Conclusions of NCRP Report No. 136:

NCRP Report No. 136 (NCRP 2001) written by NCRP Scientific Committee 1-6 (SC 1-6) concludes with a very weak affirmation of the linear-nonthreshold (LNT) model of radiation risk based on a biased review of the literature (Page 211):

"In conclusion, although the evidence for linearity is stronger with high-LET radiation than with low-LET radiation, the weight of evidence, both experimental and theoretical, suggests that the dose-response relationships for many of the biological alterations that are likely precursors to cancer are compatible with linear-nonthreshold functions. The epidemiological evidence, likewise, while necessarily limited to higher doses suggests that the dose response relationships for some, but not all, types of cancer may not depart significantly from linear-nonthreshold functions. The existing data do not exclude other dose-response relationships. Further efforts to clarify the relevant dose-response relationships in the low-dose domain are strongly warranted." (Emphasis added to words that indicate equivocation.)

In other words, they aren't sure. NCRP Report No. 136 was intended, as stated in the first sentence of the Executive Summary, to determine our "current understanding of the health effects of low doses of ionizing radiation." SC 1-6 violated this charge by discarding evidence from two important low dose rate epidemiological studies. On the other hand, it includes fabricated epidemiological data that support the LNT view point.

The words in the conclusion dealing with epidemiological evidence should have been written as "The epidemiological evidence at higher doses suggests that the dose response relationships for some, but not all, types of cancer may not depart significantly from linear-nonthreshold functions. Epidemiological data at or near background dose rates that show health benefits were not considered in reaching this conclusion."

The Report (p.197) explains the importance of epidemiological studies: "Epidemiologic data have high validity in that they represent the average risk [or benefit] in human populations resulting from an integration of all biological processes that are part of the radiation carcinogenic pathway, and no extrapolation from other species or biological systems is required. On the other hand, epidemiological data have limited precision because of an unfavorable signal to noise ratio at low dose,"

I would have written the last sentence as: "On the other hand, epidemiological data at low dose rates tend to show health benefits." The "unfavorable signal to noise ratio at low dose" relates to the difficulty of finding data to support the LNT hypothesis at low dose rates. Much good epidemiological data support the alternative hypothesis that low dose rate radiation stimulates the immune system.

The two epidemiological studies omitted from Report No. 136 have a good signal to noise ratio for decreased deaths from cancer and decreased deaths from all causes at low dose rates. Many studies show such effects. See the summary of studies in the scientific literature by Radiation, Science, and Health. (RSH 1998, 1999) These studies were provided to NCRP for use in its deliberations, along with those by many other scientists.

The report neglects the nuclear shipyard worker study (NSWS).

The report summarizes the nuclear shipyard worker study (NSWS)

on p. 196: "A large study by Matanoski (1991, 1993) of United States nuclear shipyard workers purportedly showed hormesis [reference omitted] in that there was lower total mortality in the exposed groups (both the <5mSv and >5 mSv groups) than in the nonradiation workers." (Emphasis added.) The inappropriate phrase "purportedly showed hormesis" shows the bias of SC 1-6 . The NSWS final report shows remarkable health benefits to the nuclear workers compared to their age-matched and job-matched controls.

After criticizing the NSWS report by innuendo without presenting any NSWS data, the authors of the report conclude "This interpretation [i.e., "purportedly shows hormesis"] ignores the likelihood of occupational selection factors that led some to qualify for radiation work while others did not. The fact that there was a difference for total mortality and not just for radiosensitive cancers, supports the interpretation that selection factors were operative." Since cancer induction is presumably a stochastic process, "occupational selection factors" could not explain the 4 std. dev. lower cancer mortality of the nuclear workers. The NSWS data strongly support the hypothesis that radiation stimulates the immune system. Unfortunately the NSWS are not readily available to the average reader. I published a brief summary of the data shortly after the final report was submitted in 1991. (Cameron 1992) and the Principle investigator published an abstract. (Matanoski 1993) The study has yet to be submitted to a peer reviewed journal more than 13 years after the study was completed.

The report should have presented data from the British radiologists study (Smith and Doll 1981) at this point. This study is referenced but not discussed. It showed beneficial health results similar to the NSWS. The undocumented charge of selection bias would have been seen as inappropriate. There is a better reason not to believe the charge of selection bias. The NSWS had excellent peer review including two members of SC 1-6.

All epidemiological studies have defects. The U.S. Dept. of Energy (DOE), which spent $10 million on the NSWS, went to great efforts to reduce defects. The DOE contract provided for peer review of the NSWS twice a year by a panel of eight scientists with expertise relevant to the research. Appendix 2 of the final NSWS report states: "The Technical Advisory Panel (TAP) was formed in 1980 as a standing committee of experts who would provide objective advice to the project staff on a continuing basis. In selecting its members, it was important for each [TAP member] to have had personal research experience with some of the problems related to the Shipyard Study. Disciplines which we believed to be important and which were included in the group are: radiation biology and radiation physics, medicine, genetics, industrial hygiene, epidemiology and biostatistics." (emphasis added)

The scientists who served as members of TAP were Dr. Arthur Upton, (chair); Gilbert Beebe, John Cameron (the author of this article), Carter Dennison (who resigned in 1983), Merril Eisenbud, Philip Enterline, Philip Sartwell and Roy Shore. TAP met twice a year to review data, question the scientific staff and make suggestions. Early in 1988, TAP approved the draft of the final report. Drs. Upton and Shore were TAP members who later became SC 1-6 members.

For SC 1-6 to discard the NSWS data because of a serious defect, implies that TAP scientists (including two members of SC 1-6) ignored the most basic concern of any epidemiological study—selection bias. Drs. Upton and Shore had ample opportunity to raise questions about selection bias during the eight years of TAP meetings. Presumably they approved the final report of the NSWS early in 1988 in which it states (p. 393):

"The shipyard nuclear worker population represents a large number of individuals exposed to low documented [doses] of radiation. They receive this radiation almost exclusively from gamma rays due to the decay of cobalt-60. Within the [shipyard] population there are comparable groups of workers exposed to negligible or no radiation at their shipyard jobs but who engage in similar work. Therefore this is an ideal population in which to examine the risks of ionizing radiation in which confounding variables can be controlled. " (Emphasis added.)

If either Dr. Upton or Shore suspected the serious flaw of selection bias, they were obligated to bring the weakness to the attention of the NSWS scientists and also to be sure it was mentioned in the final report. Since they raised no questions of such flaws during eight years of the NSWS study nor in the final report, the undocumented charge of selection bias is not credible.

While it is reasonable to expect some selection bias, it seems a great stretch of the scientific imagination that such a bias could explain the impressive health benefits to nuclear shipyard workers. The 28,000 nuclear shipyard workers with the highest cumulative doses had cancer death rates significantly lower (P<0.001) than the 32,500 age-matched and job-matched unexposed shipyard workers. Since the LNT hypothesis assumes that cancer induction is a stochastic event, they should explain how selection bials could lead a significant reduction of cancer deaths. A more remarkable health improvement of the nuclear workers was their 24% lower death rate from all causes. (16 std. dev., P<10^-16). That is, the nuclear workers had remarkably better health than the controls. The nuclear workers health apparently benefited from their occupational radiation. The SC 1-6 rejection of NSWS data seems based entirely on the fact that it strongly contradicts the LNT hypothesis.

British radiologists study.

NCRP Report No. 136 cites the British radiologist study (Smith and Doll 1981) but neglects to mention that it contradicts the LNT hypothesis and strongly supports the health benefits shown in the NSWS. While Table 9.1 shows some of the data from the NSWS and the British radiologist study, it does not show the impressive health benefits to radiologists who entered the field after 1920. Their cancer death rate was 37% lower than that of all men in England and Wales (P<0.001) (compared to a cancer death rate 26% higher than the same group for early radiologists). Their death rate from all causes was 24% lower (P<0.001) than the same group compared to no significant difference for the early radiologists. That is, it appears that radiation stimulation of the immune system of the early radiologists balanced out their increased cancer deaths. It is unlikely that any selection bias could have produced the significant health benefits to the radiologists who entered the profession after 1920.

I do not claim that the health benefits of nuclear shipyard workers or the British radiologists who entered the field after 1920 prove that moderate dose rates of radiation are healthy, but the data are far more consistent with the hypothesis that radiation stimulates the immune system than they are with the LNT hypothesis. If data from these two studies had been provided in the report, it would have been difficult for NCRP Council members to agree to a conclusion which supports the LNT hypothesis.

Improper statistical procedures invalidate the Cardis, et al. study.

NCRP Report No. 136 discarded all of the data of the NSWS on the basis of an undocumented flaw in selecting nuclear workers. At the same time it quotes (p. 198) without criticism the widely accepted 1995 cancer mortality study among nearly 96,000 nuclear workers in three countries which purports to show a statistically significant increase

in leukemia mortality. (Cardis et al. 1995) Although the study omitted the 70,000 workers in the NSWS from their U.S. data, my criticism of the Cardis et al. study is based on improper procedures used to produce a statistically significant increase in leukemia.

Accepting only risk increments, the report states "As there was no reason to suspect that exposure to radiation would be associated with a decrease in risk of any specific type of cancer, one sided tests are presented throughout." It also states "There was no evidence of an association between radiation dose and mortality from all causes or from all cancers. Mortality from leukemia, excluding chronic lymphocytic leukemia (CLL) .... was significantly associated with cumulative radiation dose (one-sided P value=0.046; 119 deaths."

Note the following unusual data analysis that permitted the authors to calculate a P value less than 0.05. They divided the workers into 7 dose groups. Four of the dose groups with 86 of 119 deaths were excluded because there were fewer deaths than expected. Three of the dose groups with 33 of 119 deaths had more deaths than expected and were included. The trend analysis of P=0.046 for these three groups was obtained "using computer simulations based on 5000 samples, rather than the normal approximation." (Pollycove and Feinendegen 2001). That is, the claim of a statistically significant increase in leukemia mortality with cumulative radiation dose is based on invalid statistical procedures. NCRP Report No. 136 made no criticism of the Cardis et al. data.