Update February 2007
MORE COMMENTS UPON THE REVIEW OR THE USE OF MICROWAVE THERAPY FOR THE TREATMENT OF PATIENTS WITH CANCER
Under the auspices of the NH&MRC.
INDIVIDUAL PATIENT DATA RECEIVED FROM MEDICAL PRACTITIONERS. (Page 83)
The Committee dealt with these in about ¾ page, and referred to “eight different patients”:
a. “Five of the patients had brain cancer (four glioblastoma multiforme and one grade II astrocytoma)”
b. “One had breast cancer”
c. “One had bladder cancer”
d. “One had lung cancer.”
“The. . . . patients had all received microwave therapy with glucose blocking agents in Western Australia between 1995 and 2002.”
“Of the patients with brain cancer, the patient with grade II astrocytoma was known to be alive three years after treatment. The five year survival rate of this cancer with conventional surgical and radiotherapy treatments is as high as 70% (Boyages and Tiver 1986) and therefore this result is not unexpected. One of the patients with glioblastoma multiforme was known to have died 19 months after treatment, whilst the status of the other three patients is unknown.”
The Cases Dealt-with.
The Committee has not made clear the origin of these reported cases, other than that they were (apparently) from Western Australia. Detailed under point 5. of my Submission to the Microwave Review are 5 patients with brain “cancers:” 4 males with glioblastoma multiforme (“GBM,” equivalent to an astrocytoma grade IV) and one female with an astrocytoma grade II. These were treated in Fairfield, Victoria. Reasonably, one may assume that these five patients of mine are the same as those referred-to as having been treated in Western Australia (error 1). The other 3 patients may have been from Western Australia. To recapitulate (UHF+ = UHF + GBA) :
i. GBM, Date of Birth (DoB) 13/2/1953. Failed DXRT. UHF+ October 2000. There seemed some post-treatment improvement in the degree of midline shift. Useful follow-up was precluded by surgical complications (possibly by the same surgeon who operated upon patient TU, see elsewhere on the website). Not mentioned in the Submission was the pre-treatment CT date 2/10/2000, and a follow-up CT on 14/12/2000 ~10 weeks apart.
ii. Female with Astrocytoma grade II (now termed “diffuse astrocytoma,”) DoB. Operation/diagnosed 8/2000, had UHF+GBA treatment. Had annual MRIs; November 2003 showed no change. (This was the report sent to the Committee.) After the announced closing date for the Review, she had another MRI with no change, and the most recent (November 2006) also showed no change. The Committee assumed that she had only survived 3 years (error 2), being the time to the last reported MRI – a conservative conclusion. There was no date of death or indication that she was lost to follow-up: the assumption should have been that she was still alive as at the date of the Submission (>4 years). (See below for discussion.)
iii. GBM DoB 28/8/1956. Operation/diagnosed January 2001. Had UHF+ followed within 30 min. each day by DXRT. Died 31/10/2002 - ~22 month survival.
iv. GBM DoB 20/9/49. Operation/diagnosed 22/11/2001. Had UHF+ followed within 30 min. each day by DXRT and chemotherapy. The tumour showed little change at 23/8/2002 (9 months), but relapsed thereafter. He was still alive at >1 year.
v. GBM DoB 20/11/50. Operation/diagnosed June 2001. Had DXRT and some chemotherapy, with recurrence, then UHF+ starting 6/9/2002. Felt better and foot not dragging as much, 22/9/2002-5/12/2002 (at least). No further feedback.
Comment upon cases.
I shall deal with the cases with GBM first. This condition carries an appalling prognosis, with a two year survival of 8.4% (overall, Davis et al 1999) five year survival* (aged 45-54 years) of 7%, and a median survival (patients >45 years of age) of 7 months (in an Australian series, Boyages & Tiver 1987), and ~9 months (reading from Figure 15 of Daumas-Duport et al. 1988). Any patient who can survive greater than the median survival time has achieved a bonus. *Usually taken from date of operation/biopsy.
Even without a prospective, randomized, large series, some conclusions can still be drawn from the listed cases:
Cases i. and v. are patients who had already received a course of DXRT, with progression or failure. The outlook was extremely poor. To have registered any form of remission or stable disease is noteworthy : -
Case i. had a measure of tumour size become smaller; objective stable disease, possibly partial remission and clinical improvement. Any GBM patient who can show objective improvement is noteworthy – clinical plausibility)
Case v. reported less foot dragging and feeling better, and was still alive about 18 months from diagnosis – survival greater than the median time with subjective improvement - clinical plausibility.
Case iii. He survived 22 months – over twice the median time interval – clinical plausability.
Case iv. After combined DXRT, UHF+ and later Chemotherapy, he survived with stable disease only to relapse at 11+ months, going on to survive greater than 12 months. Here, the time till progression exceeded the median life expectancy – clinical plausibility.
If we assume that the three patients who survived greater than 12 months had achieved the 30% percentile, then Fairfield should have treated some 9 patients with GBM. This seems unlikely, meaning that the results may add to clinical plausibility.
Case ii. She had a Grade II Astrocytoma, now redefined as a “diffuse astrocytoma.” This is reported to have a mean survival time after surgical intervention of 6-8 year5. Sadly, despite much being written on the treatment of this condition, no approximation of a control negative (untreated) group in a randomized study was presented until an interim report in 2002 (prior to the Committee’s commencement) and reviewed by Papagikos (2005, prior to the completion of the Committee’s Review). Until that time there had been a presumption that radiotherapy should be helpful for this condition, and studies either omitted a control negative group &/or studied survival figures retrospectively. The history of the handling of this condition is complicated by changes in radiation therapy (when relevant), the improved diagnostic imaging tools (about mid 1970s) and variations in the diagnostic terms and criteria, which were not reasonably stabilized until 19884. Even today, there are likely to be cohorts in any series: those who presented with electrical disturbances (fits), expected to present early in the tumour development, and those with physical causes of symptom (pressure, obstruction, nerve function loss), likely to present with larger tumours.
The presumption of ionizing radiation’s therapeutic value for brain tumours has been entrenched in the literature. Because of the loading of the Committee with Radiotherapists, the strange reference to the paper by Boyages and Tiver, and the comments made in it, a review of some of the landmark papers in the literature was undertaken. Some readers who have reached this poin may find the following analyses very boring, and you could jump over much of it.
χ2 Matrix Actual Predicted
Total un-irradiated = 51 10 12.75
Total irradiated = 58 17 14.25
(Total 27 27)
(Figures read from the published graph.)
P = 0.289 (CHITEST of Microsoft Exel; not significant). (This retrospective game with statistics is for illustration and is inappropriate for therapeutic conclusions). Overall, the interpretation by Andersen was that irradiation was beneficial. There was some consideration of possible factors during the study that may introduce bias. Whilst believing that irradiation may add some months to the lives of patients with GBM, the author does question whether the possible benefits are necessarily in the patients’ best interests.
· Leibel et al. (1975) notes that some earlier studies “. . lack a control group of patients treated without radiation therapy during the same time. Furthermore, how they selected patients for radiation therapy is unclear. There are numerous other reports in the literature which taken individually are unconvincing, but which in toto do suggest that radiation therapy is of benefit . . .” The study was retrospective, examining the files of patients treated between 1942 and 1967. (See earlier comments about therapy and imaging pre ~1975). “Whether or not a patient with an incompletely removed astrocytoma was irradiated depended solely upon the opinion of the neurosurgeon in charge. In reviewing the records we found no information to suggest that at the time of referral the expected prognosis for the patients referred differed from that for those not referred. There was no indication that the more favorable cases were not irradiated, and no detectable differences were noted in the age or sex of the two groups.” But then, why would a neurosurgeon document such information ? Then “Since, the primary purpose of the present study was to determine the value of irradiation in the treatment of patients with low-grade astrocytoma, the 25 patients who did not have the opportunity to be irradiated were excluded from further consideration.” (!) There is presented a table which, in part, is :
Surgery + irradiation
Interval Incomplete Incomplete
Year resection resection
1 51% (19/37) 80% (57/71)
3 27% (10/37) 59% (42/71)
5 19% (7/37) 46% (33/71)
When graphed (in their Figure 2), the labelling seems to be around the wrong way, with Surgery + irradiation showing lower values. On closer inspection of this “absolute survival rate” graph, both lines for the above percentages run close to parallel/proportion. This can be interpreted to mean that of those not treated by irradiation, some ~50% had died by the end of the first year, but the survivors then followed a relative course not very dissimilar to that of the treated group. Since we are dealing with a relatively slow-growing tumour (provided that dedifferentiation does not occur; cf GBM), the conclusion can be reached from this is that the referrals to have radiotherapy for these patients without irradiation, were not made because the surgeon felt (but may not have written) that their outlook was not good – about half were to die in a year, being a survival rate only slightly better than that for GBM: that, despite the authors’ best intentions, there had been significant bias. Not to be defeated, the authors concluded their summary with “The survival rate . . . ; in both grades [I & II], it was improved by irradiation.”
· Walker MD et al. (1976) present a study comparing the treatment of anaplastic gliomas with Mithramycin against a control negative group referred-to as receiving “best conventional care.” This group, representing a Brain Tumour Study Group, produced further papers, some of which will be examined later in the current review. It is noteworthy because :
Ø The general format of the clinical studies seems set, and will be followed subsequently
Ø There is a “Principal investigator” per institute, who applies
Ø Randomization (method not disclosed)
Ø Criteria of acceptability including,
Ø An expectation that the randomized patients are expected to have a life expectancy of two months (selection).
The starting time taken for survival assessments is from the date of randomization, which is a woman (or man) made date, dependant upon the Principal Investigator “getting around to it.” It is unlike the surgical date, which is set by the need to investigate, diagnose +/- decompress etc., and is more a disease-inherent date. Since the definition of the starting date seems omitted from subsequent studies (1978 and 1979, see later) we are left to assume that this is the definition that has been applied to the later studies. With a disease carrying a median survival in the order of seven to eight months (usually), delay (or speed) in the randomization, together with the two month survival selection criterion (at least) could have a profound influence upon the subsequently recorded “survival.” Certainly, the control negative groups, when graphed by the Kaplan-Meier method, have the “best fit*” line often (but not always) passing through 100% survival, whereas the treated groups usually have the line passing through the starting time axis above 100% and showing a shoulder, consistent with an investigator-induced effect (rather than the effects of treatment). What happens to the patients randomized and then rejected by the selection criteria is not provided. These issues will be examined in more detail with respect to the later studies. *“Best fit” is a visual assessment taken from the published graphs, some of which are quite small. These assessments are, then, approximate.
· Fazekas (1977). He starts by asking a string of rhetorical questions about Grade I & II Astrocytomas and the role of radiotherapy. His study was retrospective on 68 patients accrued between 1958 and 1974, which would have predated the CT scanning benefit. The classification was based upon Kernohan’s. When the survival interval started is not clear. The emphasis seemed to be upon the degree of surgical removal. Patients were excluded if they died within 30 day of craniotomy. Statistical evaluation was by using a computer with “. . a modified life-table program.” There seems no discussion of selection factors other than the diagnosis of astrocytoma, but “A retrospective multifactor analysis of the irradiated and non-irradiated groups implied the absence of unrecognized bias.” However, there was no patient documentation to support this. His graphs are not Kaplan-Meier type and generally have pooling of Grade I (generally considered ~benign) and Grade II (low grade malignant), so they cannot be compared with other studies. He states “. . the addition of irradiation caused dramatic improvement in the 5-year survival of patients whose tumour was incompletely excised.” The difference was “The disease-free survival rate was increased from 13% to 41% by adjuvant radiotherapy . . ,” then “Irradiation has little effect on the 10-year survival . . . Radiation therefore plays a palliative more than a curative role.” He then gives guidelines of technical type.
· Sherline (1977). After noting that the evaluation of radiotherapy for brain tumours had not been subjected to a controlled clinical trial, he then went on to conduct a retrospective review over the records from 1942-1967. He notes that “The accumulated information does suggest that irradiation increases the disease-free interval and the five- and ten-year survival rates of many, if not most, patients with brain tumours, irrespective of histologic type.” There seems no mention of criteria applied in advising radiotherapy. Regarding astrocytomas, predominantly Grade II with and without radiotherapy, there were 5 year survivals of 25% and 0% respectively, hence “Because the two groups of patients with incompletely resected astrocytomas appeared to be similar except that some were irradiated, it was concluded that radiation therapy produced the improved result.” Because of the evolution in imaging and pathology classifications, this report is of historical interest only, but illustrated the entrenched conviction that radiotherapy is clearly of benefit, and that such reviews were to affirm this.
· Walker et al. (1978). This study compares radiotherapy +/- BCNU on malignant gliomas, and includes a control negative group. It is stated to be a controlled, prospective, randomized study, and is relatively large, but does not seem to provide the accrual interval. Being in the pre-CT era, it is less relevant today but, because the Astrocytomas III & IV/GBM progress rapidly, the interval between first symptom and operation is likely to be relatively short whatever the initial presentation. The study goes into this aspect in some detail. Sadly, the details of selection and randomization are not clear: “Patients, who in the judgment of the principal investigator [per institution] met the criteria for acceptability, were randomized to one of four treatment arms by a telephone call to a Central Office. All patients were informed . . . . and the availability of continued care should they decline to enter or remain in the study. . . . The principle investigator of each institute was primarily responsible for determining the eligibility of any patient’s entrance into the study.” Required were operation, resection and histopathology. “All patients who were randomized, whether acceptable or not, were considered within the randomized population, however inclusion in the Valid Study Group (VSG) was dependent upon subsequent fulfilment of the protocol requirements of critical pathology review, receipt of at least some designated treatment, control of the use of corticosteroids, and appropriate follow-up review. The patient must have been randomized within 6 weeks of definitive surgical resection and establishment of the diagnosis. It was expected that all randomized patients would have minimum postoperative life expectancy of greater than 2 months (although none was removed from the study because of a shorted survival), . . . .” Initially, 5,000 rad was administered (=50 Gy), but after 10 months, the dose was increased to 6,000 rad. “There are uneven numbers of patients randomized to each therapeutic arm, as two participating institutions were unable to enter patients into the Supporting Care [control] arm. In addition, during the last 6 months of the study, patients were randomized only to the radiotherapy [RT] or radiotherapy plus BCNU group as interim analysis indicated that these two forms of treatment were superior.” There are a few noteworthy features in the Comparison of clinical characteristics (Table 1): - The ratios Male:female were Support 55:45, BCNU 55:45, RT 62:38, BCNU+RT 75:25, the proportion of males to females rising respectively from 1.22 through 1.63 to 3 across the groups. The incidence of left handedness ranged between 2.0% to 3.2%, which seems surprisingly low. The median interval (day) for randomization was Support 5, BCNU 9, RT 9, BCNU+RT 8. The incidence (%) of GBM in the Support group was 89%, BCNU 92%, RT 92% and RT+BCNU 90%. “The survival time of patients with anaplastic glioma is approximately exponentially distributed up to about 2 years.”
The presented Survival figures were :
Support BCNU RT RT+BCNU
Valid Study Group 14.0 week 18.5 week 36.0 week 34.5 week
Adequately Treated 17.0 “ 25.0 “ 37.5 “ 40.5 “
Now, we may examine the Kaplan-Meier graphs (Their Figure 1) :
a. The time scale markings vary (!) (The original journal was checked):
Time scale points 0-6 month 6-12 month 12-18 month 18-24 month
Measured (mm) ~22.6 mm ~19.0 mm ~21.1 mm ~21.2 mm__
This could expand the first 6 months and contract the second 6 months, the two time intervals that are critical. Whether this is reflected in the drawing of the graphs is unclear. Over the 24 month, each month averages out at ~3.5 mm on the page.
b. The makeup of the graphs is not clear: whether from the Valid Study Group or the Adequately Treated Group. From reading the graph, the Valid Study Group results seem to be the basis.
c. Overall, the shapes of the graphs are not “exponential;” – there is an initial “uncertain” phase until the survivors fall to ~80-90%, then there is an almost straight line “best fit” for all groups until about ~20-30% survival, then there is close to another straight line “best fit” to no survivors which show less steep falls.
d. The extrapolated “best fit” lines for the RT +/–BCNU are virtually identical, and cross the zero time scale at ~105%; and 90% survival is at ~ 2 month, indicating some survival advantage early &/or at the start. The “best fit” line having the formula of:
RT Survivors % ~ 105 – 6.56 x (number of Month) - - - - - - - - - - (1)
e. Both the Supporting Care and BCNU graphs show precipitous falls near the start, 5% of the Supporting Care group appearing to have died within a week ! The two graphs approximate until ~ 60% at 3 month. They then “diverge” over about a month (BCNU better), then run roughly parallel until the 6 month point, before the BCNU group has a much flatter line from ~15% at 8 month. The “diversion” may be illusory; the “best fit” line between the 80%-20% points seems to be [as presented in (1)] :
BCNUa S ~ 100 – 11.3M - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - (2)
But, if we draw a similar line through the more stable region ~58% at ~2.7 month to ~28% at ~6 month (a time with chemotherapy of uncertain intensity [see later] and a rapid decline in survivors from ~75% to ~62% over an ~18 day period), the “best fit” line is :
BCNUb S ~ 114 – 14.54M - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - (3)
The “best fit” for the Support Care (SC) 90%-20% (M based on the scale for 0-6 month as presented in Figure 1 – see above) is :
SC S ~ 112 – 18.7M - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - (4)
This slope is not very dissimilar to that for the BCNUb “best fit” (3).
So, despite a disastrous start, the SC group then was represented by a “best fit” line which indicated an initial overall survival advantage of some 1.1 month ! If the disastrous start is ignored, the slope between 90-77% survival is not very much different from the RT group.
f. The slopes for the graphs 1-3 (above) are different, the major difference shown by the line for the RT groups (essentially the same). They would seem to show an initial survival advantage of ~23 day, before the “best fit” line is followed.
g. The general points are illustrated in the rough graph below :
Expected for RT, RT+BCNU
50% (3) RT& RT+BCNU (1)
SC RT RT+BCNU
6 12 month
Comment - General:
i. The study includes a relatively large number of cases, making the assessment of survival graphs easier and more valid.
ii. The details of the accrual, selection and randomization steps are not clear; a flow diagram would have assisted.
iii. The numbers of relevant gliomas biopsied/diagnosed in all the institutions involved have not been presented, nor the numbers (percentage) who declined to enter the study, nor whether any of those who chose to decline were included in the figures for the Support Group (with or without the patients’ known participation).
iv. The “Principal Investigators” in the relevant institutions had a role not clearly defined, and could introduce bias.
v. There were judgment steps involving the Principal Investigators which were not defined clearly.
vi. The basis for two institutions seemingly unable to enter the patients into the Supporting Care group has not been presented, and how this could be processed in the randomization process is not clear.
vii. How the patients were randomized is not presented.
viii. The numbers of patients involved in the changes in RT dosage from 5000è6000 rad (50-60 Gy) is not clear, nor of any differential effect.
ix. The number of patients included in the Kaplan-Meier graphs and which groups were represented (overall, Valid Study Group or Adequately Treated Group) is not presented
x. The changing extrapolations of the “best fit” graphs back to zero time are of interest and are inadequately explained.
xi. Just when zero time was applied for any patient is not clear; was it the date of the operation, Neuro-pathologists’ report, randomization, or some other date ? (See Walker et al. 1976.)
xii. Unequal time scale points on the Kaplan-Meier graph give concern
Comment - Graphs
i. With such aggressive tumours, the events in the first month from operation and accepted diagnosis (histopathology) are crucial. There seem important features showing on the graphs to indicate that not all the groups were treated equally (the “best fit” SC graph not passing through 100% without good explanation).
ii. The graphs for RT showed a survival advantage in the first month. Whilst this may be due to the radiotherapy having a desirable effect, there may also be a bias, as for the SC group, merely reflecting the criterion to have such patients likely to survive 2 months.
iii. The slopes for the Treatment arms, particularly radiotherapy, show a less steep slope than that for the control negative group (Supporting Care). One might expect that the central part of the graphs for the treated groups should, without bias, show a slope much the same as for the control negative group. Since post therapy growth may involve clones with more aggressive growth characteristics, there is even the possibility for a slope steeper than that for the control negative graphs. Throughout this current review, no such slope has been noted. The claim may be advanced here that a central slope for a treatment group less than the control negative graph is strongly indicative of bias of some kind.
iv. Both RT graphs showed an almost amazingly straight “best fit” concurrence line between about 95% survival and 30% survival, with a slope quite different from those of the SC and BCNU groups. This seems of particular interest –
a) If the RT had an early effect upon tumour progression, this impact had largely ceased by the end of the first month and
b) The subsequent survival advantage obtained by these patients arose because the slope of the survival line was much less steep and that this slope was maintained until about 12 month from zero time. This means that, for a large number of patients, there had been demonstrated a probable difference in tumour growth characteristics (ie slower growth), and that did not change appreciably for a relatively long time. Was this difference in growth the nature of the tumour, being a continuation of the pre-study phase, or was it induced by RT ? (RT would not be expected to exert an important and sustained influence over such a long interval, as indicated by earlier studies of cytological morphology and gross kinetics,,, [although the conclusion relating to small bowel changes may need re-assessment, based upon the work of Maj and Paris, 2003. The study by Blankenberg et al. 1995 indicated that treatments did not change the tumour volume doubling time – approximating the intrinsic growth rate, and that there seemed little difference between the doubling times for Grade II and Grade III tumours ! The graphs for baseline risk factors in chronic myeloid leukaemia treated with Interferon-α is of interest, where the slopes reflect the risk factor levels, approximating the intrinsic growth rates, and the duration of the shoulders, the impact of the Interferon.]
c) The astonishingly similar Radiotherapy graphs involved different mean doses of radiotherapy, as well as the differing chemotherapy doses, the administrations of which are less than clear: the median number of doses (lots of 3 successive days) of BCNU in the BCNU only group was two, in the RT+BCNU, three. Since the lots/doses were repeated every six to eight week, those who survived to median survival at about 18.5 week could have received two to four lots/doses. There is no indication if there was a dose reduction when the patients reached the pre-terminal state.
d) There is no inflection which may indicate when the RT doses were raised, or to indicate a dual cohort behaviour (RT median 5840 rad, RT+BCNU 5500 rad.) (See a later study)
e) The group’s natural inclination was to conclude that the protocol should change “. . . as interim analyses indicated that these two forms of treatment [RT, RT+BCNU] were superior.” and cease accruing Support Care patients.
f) The BCNU “best fit” line, though showing two small deviations, seems to pass through 100% and present a reasonably straight line until ~19%. However, on closer scrutiny, between ~75%, ~2 month, and 60%, ~3 month, the “best fit” is roughly parallel with the SC line. Then it moves forward, progressing with a slightly less steep slope until ~30% at ~6 month. Again, the slope of the curve does not appear to indicate any clear early beneficial effect: rather, the reverse.
a. Despite the large size of the study, there are points in the randomization and study protocol which give rise to concerns.
b. The apparent mildly improved median survival for those treated with BCNU and the moderately improved survival for those treated with radiotherapy +/-BCNU, whilst, as presented, appear convincing, yet leave difficult points to address if the observer is to accommodate the pharmacological and biophysical effects of the treatments in the consideration (biological plausibility5):
i. The Radiotherapy “best fit” graphs show an almost coincidental straight line from about 1 month (presumably about the time of the completion of the radiotherapy course) until about 12 month. This is despite the accepted belief that radiotherapy has its major impact on cells before the next division, with the effects rapidly falling away after that. This means that one would expect to see the graph at say 6 to 12 month have a similar slope to that of the control negative group, but shifted forwards in time; indicating that after inhibition, growth of the tumour recovered and proceeded at about the same as the pre-treatment rate. This is not what has been graphed: the slope after the irradiation interval is less steep than in the control group, and maintains that slope until about 12 month – without any appreciable wobble ! The appearances would seem consistent with those deemed fit enough to undertake radiotherapy (bias, favouring a mildly improved survival at the start) surviving until RT is completed, then the growth become apparent, revealing a mean intrinsic growth rate for the tumours in these groups that probably existed before the operation and radiotherapy, which continues apparently unaffected by the radiotherapy !
ii. The BCNU alone graph shows steps, which may relate to the criteria for, the stages, and the degree of chemotherapy, for which there is a lack of certainty. Seemingly, some 15% had succumbed within the first month (possibly selection or poor risk) with a later improvement, then following reasonably closely to the control negative slope (exhibiting the intrinsic growth rate for the non-irradiated groups).
c. Whilst there may be a scientific plausibility for irradiation to benefit and, reading the graphs and figures as presented, a clinical plausibility, there seems a lack of biological plausibility, which should give rise to concern.
d. In order to assuage this concern, there needs to be much better documentation of the selection and treatment criteria.
e. If the intrinsic growth rate pre-treatment is represented in the slope for the irradiated groups (ie radiation had little effect), there needs to be an explanation as to how such apparently homogeneous groups could be selected from those chosen not to have irradiation.
f. Subsequent workers would have to be very courageous and brave to include a control negative group in future, similar, studies in the face of the “apparent and obvious” benefits of radiotherapy as presented by Walker et al. (1978).
g. Convincing proof of the benefits of radiotherapy for Astrocytomas III/IV is yet to appear.
· Walker et al. (1979). This study compares the survivals of patients with “malignant glioma” given different levels of irradiation. The diagnostic group comprise predominantly Astrocytoma grade III (9.9%) and IV (86%). The study is retrospective, based upon the patients presented in the previous studies by Walker et al. (1976 & 1978) and a study presented in abstract and which primarily set out to compare chemotherapeutic agents and radiotherapy, covering the years 1966 to 1975. As before, there was randomization, involving a “Principal investigator,” but without the method stated. “It was expected that all randomized patients would have a minimal postoperative life expectancy of greater than two months; however, no patient was removed from the study because of a shorter survival.” The “best fit” graphs for the survival figures in the Kaplan-Meier graphs can be studied and there is presented graphs for all tumours, and another for glioblastoma multiforme alone. Both express survival by week and, as before, seem to vary between time points (less than before). There are slight differences which the authors considered insignificant. Dealing with the total group, as before, the “best fit” line for the control negative group (CG) is essentially straight, seeming to arise from 100% and only deviating at ~30% survival at ~25 week (W) (as seen previously). It has a formula of :
CG S ~100 – 2.9W; expressed by Month: S ~100 – 12.4M - - - - - - - - - - - (5)
The graph for 5000 rad is S ~ 110.5 – 2.2W; S ~ 110.5 – 9.6M - - - - - - - - (6)
For 5500 rad S ~ 110 – 1.7W; S ~ 110 – 7.2M - - - - - - - - - - - - - - - - - - - (7)
For 6000 rad S ~ 109 – 1.4W; S ~ 109 – 5.8M - - - - - - - - - - - - - - - - - - - - (8)
All the radiotherapy graphs show a slight shoulder at about 7-5 – 80 % survival, and this feature is more marked in the graphs for GBM. The following estimates extrapolate the “best fit” lines for some of the straighter segments. The graphs for the GBM tumours show :
For 5000 rad, (98 – 80% survival) S ~ 113 – 2.3W; S ~ 113 – 10M - - - - - - (9)
5000 rad (80 - 45% survival) S ~ 120 – 2.6W; S ~ 120 – 11.3M - - - - (10)
For 5500 rad (75 - 45% survival) S ~ 127 – 2.3W; S ~ 127 – 10M - - - - - (11)
For 6000 rad (100 – 80% survival) S ~ 115 – 1.7W; S ~ 115 - 7.5M - - - - (12)
6000 rad (75 – 24% survival) S ~ 118 – 1.45W; S~ 118 – 7.5M - - - - (13)
So, what does all this show ? We can compare the slopes (for Months) in the earlier publication (1978) with the current one for GBM, using the figures for the parts of the graphs that cross the median level (50%) :
1978 1979 1979 1979 1978 1978 1979
rad: Nil (CG); Nil (SC); 5000; 5500; 5500; 5840; 6000
Slope: -18.7 -12.4 -11.3 -10 -5.56 -6.56 -7.5
a. The control negative groups (CG & SC) from 1978 and 1979 show conspicuous differences - the extrapolation of the graph crossing the horizontal axis at 6 month in 1978, 8.1 month (35 week) in 1979. Given the numbers involved, there would likely be some change in selection criteria or basic supporting care. Perhaps that is why the graphs changed from showing survival time from month to week !
b. The slopes for the Radiotherapy groups in 1978 were less than those in 1979.
c. The central portions of the graphs (eg survival 80 – 30%) in 1979 tended to have similar slopes to the control negative group.
d. In 1979, with each increasing increment in radiation dose, there appears an increasing latency early in the study time when there are few deaths. This would seem consistent with the selection of those able to survive the treatment, and more likely than an immediate effect of radiotherapy.
e. When one allows for the increasing latency (selection ?), the perceived benefit in increased median survival would seem to decrease appreciably.
f. At the highest dose (6000 rad) in 1979, the slope does decrease appreciably, but not to the extent reached in 1978. This is the graph which would seem to have an appreciably different inherent growth rate (see earlier) than the other groups.
g. The 1978 figures, in particular, seem consistent with the patients with the more aggressive tumours being excluded from the radiotherapy arm(s) and included in the negative control (SC) group.
Conclusion: As with the earlier study, there are points of concern over the biological plausibility and, accordingly, the conclusions reached in these studies. The authors conclude by noting that some other authors had not produced results in general agreement – “An explanation for this is not apparent.” The most likely explanation is that there were flaws in the randomization and acceptance of patients, and convincing evidence that radiation therapy for this type of malignancy contributes any real benefit is still awaited.
· Kristiansen et al., (1981). This study sets out to be a controlled, prospective and randomized investigation of the treatment of patients with Astrocytomas Grade III and Grade IV (GBM). (It is included here because Boyages & Tiver referred to it; see later.) The material presented was collected between 1974 and the fall of 1978, with AP Andersen (see earlier) one of the authors. The pathology classification was not supplied, but terms such as “astrocytic gliomas with marked cellular anaplasia, pleomorphism, endothelial proliferation and necrosis” were flags for inclusion. Inclusion was conditional upon the ability of the surgeon to adequately decompress and was decided “. . . after the neuropathologic examination of the resected tumor.” There is no indication if any patients declined to participate in the trial and, if so, how many, why, and were they included in the untreated group ? “If serious side effects necessitated interruption of treatment for more than 14 days, the patient was excluded from the trial, but the registration would be continued.” What, exactly, this meant and how many were involved seem unclear. The study group was 118 patients with “postoperative randomization,” but the technique was not provided. The numbers in the 3 groups were 45, 35 and 38: the first two groups (1 & 2) received radiotherapy (45 Gy [= 4500 rad], a low dose by the American standards] ± bleomycin, the third (3) group received neither (“surgery alone;” SA). Why the group numbers should differ by so much is not clear. However, by χ2, P = 0.189 (not significant, but of concern). A Kaplan-Meier graph gives the survivals for groups 1 & 2, and group 3, with the top survival figure for the vertical axis 99% ! Unlike the earlier study (Andersen), divergence between the graphs is apparent early, by the second month. After ~3 month (83% survival), the untreated group’s “best fit” seems to approximate a straight line until about 8 months (16% survival):
SA S ~ 118 – 11.2M - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - (14)
A similar “best fit” line for the treated groups, through the points between 5.5 month (88% survival) and about 16 month (16% survival) is:
RT +/- bleomycin S ~ 124 – 6.5M - - - - - - - - - - - - - - - - - - - - - - - - - - - - - (15)
With some reservations, a “best fit” line may be drawn between 7.7 month (75% survival) and 10 month, (35% survival) :
Central RT +/- bleomycin S ~ 145 – 9.1M - - - - - - - - - - - - - - - - - - - - - - - - - (16)
There may be considerable reservations over aspects of the statistical methodology and lack of information. The possible shoulder in the surgery alone group raises the possibility of delay or selection. The shoulder in the radiotherapy groups, with few deaths in the first two months, and a slowly steepening curve thereafter would be consistent with the effects of selection criteria and the radiotherapy over some 4 - 5 month ensuing, but the differing slopes between the central parts of the surgery alone and the radiotherapy curves raises the question of a selection reflecting intrinsic growth rate differences. Comparing the slope for the surgery alone (-11.2) with those from the studies by Walker et al. (1978-9), -18.7 and -12.4, and the Radiotherapy slopes (45 Gy, = 4500 rad) 9.1 and the Walker et al. (5000 rad) -11.3 reveal differences which raise doubts. “The difference between Groups 1 and 2, and group 3 is statistically significant.” That is all that was said of statistical analysis: no named test and no specific groups compared ! The authors felt that “The main purpose of this initial phase of the investigation has been to provide statistical proof of a benefit of postoperative irradiation in patients with astrocytomas Grade III & IV. Our results have confirmed earlier investigations. . . .” Some may be more cautious, seeking better statistical methodology and information.
· Wara (1985). In his abstract, Wara refers to “Results of radiation therapy obtained at the University of California, San Francisco over the last 25 years of various adult types of brain tumours . .” He then draws upon the publications of other authors – it is a review. Regarding Astrocytomas, he considered “Most reported data are difficult to evaluate because of the failure to include control groups of non-irradiated patients, inadequate radiotherapy, and/or pooling mixed of tumour types together.” He refers to Leibel et al. (1975 – see earlier) and Fazekas (1977), both pre-CT era and now obsolete. For Malignant gliomas, he draws upon Sheline (1977) considerably, again now obsolete for the same reason. He presents no clinical study of his own.
· Boyages and Tiver (1987)3, after reviewing some five studies on the effectiveness of radiation therapy on GBM +/- Astrocytoma Grade III stated, “Despite evidence for the value of radiotherapy in the management of cerebral astrocytomas, we have found some of our colleagues in neurology and neurosurgery reluctant to refer their patients for radiotherapy.” Effectiveness of radiotherapy was established, assumed and to be confirmed ! The study was retrospective, drawn from patients seen between January 1980 and February 1985. All grades were included. Median survivals for Grades I & II, III and IV (GBM) were 42, 12 and 7 month respectively. “Patients with grade IV tumours rarely survived beyond one year.” (Refer to the cases of GBM receiving UHF+.) Reading from the Kaplan-Meier graph (Their Figure 1) indicates that some ~36% of the Astrocytoma Grade I & II patients had died in less than 12 month. Thereafter, the death rate was much less, with some ~50% surviving 5 year. There was no attempt to construct or draw upon a control negative (untreated) group, so the figures float in isolation without relativity. With conviction, the authors conclude “External beam radiotherapy provides measurable benefits for patients with all grades of cerebral astrocytomas both in terms of quality and quantity of survival. . . . Given that radiotherapy remains the standard definitive treatment for these tumours, a case can be made for further studies . .” – a conclusion built upon faith ! Boyages, who was so convinced of the definitive benefits of radiotherapy for all grades of astrocytomas, played a prominent role in the Committee of the Review, being the fifth listed for the Review Committee (total = 12), and headed the Patient Audit Sub-committee.
In referring to his own publication in 2005 (it, apparently, being considered the definitive work and paper on the subject), the quoted date of the publication is given as 1986 in text, and in the References (p. 109) the entry is “Boyages, J. and Tiver, K.W. (1986) Cerebral astrocytoma - treatment results. Radiotherapy and Oncology, 13, 69-74.” There is no such corresponding reference in PubMed*, no reference to an earlier publication in 1986, and the names do not appear in the index of Radiotherapy and Oncology for 1986. The PubMed entry is as follows (copy & paste) : *PubMed: http://www.ncbi.nlm.nih.gov/entrez/
Boyages J, Tiver KW.
Cerebral hemisphere astrocytoma: treatment results.
Radiother Oncol. 1987 Mar;8(3):209-16.
PMID: 3033750 [PubMed - indexed for MEDLINE]
Seemingly, the title, year , volume and page numbers, as presented in the Review, are wrong ! Not bad for a well-resourced Federal Review ! (Errors 2-5.) Could you, dear reader, really believe that Boyages was completely impartial and appropriately selected for the position on the Committee ? (See following.)
· Piepmeier (1987) performed a retrospective study of 60 patients treated between 1975 and 1985. Survival was based upon time of diagnosis (unspecified). There seems no note on the reasons for providing or withholding radiotherapy. Two patients died of operative complications and were excluded from survival statistics. (Could the surgical complications be related to poor-risk status ?) There were 58 patients treated. He notes the improvements in earlier diagnoses with the CT scans, better surgery and radiotherapy standardization. He presents no graph , but concludes that “There was no significant difference in survival in the surgical subgroups, in the surgical subgroups associated with the use of radiation, or when combining surgery groups with radiation therapy.” He concludes by noting that “Prospective studies with randomized treatment groups will start to answer some basis questions about low-grade tumours and hopefully improve the management of these lesions.”
· Shaw et al. (1989). This was a randomized study of patients treated January 1960 and December 1982, using the new pathology classification of Daumas-Duport et al. (1988). Reference will be to the (new) Grade 2 (=“ordinary astrocytoma”). All patients were included “. . . .who survived at least 1 month postoperatively.” As read from the Kaplan-Meier graph in their Figure 1., “best fit” survival over the 1 – 6 year interval closely followed a straight line usually, and when extrapolated back passing through 100%, breaking away to a less steep slope at ~7 year (~40% survival point): the extrapolation crossing the time line at about 9.5 year, overall median survival ~5 year. The general patterns seen are illustrated in the Figure below. Basic features are noted in the Table :
Age group Radiotherapy (RT) Start time axis t1 (week) t2 (Year)
Age ≥ 35 Year
No/low dose RT
High dose RT
Age ≥ 35 Year New#
High dose RT
Age <35 Year
No/low dose RT
High dose RT
Age 3-27 Year
No/low dose RT
High dose RT
Age 28-34 Year
No/low dose RT
High dose RT
T1 = time in weeks before death is recorded; t2 = extrapolated survival of main cohort, usually taken as the group surviving between 1 – 6 year. #Based upon other graphs.
Rough Diagram to illustrate the main features that have been described
Extrapolation back to the start line
100% t1 Example graphs (very approximate)/age)
Red illustrates No/low dose RT ≥ 35 Year
Brown illustrates RT ≥ 35 Year
Blue illustrates RT < 35 Year
t2 Year 0% survival intercept
Only the graph for no radiotherapy or low dose radiotherapy shows deaths at or near the start. All other graphs had 100% survivors until at least 9 months from the start of graphing ! Only the “best fit” graphs for the group aged ≤ 35 year old seemed to pass through (or very close to ) 100% survival, with the radiotherapy group having no deaths for some 38 weeks, then commencing a pattern seemingly in line with the 100% start (at less than 50% survivors, there are long-term survivors, probably those who had complete removal of the tumour by surgery). Because this graph seems inconsistent with the other graphs, possibly being due to some outlier deaths early in the study, the measurements were made in accord with the other graphs, based upon points between 3 – 7 year. (marked #).
Based upon this graphical analysis, the radiotherapy groups throughout seemed to have some influencing factor(s) within the first year, which has(have) not been explained. We know that the study required the patients to survive one month post-operatively, but for those patients under the age of 35 year, to have no fatality for at least 9 months, and other patient groups to have longer survivals, should raise concern about bias. Referring Neurosurgeons (or other) would be expected to send the fittest patients for the higher doses of radiotherapy, less fit for the lower dose and those with poor prognosis for palliative care. The latter two groups are pooled. If the unusual and unexplained time interval before the first death is reduced in order to be more consistent with the “best fit” line through 100%, the significance found for the ≥ 35 year old patients’ graphs would be expected to be much reduced, and probably insignificant, as was found for the other groupings. With no well defined starting date for survival figures provided, we are left to wonder if many of the groups had their date of commencement based upon a time some months after the completion of radiotherapy !
· Shaw et al. (1989) conducted a retrospective study (1976-1983) covering all the major advances in imaging, treatment and, by review, the newer pathology classification criteria by Daumas-Duport et al (1988). This paper is similar to that of Boyages & Tiver in that there was no control negative (untreated) group. Selection criteria were vague, apparently based upon an acceptance of having both surgery and radiotherapy. “ordinary astrocytomas” (ie diffuse astrocytomas = Grade II) carried a 40% five year survival, whereas, the series by Boyages & Tiver has ~49% (from graph). The “best fit” for the Kaplan-Meier graph (their Figure 2) seems to be close to straight (a slight hump at ~ 4 year), passing through ~100% and crossing the time axis at ~9 year. Analysis of the figures to assess benefits associated with differing radiation doses produced no significant benefit, which seemed to be a disappointment, and note was made of trials that were underway.
· Whitton et al. (1990). This was a retrospective study of 88 cerebral low-grade gliomas (mainly Grade II) accrued between 1960 and 1985. Patients had been referred for radiotherapy. There was no control negative group, so the study only looked at factors that influenced survival. The “best fit” line for the astrocytomas was (Y=year) :
Astrocytoma RT S~104.3 - 14.5Y - - - - - - - - - - - - - - - - - - - - - - - - - - - (17)
The starting date used for the survival interval is not stated. The authors did not notice any “. . . plateauing of relapse[,] and the disease recurrence continued for at least 10 years.” The data for this conclusion is not supplied. “Since the role of radiotherapy. . . is still unclear, this needs to be resolved in prospective trials. . .”
· Vertosick et al. (1991). This was a retrospective review of 25 patients treated from 1978 to 1988. The Kernohan pathology classification was used. Kaplan-Meier graphs for survival and time to dedifferentiation (biopsy) were presented. Both “best fir” graphs for survival and time to dedifferentiation seemed to extrapolate through the 1.15 proportion surviving/cumulative proportion axes, indicating “lag-times” for the main group of ~8 month and ~13 month respectively. In 7 of eight deaths, the cause was dedifferentiation, median time to dedifferentiation for those who received radiation was 5.4 year, 3.7 for those without radiation. The main thrusts of the study were to confirm the impression that improved imaging had contributed greatly to improved survival from the time of diagnosis (biopsy) and the ominous implications of dedifferentiation (“recurrence” but in a more malignant form).
· McCormack et al. (1992). They start by noting that other studies had “. . examined the natural history and effect of treatment on survival.” (The “natural history” should include comparison with the untreated course – the negative control group.) They then set out to perform a retrospective review of patients from 1977 – 1999. Mentioned, were studies involving Daumas-Duport (Neuropathologist), but seemed not to compare the pathology classification; theirs being based upon Ringertz’s from 1950. Apparently three patients were excluded (one postoperative, one of AIDS and one lost), leaving 53 patients in the study. But then “Five other patients survived beyond the immediate postoperative period (30 days) but died while still hospitalized and undergoing rehabilitation therapy.” Then “Five of 53 patients did not receive RT*. A decision to defer postoperative RT was made in 3 patients who were less than 30 years of age and who had undergone gross total surgical resection of a tumour with ‘benign’ histology. Two were felt not to be candidates because of their medical condition.” (*RT = Radiotherapy) Examination of the Kaplan-Meier graphs reveals a “best fit” line extrapolated to slightly above 100% to about 68% at 31 month, then a less steep line through about 36% at 140 month. If 5 patients had died as described, one would expect there to be a rapid, early fall to about ~91%. This level was reached at over 1 year, possibly 12-13 month. So, just what made up the study group is a bit unclear. Note is made of the study by Piepmeier (1987), referring to median survival, when this latter author was reported and discussed by Vertosick et al. (1991) as having used mean survival, making findings difficult to compare with other studies. If some poor-risk patients disappeared early from the McCormack et al. study, the median survival presented (7.25 year) would be favoured favourably. “The majority of studies suggest that radiation therapy of astrocytomas improves patient survival (10 references). Other series have failed to demonstrate a statistically significant survival advantage with RT (5 references). All of these previous studies may be faulted by their methodology and statistical analysis. It has been our bias that radiation treatments do improve outcome by decreasing the incidence of, or prolonging the time to, the occurrence of dedifferentiation. This study is retrospective and does not prove this contention. . .” “Twenty-four (59%) of (these) 41 patients suffered a recurrence of symptoms . . . and had recurrent masses on CT with a median time to recurrence of 4.5 years (54 ±11.0 months). In 22 of the 24 patients, the CT demonstrated edema and contrast enhancement suggestive of transformation to a higher grade neoplasm.” Six of the seven had either Grade III or GBM. “The median time from recurrence to death was 12 months.” From all of this they conclude “. . .it is our opinion that radical surgery and RT offer the best chance for long-term survival.” Done without a control negative group. Their view favoured the hypothesis that aggressive components in the original “low-grade” tumour that enhance with contrast are missed on the tissue biopsy submitted pathology.
· Janny et al. (1993). Again, this was a retrospective study on patients treated between January 1970 and December 1990. Some patients were excluded because of incomplete records or documentation, leaving 58 patients. There were 3 operative deaths (whether resulting from poor risk status was not stated), leaving 55. Twenty three had serial CT-guided stereotactic biopsies, and no further surgery. There seems no note regarding the reason(s) for treating with or without radiotherapy. The Kaplan-Meier graph for those with ordinary astrocytomas (Kernohan Grade II) shows ~20% deaths in just over a year, thereafter the “best fit” graph is reasonably straight until ~6 year, with extrapolations to zero time at ~90% and 0% at ~ 12 year. The authors compared the Kernohan classification with that of Daumas-Dupont (Daumas was misspelt), with some 17 patients being reclassified upwards to Grade 3. “There was no difference in survival time between the 23 patients receiving irradiation and 26 not receiving irradiation. . . .” The median survival time (after surviving surgery) was 64 month, the median time for tumour recurrence was 37.5 month and the median survival time after recurrence (dedifferentiation ?) was 8 month.
· Karim et al. (1996) present a study of low dose (45 Gy) versus high dose (59.4 Gy) irradiation on cerebral low grade gliomas. There is no negative control group. “All adult patient. . . . having a definite diagnosis of low grade astrocytoma, oligodendroglioma, and mixed oligoastrocytomas of the supratentorial areas had been included in this trial. . . . The patients had to have been in reasonable to good general condition as indicated by performance score after surgery: Karnovsky index ≥ 60 and WHO score ≤ 2.” Patients with a Neurologic deficit status (defined) with “Categories 4 and 5 were excluded from this trial.” “Randomization was organized centrally . . .” after surgery, with no other details. Survival was timed from date of randomization, with therapy initiated within eight weeks of operation. (“Usually this interval was < 4 weeks.”) The Kaplan-Meier graphs used squiggly lines which, on a photocopy, makes identification of specific lines difficult to interpret clearly and with certainty. “There is no significant difference in terms of survival or the progression-free survival.” “The important question of efficacy of radiotherapy must now be awaited from the results of the other ‘non-believer’ EORTC trial 22845.” As usual, there are points in the selection, exclusion and randomization steps that give rise for concern. However, since two arms or treatment only are being compared, hopefully, one may assume that major selection factors may have been equalized, although there may be the tendency to favour the higher doses of radiation for the patients in the best condition.
· Peraud et al. (1998) presented a retrospective study on patients who had biopsy diagnosis between January 1991 and June 1993, the pathology being checked at a reference centre. The emphasis was upon the completeness of surgical removal and other prognostic factors. Fewer than 50% received radiotherapy, and it was non-standardized and not detailed. There was no control negative group. The 4 year survival for the diffuse astrocytomas with subtotal resection seemed to be !63%; there being an initial plateau to ~5 month, a steady fall to about 13 month (~75%), then another plateau to ~35 month. The extrapolated central steady fall :
Fibrillary Astrocytoma (= Grade II) S~ 120 – 2.9Y - - - - - - - - - - - - - - - - - - (18)
· Karim et al. (2002 – see earlier). This work is presented as an interim report for a near definitive assessment of radiotherapy for cerebral low-grade gliomas in adults, initiated in 1986. Whilst the histopathology make-up of Astrocytoma (n=180); grade 1 (n=7), grade 2 (n=173), Oligodendroglioma (n=72; 25%), Mixed Oligo-astrocytoma (n=29; 10%), Unknown (n=9; 3%), was presented, the graphics presented the total cohort, without separation into subtypes. Accordingly, the graphs are not completely comparable with those of studies dealing with astrocytomas alone. As usual, there was a selection, with “The patients having to be in reasonable to good general conditions as indicated by the performance scale after surgery: Karnovsky performance ≥60 . . .” “Patients with major functional impairment . . . were not eligible for this trial.” “Informed consent was obtained from each patient before randomizations” but the numbers and reasons for non-consent (if any) were not revealed. “Patients were centrally randomized at the EORTC Data Centre or the MRC Trials Office using a minimization technique” which was not explained or defined. “One group of patients was allocated for early RT within 8 weeks of the day of surgery. . . .” “A maximal permissible interval of 8 weeks was allowed between the day of surgery and the first day of RT.” “The principal end points of the study were overall survival (OS) and time to progression (TTP). The OS was computed from the day of randomization to the date of death . . . The TTP was computed from the date of randomization to the date of progression . . . . defined in the protocol as clinico-neurological deterioration confirmed by definite evidence of tumour activity clinically and on CT scan.” “The analysis was performed according to the intent to treat (ITT) policy . . .” “All eligible patients randomized in the study were included in the OS and TPP analysis (except those for whom no data were available.” There would seem to be undefined features which may be of importance. The total of “eligible and assessable” patients studied in the trial was 290 from a total of 311 patients. “Only 21 were ineligible or unassessable because of missing data (19 patients) and protocol ineligibility (2 patients, age >65 years and bad physical condition.” The treated group may have had a slight advantage until about 3 year, when the lines merge, to separate at about 4.5 year, with the untreated group appearing to perform better thereafter, when the numbers were down to 12 (control) and 9 (treated). (The long-term survivors were probably the patients with Grade 1 astrocytomas; the oligodendrogliomas also likely to survive longer.) The authors considered that there was no significant difference between the arms (P = 0.49). The review of the histology slides reduced the assessable numbers down to 172, some 52 Grade ¾ Astrocytomas were rejected. The combined “best fir” for this sample between 100% and ~80% was :
Low-grade glioma (all types) S ~ 103 – 7.5Y- - - - - - - - - - - - - - - - - - (19)
The finding of significance was the time to progression for all types. These graphs show a spindle outline, showing no patient in the treatment arm having progression in the first 6 months and a shoulder until about 1.5 year; (the graphs represent a starting point of 100% no progression, with each progression represented as was done for the survivors for the survival curves – hence similar graphical format, graphing those without progression):
TTP low-grade glioma treated TTP (initial) ~ 103 – 7.4Y - - - - - - - - - - - - (20)
TTP low-grade glioma treated TTP (late) ~ 125 – 16Y - - - - - - - - - - - - - (21)
The untreated group shows a corresponding, but mirror-image scallop, with progression registered early. There is a reasonably steady fall until about 2 year. Thereafter, the graph straightens until about 7 year :
TTP low-grade glioma untreated TTP (initial) ~ 100 – 17Y- - - - - - - - - - - - (22)
TTP low-grade glioma untreated TTP (late) ~ 90 – 9.6Y - - - - - - - - - - - - - -(23)
TTP low-grade glioma overall TTP (overall) ~ 103 ~ 12.9Y - - - - - - - - - (24)
The two graphs effectively unite at ~5.5 year. Discussion shall follow later.
The paper finishes with two memorable paragraphs of sanctimonious point-scoring, - “The significant difference demonstrated by this randomized study supports the contention of those who believe in the efficacy of early post-operative RT. Moreover, one should imagine the mental agony, as well as the physical sufferings, of the patients in the non-irradiated control arm after clinical deterioration and neuro-imaging when they are told about the recurrence of their brain tumour. We emphasize that long term follow-up of the patients in this study continues . . . . In conclusion, the significantly longer time to progression of the patients in the early irradiated group treated with conventional techniques such as were used in this study indicates that, at present, routine postoperative and early RT may be advisable for adult patients with cerebral LGG [low-grade glioma: emphasis added].” Could the research group have been biased ? Remember that this is a preliminary report (see later).
· Shaw et al. (2002). This was presented as a prospective, randomized trial of low-grade glioma treated with low- versus high-dose radiation with patients accrued between 1986 and 1994. As such, there was no control negative group. Entry was based upon a known histological diagnosis, with randomizing within 3 months of biopsy. The criteria for accrual were not provided. Proof (diagnostic histology ?) of the diagnosis was required within 3 months of study entry (randomization) which, presumably, was the starting date of the survival intervals (as in the graph legends). Randomization considered stratification factors. In that all accrued patients carried an intention to treat, randomization of those who chose to agree to the trial was probably reasonable. The Kaplan-Meier overall survival graph for all low-grade gliomas (their Figure 1), made up of two graphs – the high (64.8 Gy) and low (dose (50.4 Gy) radiation which, when combined, showed a ~straight line “best fir” between the start (through 100% survival) and ~6 year (~60% survival), the extrapolation passing through the time axis at ~17.1 year. The 5year survival ~70%. The lower dose radiation group showed a shoulder, which lasted until after ~2+ year; the patients who received 64.8 Gy survival less well in the first ~2+ years, with no real difference at 5 and 10 year. This mirror-image effect was noticed in both the survival graphs and the time to progression graphs, where it was more of a feature, with the lower dose radiation associated with the upper (better) outcome :
Survival, overall glioma S ~ 100 – 6.1Y - - - - - - - - - - - - - - - - - - - - (25)
TTP, overall gliomas, low dos TTP ~ 103 – 9.2Y - - - - - - - - - - - - - - - - - - (26)
The time to tumour progression was presented, pooled for all tumour types (their Figure 2). There seemed divergence during the first year, as with the survival graphs ( 64.8 Gy group survival falling faster over ~2 year), withclosing of the difference by about 3.5 – 4 year. The “best fit” for both groups over the interval 6 month – 18 month points showed extrapolation to slightly above 100%. The groups became close ~2.5 year, overall forming a spindle shape, apparently above and below an imaginary “pooled” group line, with the shoulder provided by the low dose less marked (not assessed here) – the initial fall in the high dose being more conspicuous and lasting about 2 year:
TTP, overall gliomas, high dose TTP (initial) ~ 101 – 15.8Y - - - - - - - - - - (27)
TTP, overall gliomas, high dose TTP (middle) ~ 94 – 8.5Y - - - - - - - - - - - (28)
This could imply that the adverse factor(s) affecting the higher dose group was(were) poorly represented in the lower dose group – some factor in selection and stratification. If radiation therapy were to have a major impact on these tumours, the expectation would be that the higher does would be more beneficial. This was not the case, raising doubts about the value of any radiation therapy. The differences in the 5 year survivals between the various irradiation treatment studies may be noted. The “best fit” survivals for the Astrocytoma + Astrocytoma-dominant mixed tumours for those aged ≥ 40 year old from 100% until ~60% at ~5 year (their Figure 3C) is given by :
Astrocytoma ≥40 year S ~ 102 – 8.2Y - - - - - - - - - - - - - - - - - - - - - (29)
Perhaps the most noteworthy point that comes from this study, is that the expected shoulder of improvement in the time to progression graph for those having radiotherapy, seen in the paper by Karim et al. (2002), is not clearly apparent: the differences in dosages administered by Karim et al. (54 Gy) was only mildly higher than the lower dose (50.4) administered by Shaw et al. (2002), and the higher dose seemed to make the patients’ outlook worse ! Something seems wrong ! – and the slopes between comparable graphs show an appreciable difference :
Karim, TTP low-grade glioma overall TTP (overall) ~ 103 – 12.9Y - - - - - - (30)
Shaw, TTP, overall gliomas, low-dose TTP ~ 103 – 9.2Y- - - - - - - - - - - - - (31)
The slopes extrapolate to the time axes at ~8 year and ~11.2 year respectively.
· Van den Bent et al. (2005). This study presents the long term results of the study already presented by Karim et al. (2002). The graphs show some interesting features :
For overall survival, the treated (early radiotherapy) and untreated (no early radiotherapy) graphs start essentially the same until about 1.5 year :
Survival treated/untreated initial S ~ 100 – 5Y - - - - - - - - - - - - - - - - - - (32)
The untreated graph then shows a step down :
Untreated, middle S ~ 98 – 6.6Y - - - - - - - - - - - - - - - - - - (33)
The treated graph also shows a step/change at about 1.8 year :
Treated, middle S ~ 102 – 6.8Y - - - - - - - - - - - - - - - - - (34)
The authors considered that there was no significant difference in the survivals. Looking at the graphs, there seems a surprising similarity in the minor variations in the two arms in the middle section. The major difference in the middle section is caused by the step in the untreated arm. Why that should occur is, of course, unknown, and may possibly be due to the chance clustering of some deaths. If it can be regarded as a chance event, an outlier group, then removal of the step would create a graph which would cross over the graph line for the treated group, with a resulting better survival (speculation).
The time to progression graphs (expressed as proportion progression free %), from which the study derives its major conclusion, shows interesting features. The untreated (no early radiotherapy) graph shows a relatively steep, straight line fall over the first 8 month :
Untreated, initial TTP ~ 103 – 30Y - - - - - - - - - - - - - - - - - - - (35)
The graph then follows a new slope over the interval 1-2 year :
Untreated, 1-2 year TTP ~ 90 – 13Y - - - - - - - - - - - - - - - - - - - - (36)
Thereafter, the graph becomes more irregular, with an overall form :
Untreated, 2-5 year TTP ~ 82 – 9.4Y - - - - - - - - - - - - - - - - - - - - (37)
The treated group shows an almost straight line through 100% until about 1.5 year, when a step commences :
Treated, initial TTP ~ 100 – 7.75Y - - - - - - - - - - - - - - - - - - (38)
At ~1.5 year, there is a step down over ~ 4 month, with the subsequent graph close to straight until ~5 year :
Treated, middle 2-5 year TTP ~ 97 – 7.76Y - - - - - - - - - - - - - - - - - - - (39)
There is no clear, specific feature in the graph of the treated patients to be considered consistent with the effects of radiotherapy. Later the arms seem to show a slow inclination to approach each-other (which will not be measured nor discussed here). The features may be depicted in an illustration :
Diagram to Illustrate features in the Time to Progression assessment
Progression-free proportion % (TTP)
phase fast Step
Middle Early Radiotherapy
Progression-free advantage claimed
~1 ~2 Year
So, dear reader, what can we make of all this ?
a. The apparent step in the treated group is unexplained. It may not be a real step. In context, it is not very important and, apart from mentioning it, it will not be considered further here.
b. Overall, the treated arm shows the type of graph we are used to – perhaps a slight shoulder over about 1.5 year that may be attributed to radiotherapy, but more likely to the selection/exclusion criteria applied prior to radiotherapy (see earlier), then there is a slightly steeper decline, with relatively straight lines.
c. The arm for those with early radiotherapy (untreated at this stage) shows a reasonably straight and rapid decline, then gradually showing a less rapid decline to approximate the slope of the treated group at an equivalent stage (slopes; treated ~9.4, untreated ~7.76).
d. These slopes in the middle section, as discussed previously, probably reflect the intrinsic growth rate of the tumours in the respective groups, showing a slight difference, with the radiotherapy arm slightly favoured.
e. The major, and statistically significant difference in the arms, occurs in the first eight month interval, when the untreated group shows a relatively steep fall. This has not been explained. Why should a group of patients (~18%), who bypass the radiotherapy department, show an apparently accelerated rate of reduction in progression-free time over about eight months after which the accelerated rate then dissipates ? It does not make sense ! Something has to be wrong !
f. Why is this difference in progression-free time not reflected in the survival graphs over the first 8 months ?
g. A shortening of the progression-free time in the early months after treatment was not shown by the lower-dose radiation group in Shaw et al. (2002)’s study.
You, dear reader, may be able to exercise your mind to fathom the flaw(s) in the trial protocol that must be there. Here are two possibilities :
a. Some patients, in association with their physicians (who could have had qualms about the delayed therapy), may have adopted a lowered threshold for assessing recurrence, so that these patients may receive the perceived “benefits” of radiotherapy early and not be “abandoned” and “denied” therapy; and/or
b. Patients come to medical attention as a result of symptoms (eg fits, headache, function loss etc.) and ultimately are investigated. This will involve (after ~1975) a CT scan ± an MRI scan (Scan 1). Surgery, biopsy ± debulking/decompression follow speedily, with subsequent selection and admission to the trial. Those admitted (unknown percentage) and allocated to the radiotherapy (RT) arm, will have another CT (Scan 2RT) in order to plan and direct the radiotherapy treatment (ie it is not a new symptom-based procedure). The untreated arm (UT) is less likely to have a repeat CT scan at this stage. When a symptom change stimulates an investigation into possible progression, the time lapse for CT comparison between the untreated (Scan 1 to Scan 2) may be greater than for the radiotherapy arm (Scan 2RT to Scan 3RT), allowing a greater observable change to fulfil the criteria for progressive disease. This would then be expected to speed measured progression changes in the untreated group, at least until they have had the second follow-up CT, as may be done routinely and without new symptoms. This suggestion may be examined in relation to the stated protocol : “At baseline, the size of the tumour and crossing of the tumour over the midline were assessed on CT scan (Scan 1) by the local investigator . . . . Clinical and radiological follow-up was done at the same intervals for both groups. In the first 2 years after randomization, clinical follow-up and contrast enhanced CT scans were done every 4 months, and thereafter once a year until tumour progression.” From Karim et al. (2002) “The local physicians were responsible for reporting the clinical and radiologic follow-up to the data centres using the forms provided . . . . The progression of the tumour was defined in the protocol as clinical-neurologic deterioration confirmed by definite evidence of tumour activity clinically and on CT scan (e.g. contrast enhancement an or markedly enhanced hypodense areas).” “Tumour activity” may be difficult to distinguish from postoperative inflammatory changes, which the radiologists in the radiotherapy arm would be inclined to ignore. Objective measurements do not seem to be parameters. There are weak undefined points in the protocol which may allow the premature assessment of tumour progression to be registered or the mild progression in the treated arm to be disregarded initially.
Conclusion: The upshot of all this, is that, from analysis of the graphs, radiotherapy offers questionable benefits (if any), and may be harmful for low grade gliomas (chiefly astrocytomas), and that the claim that the progression-free interval is significantly lengthened (improved) and proved is flawed and should be disregarded.
· Knisely (2005) was not impressed with the study by van den Bent et al. (2005), claiming 3 main points for criticism :
a. The review of the histologies resulted in 74% being low-grade glioma, and 26% high grade (eg astrocytoma III/IV/GBM), so that the study had limited relevance in clinical practice. (The reader may recall the study by Blankenberg et al. (1995 see earlier), in which the tumour volume doubling time differed little between Grade II tumours and Grade III tumours, making the fine distinction less significant.
b. The survival of those with delayed radiotherapy (which occurred in some patients who had progressive disease, and not dealt-with in the graphs presented to date) was 3.4 year, whereas the survival of those who had early radiotherapy (the patients regarded as treated with the RT in the graphs hereinabove) was 1 year. Accordingly, close follow-up and radiotherapy for those with progression seemed appropriate.
c. The data should be reviewed and represented with separate low-grade and high grade glioma cohorts and, in future, the opportunity for a second neurosurgical review with consideration for another attempt at gross total removal.
· van den Bent (2005) responded. He claimed that the patients with discrepant histology were part of a “sensitivity analysis.” (Do you know what that means ?) “Findings from sensitivity analysis were much the same as those from intention-to-treat analysis.” He notes that discrepant grades from pathologists were likely because of imprecision in diagnostic definition. “In our trial, the curves for overall survival for the two treatment groups overlap during follow-up and seem much the same. We think that increased progression-free survival might lead to a longer period in better health, but this idea is speculative.” Hedging now – so much for a near-definitive trial.
Earlier, when dealing with the malignant astrocytomas/GBMs there seemed questionable evidence that radiotherapy was beneficial (if at all), as a monotherapy, so that the argument about the higher grades requiring early radiotherapy (in the event of a pathologic underestimate of the grade) is probably unnecessary – there seems to be little (if any) hard evidence that radiotherapy as a monotherapy, has any real beneficial role for any grades of the astrocytoma series (at least) and may be harmful for some.
However, the results of flawed clinical trials are now well entrenched in the medical literature, making new and better trials which include control negative patients almost impossible, because of the inevitable reaction from ethics committees. The position of Radiotherapists in this unsatisfactory situation is, however, assured; they can expect to pay their mortgages etc. They had their chances to conduct sound trials, but did not do so.
The failure of Radiation (Ionizing radiation) Therapy to cure gliomas
So far, there has been presented a selected summary of the history relating to ionizing radiation therapy for astrocytomas over the years, which some readers may have found relevant and possibly interesting and illuminating. Mire recently, there have been scientific developments in the understanding of these tumours, for which the reader may read the reviews by Sanai et al. (2005) and Huff et al. (2006). In particular, there is the growing emphasis upon the refractoriness of the tumour stem cells to many treatments, be they ionizing radiation or chemotherapy, with the possible exception of Interferon. To keep Interferon company, there is also the possibility that UHF+ (and even Lithium, in the right protocol) may affect tumour stem cells in a therapeutically beneficial way :
The possible effect of UHF+ on the tumour stem cells
In the Submission to the Microwave Review Inquiry 2004, there were presented graphs of blood tumour markers in response to the UHF+. The series was small and the conclusions guarded, but the possibility was suggested that the UHF+ seemed to slow tumour growth (after an initial dip in the tumour marker, probably due to effects on the Golgi system). The suggestion was made that the UHF+ slowed tumour growth which, in the current scientific climate, could be regarded as slowed tumour stem cell division, and which produced the outcome, which seemed beneficial; that, being a prolongation of life, in particular (the Committee’s Review recorded some “complete remissions” and “partial remissions”). Huff et al. point out that studies which do not consider the tumour stem cell involvement in responses may miss the importance of slow changes. Such may have been the case in the Committee’s Review, because there was the traditional evaluation of CR, PR, SD and PD over three months, which came from the radiotherapy and chemotherapy literature trial criteria. The concept that the tumour stem cells play pivotal roles was recognized in the literature prior to the Committee tabling its Review in September 2005. There is no evidence that such new concepts were appreciated by the Committee.
The Case Reports – Individual Patient Data.
In this presentation there was an intermission and digression into the shabby world of clinical trials by Radiotherapists. If we return to the patient with the Astrocytoma grade2, now better informed about the condition, you may recall, she had survival without progression or any detectable MRI change up until the time of the deadline for Submission to the Review, some 4+ years. This would seem slightly better than the median to progression for most series, and could be regarded as noteworthy – clinical plausibility. We now know that her stable state persisted until the last MRI scan at the end of 2006 – some 6+ years – now more noteworthy and, as a patient possibly benefiting from the UHF+, clinically plausible.
Such single case reports do not prove the case for the UHF+, but they give a degree of comfort, in knowing that gratifying results apparently can be obtained from a patient pool that was not very large, over the ~3 years of the clinic in Fairfield. (She was the only patient treated that had an Astrocytoma Grade less than Grade IV/GBM.)
These were outlined in the Submission to the Review, but there seems to be no evidence that they were considered or, indeed, if anything written in the Submission other than the 5 case reports was read. In addressing the question as to whether the UHF+ had any effect, other than heating the skin, possibly the strongest evidence came from the development of malignant hypercalcaemia in the second week of treatment of two patients with non-Hodgkin’s lymphoma (large cell). This was dramatic and required hospitalization and pamidronate infusions for both patients. The appearance of such a clear-cut condition in patients, apparently previously asymptomatic for hypercalcaemia, under these duplicated circumstances must be very unlikely to occur by chance. These cases were received without comment.
The serum tumour markers showed interesting and seemingly repeatable changes in a proportion of patients. Whilst tumour marker changes are not accepted as adequate criteria to assess CR or PR, as a study tool, their changes give an indication as to what may be happening at a cellular level: there seemed an initial fall for two weeks, then a rise, which seems to have slope less steep than that before the UHF+, indicative of the slowed growth of the tumour stem cells.. The fall may be due to damage or interference with the Golgi secretory processes, the rise then reflects the functional/differentiated tumour bulk; the intrinsic growth rate, which may be growing more slowly because the tumour stem cells are not supplying cells and paracrine stimulation to the more differentiated cells. Again, despite interesting the tumour stem cell literature by 2005, the Review Committee was unable to bring the serum marker changes into the emerging tumour stem cell picture.
The serum and urine calcium and uric acid changes were of minor interest which by themselves mean little, but would seem worthy of follow-up, particularly since they are so cheap and easy to perform.
Radiotherapists and Holt
Holt was a surgeon, gynaecologist and radiotherapist. He was asked to assess the Tronado (original 424 MHz) machine as a means of heating tumours and augmenting the effects of ionizing radiation. He states that whilst in Germany, he noticed an absorption effect, which he then went on to study with his practice partner Alan Nelson. One would assume that a technique such as this, which might increase the sensitivity to ionizing radiation, would be viewed with interest and support by the radiotherapists of Australia. But, it would seem, that was not the case: there seemed hostility from the start, with the first NH&MRC Inquiry stacked with radiotherapists or those representing them (e.g. Prof. R Wright). The Inquiry was not looking at Holt’s professional conduct, but at the scientific issues involved with his claims. As we know, the scientific advice was scanty and of low level: most of the Committee were past or current clinicians with little or no specialized training in the relevant scientific areas. The assumption seemed to be that, because UHF is a radiation, therefore radiotherapists would know all about it. The outcome was adverse and, seemingly, Holt and Nelson were on the outer. According to Holt, in verbal and written accounts, he was barred from using a major hospital’s equipment and was effectively thrown out of his practice, with records lost or destroyed. His account may seem exaggerated, but it is supported, in part at least, by David Spall, whose Affidavit may be read on this website.
Early events could have shaped this course :
Ø Holt received public money to support the purchase of a Tronado machine in his private practice – politically risky, with professional jealousies aroused.
Ø His hypothesis had components that lacked scientific plausibility.
Ø Believing that he was onto a good thing, he may have exaggerated his claims, and that is certainly what his detractors believe, although some of their criticisms may involve misquotes or misunderstandings.
Ø Attempts to impress the non-believers may have also given rise to an exaggerated sales pitch.
Ø Having been through the first NH&MRC Inquiry in 1974-5 and its prolonged aftermath, he seemed to have developed an excessively suspicious reaction to the repeated overtures from radiotherapists to reveal his records, believing such appears to be another witch-hunt (which the second NH&MRC Review turned out to be).
Why then, would Radiotherapists, who would seem to have potential gains in terms of increased therapeutic control and knowledge, be so keen to disparage Holt’s claims ? Could it be that they have been concealing a dark, deep and nasty secret, that ionizing radiation, as a monotherapy for the astrocytoma series at least, contributes very little, if no significant benefit for patients ? Could this be one of the world’s great medical conns ? Should the NH&MRC and/or the Health Minister examine whether such therapy should be supported by Medicare funding ? Is there “value for money ?”
The analysis of the astrocytoma series, as presented above, whilst not exhaustive, would seem to show that, for the last fifty or so years, there have been studies published in peer review journals that have been fundamentally flawed. Yet some authors (e.g. Boyages & Tiver) were most enthusiastic about the undoubted benefits of ionizing radiation, and chastised those clinicians who could not see what they considered to be the obvious truth. Radiotherapists of the world have, for the most part, been unable to construct a valid trial with a negative control group to show, without doubts, that ionizing radiation has any significant beneficial effect upon the astrocytoma series. Yet the Radiotherapists have been most vocal in criticizing Holt’s claims and work, claiming that he had not conducted a prospective double blind, controlled trial - no-one seems to have bothered to advise how such a trial could be funded, established or carried through. (There is an old saying about the status of those who may feel themselves fit to throw the first stone.)
Radiotherapists, once again, were prominent in the second NH&MRC Review, despite their lack of relevant scientific training or experience; armchair theorists like Professor Lester Peters have been at the forefront of the detractor squad. Ironically, his special interest and reputation has been with irradiation of head and neck cancer, a category that seemed responsive to the sensitizing effects of the UHF, according to the Committee’s Review results ! He could have applied this potential sensitizing benefit to his patients !
If the UHF works along the lines that have been proposed – affecting the tumour stem cells and blocking the tumour stem cells’ checkpoints, a failure to follow this up with careful studies could see important potential gains, scientific development and discovery opportunities missed.
Future scientific work on the UHF topic (if ever) should not be dominated by Radiotherapists or clinicians without the relevant scientific background.
All the discussion ignores the distinct possibility that criminal elements and other pressure groups may have been involved in influencing the second NH&MRC Review. That is because these elements seem to have been involved in all other areas relating to my activity, at least. See what I wrote elsewhere about bias.
Copyright © MA Traill February 2007-02-20
Press “Back” to Return.
 Davis FG et al., Cancer, 1999; 85:485-491
 Mahaley MS Mettlin C et al., J Neurosurg 1989; 7126-36
 Boyages J & Tiver KW, Radiotherapy and Oncology 1987; 8:209-216
 Daumas-Duport C Scheithauer B et al. Cancer 1988; 62:2152-2165.
 Hoffer LJ. CMAJ 2003; 168(2):180-182
 Kleihues P et al. in “WHO Classification of Tumors.” Eds. P Kleihues & WK Cavenee, 2000. pp22-26.
 Karim AB Afra D et al. Int. J. Radiat. Oncol. Biol. Phys., 2002; 52:316-324.
 Papagikos MA Shaw E et al., Lancet Oncol. 2005; 6:240-244.
 Andersen AP. Acta Radiologica Oncology 1978; 17:475-484.
 Leibel SA Sherline G et al., Cancer 1975; 35:1551-1557.
 Walker MD, Alexander E et al. J. Neurosurg. 1976; 44:655-667.
 Fazekas JT. Int. J. radiation oncology Biol. Phys., 1977; 2:661-666
 Sherline GE. Cancer 1977; 39:873-881
 Walker MD Alexander E. J. Neurosurg. 1978; 49:333-343
 Patt HM & Quastler H. Physiological Reviews 1963; 43:357-394
 Tubiana M., Brit. J. Radiol., 1971; 44:325-347
 Lamerton LF., Brit. J. Radiol., 1972; 45:161-170
 Maj JG Paris F et al. Cancer Research, 2003; 63:4338-4341
 Blankenberg FG, Teplitz RL et al., Amer. J. Neuroradiol., 1995; 16:1001-1012
 Hasford J, Pfirrmann M et al.,; Haematologica 2005; 90:335-340
 Walker MD, Strike TA et al. Int. J. Radiation Biol. Phys. 1979; 5(10):1725-1731
 Walker MD, Strike TA. Proc. Amer. Assoc. Cancer Research. 1976; 17: 652.
 Kristiansen K, Hagen S et al., Cancer 1981; 47:649-652
 Wara WM. Cancer 1985; 55:2291-5
 Piepmeier JM. J. Neurosurg., 1987; 67:177-181
 Shaw EG Daumas-Duport C et al. 1989; 70:853-61
 Shaw EG Scheithauer B et al., Int. J. Radiation Biol. Phys. 1989; 16:663-8.
 Whitton AC Bloom HJ. Int. J. radiation Oncology, Biol. Phys., 1990; 18(4):783-86
 Vertosick FT Selker RG et al. Neurosurgery 1991; 28(4):496-501
 McCormack BM Miller DC et al. Neurosurgery 1992; 31(4):836-42
 Janny P Cure H et al. Cancer 1994; 73:1937-45
 Karim AB Maat B et al. Int. J. Radiation Biol. Phys., 1996; 36(3):549-56
Peraud A Ansari H et al., Acta Neurochirurgica (Wein) 1998; 140:1298-1222.
 Shaw EG Arusell AB et al. J. Clin Oncology 2002; 9:2267-76
 Van den Bent MJ Afra D et al. Lancer 2005; 336:985-90
 Knisely J The Lancet Oncology 2005; 6:921
 Van den Bent MJ. The Lancet Oncology 2005; 6: 922
 Sanai N Alvarez-Buylla A et al. New Engl. J. Medicine 2005; 353(8):811-22
 Huff C Matsui W et al. Blood 2006; 107(2):431-3
 Therasse P Arbuck SG et al. J. Natl. Cancer Inst. 2000; 92:205-16