horizontal lines above and
below 00 indicate a doubling or halving of disease prevalence.

in Fig. 6. When variation in fly numbers occurs near the threshold for transmission (Fig.
6 a) disease incidence is seasonally very variable and may reach peak levels higher than
the equilibrium prevalence levels that would be shown if fly density remained at the
208 D. J. ROGERS

40 (a)
Animals
s
30

20
/
^ A /\ /
10 Humans
Flies x 10
0
i i i

Animals

_ Humans

Flies X 10

I 12 000
S 8000
8 4000 I M M M D/J D/J 3 D/J D/J
Year

Fig. 6. Changing seasonal incidence of trypanosomiasis predicted by the model when tsetse
density is allowed to vary by seasonally varying daily adult mortality rate (the resulting
changes inflydensity are shown by the histograms beneath each part of the figure). When fly
numbers are near the threshold for transmission (a) seasonal disease incidence is variable
and may exceed equilibrium prevalence at the same average density. Seasonal incidence is
much less variable (at least in the more favoured host) at higher averageflydensities (6). The
graph records the changes over four years.

average of the values used; this is because during the period of lowflydensity most
vertebrates recover from infection and lose their immunity so that they are all
susceptible by the timeflynumbers increase once more. Whenflynumbers vary around
a much higher average level (Fig. 6b) seasonal incidence in the more favoured host is
much less variable (because such hosts are either incubating, infectious or immune and
there are few susceptibles to add to their number), although that in the less favoured
host may continue to vary. Peak disease incidence in the vertebrates lags behind peak
fly abundance, and that in the less favoured host lags behind that of the more favoured
host. Thus each year an increase inflynumbers initially increases the level of infection
 A model for the African trypanosomiases 209
more quickly in the more favoured hosts; this results in a higher infection rate in the
flies which in turn infect more of the less favoured host species.

DISCUSSION AND CONCLUSIONS

A model incorporating most of the features of a vector-borne disease allows us to
assess the importance of each parameter in turn in determining disease prevalence.
Whilst maximum disease prevalence in vertebrates (i.e. at high vector densities) in the
trypanosomiasis model are set by parameters determining the durations of incubation,
infectious and immune periods, prevalences at lower vector densities are rather more
sensitive to changes in other parameters, especially those connected with vector
population behaviour {plt p2 and d), survival rate (u) or tendency to develop infection
(c', T). Characteristics of the host (rl, r2) or of the host^trypanosome interaction (61; 62)
are the next most important parameters. Research effort should therefore be concen-
trated in these areas, especially to establish the variability of the various parameters
involved. Feeding cycle duration and fly mortality rate are known to vary seasonally
(Jackson, 1933; Rogers et al. 1984) and this will dramatically affect transmission rates.
Calculated disease transmission thresholds and basic rates of reproduction (Table 3)
emphasize the difficulty of eradicating a disease such as that caused by T. vivax
(threshold of 026 flies per animal) by anything other than total elimination of the tsetse
vectors; T. congolense eradication presents an almost equally severe problem. The
prospects for control of T. brucei are rather better. The much higher threshold predicted
by the model probably explains the restricted geographical distribution of this group of
trypanosomes in Africa, although the organism's ability to persist in animals other than
man in a more chronic form than the other main trypanosome species (Hoare, 1972)
again mitigates against its rapid elimination during control campaigns.
After many years of debate (Duggan, 1970; Hoare, 1972) it is now becoming clear
that both wild and domestic animals can harbour what appear to be human-infective
forms of T. brucei in West Africa (Mehlitz, 1986), although the precise significance of
this reservoir remains to be established. The trypanosomiasis model described here
suggests that human sleeping sickness in West Africa cannot be maintained without
such a reservoir. The basic reproductive rate of the disease in the human population is
very much less than 10, and the threshold for transmission (more than 150fliesper
person, Table 3) seems unlikely to be encountered in all but exceptional circumstances.
Morris (1946) surveyed the relative effectiveness of anti-tsetse and human population
surveillance and treatment campaigns in the eradication of human sleeping sickness in
West Africa and concluded that the anti-tsetse measures were more certain of success;
failures of the other approach were attributed to inaccurate diagnosis of human cases,
or to infected fly populations continuing transmission to man, although a non-human
animal reservoir now also seems a contributory cause.
The low value of the basic reproductive rate of the disease in the human population
means that infection levels in humans are determined almost entirely by characteristics
of the non-human reservoir. This implies that a greater reduction in human disease may
be achieved by treating the animal reservoir, if at all possible, rather than the humans.
Domestic animals, when infected, seem to be more frequently parasitaemic than wild
animals (Mehlitz, 1986; Ashcroft, Burtt & Fairbairn, 1959), and their proximity to man
increases the risk of transmission to humans. Village-based populations of animals such
as pigs appear to provide an ideal 'scenario' for transmission of sleeping sickness.
G. palpalis palpalis (Robineau-Desvoidy) in West African villages with pigs take only
210 D. J. ROGERS

7 % of their blood meals from human hosts (and 75% from pigs), whilst flies caught in
the surrounding coffee and cocoa plantations take 68 % of their meals from man (and
only 14% from pigs) (Azodoga Seketeli, personal communication). In either situation
the trypanosomiasis model described here suggests that human infection rates would
be of the order of 2 % (see text referring to Fig. 4). There is, however, a constant
interchange of village and plantation flies (Randolph & Rogers, 1984) so that flies
infected in the village may transmit these infections when they feed on man in
plantations. The idea of a ' protective screen' provided by pigs in villages is therefore
something of an illusion, since the 'screen' may also be the major source of infection
(Gouteux & Laveissiere, 1982; Gouteux et al. 19826).
The efficacy of sleeping sickness control campaigns may also be more readily assessed
by monitoring the non-human hosts, since it is easier to establish statistically signifi-
cant changes of higher rather than very low percentage values with any particular
sample size.
There still remains the much-neglected phenomenon of mechanical transmission of
trypanosomiasis by tsetse and other biting flies. This infection route (historically the
first to be discovered, since Bruce's original experiments must have involved mechani-
cal rather than cyclical transmission by tsetse; Hoare (1972), p. 496) is remarkably
efficient in experimental situations (Bruce, Hamerton, Bateman & Mackie, 1910;
Bailey, 1966; Gingrich, Roberts & Macken, 1983) and must also be important in at least
some field situations (Wells, 1972). No useful attempt at modelling mechanical
transmission can really be made until its importance relative to cyclical transmission is
firmly established in the field.
Finally it is hoped that a model such as the one described here will contribute to
future studies of the transmission of the African trypanosomiases from one host to
another. Models for disease transmission help to establish a correct perspective view of
an often complex situation and, with correct parameter values, will provide estimates of
the transmission thresholds that Morris (1946) saw as the key to disease eradication. As
he concluded, such problems 'will be solved only by research in the field involving first
and foremost a full appreciation of that complex of factors labelled, so conveniently,
epidemiology' (Morris, 1946, p. 244).
This paper is an extended version of an article prepared for a W.H.O. Expert Committee on the
epidemiology and control of African trypanosomiasis which met in October 1985. I thank Drs
Peter de Raadt and Felix Kuzoe for their continued interest in this work, Professor Sir Richard
South wood for providing research facilities in the Department of Zoology, and Dr Sarah Randolph for
reading the manuscript. Mike Amphlett drew all the text figures. Initial microcomputer facilities were
provided by a grant from the Overseas Development Administration (Research Scheme R3702) and
greatly improved with financial support from the U.N.D.P./World Bank/W.H.O. Special Programme
for Research and Training in Tropical Diseases (project no. 860010).

REFERENCES
ANDERSON, R. M. (1982). Population Dynamics of Infectious Diseases. London: Chapman and Hall.
ABON, J. L. & MAY, R. M. (1982). The population dynamics of malaria. In Population Dynamics of
Infectious Diseases (ed. R. M. Anderson), pp. 139-79. London: Chapman and Hall.
ASHCEOFT, M. T., BURTT, E. & FAIRBAIRN, H. (1959). The experimental infection of some African wild
animals with Trypanosoma rhodesiense, T. brucei and T. congolense. Annals of Tropical Medicine and
Parasitology 53, 147-61.
BAILEY, N. M. (1966). The mechanical transmission by Olossina morsitans of Trypanosoma brucei
subgroup trypanosomes. In Annual Report 1965, East Africa Trypanosomiasis Research Organiza-
tion, pp. 24-7. Entebbe: Government Printer.
 A model for the African trypanosomiases 211
BAILEY, N. T. J. (1982). The Biomalhematics of Malaria. London: Griffin.
BALDRY, D. A. T. (1980). Local distribution and ecology of Glossina palpalis and G. tachinoides in
forest foci of West African human trypanosomiasis, with special reference to associations between
peri-domestic tsetse and their hosts. Insect Science and its Application 1, 85-93.
BRUCE, D., HAMERTON, A. E., BATEMAN, H. R. & MACKIE, F. P. (1910). Mechanical transmission of
Sleeping Sickness by the tsetse fly. Proceedings of the Royal Society of London, B 82, 498—501.
CAWDEKY, M. J. H. (1958). Estimation of trypanosome challenge. In Annual Report 1956-57. East
African Trypanosomiasis Research Organization, p. 18. Nairobi: East African High Commission.
DUGGAN, A. J. (1970). An historical perspective. In The African Trypanosomiases (ed. H. W. Mulli-
gan), pp. xli-lxxxviii. London: George Allen and Unwin Ltd.
FAO-WHO-OIE. (1982) Animal Health Yearbook, 1981. Rome: FAO.
GEIGY, R., MWAMBU, P. M. & KAUFFMANN, M. (1971). Sleeping sickness survey in Musoma district,
Tanzania. IV. Examination of wild animals as potential reservoir of T. rhodesiense. Acta Tropica 28,
211-20.
GINGRICH, J. B., WARD, R. A., MACKEN, L. M. & SCHOENBECHLER, M. J. (1982a). Trypanosoma brucei
rhodesiense (Trypanosomatidae): factors influencing infection rates of a recent human isolate in the
tsetse Glossina morsitans (Diptera, Glossinidae). Journal of Medical Entomology 19, 268-74.
GINGRICH, J. B., WARD, R. A., MACKEN, L. M. & ESSER, K. M. (19826). African sleeping sickness: new
evidence that mature tsetseflies(Glossina morsitans) can become potent vectors. Transactions of the
Royal Society of Tropical Medicine and Hygiene 76, 479-81.
GINGRICH, J. B., ROBERTS, L. W. & MACKEN, L. M. (1983). Trypanosoma brucei rhodesiense: mechani-
cal transmission by tsetse, Glossina morsitans (Diptera, Glossinidae), in the laboratory. Journal of
Medical Entomology 20, 673-6.
GOUTEUX, J.-P. & LAVEISSIERE, C. (1982). Ecologie des glossines en secteur pre-forestier de Cote
d'lvoire. 4. Dynamique de l'ecodistribution en terroir villageois. Cahiers de I'Office de la Recherche
Scientifique el Technique Outre-Mer, Serie Entomologie Medicate et Parasitologie 20, 199-229.
GOUTEUX, J.-P., LAVEISSIERE, C. & BOREHAM, P. F. L. (1982a). Ecologie des glossines en secteur pre-
forestier de Cote d'lvoire. 2. Les preferences trophiques de Glossina palpalis 8.1. Cahiers de I'Office de
la Recherche Scientifique et Technique Outre-Mer, Serie Entomologie Medicale et Parasitologie 20,
3-18.
GOUTEUX, J.-P., LAVEISSIERE, C. & BOREHAM, P. F. L. (19826). Ecologie des glossines en secteur pre-
forestier de Cote d'lvoire. 3. Les preferences trophiques de Glossina pallicera et G. nigrofusca.
Comparaison avec G. palpalis et implications epidemiologiques. Cahiers de I'Office de la Recherche
Scientifique et Technique Outre-Mer, Serie Entomologie Medicale et Parasitologie 20, 109-24.
HABTEMARIAM, T., RUPPANNER, R., RIEMANN, H. P. & THEIS, J. H. (1983a). An epidemiologic
systems analysis model for African trypanosomiasis. Preventive Veterinary Medicine 1, 125-36.
HABTEMARIAM, T., RUPPANNER, R., RIEMANN, H. P. & THEIS, J. H. (19836). Epidemic and endemic
characteristics of trypanosomiasis in cattle: a simulation model. Preventive Veterinary Medicine 1,
137^5.
HABTEMARIAM, T., RUPPANNER, R., RIEMANN, H. P. & THEIS, J. H. (1983c). Evaluation of trypano-
somiasis control alternatives using an epidemiologic simulation model. Preventive Veterinary
Medicine 1, 147-56.
HOARE, C. A. (1972). The Trypanosomes of Mammals. Oxford: Blackwell Scientific Publications.
JACKSON, C. H. N. (1933). The causes and implications of hunger in tsetse flies. Bulletin of Entomo-
logical Research 24, 443-82.
JENNI, L., MOLYNEUX, D. H., LIVESEY, J. L. & GALUN, R. (1980). Feeding behaviour of tsetse flies
infected with salivarian trypanosomes. Nature, London 283, 383-5.
MACDONALD, G. (1952). The analysis of equilibrium in malaria. Tropical Diseases Bulletin 49, 813-28.
MACDONALD, G. (1957). The Epidemiology and Control of Malaria. London: Oxford University Press.
MACDONALD, G. (1973). Dynamics of Tropical Disease. (Collected papers, edited by L. J. Bruce-Chwatt
and V. J. Glanville.) London: Oxford University Press.
MAYNABD SMITH, J. & SLATKIN, M. (1973). The stability of predator-prey systems. Ecology 54,
384-91.
MEHLITZ, D. (1982). Trypanosomes in African wild mammals. In Perspectives in Trypanosomiasis
Research (ed. J. R. Baker), pp. 25-35. Chichester: Research Studies Press.
MEHLITZ, D. (1986). Le reservoir animal de la maladie du sommeil a Trypanosoma brucei gambiense.
Etudes et Syntheses de I'I.E.M.V.T. 18. Maisons-Alfort: Institut d'Elevage et de Medecine
Veterinaire des Pays Tropicaux.
212 D. J. ROGERS
MILLIGAN, P. J. M. & BAKER, R. D. (1988). A model of tsetse-transmitted animal trypanosomiasis.
Parasitology 96, 211-39.
MOLOO, S. K. (1983). Feeding behaviour of Glossina morsitans morsitans infected with Trypanosoma,
vivax, T. congolense or T. brucei. Parasitology 86, 51-6.
MOLOO, S. K. & DAR, F. (1985). Probing by Glossina morsitans centralis infected with pathogenic
Trypanosoma species. Transactions of the Royal Society of Tropical Medicine and Hygiene 79, 119.
MORRIS, K. R. S. (1946). The control of trypanosomiasis by entomological means. Bulletin of
Entomological Research 37, 201-50.
MURRAY, M. & GRAY, A. R. (1984). The current situation on animal trypanosomiasis in Africa.
Preventive Veterinary Medicine 2, 23-30.
MURRAY, M., GROOTENHUIS, J. G., AKOL, G. W. O., EMERY, D. L., SHAPIRO, S. Z., MOLOO, S. K., DAR,
F., BOVELL, D. L. & PARIS, J. (1981). Potential application of research on African trypanosomiases
in wildlife and preliminary studies on animals exposed to tsetse infected with Trypanosoma
congolense. In Wildlife Disease Research and Economic Development (ed. E. Karstad, B. Nestel and
M. Graham), pp. 40-5. Ottawa: International Development Research Centre.
MURRAY, M., MORRISON, W. I. & WHITELAW, D. D. (1982). Host susceptibility to African trypano-
somiasis: trypanotolerance. Advances in Parasitology 21, 1-68.
OTIENO, L. H. & DARJI, N. (1979). The abundance of pathogenic African trypanosomes in the salivary
secretions of wild Glossina pallidipes. Annals of Tropical Medicine and Parasitology 73, 583-8.
RANDOLPH, S. E. & ROGERS, D. J. (1984). Movement patterns of the tsetse fly Glossina palpalis
palpalis (Robineau-Desvoidy) around villages in the pre-forest zone of Ivory Coast. Bulletin of
Entomological Research 74, 689-705.
ROBERTS, L. W. (1981). Probing by Glossina morsitans morsitans and transmission of Trypanosoma
(Nannomonas) congolense. American Journal of Tropical Medicine and Hygiene 30, 948—51.
ROGERS, D. (1977). Study of a natural population of Glossina fuscipes fusdpes Newstead and a model
of fly movement. Journal of Animal Ecology 46, 309-30.
ROGERS, D. J. (1980). Epizootiology: the tsetse-cattle interface. Report of the Expert Consultation on
Research on Trypanosomiases, 1-5 October 1979, Appendix VI, pp. 61-7. Rome: FAO (AGA-801).
ROGERS, D. J. (1985). Trypanosomiasis 'risk' or 'challenge': a review. Acta Tropica 42, 5—23.
ROGERS, D. J. & BOREHAM, P. F. L. (1973). Sleeping sickness survey in the Serengeti area (Tanzania)
1971. II. The vector role of Glossina swynnertoni Austen. Acta Tropica 30, 24-35.
ROGERS, D. J., RANDOLPH, S. E. & KUZOE, F. A. S. (1984). Local variation in the population
dynamics of Glossina palpalis palpalis (Robineau-Desvoidy) (Diptera: Glossinidae). I. Natural
population regulation. Bulletin of Entomological Research 74, 403-23.
Ross, R. (1911). The Prevention of Malaria, 2nd Ed. London: Murray.
Ross, R. (1916). An application of the theory of probabilities to the study of a priori pathometry. I.
Proceedings of the Royal Society of London, A 92, 204-30.
SACHS, R., MEHLITZ, D. & STAAK, C. (1980). Host preference and trypanosome infection of three tsetse
species (Glossina palpalis, G. pallicera and G. nigrofusca) in rain forest zones of Liberia, West Africa.
In Tenth International Congress on Tropical Medicine and Malaria, Manila, Philippines, pp. 216-17.
VAKLEY, G. C, GRADWELL, G. R. & HASSELL, M. P. (1973). Insect Population Ecology. Oxford:
Blackwell Scientific Publications.
WATSON, H. J. C. (1963). The domestic pig as a reservoir of T. gambiense. In International Scientific
Committee for Trypanosomiasis Research, Ninth Meeting, Conakry, 1962, p. 327. Lagos: Commission
for Technical Co-operation in Africa.
WELLS, E. A. (1972). The importance of mechanical transmission in the epidemiology of nagana: a
review. Tropical Animal Health and Production 4, 74-88.
WILLETT, K. C. (1972). An observation on the unexpected frequency of some multiple infections.
Bulletin of the World Health Organization 47, 747-9.
WILSON, A. J., DAR, F. K. & PARIS, J. (1972). A study on the transmission of salivarian trypanosomes
isolated from wild tsetse flies. Tropical Animal Health and Production 4, 14-22.
W.H.O. (1986). Epidemiology and control of African trypanosomiasis. Report of a W.H.O. Expert
Committee. W.H.O. Technical Report Series, no. 739. Geneva: World Health Organization.
WIJERS, D. J. B. (1960). The importance of the age of G. palpalis at the time of the infective feed with
T. gambiense. In International Scientific Committee for Trypanosomiasis Research, Seventh Meeting,
Bruxelles, 1958, pp. 319-20. London: Commission for Technical Co-operation in Africa South of the
Sahara.
Printed in Great Britain