Acceptance speech – Alice Stewart

Antecedents and Achievements of the Oxford Survey of Childhood Cancers

After expressing my thanks to the Foundation I would like to say a few words about the Oxford Survey of Childhood Cancers.

Introduction

A lengthy association with the Oxford survey is clearly the reason why I have been honoured with one of the 1986 Awards of the Right Livelihood Foundation. This Foundation is anxious to see established “a science of permanence” in relation to many of today’s problems. Therefore, this occasion would seem to be an appropriate one for describing a survey which, whatever its other failings, could never be accused of taking too hasty a look at the problem of early cancer deaths. The name is a reminder that for 20 years the survey was an Oxford University activity, but for the last 12 years the headquarters have been in Birmingham University. Like the tortoise in the fable, the survey was never designed for sprinting. However, again like the tortoise, it has shown great endurance and an ability to outclass rival projects which once seemed certain of winning all the prizes. The credit for these achievements belongs to many, many persons, but it gives me great pleasure to acknowledge my special indebtedness to David Hewitt and George Kneale, without whom there would have been no story to tell.

Antecedents

The first English University to promote social medicine to full academic status (by establishing a Chair in the subject) was Oxford. This was a wartime move prompted by a Nuffield benefaction which was intended to keep a research and teaching Institute in funds for ten years. It led, in 1942, to a distinguished London physician (John Ryle) coming to Oxford as the first Professor of Social Medicine and accepting the following brief: (1) to Investigate the influence of social and genetic factors on the prevalence of human disease and disability; (2) to find both new ways of identifying factors which interfere with full development and maintenance of health and new ways of protecting society from the effects of these factors and (3) to provide for the teaching of the new subject to clinical students and postgraduates.

The early years of the new Institute were uneventful but by the time of Professor Ryle’s death, in 1951, both the Nuffield benefaction and the initial (wartime) enthusiasm for the subject were exhausted. Therefore, on the advice of a science faculty which naturally regarded experiments as a better research investment than surveys, the University disbanded the Institute, leaving in its place only one person with no clearly defined duties except “to do such teaching as the clinical professors require”.

Overnight I had become “University Reader it Social Medicine” with nothing to do and no special training in a subject which clearly needed statistical as well as medical expertise. All I had was considerable experience of clinical medicine and a firm belief that it only needed a few completed surveys for the University to recognise its mistake and resume full support of the subject.

During the war, while working in the Nuffield Department of Clinical Medicine, it had occurred to me that a survey might be the best way of resolving one of the problems referred to the Department by the Medical Research Council. It was a typical wartime problem (requiring quick action) and my rapidly improvised survey was so successful that similar activities became a regular accompaniment of my clinical work. It was this wartime experience which led to my being appointed as first assistant to the Professor of Social Medicine in 1946.

After crossing the Rubicon between clinical and social medicine, my first concern was a survey of workers in shoemaking factories. The original purpose of this research project was to discover how tuberculosis is spread between fellow workers but, like all good surveys, it had branched out in various directions. Therefore, when the Institute was disbanded, I still had helping me an erstwhile Medical Officer of Health (Josefine Webb) and a young statistician (David Hewitt), both of whom were paid for by an MRC grant. We were anxious to remain as a research team but in order to do so we had to find a new project and new funding.

These events were taking place at a time when, both in Europe and in North America, deaths from leukaemia were increasing at an alarming pace. During the next decade the rest of the world also exchanged low, prewar levels of leukaemia mortality for the much higher (though stable) rates of today. But in 1951 there was considerable anxiety lest we were on the brink of a leukaemia epidemic.

In spite of the rising death rate, leukaemia was still a rare disease and, except in Hiroshima and Nagasaki, there were no local epidemics. Therefore, a survey to discover the causes of the rising death rate was not regarded as a practical proposition. However, a close scrutiny of several sets of national statistics, by David Hewitt, had revealed something which made us think that perhaps one facet of the rising death rate might be identified by interviewing mothers of children who had recently died from leukaemia and comparing what they had to say with similar information from mothers of live children and children who had died from other malignant diseases.

What David had done was to show that, instead of a few leukaemia deaths being evenly distributed between infants and older children (which was the prewar pattern), there was now a conspicuous peak of mortality between 2 and 4 years caused, not by the type of leukaemia which was threatening adults (myeloid), but by the opposite type (lymphatic). Therefore, it was possible that young children were the main sufferers from an antenatal event which had escaped the notice of obstetricians but might be retrieved by systematic questioning of mothers.

We should have realised that the Medical Research Council was unlikely to be impressed by a research project which depended upon memories of lay informants and would certainly require several data collecting centres. However, although we failed to obtain an MRC grant, we were given £1000 by the Lady Tata Memorial Fund for leukaemia research. There were no conditions attached to this windfall and I was free to do what I liked. Furthermore, the University had finally agreed to my having a statistical assistant, and Dr. Webb was told that she could remain on the external staff of the MRC indefinitely. Therefore, we decided to do two things: obtain the death certificates of all children who had died from leukaemia and other forms of cancer in recent years (i.e. 1953-55), and seek the help of Public Health Departments. These Departments had an ideal source of live controls (in their registers of live births) and might agree to do paired interviewing of cases and controls since there were so few cancer deaths of children.

Achievements

We eventually persuaded all County and County Borough Health Departments to serve as data collecting centres and gradually set in motion a survey which not only became coextensive with the whole country (i.e. with England, Scotland and Wales) but was so successful that, in no time at all, over 80 per cent of the 1953-55 cancer deaths were traced and matched with live controls. We even obtained early warning of the prenatal x-ray finding (by detecting a significant case/control difference in the first batch of completed interviews). Therefore, from the onset there was systematic checking of mothers’ claims against actual records of obstetric x-rays with foetal involvement. But although we had achieved one of our goals (by finding an antenatal event with cancer associations) we soon discovered that we still had no explanation of the early peak of leukaemia mortality.

You will recollect that the original reason for the survey was to discover why young children (but not infants) were being increasingly threatened by an certain kind of leukaemia. What we had found was evidence that an uncommon event (only 7% of our controls had records of foetal irradiation) had probably caused a few deaths which were scattered among leukaemias and other malignant diseases and insufficient to leave any mark on vital statistics. Therefore, if we had not accidentally found these radiogenic cases there would have been no incentive for later searches of cancer effects of prenatal x-rays and other sources of low level radiation.

The necessarily small numbers of deaths caused by a few children being briefly exposured to a small dose of x-rays shortly before birth (and the even spread of these cases between leukaemia and other malignant diseases) were to cause us many future headaches. But, in 1958, the main problem was to convince skeptical radiobiologists that there had not been biased reporting of x-rays by mothers of live and dead children. This explanation of our finding was eventually accepted by all radiation protection committees after they discovered that neither animal experiments, nor screening of A bomb survivors for cancer deaths, nor similar screening of x-rayed children (prospective surveys) were producing any evidence of a cancer risk from foetal irradiation. Therefore, for many years faulty design of survey remained the “official” reason for our prenatal x-ray finding also a reason for giving no encouragement to case/control or retrospective surveys.

According to this assessment of OSCC data there was no reason why the x-rayed and non-x-rayed cases should have different ages. However, short of both groups consisting of cases which had been initiated during the third trimester of foetal life (i.e. during the usual period for practicing obstetric radiography) one would not expect the radiogenic and spontaneous cases to be affecting exactly the same age groups. Which group would prove to be the younger one would depend upon whether the spontaneous cases were initiated before or after the third trimester. Therefore, if we could only establish a significant difference between the ages of our x-rayed and non-x-rayed cases we could (a) confirm the existence of radiogenic cases without reference to our live controls and (b) discover whether to look for other causes of juvenile cancers among prenatal or postnatal events.

This was the reasoning behind a decision to revive the Oxford survey and put it on an on-going basis by continually adding to the original series of 1953-55 case/control pairs. Provided data collection continued until 1971 – by which time all 1953-55 births would have merged with the adult population – we would have the equivalent of an enormous prospective survey (i.e. a 15 year follow-up of more than two million live births). This would provide the correct basis for comparing ages of x-rayed and non-x-rayed cases and sufficient numbers of x-rayed cases for studying several types of cancer.

This was our main objective when we again approached the Public Health Departments and again obtained their voluntary cooperation. But we also realised that there would be other rewards. The interviews with case and control mothers had covered many topics including maternal age, birth rank and social class; also pregnancy illnesses, drugs and x-rays; postnatal infections, inoculations and x-rays, parent’s occupations, family histories of cancer and, eventually, the ultrasound replacements of prenatal x-rays. Therefore, by continuously extending the data collection period we would eventually have the necessary records for testing several hypotheses. What we did not anticipate – though it actually happened – was that during the data collection period new methods of statistical analysis would be invented which exactly suited the needs of any large data base consisting of cases with matched controls. These new techniques not only enhanced the value of OSCC data but also restored faith in retrospective surveys (i.e. surveys which often went unfunded because they had less in common with experiments than prospective surveys).

Shortly after the restart of data collection we learnt that a prospective survey of three quarters of a million live births in North Eastern States of the USA had found both a raised cancer death rate for x-rayed children and a significant difference in the ages of x-rayed and non-x-rayed cases. These findings gave us great encouragement but did not alter our plans. We were reasonably certain that the new survey design was foolproof and would eventually achieve far more than mere confirmation of the x-ray hazard. For example we already anticipated being able to test three hypotheses advanced by American scientists: first, that preconception as well as prenatal x-rays were causes of juvenile cancers; secondly, that the association between prenatal x-rays and cancers was caused not by the radiation but by the medical reasons for the x-rays and, thirdly, that the radiogenic cases were confined to children who had an inborn susceptibility to radiation, infections and allergies. Eventually, none of these theories was able to withstand tests based on OSCC data. But this is to anticipate the analytical work of George Kneale.

George was still a schoolboy when the Oxford survey began. But he began working on the project even before he had added an Oxford University diploma in Statistics to his other qualifications (i.e. an honours degree in Chemistry). He actually started work in 1962 and, ever since then, has been adding to a repertoire of new ways of analysing any survey data concerned with cancer effects of radiation or other factors with cancer associations. Even before we moved from Oxford to Birmingham University (in 1974) George had greatly enhanced the value of OSCC data by inventing a new method of analysing truncated contingency tables. This was the form necessarily taken by OSCC data when testing for cancer effects of prenatal events. With the new method we could make almost as much use of remote and recent birth cohorts as of the cohorts which could be followed from birth to 15 years of age. Therefore proof that radiogenic cases existed (and were a fraction older than spontaneous cases) was obtained well in advance of the expected date. Armed with this knowledge we were prepared for two eventualities: first, that all the OSCC cases might have moved from a safe in utero environment to a hazardous postnatal environment after cancer induction; secondly, that for most of the non-x-rayed cases this move might have coincide with a later stage of the cancer process than for most of the x-rayed cases. In short, we had learnt both the extreme importance of discovering whether there were changed reactions to other illnesses when they coincide with the latent phase of a juvenile cancer, and the importance of discovering whether these changes are the same for cancers of the immune system (i.e. leukaemia and lymphoma which are known collectively as haemopoetic neoplasms) and other malignant diseases.

While still in Oxford George had heard me discussing the possibility that the postwar increase in leukaemia mortality might have been caused by modern drugs compensating for loss of immunological incompetence and thus preventing infection deaths whose underlying cause was an early effect of the cancer. He discovered that I had been assembling data on early deaths from pneumonia and leukaemia in three countries (USA, Japan and UK) and asked to borrow these. A few weeks later he produced a statistical analysis of each data set which greatly strengthened my hypothesis since each one showed that by the time young children were on the brink of developing leukaemia they had become so infection sensitive that the risk of a pneumonia death is over 300 times greater than the normal risk. Therefore, here was both evidence of George’s analytical skills and a warning that, when a serious infection coincides with the latent phase of a juvenile cancer, only children given antibiotics are at all likely to survive, and this survival will also depend upon whether the infection coincides with an early or late stage of the malignant disease and whether this disease is a haemopoetic neoplasm or other type of cancer.

After the Birmingham move George had a short spell as consultant to the Mancuso study of Hanford workers. He used this opportunity to devise various ways of coming to grips with a problem which he was the first person to recognise. This was the problem of the same individual being exposed to a small dose of ionizing radiation not once but several times at consistently changing ages. George examined this problem from various angles and eventually decided that, although resistance to cancer induction effects of radiation is exceptionally high at 20 years of age, by 50 years it may be no greater as it was shortly before birth.

When Dr. Mancuso lost his research contract (largely as a result of George’s analysis of Hanford data producing risk estimates which were much higher than ones based on extrapolation of high dose effects) George resumed work on OSCC data and gradually assembled the evidence which underpins several important conclusions. For example, he showed that all victims of early cancer deaths are, from birth onwards, more infection sensitive than normal children, and that this undue sensitivity is most pronounced in cases of leukaemia. Therefore, when antibiotics are given to children who are (unknowingly) developing a malignant disease, an infection death may be prevented, but this “narrow escape” will shortly be followed by a cancer death. This is probably why the increase in leukaemia mortality (which provided the original excuse for the Oxford survey) gathered momentum when penicillin began to replace sulphonamides. As was noted by David Hewitt at the time, this change affected affluent before poor societies and was all but confined to infection sensitive age groups. These groups include children of under 2 years of age. However for these children low rates of leukaemia mortality were still being recorded which offered a striking contrast to the mounting death rate for 3 years olds. Hence the early peak of mortality which remains an unsolved problem.

The Oxford survey is no longer alone in showing that (a) there is a cancer hazard associated with obstetric radiography and (b) the radiogenic cancers have a distinctive age distribution. But no other survey has tried to decipher the complex relationships which exist between cancers and infections, or to make a close study of innoculations in relation to the cancers.

The national campaigns which are responsible for the fact that most children in technologically advanced countries are now innoculated against several infectious diseases are not concerned with cancer prevention. Nevertheless as a result of innumerable children being given several boosts to the immune system, some early cancer deaths may have been prevented. Presumably this is the result of immune system effects halting the pathological processes set in motion by cancer inductions. Leastways, according to OSCC data any type of innoculation can prevent any type of cancer (though this is especially likely to happen when a relatively late innoculation coincides with a slow growing tumour). These observations suggest that there may be a crucial period between induction and diagnosis when any boost to the immune system is important. Therefore it is possible that premature infection deaths are not the only consequences of infections coinciding with cancer latent periods.

Finally, the reason why the spontaneous and radiogenic cases in the Oxford survey proved to be so alike is gradually becoming clearer. This clarification is the result of comparing OSCC data with background radiation dose rates in various regions. In this way a cancer effect of the daily (and inevitable) exposures to background radiation was detected whose magnitude was compatible with this being the only common cause of early cancer deaths. This is too recent an observation for further elaboration but it shows that, even after 30 years, OSCC data still have their uses. Therefore, let me end this account of achievements with a short postscript containing George’s own description of his work and two ideas which may yet provide him with more work.

Postscript

When asked by a colleague of Dr. Mancuso to describe the work of our unit, George gave the following reply: “It is my job to prove that Dr. Stewart’s theories are wrong”. Hence the strength of our long association.

One of the ideas which is still awaiting a “George rebuttal” is that something akin to the cancer inhibitor effect of innoculations might be the reason why two unusual types of haemopoetic neoplasms (i.e. chloroma and Burkitt tumour) are found mainly in malaria infested regions. In these regions all children over 2 years of age have survived repeated attacks of malaria and related infections. Therefore, the whole population must consist of persons who, early in life, acquired immune responses which were so strong that they possibly placed constraints upon mutant cells as well as live pathogens. Whether such constraint is possible I do not know. But if it were we could account for chloromas and Burkitt tumours being unusually localized forms of haemopoetic neoplasms, and for there being exceptionally high levels of antibodies to certain viruses associated with Burkitt tumours.

My second idea is an attempt to account for the early peak of leukaemia mortality by suggesting that it might be a consequence of antibiotic intervention being (in relation to childhood infections) more successful in revealing the true prevalence of lymphatic than myeloid leukaemia. This requires the assumption that both myeloid and lymphatic cases with foetal origins prevent normal development of the immune system (which leads to premature infection deaths of children who would otherwise die from either type of leukaemia) and that the myeloid cases also prevent normal maturation of haemoglobin (which leads to early deaths from tissue anoxia of children who would otherwise die from myeloid but not lymphatic leukaemia). This sequence of events would require involvement of a common ancestral cell of erythrocytes and myelocytes in cases of myeloid leukaemia and would also require exceptionally high levels of foetal haemoglobin to have associations with three conditions, namely, very young cases of myeloid leukaemia; postnatal infections which are not responding to treatment and sudden infant deaths.

This attempt to explain the early peak of lymphatic leukaemia mortality brings my story full cycle. Therefore, it only remains for me to comment on the timeliness of your award. In Britain widespread use of ultrasound by obstetricians only began in 1975. Therefore to prove that these examinations do not have any cancer associations might have required data collection from all pre-1990 deaths. The present final year (for lack of funds) is 1982. However, your award will prevent such early closure and may help us to attract the funds which are needed to complete this project. Therefore, on behalf of all previous helpers, including our case and control mothers, I again say thank you.