Strange Glow
Page 21
This was an extraordinary chance to determine how the most feared agent on Earth affects health, and the more astute scientists recognized it as exactly that. As radiation scientist Dr. Robert H. Holmes said of the Japanese atomic bomb survivors to a journalist, “These are the most important people living.”6 And so, after some fitful starts, much controversy, and intense battles over who would pay for it, on November 26, 1946, President Harry Truman directed the National Academy of Sciences to begin a long-term study of the effects of atomic bomb irradiation on humans. This directive lead to the establishment of the Atomic Bomb Casualty Commission (ABCC), funded by the United States. The cornerstone of this research effort was the massive Life Span Study (LSS), which has been following the medical outcomes of 120,000 bomb survivor and control subjects up to the present day (i.e., over 65 years).7 The LSS is considered to be the definitive epidemiological study on the effects of radiation on human health.8
Radiation may have been a new hazard for mankind, but interest in health risks was nothing new. People have always been preoccupied with measuring their risk of death from various activities and exposures. In fact, the first known calculation of life expectancy by profession was compiled by Ulpian (170–228 AD), a Roman jurist. His calculations allowed for the determination of the risk of death from various occupations, and thus let people gauge how dangerous those occupations were. (Gladiator was a bad one!)
Then, in the seventeenth century, John Gaunt (1620–1674), a London shopkeeper with an avocation of dabbling in public health statistics, decided to use readily available public mortality data to calculate the risk of death from various health ailments. In Gaunt’s own words: “Whereas many persons live in great fear and apprehension of some of the more formidable and notorious diseases, I shall set down how many died of each; that the representative numbers, being compared with the total [mortality], those persons may better understand the hazard that they are in.”9
What Gaunt ended up producing was the first known life table, which is simply a tabulation of the probabilities of surviving to each year of age. In addition to satisfying people’s curiosity about how long they might expect to live, it was soon realized that a life table provided a rational means to price life insurance. So John Gaunt is often credited with being the originator of the modern life insurance industry. It seemed that public health statistics held much more valuable information than people had previously realized.
After a couple of centuries passed, the “bean counters” working in public health turned things up a notch, and shifted what had been merely descriptive public health statistics into the new branch of science called epidemiology. London physician John Snow (1813–1858) was their patron saint. Snow was a physician and public health advocate who believed in the value of counting and measuring. Consequently, in 1854, when a cholera epidemic swept through London and killed a large proportion of the population, it was not surprising that Snow was out counting. He was counting deaths. By counting cholera victims, calculating death rates by neighborhood, and precisely mapping the locations of victims’ residences, he was able to show that those dying at the highest rate were likely obtaining their drinking water from a residential underground water supply system that had a public pump on Broad Street.10 He had thus identified a contaminated drinking water source as the cause of the epidemic. In an apocryphal anecdote to an otherwise accurate account, Snow was reputed to have single-handedly ended the epidemic by absconding with the pump’s handle, thus cutting off that source of disease-contaminated water to the community and forcing the residents to use other, cleaner water sources.
Following in the tradition of Snow, radiation epidemiologists would count and map disease incidence in relation to the two atomic bombings’ ground zeroes; by analogy, they could show that ground zero was the epicenter for radiation sickness and various cancers, just as the Broad Street pump had been the epicenter for cholera. The radiation scientists, however, would even go further than Snow. He never measured an individual’s cholera risk in terms of the amount of water drunk (i.e., the risk per unit dose). That was a mistake that the radiation epidemiologists would not repeat.
LET’S DO LUNCH
On December 6, 1946, less than two weeks after President Truman had authorized the National Academy of Sciences to begin research, a team of American scientists was already in Hiroshima and ready to begin work. But first they attended a formal lunch that was hosted by the assistant mayor of the city, Hisao Yamamoto, and held in a primitive building near ground zero. The building had been hastily constructed to replace the totally ruined city hall.11 The assistant mayor apologized for the shabby accommodations, but he was nonetheless able to treat the Americans to a lavish three-course meal.
Dr. Masao Tsuzuki joined the lunch; he was Japan’s leading expert on the health effects of radiation and the future attending physician for the fallout-exposed fishermen from the Lucky Dragon No. 5. Local physician Dr. Ikuzo Matasubayashi was also present. Matasubayashi reported to the Americans that he and statistician Masato Kano were already gathering data on the survivors of the Hiroshima bombing. Their goal was to identify all survivors and their exact locations at the time of the bomb’s detonation, and to also include a brief medical description of each person’s injuries. During the 15 months since the bombing, with the help of volunteers, they had already registered tens of thousands of people (out of an estimated 200,000 survivors) in their study, with a file for each victim. Although they did not have the facilities to catalog the data properly, they had secured a large room with many tables and stacked the files on tables in sequential piles that corresponded to the distances the survivor had been from ground zero; this was reminiscent of Snow’s mapping the distances of cholera victims from the water pumps. They knew that any new ailments that might show up in this population over time should be traceable to the bomb’s epicenter (i.e., ground zero) if radiation were the cause. Thus, the Japanese had already broken ground on what would become one of the greatest epidemiological cohort studies ever conducted.12
A cohort study is the strongest and most reliable study design that epidemiologists have at their disposal. The reasons for its superiority are quite involved, and beyond the scope of our discussion. Having said that, we can still get a sense as to why a cohort study is superior to other study designs by comparing it to a somewhat weaker type of study—the case-control study—that is often used as a second-best alternative when a cohort study is not possible. But before doing this, we must first understand exactly what a cohort study is and how it works.
To best appreciate the strategy of a cohort study, it is useful to consider the origin of the word cohort. A cohort was the basic fighting unit of a Roman legion, starting with precisely 480 recruits, averaging about 20 years old.13 The soldiers in the cohort were all enlisted into service at the same time and usually were not replaced as they died off (unless battle casualties were very heavy), so the fighting strength of a cohort tended to diminish slowly with time.14 The soldiers in the cohort spent their entire term of enlistment together (typically 20 years), with few reassignments in or out. They lived, ate, slept, fought, and died together. And most importantly, they aged together. So the average age of the soldiers in any particular cohort was anywhere from 20 to 40, depending upon the year the cohort had originally been formed.
Over time, the weak or sickly, and the poor fighters, tended to die off first, leaving only the stronger soldiers to fulfill their enlistment terms and retire into civilian life. Any beneficial exposures or experiences that tended to increase a soldier’s fitness (e.g., a diet rich in protein or strengthening exercises) would tend to increase that soldier’s lifespan. In contrast, anything that adversely affected a soldier’s fitness to fight (e.g., drunkenness or asthma) tended to be associated with a shortened lifespan. It can be seen that by recording the different exposures and behaviors of the individual soldiers in the cohort, and then waiting to see whose names appeared on the sick call and casualty lists over time, it shou
ld be possible to identify those exposures or behaviors that put soldiers at risk for illness and premature death. For example, drunkards might be expected to enter battle with a hangover, thus decreasing their chances of surviving to fight another day. This is why any study that follows a defined group of people over a long time to determine how their exposures affect their health outcomes is said to be a cohort study. In essence, such a study emulates tracking casualty statistics of a Roman cohort.
The term “cohort” is now more broadly considered to define any collection of people that is tracked as a single defined group. With a little reflection, we can appreciate that a single individual can simultaneously be a member of multiple cohorts. For example, he or she might be 1 of 400 babies born in a certain hospital in 1955, 1 of 50,000 children vaccinated for polio in some particular state in 1962, and 1 of 200 graduates of the class of 1973 for some specific high school. By studying each of these different cohorts, we might find a higher risk of sudden infant death for children delivered by caesarian, an increased risk of contracting meningitis for people who got a polio vaccine, or a higher risk of a fatal car crash for those who binge drank in high school.15 The study of such cohorts can give us very reliable health risk information if we ask the pertinent questions and have the time and patience to wait for the data to roll in.
The LSS of the atomic bomb survivors is a classic cohort study. These survivors were recruited into the study at the time of the bombing, their exposures (in this case, their various radiation doses) were individually recorded, and their disease and death outcomes were tracked over time as the whole group (i.e., cohort) aged. If radiation doses were associated with different diseases and death, it would be readily apparent in such a study, given enough years of follow-up.
Now let’s compare a cohort study to a case-control study, in order to illustrate why the former is considered superior.16 A case-control study compares people with a certain type of disease (cases) to those without that disease (controls), and then asks the individuals in both groups what types of things they had been exposed to in their younger years. The expectation is that if exposure to a certain agent is associated with getting the disease in question, the cases will report more exposures than the controls will. Unfortunately, case-control studies often rely on the accuracy of the mental recall of the study subjects. The problem is that cases are sometimes more likely to recall an exposure than the controls, simply because cases are often seeking an explanation for their disease and have already considered various exposures that may have contributed to it. In other words, they are primed for the question. Controls, in contrast, don’t ponder their earlier exposures because they have no reason to do so, and therefore, they tend to underreport their exposures. For example, you might find it difficult to recall whether your childhood neighbors owned a dog unless you still have the leg scar where it bit you. Injury and illness tend to sharpen the memory, and this is often the doorway through which bias enters a case-control study.
Case-control studies can suffer from other biases that are too complicated to be considered here.17 Let’s just say that the well-known propensity of case-control studies to contain hidden biases makes them suspect until their findings are replicated multiple times in different study populations. Only then will case-control study evidence be considered reliable.
Cohort studies, on the other hand, have much less potential for bias because the exposures are ascertained before the disease appears, so recall of exposures is not involved. Because cohort studies are less prone to biases, they are considered the most accurate of all population studies. Cohort studies are the gold standard. This is true to the extent that even multiple case-control studies are often considered merely suggestive until they have been validated in a subsequent cohort study.
We can thus see that all studies are not equal. They differ in the strength of their scientific design. Cohort studies are much more heavily weighted than case-control studies when it comes to scientific evidence.18 Isaac Newton once famously remarked about judging scientific evidence: “It is the weight [of the experiments], not numbers of experiments, that is to be regarded.” As in so many cases, Newton was exactly right about this too.
When the ABCC (Atomic Bomb Casualty Commission) team members saw the piles of folders, they were astonished at all their Japanese counterparts had already achieved. They later noted in their report that the data already accumulated “appear as accurate and reliable as could be expected in light of the circumstances, particularly the limited facilities.”19 Furthermore, the Japanese seemed to appreciate the obstacles to completing the work in a timely fashion, and welcomed the involvement of the Americans, even providing them with English translations of all their findings thus far. So the LSS hit the ground running, thanks largely to the up-front work of a few prescient Japanese physicians and scientists, and a small army of volunteers.
Unfortunately, the honeymoon between the Japanese and American scientists would be short-lived, and petty bickering about data control and publication authorship would soon follow. As so often is the case in scientific research, professional pride, infighting, and jealousy soon marred the progress and hampered the work. The details of the multiple feuds between the American and Japanese scientists have been well documented elsewhere and are not worth relaying here.20 Nonetheless, in the end, both groups came to acknowledge that they needed the other for the LSS to succeed, and they put aside their differences enough to get the work done. This less-than-happy marriage nevertheless produced many children, in terms of scientific findings, and continues to provide new insight to this day on how radiation affects health.
In 1975, the ABCC was reorganized into the Radiation Effects Research Foundation (RERF), which now administers the ongoing LSS. The RERF, a joint venture of the Japanese and American governments, is headquartered in Hiroshima and remains the heart of radiation epidemiology to this day.21
FINDINGS OF THE LIFE SPAN STUDY
The LSS is now over 65 years old, as are its surviving study subjects. So the study has truly lived up to its name, following many subjects from nearly cradle to grave. Much has been learned. Let’s look now at who these people are and what has happened to them.
The goal of the study was to include as many survivors of the Hiroshima and Nagasaki bombings as possible. An exhaustive search was conducted and 61,984 people were identified in Hiroshima, and 31,757 in Nagasaki. It is hard to say how many people were missed, but this total of nearly 94,000 probably represents about half of the total survivors who had been within the geographical study area at the time of the bombings.
Over 97% of these people received doses less than 1,000 mSv, so relatively few had actually experienced radiation sickness, which typically requires whole-body dose levels above 1,000 mSv. In fact, 94% had received doses of less than 500 mSv, and 80% had received doses less than 100 mSv. (For comparison, annual natural background doses in the United States average about 3 mSv.)
As this cohort aged, its members succumbed to various ailments, as all of us do when we get older. Few of these natural ailments could be easily traced to the victims’ radiation exposures, with the notable exception of cancer. As of 2013, about 11,023 people in the study had died from cancer. Comparisons with cancer rates from control populations (i.e., people similar in other respects but lacking atomic bomb radiation exposure) suggested that only 10,402 cancer deaths would have been expected (by 2013), in a group comprised of unexposed 94,000 Hiroshima and Nagasaki citizens. So it is assumed that the excess deaths, 621 cases, were due to the radiation exposure. Of the 621 deaths, 94 were leukemias and 527 were other types of cancer. Therefore, the excess cancer deaths in this bomb survivor study group are about 6% more than would have been expected for a reference population with the same age and gender distribution.22
There was also a clear dose response for radiation-induced cancer risk. That is, those who had been closer to ground zero, and therefore had higher radiation doses, suffered proportionately more cancers than t
hose who were farther away. This convincing dose response relationship strongly supported the notion that the radiation was causing the additional cancer cases.
Scientists often rely on dose response data as one of several criteria to establish causal associations. The reasoning is simple. If one thing causes something else to then happen, increasing the first thing (dose) should also increase the consequence (response). For example, if eating excess food causes obesity, someone who eats large amounts of food should be more obese (all else being equal) than someone who eats less. Establishing a dose response is often considered essential evidence to support a claim of a cause and effect relationship. In other words, a study that demonstrates a dose response does not, alone, prove that the one thing causes the other, but failure to show a dose response is often considered evidence against a causal connection.
Dose response data gives us something else. It allows us to establish a rate for the response. For example, we may find that every cupcake eaten translates into one pound of weight gain, or we may find that a single cupcake results in two pounds of weight gain. A finding of “two pounds per cupcake” is twice the rate of “one pound per cupcake,” suggesting that cupcakes are twice as threatening to a person’s waistline.