Commentary: Kristian Feyer Andvord's studies on the epidemiology of tuberculosis and the origin of generation cohort analysis

The Norwegian physician Kristian Feyer Andvord (1855–1934) who had no training in epidemiology, observed differences in mortality rates from tuberculosis between counties in Norway late 19th century and started in the 1890s to speculate on why the rates differed between counties and had fallen for infants but not adults. This triggered him to spend his career on the natural history of tuberculosis by means of population data and made him suggest that primary infection with tuberculosis mainly takes place in childhood and that falling mortality rates at the time of observation was a reflection of exposure early in life.1 Wade Hampton Frost later published a similar idea citing Andvord, and in 1939 he introduced the concept of ‘cohort’.2 This seminal paper by Frost is widely known in epidemiology as it marks the birth of a concept vital to the discipline's self-identification.3–6 Less attention has been paid to Andvord predating Frost in introducing a means to elucidate the natural history of diseases, which has had such a major impact on epidemiology as an analytical discipline (Figure 1). Tuberculosis mortality per 10 000 within the various 5-years cohorts from 1896–1900 to 1926–27 in Norway. The curves for 1926 and 1927 are drawn separately, but the values were added together Source: Reference 11 The scientific community in the inter-war period when these papers were published took a keen interest in how factors early in life could result in a weakened ‘constitution’ and give rise to increased risk of ill health later in life, which has been described elsewhere in an Anglo-American context.7 A few references have been made to Andvord's 1930 publication that due to its German translation has given it some international attention as one of the first of its kind.5,6,8,9 Rather than attempting to contribute in a contest of ‘who was the first to do a cohort analysis’, it appears as if scientific progress of this method took place in different contexts. All the way through his work, Andvord's main objective was to identify the age dependent susceptibility of primary infection and explain later life clinical manifestation of the disease by relating patho-anatomical evidence and clinical accounts to macro-level patterns of mortality statistics. By drawing a line from the very start of his scientific career in 1889,10 it becomes apparent that the method itself evolved from a context much earlier than the 1930 publication alone suggests.11 More than 40 years of his working life he spent on studying the natural history of tuberculosis, in the first place by means of comparing first geographic pattern and later birth cohorts. This period coincided with major changes in the scientific conceptions of tuberculosis and diseases in general and with political and ideological notions of prevention. Here, we would like to follow the origin of this method in the work of Andvord and the causal inferences he made in this process. Andvord was born in the Southern part of Norway in 1855. His parents, brother and sister died from tuberculosis. At the age of 24, Andvord contracted tuberculosis while he was studying medicine and was admitted to the Tonsåsen sanatorium in 1879 located in the mountains in Valdres, Eastern Norway. After graduating in medicine in 1882 he worked as a general practitioner in Valdres for several years and junior doctor at The National Hospital for some months. He then moved to Kristiania (Oslo) to work as a general practitioner until his death in 1934. Every summer until 1916 he worked as a head physician at Gausdal sanatorium (Figure 2). A portrait of Kristian Feyer Andvord in front of Gausdal sanatorium Andvord went on study trips to several countries in continental Europe and Great Britain. In 1903, he went to Lille (Institut Pasteur) and London (Brompton Hospital) to study serum therapy and modern tuberculosis surgery. From early 1900, he participated at international conferences on tuberculosis, and in 1902 he was elected as one of the vice presidents in the International Union against Tuberculosis and Lung Disease. Andvord was an author of 24 publications in Norwegian on tuberculosis, the first one published in 1889.12 Around 1890, he speculated on the various health effects of climate on risk of tuberculosis and therapeutic effect of placing patients in sanatoriums in the mountains.13,14 He brought this aspect into another publication ‘On the prevalence of tuberculosis’ (1895), where he proposed several potential explanations to the geographical variability of tuberculosis mortality.15 He gradually excluded the possible effect of climate because even if mortality rates were high in the far north, in counties like Finnmark with rough weather conditions, they were also high in the more pleasant counties in the south, Agder and Rogaland (Figure 3). Rather, he suggested that the observed differences might have their origin early during the life course. In later reference to his early work he stated that: It was the frequency of the disease in Kristiansand [Agder county] which first gave me evidence that my impression was correct. Even back to the end of the 80s it appeared to be a substantial and increasing fall in childhood tuberculosis even though – despite increasing levels of prevention/prophylaxis, the mortality of pulmonary tuberculosis among adults remain high.11 In which forms and frequency does tuberculosis occur through different periods of life? In terms of frequency, how is the disease in adult age related to the disease in childhood? … Which influence do certain acute and chronic ailments have on the processes of tuberculosis and their development in various age classes? … When – within which life period – and through which pathway does the infection with tuberculosis usually take place? (p. 2) Tuberculosis seems to appear in a different way than the epidemic diseases, as it within the same area kills similar number of individuals each year – all forms included – and when there is an increase or a decrease under all circumstances this appears to follow very smooth curves; it is as if the number of deaths one year proportionally follows that in previous years. (p. 19)15 In this article he also presented rates in different age groups and showed how these rates dropped in childhood and started to increase later in adulthood—which he repeatedly through his work characterizes as the rolling movement of the disease. This pattern was similar even if the rates differed when he compared Norwegian and European cities (Figure 4). From these early observations at population level he concluded that: ‘… as the disease is more or less common in childhood this will be followed similarly with the disease in adult age’. He was particularly convinced of this relationship at that stage because of the proportional relationship between pulmonary tuberculosis mortality in adult age and total mortality in infancy. Based on these mortality data, clinical observations and patho-anatomical findings from autopsies he then proposed (1908) that a decisive immunological defence probably starts early in life making various clinical trajectories more or less likely: a fulminant deadly one in infancy, a benign glandular type or a chronic one that may have a malignant course later in life.16,17 Map of Norwegian counties with total mortality from tuberculosis. Mean values for the years 1867–91. The most heavily affected areas have darker colours. The ‘plus’ sign for a city means tuberculosis mortality is greater, ‘minus'similar and ‘division’ less than the surrounding rural areas. The table shows by county tuberculosis mortality as percentage of known causes of death (Andvord KF. Studier over Tuberkulosens Forekomst i Norge. Christiania: 1895) Tuberculosis as cause of death estimated as ‰ in the population alive for various age groups (x-axis) in Helsinki, Danish cities, Stockholm, Bruxelles, Kiel, Kristiania (Oslo), Bergen and Trondheim (Andvord KF. Studier over Tuberkulosens Forekomst i Norge. Christiania: 1895) Writing for a Norwegian speaking audience of physicians, Andvord did not use the term ‘cohort’. He rather referred to generation and ‘brood’ or ‘generation’ when he analysed mortality rates from tuberculosis in different age groups and stated that: … the frequency of infection in infancy seems to indicate the mortality death rate in later years of life, a fall in mortality in infancy being invariably followed by a similar fall 20-25 years later among adults.11 (p. 660) A ‘population frame’ based on a population of 200 000 individuals with a yearly tuberculosis mortality of 3¾‰ (750 individuals) Source: Reference 17. *Graph No. 1 (….) illustrates latent tuberculosis. Graph No. 2 (-x-x-) illustrates more or less latent and benign glandular tuberculosis. Graph No. 3 (- - - - -) illustrates frequency of latent tuberculosis bacillae detected after inoculation on animals. Graph No. 4 (-xx-xx-) illustrates according to Sophus Bang ‘undoubtedly healthy, macroscopically verified tuberculosis’ as secondary finding in 6006 autopsies. Graph No. 5 (-o-o-) illustrates according to Prof. Grancher clinically detected early stages of tuberculosis among Parisian infants and adolescents. Graph No. 6 (-.-.-) Glandular tuberculosis clinically detected. Graph No. 7 (-..-..-) Positive tuberculin reaction in healthy individuals There was a dramatic decline in tuberculosis in developed countries during the 20th century (prior to the introduction of effective drugs).19,20 This decline has been attributed to many factors, such as improvements in housing conditions and nutrition, reduced crowding and segregation of infectious cases. The extent to which the reduced morbidity was attributable to a reduction in infection transmission per case, as opposed to a reduced risk of developing disease among infected individuals, is unknown.19 All these factors (crowding, segregation, susceptibility) are potentially mediating the strong social pattern of tuberculosis morbidity and mortality observed across populations.21 For this reason, it is remarkable how little attention Andvord paid to the social environment for the evolution of the disease through time and its’ natural history in individuals, although he recognized poor people's suffering and the potentially contributing factors of ‘spiritual elasticity for the life energy’ grounded in climate and living conditions in his early works.13,22 But in his last paper, published post-mortem in 1935 (p. 607), the emphasis was much clearer: If we want to make up our opinion about the deeper regularities which seem to be governing the course of tuberculosis, it seems to me as if nature aims to increase the body's defensiveness – its resistance for the microbe, partly inborn and partly acquired through the life-course by benign processes … As second main explanation I would put the increasingly more effective prophylaxis, particularly for the children. Hygiene, therapy and the social conditions of course have some influence on the mortality – but the effects appear NB! according to my generation tables – to be of limited temporary importance, and they do not seem to any substantial degree to shift the ‘regularity’ of the generation curve. These temporary deviations might be of interest, as they point at ‘extraordinary condition’ which could have been of importance in ameliorating or deteriorating direction.23 In the author's opinion the scientific experience of the later years are in support of my opinion, also maintained by me in earlier years, that tuberculosis has turned out an infectious disease decidedly belonging to the family and the generation, becoming constantly more benign, in proportion as the resistance of the generation seem to increase as a result of a certain adaptation or, as it may be expressed – a probably inherent immunity,26 (p. 495). The Norwegian population was heavily struck by tuberculosis. Reasonably reliable statistics show that adult mortality overall in the country increased from late 1860s reaching a peak around 1900 and then declined substantially.29,30 As Newsholme noted in 1908 Norway, Ireland and France were exceptions to the general decline that had started in many other countries half a century earlier.31 In 1885 mortality was highest in the most southern counties of Norway. The increase and subsequent decrease in mortality took place in different counties in different periods. There was an increase until 1906–10 in Sør-Trøndelag and Finnmark, until 1911–15 in Sogn og Fjordane and until 1916–20 in Nordland. Even if Andvord's scientific findings and theories were controversial, he steadily gained support.32 Already in the 1890s, he became influential as parts of his statistical evidence were used as foundation for the new tuberculosis act of 1900. It was solely based on a bacteriological approach where isolation and disinfection were the primary tools for combating the disease.33 He was very active in the public debate on how to combat tuberculosis, and he inspired both private and public actors in setting up tuberculosis sanatoriums. Earlier on in his career he introduced mountain climate as medication.10 However, as Hans Jacob Ustvedt noted in 1955, Anvord never became an ‘authority’ or a ‘pope’, but his independency might also ultimately have been his advantage.34 He worked long hours as a family physician during day time and on his research in the evening. Without any influential position in the academic field, his effects on policy had to be somewhat limited. Andvord's theory on the importance of tuberculosis transmission in childhood gained increasing recognition in 1910–15.34 The theory was also supported by another colleague, Dr Frølich, who conducted a school health survey in 1913 among 3000 school children in Kristiania (now Oslo). He was able to show that 80% of the pupils at the age of 7 had a positive tuberculin test. Among working class pupils this was 90%. After 2 years follow-up this had increased to 85% in the whole sample. Transmission among the working class children had a more malign course than in other social classes. For this reason it came as a huge surprise when the two doctors Johannes Heimbeck and Olaf Scheel in the 1920s documented that primary infection in childhood might not be as important after all.35 They investigated nursing school students in their 20s at Ullevål hospital in Oslo, and found that 46% were probably carriers. After finishing their training 100% of these students were Pirquet positive. The students were mostly recruited from the middle-class. The conception of tuberculosis changed during the inter-war period. By the beginning of the 1930s when social and radical movements became stronger, the concept of disease aetiology became increasingly ‘socialized’. In Norway as elsewhere, influential physicians underlined the importance of disadvantaged economic and social factors on tuberculosis mortality; crowded houses, poor sanitary conditions, inadequate nutrition. ‘The social structure alone’, an anonymous commentator wrote in 1931, ‘is in reality decisive for tuberculosis as a mass disease … Bacteriological tools for prevention are powerless alone when confronted with these laws’. But medical doctors, including socialist doctors, were also inspired by more genetic explanations and eugenics in this period.36,37 In 1934 gipsies, alcoholics, criminals, prostitutes and mentally disabled persons were included in the national programme for sterilization, which was passed by the Parliament (Storting) against one vote. Andvord was persevering in his lifelong quest for proving that primary infection took place in childhood, despite the new evidence provided by Heimbeck and Scheel. According to Ustvedt, Andvord's theory on the importance of childhood was probably at the of the 19th particularly in areas. But and geographical were as and tuberculosis to be at from point of as Ustvedt put because its much between of exposure to the and the living conditions, which is important for the conditions may between countries and between within a the disease may different in which time point of infection may have partly because of prevention of transmission and living Andvord's scientific is related to one of this (p. The development of epidemiology as a discipline has been into different These changes took place in Norwegian medical as They did take place as of disease aetiology coincided with new from in the first part of 19th increasingly gained as an discipline in the sanitary from the to The within the was transmission of disease through and under the influence of and were After of the in the year Andvord medical became important to study transmission rather than of and geographical This shift a in from causes at Kristian Andvord this new epidemiology and the bacteriological of disease transmission and prevention. In his papers from the 1930s he referred to the of nature governing the course of This was a rather and way of by the explanations at that and which had for public health Andvord also to a Norwegian A common for population health research in Norway until has been geographic differences in health rather than and took place rather late in Norway. with the different parts of the country across a of climate and conditions, geographical have been as most In the the of Norwegian published a number of on different such as and In an on mortality in Norway, published in he proposed a on of population health and included geographical differences and climate as important potential He introduced conceptions of regularities in of much inspired by natural had impact on the of statistics in Norway. When was in the first was inspired by as Andvord have But among medical Andvord's use of population data and statistical was as and not until late in Andvord's The impact of Andvord's work a Norwegian audience was to be limited given that the 1930 publication was published in It was also in a few into Frost this and Andvord for of mortality based on which might suggest he was more in of at the generation cohort approach in a than It has been speculated that Frost had at a similar idea of Andvord based on with used a similar approach in 1934 at mortality that health of the is by the which the has and the of They concluded that was not with birth cohort as mortality in the age groups subsequent to infancy, through the life the generation by It is to that Andvord gave another of the generation cohort A similar approach was later up in of in by and by seminal on and other More of the importance of childhood on the risk of disease in used geographic between Norwegian counties with This has been followed by a of research into early life on adult diseases that on the gained from the generation cohort Andvord's period of as a and between different and of from of the in his early the of the and with the and conceptions between the His work is not in any of these in the that they periods. As it appears inspired Andvord in the first place was the of the disease in the that factors and of the could explain the But by the end of the century he put less emphasis on this theory and gradually a theory on the of acquired and inborn as tuberculosis mortality in Norway or during the early of the 20th he was able to at different and take of a generation cohort to his idea of this he moved from a geographic approach and up with comparing different birth cohorts. But the research the same although he appeared to have been inspired by different causal different through this period. Even though the decline of the disease had not in the 1890s in with most Andvord was able to and make inferences on how risk was by and time based on geographic within and between At population level such geographic may be as of how social and conditions have population risk in various and in various periods of history even if the may be context on his to the scientific Andvord his were All through his career he was as a in Norwegian tuberculosis His were presented at the Norwegian medical He was and from academic such as when they the of but late in his career he recognition and was at his by at the age of he presented the generation approach and was as But the method also some as Andvord had with his theory on age dependent susceptibility and gave the generation tables a somewhat Andvord's was in several He never to the of Norwegian as he academic research hospital a at the The approach he developed has to be a But the theories he from this was on the of the time he within as when he proposed factors for the geographic or genetic explanations for the decline in Andvord there was no inherent within this of