- Email: [email protected]
RE: SF2: Clothes for the emperor?
R
ADIOTHERAPY aO~~~~~~~
ELSEVIER
Radiotherapyand Oncology34 (1995)81-83
Letter to the Editors RE: SF*: CLOTHES FOR THE EMPEROR? To tbe Editors, We read with interest Letters to the Editors dealing with the surviving fraction at 2 Gy (SF,) [l-3]. At the risk of offending some,we wish to offer somethoughts which we think are relevant to the discussion. First of all, when the concept of SF, was initially put forth 141,the data demonstrated that groupings of cell lines could be made that showed correlation between SF2 and tumors of similar histologies. However, the correlations were crude; there was major overlap. This degree of overlap really representsthe heart of the controversy, in the sensethat if one wants to use this test for prediction, one has to ask how good a correlation can one really achieve. It is not surprising that patients who respond to treatment will have lesser values of survival at 2 Gy than patients with resistant tumors, on average.The total population of patients seenby the radiation oncologist is representedlargely by two groups, one favorable and one unfavorable. It would be surprising if the SF2values were the samein both groups. We expect some difference and we expect the patients who do well to have more favorable values on average. The issue from the standpoint of predictability is not whether the population of patients has a lower (i.e., better) SF, but whether individual patients can be predicted reliably to do well or not locally. This is the real issue, as far as we are concerned. In our opinion, the predictive value of this assay dependsnot upon the group of patients as a whole that does well versus the group that does poorly, but rather on ability to predict the individual outcome for each patient. We acknowledge that we do not expect perfection. Even so, we would be reluctant to use a test clinically that had a correlation of less than 900/o,that is to say, for us to trust the predictability of the SF, as a sole assay,we would need to be able to predict that, on the basis of that one test, the correlation between the SF, and the local outcome would be at least 9O?haccurate. Otherwise.,we believe there would be too much error in the system to trust reliably for the decision-making process about patients’ treatment. In concert with other assays,the correlation for this test might be less,but the ability to predict reliably on the basis of a full battery still requires a correlation of about 90% in order to be useful clinically. An appealing feature about the SF, is that it is a simple one point assay.On the other hand, to be truly reliable, one must be confident that the data point lies on the straight line portion of the survival curve. If the 2-Gy point is on the curving portion of the survival plot, the data may not be easily reproduci-
ble. In order to trust the SF, as an indicator of ‘sensitivity’, one must have evaluated enough points on the survival curve to be confident that the SF2 point is on the linear portion of the curve. Without that, we believe the interpretation is not necessarily reliable. The final point to emphasize concerning the SF, is that it utilizes not a cell line which has acclimatized to the in vitro environment, but rather a tumor explant. As such, many of these cells are destined to die spontaneously after plating; careful controls are necessaryto account for this variable to permit meaningful (and reproducible) interpretation. Furthermore, a heterogeneousexplant may not necessarily be representativeof the whole tumor, which may harbor multiple cell lines with diverse ‘radiosensitivities’. Sincerely, David I. Rosenthal M.D., Phuc Nguyen M.D., Dan P. Garwood M.D., Eli Glatstein M.D. (received 9 August 1994; accepted 2 September 1994) The University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX 75235-8590. USA
References
[l]
Bentzen,S.M. Predictionof radiotherapyresponseusing SF,: why it may work after all. Radiother.Oncol. 31: 85-86, 1994. [2] West,C. and Hendry, J. Predictionof radiotherapyresponse usingSF$ is it methodologyor mythology?Radiother.Oncol. 31: 86-88, 1994. [3] Raaphorst,G.P., Ng, C.E. and Bickray,T. Further on prediction of radiotherapyresposeusing SF,: is it methodologyor myth?Radiother.Oncol. 31: 88-92, 1994. [4] Fertil, B. and Malaise,E.P.Intrinsic radio sensitivityof human cell lines is correlated with radioresponsiveness of human tumors. Analysis of 101 published survival curves. Int. J. Radiat.Oncol., Biol. Phys.9: 1699-1707,1985.
DOSE ESCALATION, TREATMENT PROTRACTION, AND THE STEEPNESS OF THE DOSERESPONSE CURVE FOR SQUAMOUS CELL CARCINOMA OF THE HEAD AND NECK To the Editor,
Bentzen [l] has discussedthe steepnessof the dose-response curve for tumor control and its relation to the design of clinical
0167-8140/95/$09.50 0 1995ElsevierScienceIreland Ltd. All rights reserved SSDI 0167-8140(94)01447-B
ADIOTHERAPY aO~~~~~~~
ELSEVIER
Radiotherapyand Oncology34 (1995)81-83
Letter to the Editors RE: SF*: CLOTHES FOR THE EMPEROR? To tbe Editors, We read with interest Letters to the Editors dealing with the surviving fraction at 2 Gy (SF,) [l-3]. At the risk of offending some,we wish to offer somethoughts which we think are relevant to the discussion. First of all, when the concept of SF, was initially put forth 141,the data demonstrated that groupings of cell lines could be made that showed correlation between SF2 and tumors of similar histologies. However, the correlations were crude; there was major overlap. This degree of overlap really representsthe heart of the controversy, in the sensethat if one wants to use this test for prediction, one has to ask how good a correlation can one really achieve. It is not surprising that patients who respond to treatment will have lesser values of survival at 2 Gy than patients with resistant tumors, on average.The total population of patients seenby the radiation oncologist is representedlargely by two groups, one favorable and one unfavorable. It would be surprising if the SF2values were the samein both groups. We expect some difference and we expect the patients who do well to have more favorable values on average. The issue from the standpoint of predictability is not whether the population of patients has a lower (i.e., better) SF, but whether individual patients can be predicted reliably to do well or not locally. This is the real issue, as far as we are concerned. In our opinion, the predictive value of this assay dependsnot upon the group of patients as a whole that does well versus the group that does poorly, but rather on ability to predict the individual outcome for each patient. We acknowledge that we do not expect perfection. Even so, we would be reluctant to use a test clinically that had a correlation of less than 900/o,that is to say, for us to trust the predictability of the SF, as a sole assay,we would need to be able to predict that, on the basis of that one test, the correlation between the SF, and the local outcome would be at least 9O?haccurate. Otherwise.,we believe there would be too much error in the system to trust reliably for the decision-making process about patients’ treatment. In concert with other assays,the correlation for this test might be less,but the ability to predict reliably on the basis of a full battery still requires a correlation of about 90% in order to be useful clinically. An appealing feature about the SF, is that it is a simple one point assay.On the other hand, to be truly reliable, one must be confident that the data point lies on the straight line portion of the survival curve. If the 2-Gy point is on the curving portion of the survival plot, the data may not be easily reproduci-
ble. In order to trust the SF, as an indicator of ‘sensitivity’, one must have evaluated enough points on the survival curve to be confident that the SF2 point is on the linear portion of the curve. Without that, we believe the interpretation is not necessarily reliable. The final point to emphasize concerning the SF, is that it utilizes not a cell line which has acclimatized to the in vitro environment, but rather a tumor explant. As such, many of these cells are destined to die spontaneously after plating; careful controls are necessaryto account for this variable to permit meaningful (and reproducible) interpretation. Furthermore, a heterogeneousexplant may not necessarily be representativeof the whole tumor, which may harbor multiple cell lines with diverse ‘radiosensitivities’. Sincerely, David I. Rosenthal M.D., Phuc Nguyen M.D., Dan P. Garwood M.D., Eli Glatstein M.D. (received 9 August 1994; accepted 2 September 1994) The University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX 75235-8590. USA
References
[l]
Bentzen,S.M. Predictionof radiotherapyresponseusing SF,: why it may work after all. Radiother.Oncol. 31: 85-86, 1994. [2] West,C. and Hendry, J. Predictionof radiotherapyresponse usingSF$ is it methodologyor mythology?Radiother.Oncol. 31: 86-88, 1994. [3] Raaphorst,G.P., Ng, C.E. and Bickray,T. Further on prediction of radiotherapyresposeusing SF,: is it methodologyor myth?Radiother.Oncol. 31: 88-92, 1994. [4] Fertil, B. and Malaise,E.P.Intrinsic radio sensitivityof human cell lines is correlated with radioresponsiveness of human tumors. Analysis of 101 published survival curves. Int. J. Radiat.Oncol., Biol. Phys.9: 1699-1707,1985.
DOSE ESCALATION, TREATMENT PROTRACTION, AND THE STEEPNESS OF THE DOSERESPONSE CURVE FOR SQUAMOUS CELL CARCINOMA OF THE HEAD AND NECK To the Editor,
Bentzen [l] has discussedthe steepnessof the dose-response curve for tumor control and its relation to the design of clinical
0167-8140/95/$09.50 0 1995ElsevierScienceIreland Ltd. All rights reserved SSDI 0167-8140(94)01447-B