The interruption of the transmission of soil
transmitted helminths by repeated mass
chemotherapy targeted at school aged children
James Truscott1, T. Déirdre Hollingsworth234 and Roy Anderson1
1
London Centre for Neglected Tropical Disease Research, Department of Infectious
Disease Epidemiology, Faculty of Medicine, Imperial College London, St Marys
Campus, Norfolk Place, London W2 1PG
2
Mathematics Institute, University of Warwick, Coventry, CV4 7AL, UK
3
School of Life Sciences, University of Warwick, Coventry, CV4 7AL, UK
4
Department of Clinical Sciences, Liverpool School of Tropical Medicine, Pembroke
Place, Liverpool, L3 5QA, UK
Supporting Information
A: Calculating the growth rate of the parasite population under
regular treatment
The application of the concept of reproduction number to STHs is a little problematic.
The mechanism of sexual reproduction generates non-linear behaviour for the very
low worm burdens at the very threshold where linear behaviour is necessary for the
definition of classical reproduction numbers (the ‘breakpoint’). In this first analysis,
we neglect the issue of sexual reproduction and assume that all eggs produced by
female worms are fertilised. The effects of this assumption are analysed in more detail
in the main text.
The equations governing the two-component plus reservoir system are
dM c
 cl   M c ,
dt
dM a
 al   M a ,
dt
R0 
d
l
 f  M c ; k , z  nc p  f  M a ; k , z  (1  nc )(1  p)   l
dt
 z   c nc p   a na q  
Where Mc and Ma are the mean worm burden in children and adults respectively and l
is the level of infection in the environment. Further details are given in the main text.
Linearised around zero worm burden in both adults and children and scaling time by
worm lifespan σ, we get
0
c  M c 
 M c   1
d
 

Ma   
0
1
 a 
Ma 


dt 
 


 l   R0 A /  D R0 B /  D 1/   l 
where    /  . The contact rates βc and βa are scaled by σ and A  nc p ,
B  na (1  p) and D  c A  a B . The eigen-value/eigen-vector pairs for this system
are i , v i , where 1  2  3 . The eigen-values and vectors are given by
1 
(1   )  (1   ) 2  4 ( R0  1)
2  1
3  (1  1/  )  1
2
and
 c 
 B
 c 


 


v1    a  , v 2    A  , v3    a 
   1
 0 
   1
 1 
 
 3 
The evolution of the system can be described by
 M c (t ) 


wi eit vi
 M a (t )   X(t )  
i
 l (t ) 


We now let,
M   v1 , v2 , v3 
Then M transforms coordinates in the eigen-vector space into real space (Mc, Ma and
l) and M-1 back the other way. To describe the dynamics of the periodic treatment
regime, we need two other operators. Firstly a treatment operator, T;
T  diag(1   ,1,1)
which acts on real state vector and describes the reduction of children’s worm burden
by 1   . Also a growth operator,
G  diag(exp(
1 ),exp(  ),exp(3 ))
which acts on the eigen-state vector and gives the state after a period of growth of
length τ. So the sequence of a complete cycle of treatment is
ω n 1  M 1TMG (tl )ω n  Hω n
where ω is the state of the system in eigen-state space. The parameter tl is the interval
between treatment, scaled by worm lifespan, σ. Hence the dynamics of the system is
dependent on the largest eigen-value/vector pair of the mapping matrix H.
Hu  qu
When k  1 , the parasite will be eliminated from the system. The matrix, H, is given
by
Br 


 c
 rc
 1  rc

c


  a  c
 a  c 
H 
1  (1  rc )

 G ( )
D
D


 r  (1   )  Brc (1   ) 1  r  (1   ) 
c
 c

c


where rc  c A / D and   (1  3 ) / (1  3 ) . The eigenvector q tells us the growth
in the parasite population over a single treatment interval. One interpretation of q is as
a crude effective reproduction number, with each worm generating q off-spring
worms after an interval of τ years. In terms of an equivalent continuous growth rate,
r  ln(q) / tl or in dimensional units
ln( q )
r
Tint
At a time 0  t   , the state of the system in real coordinates is proportional to
MG (t )u . Hence
3
3
i, j
i
M n (t )   M niGij (t )u j   M ni exp(it )ui ,
where n=1 for children and 2 for adults.
Calculating the effective reproduction number, Re
If we know the number of eggs generated by worms in both types of hosts and the
relative worm populations within each type of host, we can calculate the total number
of new worms generated and hence the effective reproduction number. Under
treatment, the number of eggs generated by a worm will depend on when it is ‘born’
in the treatment cycle (between 0 and tl). Egg production is assumed constant with
time but the probability of a worm surviving treatment for time t is exp(t ) , with a
factor of 1-γ included for each treatment round. The mean life of a worm is the timeintegral of the survival function,
 exp(tl ) exp(t0 )
Sc (t0 )  1 
 1  L(tl ) exp(t0 )
[1  (1   )etl ]
where 0  t0  tl is the time of birth of the worm in the treatment cycle. For worms in
adults, γ=0, and S a (t0 )  1 . The output from each worm can then develop into new
worms of either type. Hence we can define an effective R for a worm born at time t 0
in either a child or adult
 R0 A(  c   a )
  R0 rc (  c   a ) (1  L(t ) exp(t )) 
l
0
(1

L
(
t
)
exp(
t
))

l
0

 
c
D


R (t0 )  



R
B
(



)
R
(1

r
)(



)


0
c
a
0
c
c
a

 


D
 
a

The number of worms born at any time is the incidence, I (t ) , and depends on the
time within the treatment cycle.
dM k
 Ik  M k
dt
where k is either c or a (child or adult). So
d
I k (t )  e  t et M k (t )
dt
Substituting the expression for the evolution of the worm burdens gives
3
I k (t )   M ki ui (1  i ) exp(i t )   M kiui (1  i ) exp(it )
i 1,3
i
for host type k. Note that the i=2 term doesn’t contribute since 2  1. As a result,
the above expression for Ik(t) can be written in terms of vectors and matrices of degree
two. That is
2
I k (t )   M ki* ui* (1  i* ) exp(i*t )
i
where M* is made up of the first and second row and first and third columns of M and
u* comprises the first and third elements of u. The vector of eigen-values is now made
up of λ*   1 , 3  .
The overall reproductive number, Re, is the average R over a treatment cycle,
weighted by the incidence of new worms in each type of host at any given time.
tl
2
s 0
j
  R ( s) I
j
Re 
tl
j
( s )ds
2
 I
k
( s )ds
s 0 k
This expression can be considerably simplified. The definition of R(t) can be split into
a constant and a time-dependent term,
 r /    r L(t ) /  c  
a
b
R (t0 )  R0 (  c   a )  c c    c l
 exp(t0 )  R  R (t0 )

0
 
 (1  rc ) /  a  
Note also that from the definition of M above, M ij*  i . As a result, effective R can
be expressed
2


*

i ui* e(1i )tl  1 

Re  R0 1  rc L(tl ) 2
*
*
(1


)


i ui*  * i ei tl  1 

i


Finally, we need to calculate u*, which comprises the first and third elements of u
defined by Hu  qu . By scaling u, the simplest form for u* is




  (e2 tl  q) 
.
u*  
 (1   )(e1*tl  q) 


The key stability indicators for this system are therefore expressible purely in terms of
5 non-dimensional parameters; R0, γ, rc, tl and ε.
*
B: Comparing largest eigenvector q with Re
In Section A, above, we calculated two complementary measures for the stability of
the parasite population in the absence of sexual reproduction. The effective
reproduction number, Re, captures the ability of the worms to generate new adults
within their reproductive lifetime while being subject to regular treatment. It captures
the essential natural history of the parasite but no information about time scale of its
constituent stages. The parameter q represents the growth of the parasite population as
a whole in the interval between treatments. It captures the bulk behaviour of the
parasite population and the timescale on which the changes happen. As shown in
Error! Reference source not found., Re and q match qualitatively for variation in
parameters that don’t affect relative timescales of parasite development, such as R0,
level of treatment (γ) and relative contribution of children (rc). However, for the
parameter ε, that governs the relative lifespans of the adult and infectious phase of the
parasite, there are significant differences. Error! Reference source not found.A and
B compare the values q and Re against ε and treatment level. Both measures pick up
stabilising effect of a longer reservoir timescale (larger ε). This is a consequence of
the changing distribution of infectious material among the different forms in the
parasite lifecycle and is a feature of the natural history of the parasite. However, only
q picks up on the increasing growth-rate for short reservoir timescales (lower ε) at
lower levels of treatment. Thus the parameter q picks up the increased resilience in
the parasite population; what might be thought of as the faster overall bounce-back
time of the population under regular treatment. This information is absent from Re
since it is purely a timescale phenomenon.
Figure Captions:
Figure 1: Dependence of q (Panel A) and Re (Panel B) on R0 and the effective fraction treated.
Figure 2: Dependence of q (Panel A) and Re (Panel B) on the timescale parameter ε and the effective fraction
treated.