Radiotherapy Elsevier RADION

and Oncology,

17 (1990) 95-102 95

00656

Dose-time considerations of head and neck squamous cell carcinomas treated with irradiation J. M. G. Taylor

‘, H. R. Withers

’ and W. M. Mendenhal12

‘Departmentof Radiation Oncologv and Jonsson Comprehensive Cancer Center, UCLA, Los Angeles. CA 90024, and 2 Department of Radiation Oncology, University of Florida College of Medicine, Gainesville, FL 32610,

(Received

15 December

Key words: Squamous

1988, revision received 18 July 1989, accepted

cell carcinoma;

Irradiation;

U.S.A.

1 August 1989)

Tumor control; Accelerated

repopulation

Summary The dose-time factors in the external beam treatment of 473 patients with squamous cell carcinoma of the pharyngeal wall, vocal cord, pyriform sinus or supraglottic larynx were considered. The effect of overall treatment time on the tumor response was quantified by estimating the increment in dose per day needed to achieve a constant rate of local control, that is, the dose required to counterbalance the effect of growth of the tumor during irradiation. The estimated increment in isoeffect dose per day varies between sites, however, the increments, although estimated with considerable uncertainty, are in general larger than 1 Gy per day. These estimates are consistent with accelerated tumor clonogen repopulation during irradiation.

Introduction A series of papers [2-71 describes the experience in the radiotherapy of squamous cell carcinoma of the pharyngeal wall, vocal cord, pyriform sinus and supraglottic larynx at the University of Florida. This paper describes further analyses of these data sets with emphasis on the effect of overall treatment time on the local control rate. Recent articles [ 1,8] considered the time effect in head and neck radiotherapy from a series of

patients treated in Poland [l] and 20 additional published data sets [ 81. These articles concluded that an extension of the overall treatment time by 1 day needed to be balanced by an extra 0.6 Gy per day when treatment durations exceeded 28 days. This dose is larger than that necessary to balance the tumor growth if it maintained its preirradiation rate. Thus, this extra dose can be interpreted as an accelerated growth of the tumor.

Patients and methods Address for correspondence

: J. M. G. Taylor, Department of Radiation Oncology and Jonsson Comprehensive Cancer Center, UCLA, Los Angeles, CA 90024, U.S.A.

0167-8140/90/$03.50

0 1990 Elsevier Science Publishers

This analysis is based on 473 patients treated with external beam irradiation only for squamous cell

B.V. (Biomedical

Division)

96 carcinoma of the pharyngeal wall, vocal cord, pyriform sinus or supraglottic larynx. There were 133 local failures from these 473 patients. The details of the patients and treatment variables are given in Table I. A small fraction of these patients were in a previous analysis [8] of the dose-time effects. The present material represents a considerable update on the previously published work. All patients were followed for at least 2 years ; patients who died within 2 years of treatment with the primary site continuously diseasefree have been excluded. Patients with pharyngeal wall or pyriform sinus lesions treated with the split-course technique were excluded from previous publications but are included in this analysis [4,5]. For T, vocal cord, the patients are further divided into two groups : (1) T,, suitable for hemi-

TABLE

laryngectomy (n = 27) and (2) others (n = 81). Further details of the patients and irradiation techniques can be found elsewhere [2-71. The variables used in this analysis were total dose, dose per fraction, overall treatment time, tumor stage, and local tumor control. Patients with overall treatment times greater than 80 days were excluded from the analysis. The statistical methods used and their justification are described elsewhere [ 81. Briefly, it is assumed for the pharyngeal wall that the probability of local failure (p) is related to the other variables through the statistical model : hT(Pl(l - PI> = &I + A, (a/bd (a//l + 2) - A,(Time-28)’ +

A, log((tumor

+ nd2Y

diameter)*).

I

Description

of patients and treatment

variables. Site and stage

Total

Pharyngeal wall

Vocal cord

Supraglottic larynx

Pyriform sinus

T, T, T, T,

4 17 33 14

170 108 18

12 26 12 9

14 21 7 8

473

68

296

59

50

40 13 15

290 4 2

43 3 13

30 8 12

vocal cord

Supraglottic larynx

Pyriform sinus

Fractionation

pattern

Standard continuous course 2 Fractions per day Split course

Fractionation Pharyngeal wall Treatment time (days) Total dose (Gy) Dose per fraction (Gy)

variables

44-77 61-79

33-61 54-75

1.13-2.14

1.20-2.57

36-72 56.1-80 1.20-2.26

42-79 47-80 1.20-2.03

The tumor diameters were assumed to be 1.5, 3.0, 4.5 and 5.5 cm for T,, T,, T, and T, tumors, respectively. The use of (tumor diameter)2 rather than (tumor diameter)3 is based on the assumption that tumors in the pharyngeal wall are disk shaped rather than spherical. For the other three sites, a similar model was used except that the tumor diameter was replaced by an indicator variable denoting the various T-stages. That is, :

WPl(l -P>) = A, + A,(dbd + nd2)/ (a/j? + 2) - A,(Time-28)’ + A,Z(2) + A,Z(3) + A,Z(4) where Z(x) = 1 if T-stage = x and Z(x) = 0 otherwise. This parametrisation was chosen for sites other than pharyngeal wall because for these sites the T-stage is determined not only by the size of the primary, but also by its location. The term (Time-28) + = Max(O,(Time-28)) represents the situation where accelerated repopulation begins 28 days after the start of radiotherapy. The choice of 28 days is based on a previous analysis [8]. However, since there are no patients treated in less than 28 days, this assumption has no effect in the analysis; exactly the same conclusions would be reached if any number of days less than 28 days was used instead. The unknown parameters of the model (A,, A 1, A,, A,, A,, A5) are estimated by maximum likelihood. Confidence intervals are obtained using the profile likelihood method. The quantity (cc/p& + nd2)/(a/b + 2) is the normalized total dose, which is the equivalent total dose if all fractions had been of size 2 Gy. The LX//I ratio was fixed at 10 Gy as there was not enough variation of dose per fraction to estimated accurately its value. Attempts to estimate the U/B ratio using a logistic regression model suggested that a high value (10 Gy) had a greater likelihood than a low value (2 Gy) for 3 of the 4 data sets. However, the confidence intervals for the estimated a/P ratio were very wide covering the range 0 to cc for all the 4 data sets. The ratio A,/A, is the estimate of the dose necessary to balance 1 day’s extension of the treatment time. The effect of extending the

0

P

h.

60

I

50

I

70

60

TREATMENT

TIME

(DAYS)

Fig. 1. Scattergram of normalized total dose (a/j = 10 Gy) versus overall treatment time for pharyngeal wall T, (triangles) and T, (circles). Open symbols represent local control, closed symbols represent local failure.

overall treatment time on the local control rate is also obtained from the coefficient A,. In particular, 7A, is the estimated change in the logit of the local control probability if the overall treatment time is extended by 7 days without any accompanying change in the total dose or dose per fraction. This change in logit is re-expressed as a change in control probability when the initial control rate is assumed to be either 0.9 or 0.6.

Results Figures l-6 show the scattergram of normalized total dose versus overall treatment time for the I

=

55

I’

I

P 35

I 40

,

I

I 45

I

I 50

1

1

I 55

TREATMENT TIME (DAYS)

Fig. 2. Scattergram of normalized total dose versus overall treatment time for vocal cord T,.

98

7D-

65-

BO-

55 TREATMENT

TIME

40

(DAYS)

Fig. 3. Scattergram of normalized total dose versus overall treatment time for vocal cord T,A, suitable for hemilaryngectomy.

majority of the patients. (Data for patients in small T-stage subgroups are not shown, but are given in the Appendix so that others may do their own analysis.) The lines indicate the isoeffect lines for 50% control rate for the various stages, estimated from the statistical model. The slope of the line on each graph is determined by the data from all stages for that site, but the position is determined by each specific stage. The slope of the lines (Table II) indicates that the dose to balance 1 day’s extension of treatment is in general larger than 1 Gy per day; the wide confidence intervals preclude more accurate estimates. Each line is the best estimate of the 50% control rate isoeffect, but as there is considerable uncertainty

45

50

TREATMENT

55 TIME

60

65

70

(DAYS)

Fig. 5. Scattergram of normalized total dose versus overall treatment time for supraglottic larynx T, (triangles) and T, (circles).

in the position of the lines, many other isoeffect lines are consistent with the data. The reduction in control rate (Table II) if the overall treatment time is extended one week without any accompanying increase in dose is from 90% to approximately 75 y0 or from 60 % to approximately 35 %, averaging over the four tumor sites. These changes correspond to approximately l-l.5 doublings per week of clonogen in tumor cells, values similar to previous estimates [8]. The estimates and standard errors for the coefficients A,-&, are given in Table III. The four time coefficients (AZ) are all negative and statistically significant, whereas only 3 of the 4 NTD coefficients (A ) are negative and none are statistir

0 45TREATMENT

TIME (DAYS)

Fig. 4. Scattergram of normalized total dose versus overall treatment time for vocal cord T,, not suitable for hemilaryngectomy.

, 45

I I 50 55 TREATMENT TIME

I 60 (DAYS)

h 65

I 70

J

Fig. 6. Scattergram of normalized total dose versus overall treatment time for pyriform sinus T, (triangles) and T, (circles).

99 TABLE II Effect of overall treatment

time.

(A) Dose to balance one day’s extension

in treatment

time.

Pharyngeal wall

Vocal cord

Supraglottic larynx

Pyriform sinus

1.04 Gy ( - co, - 2.05) (0.15, “c)

1.92 Gy (-co,-7.70) (0.91, co)

1.76 Gy (- co, -2.50) (0.55, co)

- 40.48 Gy ( - co, - 0.45) (0.53, co)

(B) Effect on local control rate of extending treatment

time by 1 week.

Initial control rate

Pharyngeal wall

Vocal cord

Supraglottic larynx

Pyriform sinus

0.90

0.82b (0.72, 0.89) 0.43 (0.30, 0.58)

0.68 (0.53, 0.80) 0.26 (0.16, 0.39)

0.72 (0.50, 0.87) 0.30 (0.14, 0.53)

0.82 (0.71, 0.89) 0.43 (0.29, 0.58)

0.60

X The values in parentheses are 95% confidence intervals. b Predicted new control rate if treatment time is increased

by 1 week without any accompanying

change in dose.

TABLE III Coefficient estimates

A, 0-D) A, (Time) A, (T,) A, (TX) A, (T,)

and standard

errors from best fitting model.

Pharyngeal wall

Vocal cord”

Supraglottic larynx

- 0.092 (0.071) - 0.096 (0.042) 1.06 (0.51 )

- 0.108 (0.067) - 0.207 (0.044) 0.28 (0.69 ) 1.17 (0.44 ) 0.81 (0.73 )

- 0.100 - 0.0176 0.73 0.84 4.53

a For vocal cord, A, corresponds

to T,,

suitable for hemilaryngectomy,

tally significant different from zero. The coefficients for T-stage (A3, A,, A,) are all positive, indicating increased failure rate for other T-stages relative to stage 1. The large confidence intervals prevent any definite statements concerning the relative prognosis for different stages, however the analysis does suggest for pyriform sinus and supraglottic larynx that T, , T,, T,, T, is the most appropriate order, whereas for vocal cord the best order appears to be T, , T,, suitable for hemilaryngectomy, T,, other T,.

A, corresponds

Pyriform sinus 0.002 - 0.099 0.86 1.91 3.15

(0.064) (0.042) (1.00 ) (1.15 ) (1.41 )

to other T,, A, corresponds

to T,

(0.079) (0.068) (1.29 ) (1.91 ) (1.70 )

Discussion In this article we have focused on the ratio of rather than the individual coefficients AJA, coefficients because this ratio has a nice interpretation in terms of the dose to balance 1 day’s extension of treatment time and for comparison with previous analysis [ 81. However, the fact that A, is not statistically significantly different from zero does detract from the interpretation and results in wide confidence intervals, which include

100 infinity, for&/A,. Despite these reservations, it is clear from the analysis that a value of 0 or 0.2 Gy per day is not consistent with the data (Table II), but values of 1 Gy per day or greater are consistent with the data. The negative values of AZ/A,, which are included in the confidence intervals, can be discounted on radiobiological grounds. The estimates of the values for dose to balance 1 day’s extension of the treatment time were based on the assumption that IX//~= 10 Gy. When other values of a//? in the range 10 Gy to infinity were assumed, the estimate of the dose per day to balance 1 day’s extension of the treatment changed by less than 5%. The equation used for estimating the isoeffective control rate lines in Figs. 1-6 is at best only an approximation. However, from the figures it is clear that high doses in short overall treatment durations were associated with high rates of local control and that values larger than 1 Gy per day for the loss of effective dose are reasonable. Considering the wide confidence intervals, the estimated value of larger than 1 Gy per day is not inconsistent with the estimate of 0.6 Gy per day obtained from many published data sets [8]. These values are larger than would be expected based on the pre-irradiation growth rate of tumors and suggests an accelerated growth rate of tumors in response to radiation injury. This is supported by the statistically significant value for the time coefficient. The confidence intervals for the dose to balance an extra day are wide, particularly for pyriform sinus because of the small sample size and the fact that the dose-response curves were surprisingly shallow. In fact, the effect of dose was not statistically significant at the 5 y0 level for any of the four data sets. It could be that there is no real effect of dose on the control probability, however, this is unlikely on radiobiological grounds. A more reasonable explanation is that the finding is a result of the small sample size and general heterogeneity of clinical data. The result may also reflect the “tailoring” of total doses for differences in the predicted outcome of treatment, e.g. by giving a higher total dose to a larger tumor

within a given T-stage. Thus, good clinical practice is not necessarily the most useful for accurate estimation of radiobiological parameters. Increasing awareness of the importance of accelerated repopulation may lead to increasing difficulty in its investigation. This suggests the need for carefully designed and executed clinical trials to obtain the type of clinical data in which the outcome is determined only by the pre-determined treatment schedule and not by factors which cannot easily be accounted for in the analysis. One point which this paper does emphasize is that, despite the relatively large sample size, it is difhcult to make strong conclusions from retrospective clinical data in which the treatment regimen is systematically determined by, for example, the prognostic characteristics of the patients and the early response to treatment. The radiotherapy implications of the accelerated growth of tumors as a response to radiation are more fully discussed elsewhere [ 81, however, they do suggest that unnecessary extension of the overall treatment time should be avoided.

References Maciejewski, B., Withers, H. R., Taylor, J. M. G. and Hliniak, A. Dose fractionation and regeneration in radiotherapy for cancer of the oral cavity and oropharynx. 1. Tumor dose-response and regeneration. Int. J. Radiat. Oncol. Biol. Phys. 16: 831-843, 1989. Mendenhall, W. M., Million, R. R. and Cassisi, N. J. Squamous cell carcinoma of the supraglottic larynx treated with radical irradiation: analysis of treatment parameters and results. Int. J. Radiat. Oncol. Biol. Phys. 10: 2223-2230, 1984. Mendenhall, W. M., Million, R. R., Sharkey, D. E. and Cassisi, N. J. Stage T3 squamous cell carcinoma of the glottic larynx treated with surgery and/or radiation therapy. Int. J. Radiat. Oncol. Biol. Phys. 10: 357-363, 1984. Mendenhall, W. M., Parsons, J. T., Cassisi, N. J. and Million, R. R. Squamous cell carcinoma of the pyriform sinus treated with radical radiation therapy. Radiother. Oncol. 9: 201-208, 1987. Mendenhall, W. M., Parsons, J. T., Mancuso, A. A., Cassisi, N. J. and Million, R. R. Squamous cell car-

101 cinema of the pharyngeal wall treated with irradiation. Radiother. Oncol. 11: 205-212, 1988. 6 Mendenhall, W. M., Parsons, J. T., Million, R. R. and Fletcher, G. H. Tl-T2 squamous cell carcinoma of the glottic larynx treated with radiation therapy: relationship of dose-fractionation factors to local control and complications. Int. J. Radiat. Oncol. Biol. Phys. 15: 1267-1273, 1988.

7 Parsons, J. T., Cassisi, N. J. and Million, R. R. Results of twice-a-day irradiation of squamous cell carcinomas of the head and neck. Int. J. Radiat. Oncol. Biol. Phys. 10: 2041-2051, 1984. 8 Withers, H. R., Taylor, J. M. G. and Maciejewski, B. The clinical hazard from accelerated tumor growth during radiotherapy. Acta Oncol. 27: 131-146, 1988.

Appendix Listing of data for small subgroups not shown in the jgures Pharyngeal

Supraglottic

wall T,

larynx T,

Overall time

Local controla

Total dose

(GY)

No. of fractions

(GY)

No. of fractions

Overall time

Local controla

66.00 65.00 75.00 77.80

32 34 40 65

44 51 57 64

C C F C

70.00 56.40 61.00 61.30 61.00 56.11 56.58 65.00 71.29 61.25 56.58 56.58

36 25 28 32 34 25 25 33 32 31 28 26

53 36 43 46 50 36 36 45 59 58 52 50

C C

Total dose

Pharyngeal

wall T,

(GY)

No. of fractions

Overall time

Local control”

74.00 70.00 75.00 75.00 76.00 75.00 75.40 77.20 66.00 70.73 66.00 66.00 66.00 62.23

37 41 43 38 44 44 61 64 34 37 34 36 40 32

54 61 63 57 76 63 50 57 60 69 61 71 71 62

F F F F F C F C F F F F F F

Total dose

il C = cure; F = fail.

Supraglottic

C C

C C

C C C

F C C

larynx T,

~_ (GY)

No. of fractions

Overall time

75.00 66.01 70.00 75.00 70.00 63.65 75.00 61.30 61.30

37 35 40 36 42 33 44 33 33

55 47 66 50 65 48 60 59 63

Total dose

Local control”

102 Vocal cord T,

Pyriform sinus T, Overall time

Local controla

Total dose

(GY)

No. of fractions

(GY)

No. of fractions

Overall time

Local controla

60.00 60.00 70.00 56.58 61.30 65.00 61.30 65.00 60.00 61.00 67.00 61.30 61.30 66.00

33 33 40 29 32 33 32 32 30 33 34 32 36 40

51 48 58 50 46 45 51 44 45 79 60 58 67 74

C C C C C C F C C F C C C F

70.00 61.30 74.00 70.00 63.65 70.00 60.35 70.00 71.32 67.80 65.80 70.00 74.40 74.40 65.70 74.40 62.33 65.80

35 33 37 34 31 38 32 35 43 30 32 34 62 62 29 62 31 34

51 47 55 45 44 53 46 46 61 41 48 45 43 44 46 43 60 61

C F F C C F F C F C F C C C C C C F

(GY)

No. of fractions

Overall time

Local control”

76.80 80.40 74.00 75.00 65.06 66.95 61.30 61.30

64 67 43 43 36 34 39 33

50 51 64 60 67 65 70 59

Total dose

Pyriform sinus T, Total dose

a C = cure; F = fail.