Home Search Collections Journals About Contact us My IOPscience

The effects of delays in radiotherapy treatment on tumour control

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2003 Phys. Med. Biol. 48 139

(http://iopscience.iop.org/0031-9155/48/2/301)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 132.203.227.62
This content was downloaded on 14/06/2014 at 07:38

Please note that terms and conditions apply.

INSTITUTE OF PHYSICS PUBLISHING PHYSICS IN MEDICINE AND BIOLOGY
Phys. Med. Biol. 48 (2003) 139–155 PII: S0031-9155(03)52562-3

The effects of delays in radiotherapy treatment on

tumour control
R M Wyatt1, A H Beddoe1 and R G Dale2
1 Department of Radiotherapy Physics, Queen Elizabeth Hospital, Edgbaston, Birmingham
B15 2TH, UK
2
Department of Radiation Physics and Radiobiology, Charing Cross Hospital, London W6 8RF,
UK

Received 20 August 2002, in final form 7 November 2002

Published 23 December 2002
Online at stacks.iop.org/PMB/48/139

Abstract
There is often a considerable delay from initial tumour diagnosis to the start
of radiotherapy treatment, which may be due to factors such as waiting lists
and referral delays. This paper uses widely published models and clinical
parameters to calculate the effect of delays in treatment on local tumour
control for four different types of tumour—squamous cell carcinoma (head
and neck), breast, cervix and prostate. The Poisson model for tumour control
probability (TCP), an exponential function for tumour growth and the linear
quadratic model of cell kill are used to calculate the change in TCP for
delays between diagnosis and treatment of up to 100 days. Typical values
of the clinical parameters have been taken from the literature; these include
α and β, σ α , tumour size at diagnosis, pre-treatment doubling time, delay
in onset of accelerated repopulation and doubling time during treatment. It
is acknowledged that there are limitations in the reliability of these data for
predicting absolute values of tumour control, but models are still useful for
predicting how changes in treatment parameters are likely to affect the outcome.
It is shown that for fast-growing tumours a delay of 1–2 months can have a
significant adverse effect on the outcome, whereas for slow-growing tumours
such as Ca prostate a delay of a few months does not significantly reduce
the probability of tumour control. These calculations show the importance
of ensuring that delays from diagnosis through to treatment are minimized,
especially for patients with rapidly proliferating tumours.
(Some figures in this article are in colour only in the electronic version)

1. Introduction

Recent increases in waiting times for radiotherapy treatment are a cause for concern in many
UK oncology departments. There have been several clinical studies on the effects of delay
0031-9155/03/020139+17$30.00 © 2003 IOP Publishing Ltd Printed in the UK 139
140 R M Wyatt et al

on outcome in radiotherapy. Most of these have failed to demonstrate a convincing effect of

waiting times on treatment outcome, but a recent paper on the effect of waiting for radiotherapy
in rapidly growing grade III/IV gliomas found that the risk of death increased by 2% for each
day of waiting for radiotherapy (Do et al 2000).
This paper uses widely accepted models and associated clinical parameters to calculate
the effect of delays between tumour diagnosis and treatment on local control probability for
four different types of tumour—squamous cell carcinoma (SCC) of the head and neck, breast,
cervix and prostate. The initial number of clonogens is estimated for typical tumour sizes
at diagnosis and increase in the number during the waiting period is calculated using typical
pre-treatment growth rates as obtained from the literature. In the case of breast cancer, for
which the normal treatment regime is breast-conserving surgery followed by radiotherapy, the
waiting times considered are those between surgery and start of radiotherapy. Tumour growth
following surgery is likely to be considerably faster than undisturbed growth due to improved
oxygenation and this is taken into account for the breast cancer calculations. The Poisson
model for tumour control is then used in combination with the linear quadratic model for cell
survival to investigate how the probability of local control changes with delays in the start of
treatment.

2. Method

We consider here the effect of treating the tumour at different times after positive diagnosis.
Generally, tumours grow more slowly before treatment than afterwards, due to poor
oxygenation. This leads both to high cell loss and a low growth fraction (i.e. a low proportion of
viable cells, which are actually proceeding through the cell cycle and multiplying). Following
irradiation, substantial cell kill occurs, the remaining cells become better oxygenated, and
tumour growth rates increase.
The existence of a lag period before the onset of this accelerated repopulation is probably
due to a slow exponential reduction in the cell loss factor due to improved perfusion during
treatment (Jones and Dale 1995). This delay time, Tdel, has been fairly extensively investigated
from clinical data for head and neck cancers and, where sufficiently long treatment times are
available for analysis, a lag can be detected (Roberts and Hendry 1998). Estimates for the lag
time are mostly in the range of 20–32 days (Roberts and Hendry 1998, 1999). In this paper
a lag time of 27 days is assumed for SCC head and neck, which is an average of the lag time
estimated by Roberts and Hendry (1999) for 2225 patients from four centres, both with and
without taking heterogeneity into account. Published data for cervix cancer do not include
analysis of delay to onset of accelerated repopulation, but a deleterious effect of prolongation
of overall treatment time on the outcome has been reported (Petereit et al 1995). As tumour
doubling times for Ca cervix are similar to SCC head and neck, it seems reasonable to assume
the same lag time to the onset of accelerated repopulation. There is currently no evidence
suggesting the existence of delay time in prostate cancer but, for the sake of commonality with
the head and neck and cervix calculations, a delay time of 27 days is also assumed here. It
should be noted that as prostate tumours are very slow growing, the value assumed for delay
time has almost no influence on the numerical results. There are few data in the literature
relating to the repopulation rate for breast tumours following surgery. Animal models have
shown that partial (Braunschweiger et al 1982) or total (Gunduz et al 1979) tumour resection
resulted in stimulation of cell proliferation. Some authors have suggested that subsequent to
surgery there is a period of tumour dormancy followed by a growing phase (Demichelli et al
1998). However, this is questioned by others (Koscielny and Tubiana 1999). In this paper, the
repopulation rate for breast tumours is assumed to accelerate following surgery due to reduced
Effect of radiotherapy treatment delays 141

cell loss. No time lag is assumed as all tumour cells are removed at once, unlike radiotherapy,
where cell kill is gradual. For all tumour types, the effect of varying the time lag to the onset
of accelerated repopulation is also investigated.
Tumours grow due to the presence of clonogens, which are stem cells able to proliferate
indefinitely to produce a large colony. Tumour growth in the time period between diagnosis
and treatment can be modelled by simply using an exponential model as follows. If the
number of clonogens at diagnosis is N0, then the number at any time afterwards (prior to
medical intervention) is given by
N = N0 2T /Teff (1)
where Teff is the effective doubling time of the clonogens. It should be noted that other
tumour growth models have been considered including the Grompertzian model (Laird 1965).
However, it has been shown that the different models produce growth curves which are very
similar over the restricted growth ranges relevant to this paper, and that in practice it might
be difficult to choose between them (Brown and Rothery 1993). We have therefore chosen to
use the simple exponential growth model in this work.
The tumour control probability, TCP, is the probability, P0 of there being no surviving
clonogens after a course of radiotherapy. The Poisson model may be used to calculate the
TCP in individuals via
TCP = P0 = exp(−
a) (2)
where
a is the expectation value for the number of surviving clonogens (Porter 1980). In
this paper, TCP is expressed as a percentage.
The expected number of surviving clonogens is proportional to the initial number of
clonogens N at the commencement of treatment multiplied by a reduction factor determined
by applying an appropriate clonogen cell kill model. The following calculations use the
widely accepted linear quadratic (LQ) model of cell kill, with an extension to allow for
clonogen proliferation during treatment (Jones and Dale 1999). The surviving fraction of
cells, S, after a course of external beam radiotherapy is given by
S = exp(−{n(αd + βd 2 ) − γ (Te − Tdel )}) (3)
where n is the number of fractions of external beam radiotherapy, d is the dose per fraction
for external beam radiotherapy, α and β are the parameters which determine, respectively, the
initial slope and degree of curvature of the underlying cell-survival curve, Te is the overall
time for external beam radiotherapy, Tdel is the delay time to onset of accelerated repopulation,
and γ is the time factor for accelerated repopulation (days−1). The factor γ is related to the
reciprocal of the tumour doubling time (days) following onset of accelerated repopulation.
Fits of the linear quadratic model to clinical head and neck data (Roberts and Hendry 1998)
give values of tumour doubling time during treatment which are very similar to the tumour
cell doubling time in the absence of cell loss, Tpot, as estimated from assay data. Prolongation
of treatment duration in Ca cervix also has a detrimental effect on the outcome, with a similar,
although slightly lower, loss in tumour control per day to head and neck cancer (Petereit et al
1995). Values of Tpot for Ca cervix are comparable with SCC head and neck (around 3–5 days).
Hence, although Tpot does not appear to be a good predictor of radiotherapy tumour control for
individual patients (Begg et al 1999), it may be considered representative of typical tumour
doubling times during treatment for a particular tumour type. Therefore, the time factor γ was
calculated using γ = 0.693/Tpot.
The LQ model was further extended (Jones and Bleasdale 1994) to incorporate the effects
of low dose rate cervix brachytherapy treatments (assuming brachytherapy is given post
external beam therapy) as
S = exp(−{n(α d + β d 2 ) + (α D + β D 2 f (µTb )) − γ (Te + t − Tdel )}) (4)
142 R M Wyatt et al

where D is the total dose from brachytherapy, t is the time interval between completion of
external beam radiotherapy and brachytherapy, Tb is the duration of brachytherapy, f (µTb )
is a correction for β-mediated cell kill, given by 2/µTb {1 − (1 − e−µTb )/µTb }, and µ is the
recovery constant related to the repair half-time T1/2 by µ = 0.693/T1/2.
Four types of tumour were investigated—SCC (head and neck), breast, cervix and prostate.
The effect on TCP of delays between tumour diagnosis and treatment of up to 100 days was
estimated by assuming a typical number of clonogens at diagnosis (or immediately following
surgery in the case of breast tumours), then calculating the number of clonogens, N, on the
first day of radiotherapy treatment from equation (1). It is not possible to measure directly
the total number of clonogenic cells in a tumour, but there have been several estimates made.
The proportion of clonogenic cells in a tumour has been measured using in vitro assays as
between 1/100 to 1/100,000 (Hill and Milas 1989, Peters 1990) and has also been estimated
from TCP data as being around 1/1000 (Brenner 1993). Hermens and Barensden (1969)
estimated total tumour cell density by DNA content to be approximately 5 × 108 cm−3,
whilst Brenner et al (1989) state that the total cell density in a tumour is probably 108–
109 cm−3.. Thus, assuming a clonogenic proportion of between 1/100 and 1/100,000,
reasonable upper and lower limits of clonogen density are 103–107 clonogens cm−3. The
clonogenic cell density in human tumours has been estimated as 107 cm−3 from analysis of
a large series of breast cancer, SCC upper respiratory and digestive tracts, and malignant
melanomas (Webb 1994). A similar value of 107 cm−3 for bladder cancers was also estimated
from clinical data (Nahum and Tait 1992). For head and neck tumours, clonogenic cell density
has been estimated as between 1.8 ×103 and 6.6 ×105 from fits to clinical data (Bentzen
1992) and the maximum number of clonogens was estimated as 5 × 106 (Tepper 1981).
Data from experimental mouse mammary tumours (Suit et al 1992) gave an approximate
value of >105.5 clonogens per tumour of 8 mm diameter, equivalent to >1.17 × 106 cm−3
(Wigg 2001).
It is likely that the clonogen density will be non-uniform, but there are virtually no in vivo
data quantifying this. However, it has been shown that inter-patient heterogeneity is dominated
by heterogeneity in the radiosensitivity parameter α—TCP curves obtained using biologically
realistic parameter values incorporating heterogeneity in α alone are almost identical to those
incorporating heterogeneity in α, clonogen number and time factor γ (Roberts and Hendry
1998). Furthermore, σ α values have been estimated from clinical data assuming uniform
clonogenic cell density (Webb 1994, Roberts and Hendry 1998).
The TCP was calculated using equation (2), assuming N initial clonogens and a cellular
survival (S) at the end of treatment calculated from equations (3) or (4). For all calculations
(except breast treatments where accelerated repopulation starts after surgery) it was assumed
that there was no repopulation during the delay time between start of treatment and onset of
accelerated repopulation. (The maximum absolute difference in reduction in TCP, which is
introduced for Ca prostate, head and neck or cervix if repopulation at the pre-treatment rate is
included in the lag period, is 0.1% for 100 days delay—a negligible effect.)
Representative parameters α, α/β, Teff, Tpot, Tdel, T1/2 and clonogen number from the
literature were used for each tumour type, as shown in table 1. It is accepted that to be generally
more reliable for modelling purposes the inter-related parameters for any one tumour type need
to be derived simultaneously (Dale et al 2002). However, unfortunately, this approach has
not been used for many sites other than head and neck. In the case of breast cancer, it was
assumed that the number of clonogens immediately after surgery is reduced to 1% of the
number at diagnosis. The treatment details are shown in table 2; these are typically used at the
Queen Elizabeth Hospital, Birmingham, and are considered to represent treatment to clinically
acceptable normal tissue tolerance.
Effect of radiotherapy treatment delays 143

Table 1. Radiobiological parameters used in equation (1).

T1/2 N
Tumour type α α/β Teff (days) Tpot (days) Tdel (days) (hours) (at diagnosis)

Cervix 0.3 14a 80–160 4.5b 27 1.0 2 × 108

SCC head/neck 0.3c 10.5d 33–150e (median 5f 27 – 5 × 106
45 days)
Prostate 0.25 5 200–1000 30g 27 – 2 × 108
(typically 1 year)
Breast 0.27c 4h 4–74 daysi,j 4–74 daysi 0 – 8 × 104
(post surgery) (median 15 days) (median 15 days)
Notes:
(i) The value of 5 Gy for α/β used for Ca prostate is an intermediate value, between the commonly as-
sumed value 10 Gy taken to be representative of most tumours, and the recently proposed value of 1.5 Gy
(Brenner and Hall 1999), which is estimated from combined LDR brachytherapy and EBRT data for Ca prostate.
A recent publication confirms that a low value of around 1.5 is probably correct. However, the effect of
changing α/β from 5 to 1.5 is insignificant providing α is also adjusted to give the same level of cell kill for
no delay. The value for α is taken as 0.25 Gy, which is within the range of values derived for in vivo and
in vitro tumours.
(ii) The value of α for Ca cervix is taken as 0.3 Gy−1, which is within the range of values derived for in vivo and
in vitro tumours.
(iii) The Teff values (pre-treatment doubling times) used for Ca prostate are derived from PSA measurements which
indicate tumour doubling times <1 year to >5 years (Brenner and Hall 1999).
(iv) The Teff values (pre-treatment doubling times) used for Ca cervix are taken to be those typical for SCC and
adenoCa, i.e. 80–160 days (Hall 2000).
(v) There is a large variation in the estimated numbers of clonogens in a typical tumour (between 103 and 107 cm−3.
Webb (1994) found that the best fit to clinical data for Ca breast, SCC upper respiratory and digestive tracts, nodal
metastases and malignant melanomas occurred for a clonogenic cell density of 107 cm−3. The number of clonogens
in prostate and cervix tumours assumes a tumour volume 20 cm3 and clonogenic cell density 107 cm−3. For head
and neck tumours, the total number of clonogens is taken as 5 × 106 (Tepper 1981). The number of clonogens in a
typical T1–2 breast tumour is taken as 8 × 106 before surgery. It is assumed that 99% of the clonogens are removed
by surgery, leaving 8 × 104 clonogens.
(vi) Repair constants (T1/2) for tumours are not well known. A representative time for early responding
tissues/tumours of 1.0 h (Steele 1997) is used in the following calculations.
a Thames et al (1990), b Tsang et al (1995), c Webb (1994), d Stuschke and Thames (1999), e Steele (1997), f Bourhis

et al (1996), g Trott and Kummermehr (1985), h Jones and Dale (1999), i Haustermans et al (1998), j Doubling times
post surgery assumed similar to Tpot values.

Table 2. External beam and brachytherapy treatment parameters used at QEH, Birmingham.

Dose per Brachytherapy Time interval between

fraction Number of Total treatment Brachytherapy treatment external beam and
Tumour type d (Gy) fractions (n) time Te (days) dose D (Gy) time S (hours) brachytherapy t (days)
Cervix 1.852 27 38 25.0 13.0 14
SCC head/neck 3.0 15 21
Prostate 2.75 20 28
Breast 2.667 15 21

For each tumour type, the change in TCP with delay time to treatment was calculated
for a range of tumour pre-treatment doubling times representative of the particular tumour
type, as shown in table 1. Equation (2) is appropriate for an individual patient since it
assumes fixed values for all parameters. Dose–response curves calculated using the Poisson
model with fixed radiobiological parameters as in equation (2) are considerably steeper
than those observed in clinical practice, probably due to inter-tumour heterogeneity in the
144 R M Wyatt et al

radiobiological parameters for a patient population (Tucker et al 1990). These authors also
showed that when tumour cell proliferation occurs during treatment, the Poisson model
underestimates the probability of tumour cure compared with ‘true’ control rates from
numerical simulations. However, the Poisson model still provides a good empirical description
of the dose–response curve when fitted to data although parameter estimates are incorrect
(Tucker 1999). Since most radiobiological parameter estimates have been made by fitting
a Poisson model incorporating heterogeneity to clinical data, it is consistent to use this
model here.
Analysis of clinical data has shown that inter-patient heterogeneity is dominated by
heterogeneity in the radiosensitivity parameter α (Roberts and Hendry 1998), which varies
typically by 20–50% (Roberts and Hendry 1999, Webb 1994). For further calculations we
therefore chose to incorporate a heterogeneity of 30% in α, which is in the middle of this
range, to estimate the effect of treatment delays averaged over an entire patient population.
This was modelled assuming a Gaussian distribution of α, where the fraction gi of patients
have α = α i. The overall mean TCP was then calculated using


K
TCP = gi TCPαi (5)
i=1

where
  

gi ∝ exp −(αi − αm )2 2σα2 and gi = 1
The ratio α/β was kept constant for these calculations (i.e. changes in β are assumed to be
proportional to changes in α).

3. Results

The results of TCP calculations over a range of typical pre-treatment doubling times are shown
for each tumour type in figure 1. This shows that delays in treatment can have a considerably
detrimental effect on outcome. Breast tumours show the largest decrease in TCP, due to
the fast tumour growth rates post surgery. The maximum decrease in absolute TCP when
heterogeneity is taken into account is 67% for 100 days delay when a volume-doubling time
of 4 days post surgery is assumed. This very fast doubling time will not apply to the majority
of patients, but is within the range of measured values of potential doubling times from human
breast tumours. As expected, the detrimental effect is considerably smaller for slowly growing
tumours. Ca prostate in particular is slow growing (typical effective doubling times around
1 year) and a delay of 100 days in treatment gives an absolute decrease of only 1% in TCP
(assuming doubling time of 1 year, taking heterogeneity into account).
Where the TCP for an individual (i.e. no heterogeneity) is very high for no delay in
treatment (breast patients after surgery—most tumour cells already removed— and cervix
patients) the effect of adding heterogeneity in α is to increase the reduction in TCP with delay
to the start of treatment. This is due to the fact that at the extremes of very high (or very
low) TCP, the associated portion of the sigmoid TCP curve is very ‘flat’. When heterogeneity
is included, the TCP for no delay decreases so that the results now lie on a steeper portion
of the curve. For other disease categories (head and neck, prostate, breast with a very short
doubling time) the effect of adding heterogeneity is to cause a smaller reduction in TCP with
delay in treatment. For head and neck cancers, the maximum effect reduces from a 57% TCP
reduction with 100 days delay (no heterogeneity) to a 15.5% reduction (30% heterogeneity
in α). Conversely, for breast cancer with a 15-day doubling time, the effect of 100 days
Effect of radiotherapy treatment delays 145

100% 100%

80% 80%

60% 60%
TCP

TCP
40% 40%

4 4

15 15
20% 20% 50
50

0% 0%
0 20 40 60 80 100 0.0 20.0 40.0 60.0 80.0 100.0
Delay to treatment (days) Delay to treatment (days)
(a) (b)

100% 100%

90% 90%
TCP

80% 80%
80
TCP

120

160
70% 80 70%
120
160

60% 60%
0 20 40 60 80 100 0 20 40 60 80 100
Delay to treatment (days) Delay to treatment (days)
(c) (d)

Figure 1. TCP versus treatment delay for Ca breast with pre-treatment doubling times of 4, 15 or
50 days (a) without and (b) with 30% heterogeneity in α; Ca cervix with pre-treatment doubling
times 80, 120 or 160 days (c) without and (d) with 30% heterogeneity in α; SCC head/neck with
pre-treatment doubling times of 33, 45, 90 or 150 days (e) without and (f) with 30% heterogeneity
in α; Ca prostate with pre-treatment doubling times of 200, 365 or 1000 days (g) without and (h)
with 30% heterogeneity in α.

delay increases slightly from 27% reduction (no heterogeneity) to 28.4% reduction (30%
heterogeneity in α).
Table 3 shows the expected reduction in TCP for a 50- or 100-day delay from diagnosis to
treatment, and a range of typical pre-treatment doubling times incorporating 30% heterogeneity
in α.
146 R M Wyatt et al

90% 90%

80% 80%

70% 70%

60% 60%
TCP

TCP
50% 50%
33
45 33
40% 40%
90 45
150 90
30% 30% 150

20% 20%
0 20 40 60 80 100 0 20 40 60 80 100

Delay to treatment (days) Delay to treatment (days)

(e) (f)

90% 90%

200

365

80% 80% 1000

70% 70%
TCP
TCP

60% 200 60%

365

1000

50% 50%
0 20 40 60 80 100 0 20 40 60 80 100
Delay to treatment (days) Delay to treatment (days)
(g) (h)

Figure 1. (Continued.)

The predicted TCP values are highly dependent on some of the assumed radiobiological
parameters but not on others. For example, varying the lag time to onset of accelerated
repopulation between 7 and 54 days has little effect on the results for Ca head and neck, cervix
or prostate (the maximum change in the reduction in TCP for 100 days delay was 1.3% when
incorporating 30% heterogeneity in α). However, the introduction of a lag time of 27 days to
onset of accelerated repopulation following surgery for Ca breast has a larger effect. When a
doubling time of 100 days prior to accelerated repopulation is assumed, the reduction in TCP
for 100 days delay to treatment and 30% heterogeneity in α changes from 67.3 to 83.1%, 28.4
to 24.5%, and 6.4 to 6.1% for an accelerated doubling time of 4, 15 or 50 days respectively.
Effect of radiotherapy treatment delays 147

Table 3. Absolute reduction in TCP for Ca breast, cervix, prostate and SCC head/neck assuming
(a) 50- and (b)100-day delay to treatment, incorporating 30% heterogeneity in the α component.

Absolute reduction in TCP (%)

(30% heterogeneity in α)
Pre-treatment
Tumour type doubling time (days) 50-day delay 100-day delay
Breast 4 54.2 67.3
15 12.7 28.4
50 3.0 6.4
Cervix 80 1.9 3.9
120 1.3 2.6
160 1.0 1.9
SCC head/neck 33 7.7 15.5
45 5.6 11.4
90 2.8 5.6
150 1.7 3.4
Prostate 200 1.0 2.1
365 0.6 1.1
1000 0.2 0.4

Note that when lag time is introduced, the reduction in TCP actually increases for very short
(4-day) doubling times following onset of accelerated repopulation. This is due to the fact
that the TCP for no delay increases from 67.5% to 86.1%, giving a larger range.
The effect of varying the heterogeneity parameter σ α between 20% and 50%, i.e. over the
range of values quoted in the literature, has also been investigated. As expected, in general,
the larger the value of σ α, the smaller is the observed reduction in TCP caused by treatment
delays, due to flattening of the TCP curve. The magnitude of this change is largest for tumours
with short pre-treatment doubling times, e.g. for Ca breast with a doubling time of 4, 15 or
50 days the estimated reduction for a 100-day delay varies between 55.8%, 19.1% and 5.2%
(σ α = 50%) to 74.6%, 35.2% and 5.8% (σ α = 20%) respectively, i.e. absolute changes in TCP
of 18.8%, 16.1% and 0.6%. However, the general trend and the order of magnitude of TCP
reductions remains the same.
The effect of varying clonogenic cell density, α and α/β was investigated for a head and
neck tumour treated with 45 Gy in 15 fractions over 21 days. Either α and α/β were varied
for a fixed number of 1 × 106 clonogens, or α and number of clonogens were varied for a
fixed α/β ratio of 10 Gy. The results are shown in figures 2 and 3. They show that when no
heterogeneity in parameters is assumed, the TCP shows a sharp rise between 0 and 1 for certain
combinations of α, the number of clonogens and α/β, whereas when 30% heterogeneity in α
is incorporated, the change is much more gradual and TCP does not reach 0% or 100%.
The calculations were repeated for a 50-day delay to treatment (assuming a 45-day
pre-treatment doubling time) and the variation in TCP difference with α, α/β and number
of clonogens was calculated with and without incorporating the 30% heterogeneity in α
(figures 4 and 5). These graphs show that when there is no heterogeneity in parameters,
the reduction in TCP is particularly rapid for particular combinations of α, number of
clonogens and α/β corresponding to the points where the TCP curves change from 0%
to 100%. The maximum reduction in TCP is around 27%. When 30% heterogeneity in
α is incorporated in the calculations, the reduction in TCP is considerably less (maximum
reduction 8.1%) and shows a much more gradual change. This maximum value is substantially
148 R M Wyatt et al

100%
100%
80%
80%
60%

TCP
60%
TCP

40% 40%

20% 20% 0.36

0.36
0% 0.28 0% 0.28

4.0E+04
4.0E+04

3.2E+05
3.2E+05

0.2 0.2 Alpha

2.6E+06
2.6E+06

2.0E+07

2.0E+07
Alpha

Number of clonogens Number of clonogens

(a) (b)

Figure 2. TCP for head and neck plotted against α and number of clonogens assuming (a) no
heterogeneity in parameters and (b) 30% heterogeneity in α.

100% 100%
80% 80%
60% 60%
TCP

TCP

40% 40%
5 5
20% 10 20% 10
0% 15 0% 15
0.2

0.2
0.24

0.24

25 25
0.28

0.28
0.32

0.32
0.36

alpha/beta 0.36
0.4

0.4
alpha/beta
alpha alpha
(a) (b)

Figure 3. TCP for head and neck plotted against α and α/β assuming (a) no heterogeneity in
parameters and (b) 30% heterogeneity in α.

0%
0%
TCP Reduction
TCP Reduction

-10%
-10%

-20% -20%
25 25
15 15
-30% 10 -30% 10
0.2 0.24 5 0.2 0.24 0.28 5
0.28 0.32 Alpha/Beta Alpha/Beta
0.36 0.32 0.36
0.4 0.4
Alpha Alpha
(a) (b)

Figure 4. Absolute reduction in TCP plotted against α and α/β with a 50-day delay in the
treatment of SCC head and neck (45-day pre-treatment doubling time). (a) No heterogeneity in
parameters and (b) 30% heterogeneity in α.

greater than the 5.6% reduction shown in table 3 for SCC head and neck with 45-day pre-
treatment doubling time (incorporating 30% heterogeneity in α).
Effect of radiotherapy treatment delays 149

0%
0%
TCP reduction

TCP reduction
-10%
-10%

-20% -20%
0.36 0.36
-30% 0.28 -30% 0.28

4.0E+04
0.2 Alpha
8.0E+04

0.2

3.2E+05
6.4E+05

2.6E+06
5.1E+06 Alpha

2.0E+07
4.1E+07
Number of clonogens Number of clonogens
(a) (b)

Figure 5. Absolute reduction in TCP plotted against number of clonogens and α with a 50-
day delay in treatment of SCC head and neck (45-day pre-treatment doubling time). (a) No
heterogeneity in parameters and (b) 30% heterogeneity in α.

4. Discussion

This theoretical study has shown that long waiting times for radiotherapy are expected to
have a detrimental effect on outcome, particularly for category 1 and post-operative breast
tumours. The effect is predicted to vary widely for individual patients, depending on factors
such as initial tumour size, tumour radiosensitivity and repopulation rate. On the other hand,
retrospective clinical studies of treatment delay on outcome in undisturbed larynx cancer
(Barton et al 1997) and nasopharyngeal carcinoma (Lee et al 1994) have failed to demonstrate
a convincing effect of delay. However, the nasopharyngeal study did show a trend towards
increase in distant metastasis with delay time. Nasopharyngeal cancer is not typical of other
head and neck cancers due to its high metastatic potential and controversy over the time factor,
which may account for the effects seen. The reason for the apparent lack of effect of delay
time on outcome is probably that the waiting times were not very long. The larynx cancer
study included 581 patients with T1, T2a and T2b tumours and 90% of the waiting times were

31 days. The expected average reduction in TCP for a waiting time of 31 days in head and
neck cancer (assuming 45-day pre-treatment doubling time and 30% heterogeneity in α) is
only 3.5%, so that most of the changes will be within a range of 0–3.5%, which would be
difficult to detect given all the other variables. However, as shown here, the reduction in TCP
for some individual patients with rapidly proliferating tumours is expected to be large—up
to around 13% or more. The study of nasopharyngeal carcinoma included 290 patients with
T1N0–3M0 disease, and >90% of the waiting times in the range of 8–42 days. As in the larynx
study, this range is expected to be too small for a measurable effect of waiting time given the
other variables.
A large reduction in TCP with waiting time between surgery and radiotherapy is predicted
for Ca breast patients—up to 54% reduction for 50 days delay when a 4-day doubling time
is assumed (incorporating heterogeneity). A clinical analysis of 400 patients with early
breast cancer, who underwent lumpectomy and post-operative radiotherapy without adjuvant
systemic therapy, showed a trend towards increased local recurrence for patients treated
between 8 and 12 weeks after surgery compared to those treated within 8 weeks of surgery,
although this was not statistically significant (Whelan et al 1996). However, if adjuvant
chemotherapy is used in addition to radiotherapy, a delay between surgery and radiotherapy
150 R M Wyatt et al

is required in order to deliver the full course of chemotherapy. A clinical study of 74

patients treated with radiation therapy and adjuvant chemotherapy following surgery showed
no significant influence on local control, but a trend towards increased regional and distant
relapse for surgery–radiotherapy intervals
20 weeks versus >20 weeks, due to a delay in the
delivery of the full course of chemotherapy, which is consistent with other authors (Willers et al
1996). Conversely, a study of 105 patients treated with radiotherapy and chemotherapy alone
(no surgery) showed that a delay from diagnosis to treatment of >6 months in order to
maximize chemotherapy resulted in a significantly lower local control and overall survival
than for patients who received radiotherapy within 6 months of diagnosis (local control 76%
versus 98%; overall survival 52% versus 80%) (Bucholz et al 1993).
A recent study in patients with high-grade glioma (Do et al 2000) showed that the waiting
time for radiotherapy was a significant predictor of overall survival, and risk of death increased
by 2% per day for each day of waiting. Glioma is a particularly rapidly proliferating tumour;
a pre-treatment doubling time of 3 days was estimated by these authors, which is similar to
doubling times measured in vitro.
Mackillop et al (1996) used a similar technique to that used here for modelling the
effects of treatment delays, although only cancer of the tonsil was considered. In addition,
although heterogeneity in tumour doubling time and size was taken into account, a fixed
cell survival fraction Q was assumed, i.e. heterogeneity in the radiosensitivity parameter α
(which dominates the effect on results) was ignored. The effect of delays in treatment of
tonsillar cancer assuming a mean pre-treatment doubling time of 58 days was calculated as
approximately 10% reduction in TCP per month, i.e. considerably larger than the effect found
here for head and neck cancers (approximately 3.4% per month assuming a 45-day doubling
time and 30% heterogeneity in α). The large reduction calculated by Mackillop et al is closer
to our values calculated ignoring heterogeneity in α, and is unlikely to be seen in clinical
practice.
In addition to tumour doubling times, the reduction in TCP resulting from a certain delay
in treatment is highly dependent on the TCP expected when there is no delay. Tumours
with very high or very low expected TCPs are affected to a much smaller extent by delay
in treatment. Figure 6 shows the expected reduction in TCP for a 50-day delay in treatment
versus TCP with no delay for SCC head and neck with a pre-treatment doubling time of 45
days (a) assuming no heterogeneity in parameters and (b) assuming 30% heterogeneity in
α. The curves are produced for different numbers of clonogens at diagnosis, with α varying
to alter the TCP. All other radiobiological and treatment parameters are as given in tables 1
and 2. These curves show that when there is no heterogeneity in the parameters, reduction in
TCP is independent of the actual values of α and number of clonogens at diagnosis, being a
function of TCP for no delay only; a maximum absolute reduction in TCP of 28% is obtained
when the TCP for no delay is 50%. This compares with a prediction of 13.9% when using
radiobiological parameters from the literature. When heterogeneity in α is incorporated the
reduction in TCP is dependent on the actual values for number of clonogens at diagnosis
and α in addition to TCP for no delay. The maximum predicted absolute reduction in TCP
always occurs when the TCP for no delay is 50% and varies between approximately 8.1%
(2 × 104 clonogens at diagnosis, α = 0.20) and 5.5% (1 × 107 clonogens at diagnosis,
α = 0.30). This compares with a predicted reduction of 5.7% for typical radiobiological
parameters.
In order to predict the effect of delay in treatment, the average reduction in TCP per
week’s delay of treatment was calculated using the radiobiological and treatment parameters
in tables 1–3. The reduction per week showed little variation up to 20 weeks for head and
neck, cervix and prostate tumours (a maximum difference of approximately 3% in relative
Effect of radiotherapy treatment delays 151

0% 0%
0.00 0.20 0.40 0.60 0.80 1.00 0.00 0.20 0.40 0.60 0.80 1.00

-5% -5%
4.0E+04
3.2E+05
-10% 2.6E+06 -10%
TCP reduction

1.0E+07

TCP reduction
-15% -15%

-20% -20%
2.0E+04
4.0E+04
-25% -25% 3.2E+05
2.6E+06
1.0E+07
-30% -30%
Initial TCP Initial TCP
(a) (b)

Figure 6. Absolute reduction in TCP due to a 50-day delay in treatment of SCC head and
neck plotted against initial TCP (no delay) for varying clonogenic numbers (45-day pre-treatment
doubling time). (a) No heterogeneity in parameters and (b) 30% heterogeneity in α.

Table 4. Absolute reduction in % TCP per week for Ca breast, cervix, prostate and SCC head/neck
for various pre-treatment doubling times, using radiobiological and treatment parameters in tables 1
and 2 and assuming 30% heterogeneity in α.

Pre-treatment Absolute reduction in % TCP

Tumour type doubling time (days) per week of delay
Cervix 60 −0.4
90 −0.2
120 −0.2
Prostate 200 −0.2
365 −0.1
1000 −0.0
SCC head and neck 35 −1.0
45 −0.8
90 −0.4
150 −0.2
Breast 4 1st ten weeks 7.3 [5.3–8.7]
2nd ten weeks 1.4 [4.3–0.0]
15 1st ten weeks 1.7 [1.3–2.0]
2nd ten weeks 2.2 [2.0–2.3]
50 1st ten weeks 0.4 [0.4–0.4]
2nd ten weeks 0.5 [0.4–0.5]

terms) and therefore the use of mean values is acceptable. Breast tumour showed a larger
variation since the fast doubling times lead to larger changes in TCP and the position on the
TCP curve. The results are given in table 4 and may be used as a quick reference. For patients
with breast tumours, the average value over the first and second 10-week periods is given,
together with the range of values.
152 R M Wyatt et al

Table 5. Recent data for waiting time between surgery and start of radiotherapy for radical breast
treatments at QEH.

Weeks Total
delay 0 1 2 3 4 5 6 7 8 9 10 11 12 13 >14 patients
Number of 0 0 2 1 3 4 9 10 17 14 15 22 5 1 0 103
patients

Table 6. Calculated average reduction in TCP breast due to waiting times in table 5 summed over
a log-normal distribution of pre-treatment doubling times.

Average reduction Average reduction

Mean doubling in absolute% TCP in absolute% TCP
time µ (days) Standard deviation σ (no heterogeneity) (30% heterogeneity in α)
15 0.588 [0.5∗log (15)] 8.4 9.8
15 0.941 [0.8∗log (15)] 7.8 9.1

Finally, waiting times between surgery and treatment for recent radical breast treatments
in this department (103 patients—see table 5) were used to estimate the mean overall reduction
in TCP to be expected using the radiobiological and treatment parameters for breast in tables 1
and 2. The population distribution of tumour doubling times is approximately lognormal; the
coefficient of variation for breast Tpot measured in vitro for 35 breast tumours was 80% (range
4–74 days, mean 15 days) (Haustermans et al 1998). Fairly similar coefficients of variation for
Tpot (62%–68%) have been measured in head and neck tumours (Begg et al 1992, Bourhis et al
1993). The reduction in TCP for Ca breast was averaged over a lognormal distribution of
doubling times (mean µ = 15 days, range 0–100 days, σ = 50% or 80% of log (µ)) and
over the actual range of waiting times encountered in this department. If Ni is the number of
patients with delay time i weeks and TCP ( j, i) is the TCP for a doubling time of j days and a
delay time of i weeks, then the average reduction in TCP is given by
j =100 j =100
1
1

n
Ni

j =100 A(j ) TCP (j, i = 0) −  j =100 A(j ) TCP(j, i)
j =1 A(j ) j =1 i Ni i=1 j =1 A(j ) j =1

where
exp(log(j ) − log(µ))2
A(j ) =
2σ 2
where TCP ( j, i = 0) is the expected TCP when there is no delay.
The results are shown in table 6, which indicate that the expected average reduction in TCP
for current waiting times is probably around 9%. This represents a considerable detrimental
effect and highlights the importance of minimizing waiting times for radiotherapy in order to
achieve the best possible cure rates, especially where the tumour has been disturbed in some
way which is likely to enhance tumour growth. However, if adjuvant chemotherapy is used,
a delay between surgery and radiotherapy is necessary and the adverse effect appears to be
lessened.
Another adverse effect of delaying treatment, which has not been modelled here, is the
increasing probability of distant metastases, as was suggested by the study of nasopharyngeal
cancer. This effect has recently been quantified for breast tumours by Thames et al (1999),
who showed that the likelihood of first metastasis increased by 1–2%, 2–4% and 3–6% per
month for stages T1, T2 and T3 respectively. The net likelihood of being free of any cancer
(primary + metastasis) therefore falls faster than indicated by table 4.
Effect of radiotherapy treatment delays 153

5. Conclusion

This theoretical study uses the widely accepted linear quadratic model together with
representative radiobiological parameters from the literature to assess the likely effect of delays
in the start of radiotherapy treatment on tumour control. For the most reliable results, the inter-
related parameters for any one tumour type should be derived simultaneously, but unfortunately
parameters derived in this way are not available for most treatment sites. However, modelling
in this area is still a useful tool for predicting the likely effects of treatment delays.
The results suggest that delays in the start of radiotherapy treatment do have an adverse
effect on tumour control. This is especially true following surgery, when tumour growth rates
are likely to increase due to reduced cell loss. The expected reduction in local tumour control
depends mainly on the pre-treatment growth rate of the tumour and on the expected tumour
control probability for no delay. Slowly growing tumours such as Ca prostate are likely to
be affected only to a very small extent by delays in treatment (around 0.1% reduction in
TCP per week of delay). However, rapidly growing tumours such as Ca breast post-surgery
and SCC head and neck are affected to a much larger extent (up to about 7% reduction for
each week’s delay for breast, 1% reduction per week for head and neck). It must be stressed
that these figures do not include the increase in probability of distant metastasis, which will
add to the detrimental effect of treatment delays. These calculations show the importance
of reducing waiting times especially between surgery and radiotherapy and for patients with
rapidly growing tumours.
Guidelines already exist on how to manage treatment interruptions once patients have
started treatment—local guidelines should also be developed to minimize delays at all stages
from diagnosis through to treatment for critical groups of the type identified here.

Acknowledgments

The authors would like to thank Dr D R Wigg whose synthesis of radiobiological information
has been of great help in this work.

References

Barton M, Morgan G, Smee R, Tiver K W, Hamilton C and Gebski V 1997 Does waiting time affect the outcome of
larynx cancer treated by radiotherapy? Radiother. Oncol. 44 136–41
Begg A et al 1999 The value of pretreatment cell kinetic parameters as predictors for radiotherapy outcome in head
and neck cancer: a multicenter analysis Radiother. Oncol. 50 13–23
Begg A C, Hofland I, Van Glabekke M, Bartelink H and Horiot J C 1992 Predictive value of potential doubling time
for radiotherapy of head and neck tumour patients: results from the EORTC cooperative trial 22851 Semin.
Radiat. Oncol. 2 22–5
Bentzen S 1992 Steepness of the clinical dose–control curve and variation in the in vitro radiosensitivity of head and
neck squamous cell carcinoma Int. J. Radiat. Biol. 61 417–23
Bourhis J et al 1996 Potential doubling time and clinical outcome in head and neck squamous cell carcinoma treated
with 70 Gy in 7 weeks Int. J. Radiat. Oncol. Biol. Phys. 35 471–6
Bourhis J, Wilson G, Wibault P, Bosq J, Chavaudra N, Janot F, Luboinski B, Eschwege F and Malaise E P 1993 In vivo
measurement of the potential doubling time by flow cytometry in oropharyngeal cancer treated by conventional
radiotherapy Int. J. Radiat. Oncol. Biol. Phys 26 793–9
Braunschweiger P G, Schiffer L M and Betancourt S 1982 Tumour cell proliferation and sequential chemotherapy
after partial tumour resection in C3H/HeJ mammary tumours Breast Cancer Res. Treat. 2 323–9
Brenner D 1993 Dose, volume and tumour control predictions in radiotherapy Int. J. Radiat. Oncol Biol. Phys. 26
171–9
Brenner D, Geard C and Hall E 1989 Mossbauer cancer therapy doubts Nature 339 185–6
154 R M Wyatt et al

Brenner D and Hall E 1999 Fractionation and protraction for radiotherapy of prostate carcinoma Int. J. Radiat. Oncol
Biol. Phys. 43 1095–101
Brown D and Rothery P 1993 Models in Biology: Mathematics, Statistics and Computing (Chichester: Wiley)
pp 48–50
Buchholz T A, Austin-Seymour M M, Moe R E, Ellis G K, Livingston R B, Pelton J G and Griffin T W 1993 Effect
of delay in radiation in the combined modality treatment of breast cancer Int. J. Radiat. Oncol. Biol. Phys 26
23–5
Dale R G, Hendry J H, Jones B, Robertson A G, Deehan C and Sinclair J A 2002 Practical methods for compensating
for missed treatment days in radiotherapy, with particular reference to head and neck schedules Clin. Oncol. 14
382–393
Demichelli R, Terenziani M and Bonadonna G 1998 Estimate of tumour growth time for breast cancer local
recurrences: rapid growth after wake-up? Breast Cancer Res. Treat. 51 133–7
Do V, Gebski V and Barton M. 2000 The effect of waiting for radiotherapy for grade III/IV gliomas Radiother.
Oncol. 57 131–6
Gunduz N, Fisher B and Saffer E A 1979 Effect of surgical removal on the growth and kinetics of residual tumour
Cancer Res. 39 3861–5
Hall E 2000 Radiobiology for the Radiologist 5th ed. (Philadelphia: Lippincott Williams and Wilkins) 390 pp
Haustermans K, Fowler J, Geboes K, Christiaens M, Lerut A and Schueren E 1998 Relationship between potential
doubling time (Tpot), labelling index and duration of DNA synthesis in 60 esophageal and 35 breast tumours: is
it worthwhile to measure Tpot? Radiother. Oncol. 46 157–67
Hermens A and Barensden G 1969 Changes of cell proliferation characteristics in a rat rhabdomyosarcoma before
and after irradiation Eur. J. Cancer 5 173–89
Hill R and Milas L 1989 The proportion of stem cells in murine tumours Int. J. Radiat. Oncol. Biol. Phys. 16 513–8
Jones B and Bleasdale C 1994 Effect of overall time when radiotherapy includes teletherapy and brachytherapy: a
mathematical model Br. J. Radiol. 67 63–70
Jones B and Dale R 1995 Cell loss factors and the linear quadratic model Radiother. Oncol. 37 136–9
Jones B and Dale R 1999 Mathematical models of tumour and normal tissue response Acta Oncol. 38 883–93
Koscielny S and Tubiana M 1999 The link between local recurrence and distant metastases in human breast cancer
Int. J. Radiat. Oncol. Biol. Phys. 43 11–24
Laird A 1965 Dynamics of relative growth Growth 29 249–63
Lee A, Chan D, Fowler J, Poon Y, Law S, Foo W, Tung S Y, Cheung F K, Ho J H C and Chappell R 1994 T1
nasopharyngeal carcinoma: the effect of waiting time on tumour control Int. J. Radiat. Oncol. Biol. Phys. 30
1111–7
Mackillop W, Bates J, O’Sullivan B and Withers H 1996 The effect of delay in treatment on local control by
radiotherapy Int. J. Radiat. Oncol. Biol. Phys. 34 243–50
Nahum A and Tait D 1992 Maximising control by customised dose prescription for pelvic tumours Advanced
Radiation Therapy: Tumour Response Monitoring and Treatment Planning ed A Breit (Heidelberg: Springer)
pp 425–31
Petereit D G, Sarkaria J N, Chappell R, Fowler J F, Hartmann T J, Kinsella T J, Stitt J A, Thomadsen B R and Buchler
D A 1995 The adverse effect of treatment prolongation in cervical cancer Int. J. Radiat. Oncol. Biol. Phys. 32
1301–7
Peters L 1990 Inherent radiosensitivity of tumour and normal tissue cells as a predictor of human tumour response
Radiother. Oncol. 17 177–90
Porter E H 1980 The statistics of dose/cure relationships for irradiated tumours Br. J. Radiol. 53 210–27
Roberts S and Hendry J 1998 A realistic closed-form radiobiological model of tumour-control data incorporating
intertumor heterogeneity Int. J. Radiat. Oncol. Biol. Phys. 41 689–99
Roberts S and Hendry J 1999 Time factors in larynx tumour radiotherapy: lag times and intertumour heterogeneity
in clinical datasets from four centres Int. J. Radiat. Oncol. Biol. Phys. 45 1247–57
Steele G (ed) 1997 Basic Clinical Radiobiology 2nd edn (New York: Oxford University Press)
Stuschke M and Thames H 1999 Fractionation sensitivities and dose–control relations of head and neck carcinomas:
analysis of the randomised hyperfractionation trials Radiother. Oncol. 51 113–21
Suit H, Skates S, Taghian A, Okunieff P and Efird J 1992 Clinical implications of heterogeneity of tumour response
to radiation therapy Radiother. Oncol. 25 251–60
Tepper J 1981 Clonogenic potential of human tumours Acta Radiol. Oncol. 20
Thames H D, Bentzen S M, Turesson I, Overgaard M and Van Den Bogaert W 1990 Time-dose factors in radiotherapy:
a review of the human data Radiother. Oncol. 19 219–35
Thames H, Bucholz T and Smith C 1999 Frequency of first metastatic events in breast cancer. Implications for
sequencing of systemic and local-regional treatment J. Clin. Oncol. 18 2649–58
Effect of radiotherapy treatment delays 155

Trott K and Kummermehr J 1985 What is known about tumour proliferation rates to choose between accelerated
fractionation or hyperfractionation? Radiother. Oncol. 3 1–9
Tsang R W 1995 Proliferation measurements with flow cytometry Tpot in cancer of the uterine cervix: correlation
between two laboratories and preliminary clinical results Int. J. Radiat. Oncol. Biol. Phys. 32 1319–29
Tucker S L 1999 Pitfalls in estimating the influence of overall treatment time on local tumour control Acta Oncol. 38
171–8
Tucker S, Thames H and Taylor J 1990 How well is the probability of tumour cure after irradiation described by
Poisson statistics? Radiat. Res. 124 273–82
Webb S 1994 Optimum parameters in a model for tumour-control probability including interpatient heterogeneity
Phys. Med. Biol. 39 1895–914
Whelan T J, Clark R M, Levine M N, Willan A, McCulloch P, Lipa M and Wilkinson R H 1996 The effect of delay in
initiating radiotherapy post-lumpectomy on local breast recurrence Int. J. Radiat. Oncol. Biol. Phys. 36 (Suppl)
280
Wigg D 2001 Applied Radiobiology and Bioeffect Planning (Madison, WI: Medical Physics Publishing) 268 pp
Willers H, Wurschmidt M D, Janik M D, Bunemann H and Heilmann H P 1996 Combined conservative surgery,
chemotherapy and radiation therapy in treatment of the breast cancer patient: the influence of the interval
between surgery and start of irradiation Int. J. Radiat. Oncol. Biol. Phys. 36 (Suppl) 280