Correspondence 989

estimated by the Kaplan-Meier curve, is usually of greater biological interest. This distinction hinges on the status of the ultimate competing risk, death: To a patient, it is obviously of great interest; to a radiation oncologist who is concerned about the probability of late complications if he should, by some means, be able to improve the survival rate, or to a radiobiologist who wants to fit the linear-quadratic or other equation to late complications data, it is only a nuisance to be factored out of the equation.

Which probability do we want to estimate? The choice of question can only be addressed by the researcher. Once it is set, Bentzen et al. ( 1) , Caplan et al. (2 ) , and references in these works provide the means of achieving an answer.

RICK CHAPPELL, Ph.D. University of Wisconsin Medical School 600 Highland Avenue, K6/430 Madison, WI 53705

Bentzen, S. M.; Vaeth, M.; Pedersen, D. E.; Overgaard, .I. Why ac- tuarial estimates should be used in reporting late normal-tissue ef- fects of cancer treatment. Int. J. Radiat. Oncol. Biol. Phys. 32:1531- 1534; 1995.

Caplan. R. J.; Pjak, T. F.; Cox, J. D. Analysis of the probability and risk of cause-snecific failure. Int. J. Radiat. Oncol. Biol. Phvs. 29:1183-1186;‘1994. Caplan, R. J.; Pajak, T. F.; Cox, J. D. Letter to the editor. Int. J. Radiat. Oncol. Biol. Phys. 32:1547; 1995. Kaplan, E. L.; Meier, P. Non-parametric estimation from incomplete observations. J. Am. Stat, Sot. 53:457-481; 1958.

RE: KAPP EDITORIAL ZJROBP 35189-194; 1996

To the Editor: We appreciate the editorial comments of Dr. Kapp (Thermal dose response, systemic hyperthermia, and metastases: Old friends revisited) with regard to our article “Radiation plus local hyperthermia versus radiation plus the combination of local and whole body hyperthermia in canine sarcomas.” Dr. Kapp’s assess- ment of our findings was correct, and many points iterated in the article were repeated in the Editorial. However, we take exception to one of Dr. Kapp’s comments: “the lack of any significant enhance- ment of local control at the levels of hyperthermia achieved in either hyperthermia arm of the current study could have been anticipated based on prior investigations at Duke University.” This is incorrect. it is true that there are data from human trials at Duke University that provide some appreciation for the magnitude of thermal dose nec- essary for a hyperthermia effect. However, this information was not available in the late 1980s when the canine study was developed. Considerable time was spent in development of the trial, and the se- lection of number and duration of hyperthermia treatments was care- fully considered. With hyperthermia dose-response knowledge available at that time, the selection of thermal parameters was rea- sonable. Although one might design the study differently now, to state that we should have known at that time that the hyperthermia dose was insufficient is incorrect.

This points to a general problem with the conduct of clinical trials in this field. Since our concepts of thermal dose are evolving and it takes many years to conduct a trial, there will inevitably be cases where trials are designed inappropriately, given the knowledge of the field at the time that the paper is finally published. On the positive side, however, one could take the negative effect of I-IT observed in the radiation plus local hypertheimia group as evidence that the current concept regarding ther- mal dose are correct. I f the response in the LH group had been better than expected for radiation alone, we would have been required to re- evaluate our current paradigm on this issue.

DONALD E. THFXLL, D.V.M., Ph.D. Professor of Radiology North Carolina State University Raleigh, North Carolina 27606 THADDEUS V. SAMULSKI, Ph.D. Professor of Radiation Oncology MARK W. DEWHIRST, D.V.M., Ph.D. Professor of Radiation Oncology

DEBORAH M. PRESCCY~-r, D.V.M., Ph.D. Assistant Professor of Radiation Oncology Duke University Medical Center North Carolina

1. Kapp, D. S. Thermal dose response, systemic hyperthermia, and me- tastases: Old friends revisited. Int. .l. Radiat. Oncol. Biol. Phys. 35:189-194; 19%.

2. Thrall, D. E.; Prescott, D. M.; Samulski, T. V.: Rosner, G. L.; Den- man, D. L.; Legorreta, R. L.; Dodge, R. K.; Page, R. L.; Cline, J. M.; Lee, J.; Case, B. C.; Evans, S. M.; Oleson, J. R.; Dewhirst, M. W. Radiation plus local hyperthermia versus radiation plus the combination of local and whole body hyperthermia in canine sar- comas Int. J. Radiat. Oncol. Biol. Phys. 34: 10X7--- 1096: 1996.

IN RESPONSE TO THRALL ET AL: BXNDSHXT IS ALWAYS PERFECT: THFBMAGDOSE RESPONSE,

SYSTEMIC WPERTHERMU, AND MFXASTASES: OLD FRIENDS REVISITED

To the Editors: I welcome the comments of Dr. Thrall and his col- leagues concerning the editorial ( 1) on their thorough investigation, ‘*Radiation plus local hyperthemua versus radiation plus the combina- tion of local and whole body hypetthermia in canine sarcomas” (4). The statement that “the lack of any significant enhancement of local control at the levels of hyperthennia achieved in either hyperthermia arm” of their study “could have been anticipated based on prior investigations at Duke University” was not meant to imply mat the data required to make such a prediction was available at the time of the initial design of their trial. However, this information did become available during the years of execution of this trial and the negative results could therefore have been “anticipated” by the time of my review of their completed study. Since hyperthermia trials such as theirs are often of long duration and labor intensive, the use of a standard treatment control arm and interim analyses with built in early stopping rules I as was used in their study) may be beneficial (2, 3).

DANIEL S. KAPP, Ph.D., M.D. Professor, Radiation Oncology Stanford University Medical Center Stanford, CA 94305-5105

1.

2.

3.

4.

5.

Kapp, D. S. Thermal dose response, systemic hyperthermia, and me- tastases: Old friends revisited. Int. 1. Radiat. 0~01. Biol. Phys. 35:189-194; 1996. Simon, R. M. Design and conduct of clinical trials. in: DeVita, V. T.; Hellman, S.; Rosenberg, S. A., eds. Cancer: Principles and practice of oncology. 2nd ed. Philadelphia, PA: Lippincott Co: 19&X5:329-350. Souhami, R. L.; Whitehead, J., eds. Workshop on early stopping rules in cancer clinical trials. Stat. Med. 13: 1293 - 1499: 1994. Thrall, D. E.; Prescott. D. M.; Samulski, T. V.; Rosner, G. L.; Den- man, D. L.; Legorreta, R. L.; Dodge, R. K.; Page, R. L.; Cline, J. M.; Lee, J.; Case, B. C.; Evans. S. M.; Oleson, J. R.; Dewhirst, M. W. Radiation plus local hyperthermia versus radiation plus the com- bination of local and whole body hyperthermia in canine sarcomas. Int. J. Radiat. Oncol. Biol. Phys. 34: 1087- 1096; 1996. Whitehead, J. Interim analyses and stopping rules in cancer clinical trials. Br. J. Cancer 68:1179-l 185: 1993.

RE: LO ET AL. ZJROBP 34(5):1113-1119; 1%

To the Editor: The article by Lo eb al. I, 3 ) offers an interesting pre- diction that fractionated stereotactic radiotherapy may not yield im- proved results over radiosurgery because of an increase in volume that may need to be irradiated to accommodate a greater setup uncertainty in the fractionated treatments. There are two possibilities that work against this conjecture:

1. The statistical effect of fractionation on the margin is not dis- cussed. Several workers (2. 5) have suggested that given a

990 I. J. Radiation Oncology 0 Biology 0 Physics Volume 36, Number 4, 1996

random setup error, the margin is related to the reciprocal of the square root of the number of fractions. We readily admit that these reports do not address the important range of l- 10 fractions, but the potential for improvement in the like- lihood of hitting the target simply by repetition should be considered.

2. There are three classes of stereotaxic radiotherapy. The authors con- sider invasive head frames and remountable frames but do not mention implanted fiducials, a technique described in 1982 by Columbo et nl. ( 1) . As observed by Maciunas (4)) there are fundamental advantages obtained with this approach that make it ideal for fractionated treatment since the target is localized with respect to the trearnrent beam, not a frame, in the same manner for each treatment.

The subject of margins and development of the PTV from the CTV requires much more discussion. It will be difficult to review radio- surgery data retrospectively because of the diversity of opinion on the allowance for uncertainty. It is commonplace in radiosurgery not to allow any margin for setup or, as important, the uncertainty that arises in the imaging study. Even the Radiation Therapy Oncology Group in its radiosurgery trials has specifically not allowed for un- certainty which must have resulted in partial target irradiation in all patients in the trial.

DOUGLAS JONES Director N. W. Medical Physics Center Roentgen Hall 2 103 1 67th Ave. W. Lynuwood, WA 98036 MARK D. HAFEFWANN, M.D.

JOHN RIEKE, M.D. Virginia Mason Medical Center

1. Columbo, F.; Angrilli, F.; Zanard, 0. A universal method to employ CT scanner information in stereotactic surgery. App. Neurophysiol. 45:352-364; 1982.

2. Leong, J. Implementation of random positioning error in computer- ised radiation treatment planning systems as a result of fractionation. Phys. Med. Biol. 32:327-334; 1987.

3. Lo, Y. C.; Ling, C. C.; Larson, D. A. The effect of setup uncertainties on the radiobiological advantage of fractionation in stereotaxic ra- diotherapy. Int. J. Radiation Oncology Biol. Phys. 34: 1113- 1119; 1996.

4. Maciunas, R. J.; Fitzpatrick, J. M.; Galloway, R. L.; Allen, G. S. Beyond stereotaxy: extreme levels of application accuracy are pro- vided by implantable fiducial markers for interactive image-guided neurosurgery. In: Maciunas, R. J., ed. Interactive image-guided neu- rosurgery. American Association of Neurological Surgeons; 1993:259-270.

5. Svensson, G. K. Quality assurance in radiation therapy: Physics ef- forts. Int. J. Radiation Oncology Biol. Phys. 10:23-29; 1984.

IN RESPONSE TO JONES ET AL

To the E&or: We appreciate the interest of and comments from Jones et at. ( 1) concerning our article. As stated in that study, our purpose was “to assess the effect of setup uncertainty on FSRT (frac- tionated stereotactic radiotherapy) without prejudice as to its mag- nitude and as to whether planning target volume should be enlarged. Indeed, the quality assurance (QA) procedure and the type of frames used may affect the magnitude of setup uncertainty. Thus, each in- stitution should evaluate the associated setup uncertainty for its own QA procedure and frame when using FSRT. It is then a clinical judg- ment as to how much margin should be included in the planning target volume to account for the effect. Relatedly, whether prescribed dose should be adjusted is a matter of clinical judgment.” I f a larger plan- ning target volume is judged to be necessary for FSRT due to in- creased setup error relative to that for single-fraction stereotaxic ra- diosurgery (SSRS ), and if a reduction in dose is deemed advisable, then the potential radiobiological advantages of fractionation may be partly offset. The following pertains to the two specific points in their

letter. The main issue is, if different immobilization methods are used in SSRS and FSRT, whether setup error would be larger for the latter. Increased setup error could arise from the reduced rigidity and pre- cision of the “removable” head frames. In addition, the stringent QA procedure of SSRT may be impractical to be applied for a large frac- tional number. Related to the use of multiple fractions, Jones et al. suggested, based on statistical consideration, “the potential for im- provement in the likelihood of hitting the target simply by repeti- tion.” However, given that the random setup error remains the same, the margin should be set with a view to avoid marginal misses in the treatment of the tumor planning target volume. The increase in frac- tion number does not reduce the magnitude of setup error, although it does lessen the likelihood of errors of significant consequence due to the smaller dose per fraction.

We do agree that the use of implanted fiducials may increase the pre- cision of repeated setup. If such an approach yields the same setup ac- curacy as that achieved by invasive head frames in SSRT, then the ra- diobiologic advantage of fractionation would not be compromised by the need to increase the irradiated volume.

YEH-CHI Lo, Ph.D. C. CLIFTON LING, Ph.D. Department of Medical Physics Memorial Sloan-Kettering Cancer Center New York, NY 10021 DAVID LARSON, M.D., Ph.D.

Department of Radiation Oncology University of California at San Francisco San Francisco, CA 94145

1. Lo, Y.-C.; Ling, C. C.; Larson, D. A. The effect of setup uncertain- ties on the radiobiological advantage of fractionation in stereotactic radiotherapy. Int. J. Radiat. Oncol. Biol. Phys. 34: 1113- 1119; 1996.

RE: KONSKI AND SOWERS IJROBP 35(2):361-367; 19%

To the Editor: Konski and Sowers have recently reported in IJROBP 35 (2) :361-367; 1996 (3) an association of pubic fractures with pelvic radiotherapy for endometrial cancer. Radiation has been demonstrated to cause accelerated bone loss, and radiation osteonecrosis in other sites is well described. However, epidemiologic data to demonstrate the contri- bution of radiotherapy to the risk of pelvic fractures are difficult to in- terpret. There is virtually no literature on the incidence of pubic fractures in the general, unirradiated population. In addition, patients with endo- metrial cancer commonly have many risk factors for osteoporosis and fractures, such as older age, postmenopausal status, and, of course, fe- male gender. Therefore, it is nearly impossible to produce a meaningful relative risk.

As these are relatively rare, sporadic events and detection is usually fortuitous rather than systematic, the quoted incidence rate of 2% needs to be interpreted cautiously (what was their incidence in other years?).

We observed postradiation pubic fractures in the past, and there- fore, we have recently been carefully investigating and documenting new cases. In total we have observed 11 cases in four endometrial and seven cervical cancer patients occurring between 1985 and 1996. In the last 5 years, when we have been more vigilant, we have diag- nosed O-4 such fractures per year while treating an average of 162 endometrial and cervical cancer patients with pelvic radiotherapy an- nually (approximate risk = 1% of treated patients). Fractures oc- curred at median 14.8 months (7-54) after radiotherapy. Dual ab- sorption bone density scans have been done on six patients and confirm low bone density in all patients (Ll bone density 63- 104% of age matched normals with one value spuriously above 100% be- cause of compression fracture). Patients developing fractures were significantly older than the average patient receiving pelvic radio- therapy (mean 69.2 vs. 58.9 years) and had earlier menopause than the general population (mean 48.3 vs. 52 years). One patient had been taking glucocorticoids chronically for Addison’s disease, 1 pa- tient had bilateral total hip replacement and two patients had previous history of peripheral vascular disease. Body habitus, previous history of fractures or concomitant medical problems were not consistent risk