Next | ToC | Prev


The Duration of Radiation's Carcinogenic Effect

This chapter is arranged in five parts:

  1. A-Bomb Study:   The Evidence from Combined Age-Bands, p.1
  2. A-Bomb Study:   The Evidence from Separate Age-Bands, p.3
  3. Evidence from Other Human Studies, p.5
  4. The A-Bomb Study versus the Ankylosing Spondylitis Study, p.7
  5. The Bottom Line, p.9
    Then tables.

Introduction :

          In Chapters 13 and 16, we emphasized that a most crucial determinant of realistic Lifetime Fatal Cancer-Yields is the duration of the radiation's effect in producing cancers, additional to those occurring spontaneously.

          In making our estimates of Lifetime Fatal Cancer-Yield in Chapter 16, we made the approximation that the K-values determined for the period 1950-1982 will continue to operate beyond 1982, for the remainder of the lifespan of the residual survivors. (K is defined as the fractional increase in the spontaneous cancer death-rate per centi-sievert of exposure. Multiplied by 100, K becomes the percent increase in the spontaneous rate, per cSv.)

          In this chapter, we shall examine the basis for our approximation about duration. First, we shall quantify and evaluate the trends within the A-Bomb Study itself. Then we shall consider some of the other data reported in the literature. Radiation studies having over forty years of follow-up beyond the time of radiation exposure are very rare indeed. Studies with insufficient quantitative information, and anecdotal reports, are not considered in our evaluation.

          There are two ways in which we shall analyse the A-bomb data. First, we shall compare the behavior of cancer-risk -- with the passage of time -- in all the exposed survivors combined, versus the unexposed survivors. Second, we shall make the same comparisons for each of the five age-bands separately. (As always, we will make the approximation that the Reference Group is an unexposed group.)

1.   A-Bomb :   The Evidence from Combined Age-Bands

          Table 17-A provides the analysis for all ages combined, both cities, both sexes, for the study of cancer-induction by radiation, over the period 1950-1982. In this table, the initial sample of 91,231 persons has been divided into two groups:   The 25,203 exposed survivors (Dose-Groups 3, 4, 5, 6, 7, and 8), and the 66,028 virtually unexposed survivors (Dose-Groups 1+2, our familiar reference group).

          During 8 four-year follow-ups, the 91,231 persons have accumulated 6,050 cancer-deaths (Column B plus Column C of Row 9). Rows 1 through 8 show when the 6,050 cancer-deaths occurred. This detailed information comes from the diskette R10ALL (R-Ten-All) provided by RERF;   see Chapter 7. These are the "raw" numbers. There is no need for age- normalization here, for we will look at the age- bands separately in Part 2.

          Within any specified time-period after the bombing, some new cancer-deaths occur in each of the two groups (exposed and reference). In Columns D and E, the cancer death-rates per 10,000 initial persons apply to the rates of new cases which occurred during a particular period (Note 4, Table 17-A). This chapter is the only one in which we are not dealing with the aggregate observations for all eight follow-up periods combined.

          The ratio of the cancer death-rates (rate in exposed / rate in reference), or R2/R1 -- during a specified set of years -- is provided in Column F.

          Because the difference in mean dose received by the two groups, from the bombings, was fixed forever in August 1945, the dose-difference is the same in all successive follow-ups. Therefore, we can compare the ratios in Column F directly, as indicators of how radiation's carcinogenic effect behaves over time.

          In Column G, we also provide the indicator commonly called "excess relative risk." The quantity (R2/R1)-1 is equivalent in this case to (O-E) / E, or (O/E)-1, because R2 designates cancer-rate in the exposed group, or the "Observed" (O) value, while in this case R1 designates the cancer-rate in the virtually unexposed group, or the spontaneous "Expected" (E) value.

          It should be noted that the quantity (R2/R1)-1 is also the numerator of the equation from which we derive K-values in Chapter 15, namely Equation (7). It is an excellent indicator with which to compare the carcinogenic potency of radiation, during one set of years versus another set of years.

Findings from Our Analysis :

          With respect to predicting the duration of the radiation effect beyond 1982, Table 17-A incorporates an important confounding variable. We shall discuss it after examining what the table does show, and what a similar analysis by RERF analysts shows.

          Readers who inspect Table 17-A, Column F, Rows 1-14, can see for themselves that a consistent picture emerges from the successive follow-ups, in spite of fluctuations.

          1.   In Rows 1 through 8, the Cancer-Rate Ratios and the excess relative risks are in a general rising trend, with some recent entries notably higher than the overall value (Row 9) for the 1950-1982 follow-up as a unit. For specific four-year periods, there are some deviations which are almost surely the result of sampling variation.

          2.   The rising trend is confirmed when we divide the entire 32 years of observation into two equal intervals of 16 years (Rows 10 and 11). The Cancer-Rate Ratio rises from 1.104 to 1.263. The term, (R2/R1)-1, rises from 0.104 to 0.263. This 2.53-fold increase describes the increase in effective K-value over the two halves of the follow-up, or the increase in excess relative risk.

          3.   If we divide the observations at 25 years post-bombing (Rows 13 and 14), the excess relative risk beyond 25 years does not vanish, fall, or even level out. On the contrary, it rises by the factor of (0.266 / 0.139), or 1.91. This is a striking contrast with reports (which we will examine in Part 3 of this chapter) that the radiation effect is finished after 25 years, in the Ankylosing Spondylitis Study.

          4.   The most recent four-year follow-up period, 1979-1982, shows the highest Cancer-Rate Ratio of any interval of the first 32 years, 1.308, and the highest value of (R2/R1)-1, namely 0.308. However, the findings from any isolated four-year follow-up are unstable relative to values from eight-year, twelve-year, or sixteen-year periods. A four-year follow-up deserves noting, but not too much weight.

Findings by RERF Analysts :

          The findings in our Table 17-A are confirmed in a totally separate study by RERF analysts, Shimizu and co-workers (TR-5-88). Although they are looking at RERF's DS86 "sub-cohort" of 75,991 persons (not the full 91,231 sample), and although RERF has moved many persons into and out of the unexposed sample (see our Tables 10-A and 10-B), their results and ours should be in general agreement, and they are.

          Shimizu and co-workers present the following data, for excess relative risk as a function of follow-up time in the A-Bomb survivors (Shi88, Table 2-4). The data are for all cancers combined, leukemia omitted.

    Follow-Up               Excess Relative Risk
    Interval                per Gray
    ---------               --------------------
    1950-55                 0.2372
    1956-60                 0.1230
    1961-65                 0.2264
    1966-70                 0.2576
    1971-75                 0.2900
    1976-80                 0.3958
    1981-85                 0.4301
    1983-85                 0.4717

          Their results are in very close agreement with those in our Table 17-A. Shimizu and co-workers also show an "aberrant" point at 1956-60, and also show the highest excess relative risk occurring in the most recent follow-up -- which means through the 40th post-bombing year (1985) in their DS86 "sub-cohort."

          In both their study and ours, a strong trend exists over a 32-35 year period toward an increasing excess relative risk, and there is no sign of a downturn in the trend. The trend for the combined age-groups is still upward at 1985.

          Because both the Shimizu analysis and our Table 17-A examine all five age-bands combined, the results are exactly what we would expect to see if we assume that the individual K-value for each age-band remains constant through time. The expectation, of rising Cancer-Rate Ratio and rising excess relative risk, rests on two observations.

          (1)   Our study of K-values (Table 15-L) shows that the highest K-values occur for those who were young ATB. K-values decrease progressively (with negligible exceptions) for persons of successively older age ATB. And K-values are directly related to the term (R2/R1)-1.

          (2)   From the very beginning of the follow-up in 1950, some people in every age-band die during each follow-up period, but obviously, people die off fastest from the oldest age-band (50+ years of age ATB). In the oldest age-band, the mean age ATB in 1945 was 58.5 years.

          By late 1982, Table 4-B shows that only 4.8 % of the oldest age-band ATB was still alive, whereas 96 % of the youngest age-band ATB was still alive. The effect of this general selective mortality, according to age itself, means that the effective age ATB of the residual survivors keeps decreasing with each passing year of follow-up. (Note that age ATB is a very different number than age attained at some particular year of the follow-up.)

          With lower age ATB, we know that we find greater cancer-induction by radiation (higher K-values). And so the rising values in Columns F and G of Table 17-A are what we would expect, if K-values for each age-band do not change with time but the average age ATB of the residual survivors is constantly shifting downward between 1950-1982.

          The steady "replacement" of the study-sample by one with an ever younger effective age ATB -- and thus an ever rising effective K-value -- is the confounding variable which we mentioned at the beginning, with respect to the duration of the effect.

          Although both Table 17-A and the Shimizu analysis can tell us directly that the radiation effect on cancer-induction is certainly not finished at 25, 30, 35, or 40 years after the bombing, neither study can tell us whether the K-values for individual age-bands are showing any increase or decrease, with the passage of time, which might help predict the duration and magnitude of the radiation effect for the ages separately. Therefore, we will explore the issue further in Part 2.

2.   A-Bomb:   The Evidence from Separate Age-Bands

          The data and findings for the study's five separate age-bands ATB are in Table 17-B (cities and sexes combined).

The Small-Numbers Problem:

          As usual whenever analysts subdivide these data, we face the small-numbers problem to a serious degree.

          The problem is particularly bad when we isolate a single four-year time-period, such as the most recent one (1979 through 1982). Nonetheless, in order to make Tables 17-A and 17-B comparable, we do isolate the most recent follow-up period.

          The instability of small numbers, in any of the isolated four-year periods, needs to be recognized. Readers inspecting such entries, above the dotted line in each age-band, will see how their irregular progression results in irregular progression ("bounciness") of the Cancer-Rate Ratio. It would be preposterous and misleading to "interpret" each wiggle as anything other than sampling variation, in data which have been excessively subdivided. To look for believable trends in the current data, one must re-combine some intervals -- as shown.

The 3 Groups in the Middle of the Age-Range :

          In the middle of the age-range are three age-bands:   10-19, 20-34, and 35-49 years of age ATB. In all three of these age-bands, the Cancer- Rate Ratio in Column F is well above 1.0 in the follow-up period 1971-1982, which shows that the radiation effect was not finished by 25 years after the bombing. Also in all three, the Cancer-Rate Ratio is higher in the second 16-year period than in the first 16-year period.

"Relative" versus "Absolute" Risk-Models :

          It might be noted that, in the Reference Group of all three of these age-bands, the spontaneous cancer death-rate per 10,000 initial persons (Column E) increases with each four-year follow-up. In other words, the spontaneous frequency grows as age advances. (In the 35-49 year group, this trend ends in the 1979-1982 follow-up, when the youngest member of that cohort was about 70 years old.)

          Yet even while the spontaneous risk (R1) is tripling or more within these age-bands, the ratio R2/R1 (which is equivalent to Observed/Expected) is approximately constant or even growing, when one divides the observations into two equal intervals.

          These observations are consistent with the so-called "relative risk" model of radiation carcinogenesis, and inconsistent with the so-called "absolute risk" model.

Oldest and Youngest Age-Bands ATB :

          The remaining two-age bands are represented by those very young ATB (average age ATB = 4.1 years) and those in the oldest age-band ATB (average age ATB = 58.5 years).

The Oldest Age-Band :

          For the oldest group, Table 17-B shows that the radiation effect appears to have declined in the second half of the study (1.103 falling to 1.038). During the most recent follow-up, 1979-82, the ratio fell below 1.0. The ratio of 0.669 is based on rapidly dwindling numbers of cases. The total cancer-deaths (exposed plus reference) were 63 in the years 1979-1982, and by late 1982, Table 4-B shows that only 720 persons out of the original 15,011 persons were still alive (the youngest would be 87 years old).

          It just does not matter whether or not the ratio in this age-band ever climbs above 1.0 again, in terms of properly estimating a Lifetime Fatal Cancer-Yield. The K-value for this age-band, used in Table 16-B, is based on all the observations through 1982, and is valid for the 1950-1982 period. The presumption in Table 16-B, that the same K-value will persist beyond 1982, adds almost nothing to the Lifetime Fatal Cancer-Yields. Readers can see this for themselves in Table 16-B by comparing Columns G and H in the two rows (male and female) for the oldest age-band.

The Youngest Age-Band :

          What does matter a great deal, in estimating Lifetime Fatal Cancer-Yield, is what will happen beyond 1982 in the youngest age-band ATB (0-9 years).

          By late 1982, the oldest member of this cohort would be about 46 years old (9 years ATB + 37 years since the bombing). The youngest would be only 37. In Table 17-B, Column E, the low spontaneous cancer death-rates in this age-band's Reference Group confirm that the decades of high spontaneous cancer-mortality are still ahead for this age-band.

          What, if anything, can Table 17-B tell us about the likely behavior of the Cancer-Rate Ratio in this age-band, during the decades ahead?

          For the age-band 0-9 years ATB, the total number of cancers (exposed plus reference classes) in the first half of the follow-up period, 1950-66, was only six. There were five cancer-deaths in the Exposed Group, and one in the Reference Group. The problem of statistical instability for this first half of the whole follow-up is self-evident. Sampling variation in the early period is such that a comparison of the early versus late follow-up periods cannot be taken seriously.

          For the entire period beyond 1970 (the period more than 25 years post-irradiation), with 75 total cancer-deaths, the Cancer-Rate Ratio is 1.228, which suggests that the carcinogenic effect of radiation did not cease by 25 years after exposure.

          It can be noted that the highest Cancer-Rate Ratio of any follow-up period occurs in the most recent period (1979-82), where the ratio of 1.690 is based on a total of 33 cancer-deaths. Our statement knowingly ignores the ratio of 5.917, observed in the 1959-62 period, because it was based on a total of three cancer-deaths.

          Although in the 1979-82 period, the numbers are still small and unstable, the observed ratio of 1.690 is very different indeed from the three preceding periods. On the other hand, when the data are so very bouncy, no one should be surprised if the ratio is appreciably lower than 1.690 in the next follow-up period.

Summary on the Separate Age-Bands :

          Examination of the five age-bands, separately, leads us to two statements.

          (1)   With the one exception of the oldest age-band ATB, the carcinogenic effect did not cease 25 years after the exposure. In the two age-bands (20-34 ATB, 35-49 ATB) which are currently contributing most heavily to the cases, there was not even any detectable decline in the effect after 25 years.

          (2)   The evidence at hand provides no basis for assuming that the average K-values, which will characterize the post-1982 period, will be either higher or lower than the K-values which characterize the 1950-82 period as a whole. Within the data, the reasonable assumption is that average K-values will be about the same in the future as in the past -- approximately constant K-values.

3.   Evidence from other Human Studies

The Ankylosing Spondylitis Study

REFERENCE:   DARBY, 1987 (Dar87).

TITLE:   "Long Term Mortality after a Single Treatment Course with X-Rays in Patients Treated for Ankylosing Spondylitis."

          To begin, we will cite the summary of conclusions from the study of the ankylosing spondylitis patients (Dar87, p.179):

          "Mortality up to 1 January 1983 has been studied in 14,106 patients with ankylosing spondylitis given a single course of X-ray treatment during 1935-1954. For neoplasms other than leukemia or colon cancer, mortality was 28 % greater than that of members of the general population of England and Wales, and this increase is likely to have been a direct consequence of the treatment. The proportional increase reached a maximum of 71 % between 10.0 and 12.4 years after irradiation and then declined. There was only a 7 % increase in mortality from these tumours more than 25.0 years after irradiation and only for cancer of the esophagus was the relative risk significantly raised in this period. Neither the magnitude of the relative risk, nor its temporal pattern following treatment, were [sic] greatly influenced by the age of the patient at first treatment."

          Before continuing, we would like to correct the impression that this study is a study of 14,000 patients. It is really a study of half that number. While it is true that some 14,000 persons entered the study, the authors reveal that "By 1 January 1983 just over half the patients had been re-treated . . . " After 18 months following re-treatment, the re-treated patients were deleted from further follow-up study of cancer. Since the re-treatment constitutes a possible confounding variable, it is not clear why the authors include these patients at all in the results.

          Readers may wonder why Darby and co-workers exclude colon-cancer in the statement above. They report on colon-cancer separately (Dar87, p.179):   "For colon cancer, which is associated with spondylitis through a common association with ulcerative colitis, mortality was increased by 30 %." For this reason, they treat it separately. Among the ankylosing spondylitics, the mortality-rate from ulcerative colitis was 12.8 times the rate in the general population (Dar87, Table 10).

          (People with ulcerative colitis are said to have an elevated rate of fatal colon-cancer. It is my opinion that analysts and physicians need to consider the possibility that an elevated rate of fatal colon-cancer in such people may result, partly or even largely, from the many diagnostic X-ray exams invoked by the presence of ulcerative colitis.)

Duration of the Effect from Radiation :

          On the issue of duration of carcinogenic effect, from the X-ray exposure, Darby and co- workers state (Dar87, p.188):

          "This is the first large study to suggest an apparent end to the effects of exposure to radiation for neoplasms other than leukemia and the possibility must be considered that the findings are spurious . . . "

          We shall try to evaluate the possibility in this chapter, Parts 3 and 4.

Division of Results at 25 Years :

          From Table 5 of the Darby paper, we extract the summary values for observed and expected deaths from neoplasms, other than leukemia or colon cancer, occurring before age 85 years. Darby and co-workers obtained their expected values from population statistics for England and Wales. The data below are for all ages at first treatment.

     Time since First Treatment, in Years.
5 to 24.9 Years         | 25 Years or More
Observed Expected  O/E  | Observed Expected   O/E
   385    279.39   1.38 |   178     166.56    1.07

          We should note here that Darby and co-workers report that approximately 40% of the cancer-deaths in each of the time periods were from lung-cancer. We shall return to this point, in Part 4 of this chapter.

          Darby and co-workers state that the value 1.38, for O/E in the earlier period, is significantly different from the value 1.07, for O/E in the later period, and that neither in males nor females was the O/E value significantly elevated in the later period.

          The data summarized above are the data which lead to the suggestion that the radiation-induction of cancer is at "an apparent end" by the 25th year post-irradiation.

Contrast with the A-Bomb Study :

          With respect to the duration of radiation's carcinogenic effect, this report from the Ankylosing Spondylitis Study is clearly at variance with the A-Bomb Study.

Other Studies on the Duration Question

          In view of the apparent conflict between the A-Bomb Study and the Spondylitis Study, analysts have to consider which data are more reliable.

          One approach to the question is to examine other studies in the literature. While such studies will be much less comprehensive than the A-bomb or the spondylitis studies, some of them are capable of addressing the particular question at hand here:   Is the carcinogenic effect finished at 25 years, or is it appreciable beyond 25 years?

(1) ----------------------------------------------------
REFERENCE:   BOICE, 1985 (Boice85).
Irradiation:   X-ray Therapy for Cervical Cancer.

          Cancers Evaluated:
Cancers arising in sites (in or near the radiation field) which "likely" received over 100 rads, including stomach, small intestine, colon, rectum, gallbladder, pancreas, uterine corpus, ovary, other genital organs, kidney, bladder, bone, and connective tissue. O/E was measured for all these sites as a function of time after irradiation. The comparison was made between those cervical cancer cases exposed to radiation and those cervical cancer cases not exposed to radiation.

          Effect Beyond 25 Years:
Radiation-induction of cancer showed no diminution of risk even after 30 years of observation. The authors' exact statement:   "Apparently, after a period of about 10 years, the risk of radiation-induced solid tumors following exposure in adult life will persist, if not increase further, for at least 30 years and possibly throughout the remainder of life."

Consistent with A-Bomb Study.
Inconsistent with Spondylitis Study.

(2) ----------------------------------------------------
REFERENCE:   HILDRETH, 1983, 1985, 1989 (Hild83, Hild85, Hild89).
Irradiation:   Thymus Irradiation in Infancy.

          Cancers Evaluated:   Breast-Cancer.
This is a study of females whose breast irradiation occurred in early infancy. Excess cancer has been ob- served in adulthood, as predicted elsewhere (Go81, p.260).

          Effect beyond 25 Years:
The earliest effect was observed 28 years post- irradiation. In the exposed group, the median age at diagnosis is 39 years. The excess relative risk is 2.60, and has been essentially constant since the effect was first observed (Hild89, p.1283).

Consistent with A-Bomb Study.
Inconsistent with Spondylitis Study.

(3) ----------------------------------------------------
REFERENCE:   HOWE, 1984 (Howe84). Also MILLER, 1989 (Mi89).
Irradiation:   Fluoroscopic exposure of patients being treated for tuberculosis.

Cancers Evaluated:   Breast-Cancer.
          Effect Beyond 25 Years:

Howe reported (Howe84) that "All age at first exposure groups continue to show increased risk of breast cancer mortality up to 40 years after first exposure, the data beyond 40 years being too sparse for for meaningful interpretation." Miller and co-workers report (Mi89, Table 4) the temporal distribution of risk as follows:

     5-14 years post-irradiation :  1.47
    15-24 years post-irradiation :  1.40
    25-34 years post-irradiation :  1.48
    => 35 years post-irradiation :  1.24
Consistent with A-Bomb Study.
Inconsistent with Spondylitis Study.

(4) ----------------------------------------------------
REFERENCE:   MARTIN, 1970 (Mart70).
Irradiation:   X-ray therapy for benign skin disorders.

Cancers evaluated:   Skin-cancers in or adjacent to irradiated sites.
          Effect beyond 25 Years:

Definite persistence of effect well beyond 25 years post-irradiation. In 73 of total observed 368 cases, the skin-cancers appeared after latent periods of 31 to 50 years post-irradiation.

Consistent with A-Bomb Study.
Inconsistent with Spondylitis Study.

(5) ----------------------------------------------------
REFERENCE:   MODAN, 1989 (Modan89).
Irradiation:   X-ray Therapy for Tinea Capitis (Ringworm of the Scalp). Israel Series.

Cancers Evaluated:   Breast-Cancer.
          Effect Beyond 25 Years:

The exposure occurred between 1949 and 1959, so 1954 was the midpoint. The follow-up recently accomplished is 1982 through 1986, so 1984 was its midpoint. Thus, about 30 years have passed since exposure. During the 1982-1986 follow-up, a total of 22 cases of breast-cancer occurred (exposed plus unexposed) and the relative risk of breast-cancer for the exposed versus the controls was 2.11, for an average dose of 1.6 rads to the breast.

Consistent with A-Bomb Study.
Inconsistent with Spondylitis Study.

(6) ----------------------------------------------------
REFERENCE:   RON, 1988 (Ron88).
Irradiation:   X-ray Therapy for Tinea Capitis (Ringworm of the Scalp). Israel Series.

Cancers Evaluated:   "Head and Neck Tumors" -- a conglomerate of many separate sites.
          Effect Beyond 25 Years:

Cannot demonstrate excess of head and neck tumors beyond 25 years. Total cancers observed beyond 25 years = 7 cases, including both exposed and unexposed.

Consistent with Spondylitis Study.
Inconsistent with A-Bomb Study.

(7) ----------------------------------------------------
REFERENCE:   SHORE, 1984 (Sho84).
Irradiation:   Treatment of Tinea Capitis in Children.

Cancers Evaluated:   Skin.
          Effect Beyond 25 years:   O/E still rising at 35 years post-irradiation.

Consistent with A-Bomb Study.
Inconsistent with Spondylitis Study.

(8) ----------------------------------------------------
REFERENCE:   SHORE, 1985 (Sho85).
Irradiation:   Thymus Irradiation in Infancy.

Cancers Evaluated:   Malignant & Benign Thyroid Tumors.
          Effect Beyond 25 Years:

Thyroid carcinoma:   Excess cancer persists at least out to 40 years. Thyroid adenoma:   Excess benign tumors at least out to 45 years.

Consistent with A-Bomb Study.
Inconsistent with Spondylitis Study.

(9) ----------------------------------------------------
REFERENCE:   SHORE, 1986 (Sho86).
Irradiation:   X-ray Therapy for Post-Partum Mastitis.

Cancers Evaluated:   Breast-Cancer.
          Effect Beyond 25 Years:   Excess breast-cancer occurring beyond 30 years.

Consistent with A-Bomb Study.
Inconsistent with Spondylitis Study.

Conclusion, from the Survey Above

          With only one exception (entry 6 above), the evidence from these other studies, of irradiation for medical reasons, is consistent with the A-bomb evidence and not consistent with the spondylitis evidence. On the duration issue, our survey confirms that the warning from Darby and co-workers was certainly appropriate:   " . . . the possibility must be considered that the findings are spurious . . . " (Dar87, p.188).

4.   A-Bomb Study versus Spondylitis Study

          Both the A-Bomb Study and the Ankylosing Spondylitis Study are studies of reasonably large size, in terms of numbers of observations, and both are of reasonably long duration. The question is, "Which is the more reliable study, with respect to the duration of radiation's carcinogenic effect?"

          In my judgment, the credibility of the A-Bomb Study is clearly superior to that of the Spondylitis Study on this issue. Not only is the A-Bomb Study supported by, and the Spondylitis Study at variance with, almost all other human evidence, but there are at least two possibly serious confounding variables in the Spondylitis Series, and they raise real questions about studying temporal behavior of radiation in that series.

          In addition, confidence in any database is necessarily undermined severely, when its original input disappears from the on-going analysis. The dosimetry of the Ankylosing Spondylitis Study has been retroactively altered more than once -- in Beir72, again in Beir80, and now it is undergoing yet another retroactive alteration. According to Darby and co-workers (Dar87, p.181), for "many organs previously classed as lightly irradiated" (including liver, kidney, bladder, uterus), the new revised doses will be high enough to re-classify them as heavily irradiated. In a study with two classes of dose -- heavy and light -- it is no minor matter when the distinction is retroactively obliterated for many organs. Yet for this study, I am unaware of any effort to maintain continuity with a "constant-cohort, dual-dosimetry" approach -- an approach which we have demonstrated for the A-Bomb Study through its 1982 follow-up.

(1)   Health of Exposed versus Controls

          For the A-bomb survivors, the exposed and reference groups are matched except for the variable, radiation. And in Table 11-H, Column S, we have positive evidence that the non-cancer mortality rate is the same for exposed and reference groups. Thus, there is no force of mortality that might alter the temporal response of radiation-induced cancer.

          For the spondylitics, the exposed and reference groups are far from alike. The reference group is the general population. The exposed group is a patient-group suffering from a severe chronic malady associated with a 51 % excess non-neoplastic mortality-rate (Dar87, p179). And this excess mortality-rate covers a wide range of diseases, in addition to a small number previously expected to be associated with ankylosing spondylitis.

          Can anyone be confident that the presence of a serious increase in force of mortality (evidenced by the 51 % increase in non-malignancy death-rate) has no effect in altering the temporal course of radiation-induced cancer-mortality in the spondylitics? If the temporal course were shifted, then the use of national rates for the "expected" cancer-rates may be totally unwarranted.

          It would seem rational that any challenge, to the temporal course of radiation-induction of cancer observed in the A-bomb survivors, should not be made from a series of persons with a serious increase in mortality-rate from a wide assortment of non-malignant diseases.

(2)   Smoking among Exposed versus Controls

          In the spondylitic data, the case for a drastic decline in radiation-induced cancer, beyond the 25th year after exposure, rests very heavily on the lung-cancer rates for the early and late periods.

          In the early period, (5.0-24.9 years post- therapy), 155 out of a total of 385 cancer-deaths are lung-cancer deaths, with an O/E of 1.37. In the late period (25.0 years or more post-therapy), 69 out of a total of 178 cancer-deaths are lung-cancer deaths, with an O/E of 0.97. The data on lung-cancer, below, are from Table 4 of the 1987 Darby paper.

               Cancer of the Lung
Follow-Up Period  Observed  Expected    O/E

 5 - 24.9 Years      155      113.08    1.37
 25 or more Yrs       69       71.41    0.97

          Darby et al acknowledge that the exclusion of lung-cancer makes it impossible to prove any significant difference in O/E for the early period versus the late period. And of course, analysts should never be excluding data frivolously in any case. We simply point out that, for the reason which follows, there must necessarily be some serious reservations about the heavy dependence on lung-cancer, for their finding about duration. There may be a reasonable alternate explanation for the change (above) in the O/E ratio.

          It is an interesting fact that the spondylitis cases had been treated with X-rays at some time between 1935 and 1954. It so happens that approximately 1954 is the time that the worldwide recognition of the relationship between cigarette-smoking and bronchogenic (lung) cancer occurred.

          During the latter (and still large) part of the follow-up beyond 1954, it is extremely likely that patients with spondylitis -- who fare badly from a variety of respiratory ailments -- were placed under special pressure to stop smoking cigarettes. It is reasonable to consider the possibility that lung-cancer may have been cut down as a result of less smoking among the spondylitics, during the latter part of the follow-up period.

          But among the general population (the control group), the pressures to stop smoking in the population-at-large would have been less. Thus, if the spondylitics reduced their smoking more than did the general population, the result expected in the latter part of the follow-up would be a decline in the O/E value for lung-cancer -- a decline wholly unrelated to radiation exposure.

          Obviously, we do not know that the spondylitics of the British series were in fact reducing their cigarette-use after 1954. All we say is that it is reasonable to worry about smoking as a possible confounding variable, and that it might explain the anomalous results on duration.

Conclusion on Reliability

          For all the reasons given above, we think analysts should place far more credence in the A-Bomb Study (1950-1982) on the issue of duration, than in the Ankylosing Spondylitis Study. We have shown several grounds for heeding the warning about the Spondylitis Study that " . . . the possibility must be considered that the findings [on duration] are spurious . . . " (Dar87, p.188).

          Nonetheless, the Spondylitis Study is sure to be cited by some as if it were the scientifically weighty evidence on duration. In fact, Robin Mole (a former member of the ICRP) has already done so in Lancet. The parenthesis is his own:

          "The data from the X-rayed spondylitics suggested that the risk of excess cancer other than leukemia was possibly not continued to the end of life (as commonly assumed for purposes of radiological protection) but may be limited to 2-3 decades after exposure. Direct observation on man is more to be relied upon than hypotheses based on extrapolation from the bomb survivors' experience" (Mole87).

          We cannot understand why Mole characterizes the Spondylitic Study as "direct observation on man," and then dismisses direct observations on the A-bomb survivors -- of comparable duration -- as merely the basis for "hypotheses" about duration. Nor can we understand Mole's failure to mention findings from the other studies cited in Part 3 of this chapter.

          By contrast, we think the issue is too important for casual treatment.

          Having examined the available evidence in Parts 1, 2, 3, and 4 of this chapter, we think it would be scientifically indefensible to estimate Lifetime Fatal Cancer-Yields on the basis of an end to the carcinogenic effect two to three decades after exposure, or to drop the presumption of lifelong duration on the basis of the Spondylitis Study. Virtually all the non-spondylitic evidence (1) is overwhelmingly against an end to the effect by 25 years post-irradiation, and (2) is pointing to a lifelong effect.

          Within the A-Bomb Study, which merits far more credence than the Spondylitic Study on this issue, the evidence to date not only supports the presumption of a lifelong effect out to the age of about 80 years, but the evidence for most age-bands does not even show any meaningful decline in the average intensity of the effect, when the periods before and after 25 years post-irradiation are compared. Indeed, in some age-bands, the data show the intensity increasing in the period beyond 25 years (Table 17-B).

5.   The Bottom Line

          1.   For reasons given in detail above, evidence from the A-bomb survivors is inherently more reliable on the issue of duration of effect, than evidence from the ankylosing spondylitic patients. Moreover, evidence from other studies is -- almost without exception -- consistent with the A-Bomb Study and inconsistent with the Spondylitic Study.

          2.   In the A-Bomb Study, we have demonstrated (Tables 17-A, 17-B) that the induction of cancer by radiation is clearly not over by 25 years after the irradiation. In some age-bands, the average effect appears to be even greater in the period beyond 25 years after exposure, than in the period before the 25th year.

          3.   Therefore, the only reasonable presumption from the evidence at hand is that the radiation effect in cancer-induction will be lifelong, and that its intensity will remain approximately constant, at the average level observed so far in the A-Bomb Study. (As noted in Chapter 13, Part 6, and Chapter 16, Part 2, RERF analysts are using the same presumption in estimating their Lifetime Fatal Cancer-Yields.)

          4.   Only time, and preservation of the A-Bomb Study as a legitimate prospective study, can validate or invalidate the presumption. Meanwhile, if we were to use any other presumption, unsupported by the available evidence, it would be a clear sign of some sort of bias.

Table 17-A
Change in Cancer-Rate Ratio, and in Excess Relative Risk,
with Time after Exposure.
|           Col.A  |    Col.B             Col.C  |    Col.D      Col.E   |      Col.F      Col.G  |
|                  |  Exposed         Reference  |                       |                Excess  |
|       Follow-Up  |    Group             Group  |  Exposed   Reference  |    Cancer-  Rel. Risk  |
|        Interval  |   25,203            66,028  |    Group      Group   |       Rate  ---------  |
|                  |  Persons           Persons  |                       |      Ratio  (R2 / R1)  |
|       --------   |  -- CANCER-DEATHS, RAW ---  |  -- CANCER-RATES ---  |   --------      Minus  |
|                  |                             |     (R2)        (R1)  |  (R2 / R1)       1.0   |
|                  |                             |                       |                        |
|  Row 1  1950-54  |      150               353  |   59.517      53.462  |      1.113      0.113  |
|  Row 2  1955-58  |      150               423  |   59.517      64.064  |      0.929     -0.071  |
|  Row 3  1959-62  |      216               478  |   85.704      72.394  |      1.184      0.184  |
|  Row 4  1963-66  |      236               531  |   93.640      80.420  |      1.164      0.164  |
|  Row 5  1967-70  |      257               536  |  101.972      81.178  |      1.256      0.256  |
|  Row 6  1971-74  |      268               588  |  106.337      89.053  |      1.194      0.194  |
|  Row 7  1975-78  |      311               631  |  123.398      95.566  |      1.291      0.291  |
|  Row 8  1979-82  |      307               615  |  121.811      93.142  |      1.308      0.308  |
|                  |                             |                       |                        |
|  Row 9  1950-82  |     1895              4155  |  751.895     629.278  |      1.195      0.195  |
|===============================================================================================  |
|        Division of Follow-Up into Two Equal Parts :                                             |
|                                                                                                 |
| Row 10  1950-66  |      752              1785  |  298.377     270.340  |      1.104      0.104  |
| Row 11  1967-82  |     1143              2370  |  453.517     358.939  |      1.263      0.263  |
|===============================================================================================  |
|        Most Recent Follow-Up Period :                                                           |
|                                                                                                 |
| Row 12  1979-82  |      307               615  |  121.811      93.142  |      1.308      0.308  |
|===============================================================================================  |
|        Division of Follow-Up at 25 Years Post-Bombing :                                         |
|                                                                                                 |
| Row 13  1950-70  |     1009              2321  |  400.349     351.518  |      1.139      0.139  |
| Row 14  1971-82  |     886             1834    |  351.545     277.761  |      1.266      0.266  |

  1. The exposed group, in Columns B and D, represents all individuals for both cities in Dose-Groups 3 through 8. In this combined group, 25,203 persons have been followed-up since 1950.
  2. The reference group, in Columns C and E, represents all individuals for both cities in Dose-Groups 1 + 2. In this nearly unexposed group, 66,028 persons have been followed-up since 1950.
  3. The entries in Columns B and C, for each four-year follow-up period separately, are obtained from RERF's diskette "R10ALL" (R-Ten-All);   see Chapter 7. Cumulative cancer-deaths, 1950-1982, amount to 6,050 cases;   Columns B and C show when they occurred.
  4. In Columns D and E, all cancer-rates per 10,000 initial persons apply to the rate produced by the new cancer-deaths which occurred during the particular time-interval indicated in Column A. Thus the rate from 1950-1966 (Row 10), in the exposed group, is (752 cases / 25203 persons) times (10000 persons) = 298.377 .
  5. (Column F) = (Column D entry) / (Column E entry).
  6. (Column G) = (Column F minus 1.0). As noted in the text, Part 1, entries in Col. G are commonly called "excess relative risk" when R1 is the rate observed in the unexposed group. The quantity (R2/R1)-l is also the numerator of Equation (7) in Chapter 15, for calculating K-values. In this table, it provides a direct basis for comparing strength of cancer-risk, during one set of follow-up years, with strength of cancer-risk during any other set of years.

Table 17-B
The Time-Course of Cancer-Rate Ratios, for the Individual Age-Bands.
|                 CANCER-DEATHS         CANCER-RATES   Cancer  |                  CANCER-DEATHS       CANCER-RATES  Cancer |
|  Follow-up  Exposed  Reference   Exposed  Reference   Rate-  |  Follow-up  Exposed  Reference  Exposed  Reference  Rate- |
|   Interval    Group      Group     Group      Group   Ratio  |   Interval    Group      Group    Group      Group  Ratio |
|          A        B          C         D          E       F             A        B          C        D          E      F |
|                                                                                                                          |
| AGE-GROUP = 0-9 years ATB.                                   | AGE-GROUP = 10-19 years ATB.                              |
|  Exposed Persons = 4649.  Reference = 13753.                 |  Exposed Persons = 5443.  Reference = 13781.              |
|    1950-54        0          0         0          0    Ind.  |    1950-54        1          1    1.837      0.726  2.532 |
|    1955-58        1          0     2.151          0    Ind.  |    1955-58        2          9    3.674      6.531  0.563 |
|    1959-62        2          1     4.302      0.727  5.917   |    1959-62        3          8    5.512      5.805  0.949 |
|    1963-66        2          0     4.302          0    Ind.  |    1963-66        6         12   11.023      8.708  1.266 |
|    1967-70        3          9     6.453      6.544   0.986  |    1967-70       17         24   31.233     17.415  1.793 |
|    1971-74        4         11     8.604      7.998   1.076  |    1971-74       19         37   34.907     26.849  1.300 |
|    1975-78        6         21    12.906     15.269   0.845  |    1975-78       43         61   79.001     44.264  1.785 |
|    1979-82       12         21    25.812     15.269   1.690  |    1979-82       29         77   53.279     55.874  0.954 |
| Division of Follow-Up into Two Equal Parts                   | Division of Follow-Up into Two Equal Parts                |
|    1950-66        5          1    10.755      0.727  14.791  |    1950-66       12         30   22.047     21.769  1.013 |
|    1967-82       25         62    53.775     45.081   1.193  |    1967-82      108        199  198.420    144.402  1.374 |
| Most recent Follow-Up Period                                 | Most recent Follow-Up Period                              |
|    1979-82       12         21    25.812     15.269   1.690  |    1979-82       29         77   53.279     55.874  0.954 |
| Division of Follow-Up at 25 Years after Exposure             | Division of Follow-Up at 25 Years after Exposure          |
|    1950-70        8         10    17.208      7.271   2.367  |    1950-70       29         54   53.279     39.184  1.360 |
|    1971-82       22         53    47.322     38.537   1.228  |    1971-82       91        175  167.187    126.986  1.317 |
| AGE-GROUP = 20-34 years ATB.                                 | AGE-GROUP = 35-49 years ATB.                              |
|  Exposed Persons = 5138.  Reference = 12553.                 |  Exposed Persons = 6066.  Reference = 14837.              |
|    1950-54        8         24    15.570     19.119  0.814   |    1950-54       57        107   93.966     72.117  1.303 |
|    1955-58       10         28    19.463     22.305  0.873   |    1955-58       59        153   97.263    103.121  0.943 |
|    1959-62       15         42    29.194     33.458  0.873   |    1959-62       98        186  161.556    125.362  1.289 |
|    1963-66       31         71    60.335     56.560  1.067   |    1963-66      101        226  166.502    152.322  1.093 |
|    1967-70       47         69    91.475     54.967  1.664   |    1967-70      126        276  207.715    186.021  1.117 |
|    1971-74       58         97   112.884     77.272  1.461   |    1971-74      134        300  220.903    202.197  1.093 |
|    1975-78       77        133   149.864    105.951  1.414   |    1975-78      154        330  253.874    222.417  1.141 |
|    1979-82       83        156   161.541    124.273  1.300   |    1979-82      171        310  281.899    208.937  1.349 |
| Division of Follow-Up into Two Equal Parts                   | Division of Follow-Up into Two Equal Parts                |
|    1950-66       64        165   124.562    131.443  0.948   |    1950-66      315        672  519.288    452.922  1.147 |
|    1967-82      265        455   515.765    362.463  1.423   |    1967-82      585       1216  964.392    819.573  1.177 |
| Most recent Follow-Up Period                                 | Most recent Follow-Up Period                              |
|    1979-82       83        156   161.541    124.273  1.300   |    1979-82      171        310  281.899    208.937  1.349 |
| Division of Follow-Up at 25 Years after Exposure             | Division of Follow-Up at 25 Years after Exposure          |
|    1950-70      111        234   216.037    186.410  1.159   |    1950-70      441        948  727.003    638.943  1.138 |
|    1971-82      218        386   424.290    307.496  1.380   |    1971-82      459        940  756.677    633.551  1.194 |
| AGE-GROUP = 50+ years ATB.                                   |                                                           |
|  Exposed Persons = 3907.  Reference = 11104.                 | Notes 1 through 5 of Table 17-A apply here, too.          |
|    1950-54       84        221   214.999    199.027  1.080   |      Total persons = 91,231.                              |
|    1955-58       78        233   199.642    209.834  0.951   |      Total cancer-deaths = 6,050.  These are the "raw"    |
|    1959-62       98        241   250.832    217.039  1.156   |           values from Tables 11-B and 11-D.               |
|    1963-66       96        222   245.713    199.928  1.229   | Each age-band includes both sexes in both cities.         |
|    1967-70       64        158   163.809    142.291  1.151   | "Exposed" class includes all Dose-Groups 3 through 8.     |
|    1971-74       53        143   135.654    128.782  1.053   | "Reference" class includes Dose-Groups 1 + 2.             |
|    1975-78       31         86    79.345     77.450  1.024   | ------------------------------------------------------    |
|    1979-82       12         51    30.714     45.929  0.669   | Initial persons in each class are stated, near the top    |
| ----------------------------------------------------------   |      of each age-band.                                    |
| Division of Follow-Up into Two Equal Parts                   | Cancer Death-Rates, per 10,000 persons initially in a     |
|    1950-66      356        917   911.185    825.829  1.103   |      class, are for the specified interval.               |
|    1967-82      160        438   409.521    394.452  1.038   | Cancer Death-Rates are (number of deaths                  |
| Most recent Follow-Up Period                                 |      from Col.B or Col.C / initial persons) x (10,000).   |
|    1979-82       12         51    30.714     45.929  0.669   | "Ind." abbreviates "indeterminate."                       |
| Division of Follow-Up at 25 Years after Exposure             | ------------------------------------------------------    |
|    1950-70      420       1075  1074.994    968.120  1.110   | COMMENTS ON THIS TABLE ARE IN TEXT, PART 2.               |
|    1971-82       96        280   245.713    252.161  0.974   |                                                           |

Next | ToC | Prev
back to RIC | CNR | radiation | rat haus | Index | Search