Background: Armitage recollections (2024)

Abstract

1 Introduction

In a companion article (Armitage, 2024), Peter Armitage presents comments on the early history of the Medical Research Council (MRC) Statistical Research Unit along with personal recollections of his time working in the Unit. This material was originally presented in 2003 at the time of the inaugural Armitage Lecture which is an annual event in the MRC Biostatistics Unit. The MRC Statistical Research Unit was the direct descendant of the original MRC Statistical Department which was later renamed the MRC Biostatistics Unit. The publication of these remarks helps to mark the 20th anniversary of the event.

For the purposes of publication, after consultation with Peter Armitage, I added various references to the original 2003 typescript of his remarks. In this note, I provide some more specific background to these remarks focusing on the early history of the MRC Statistical Research Unit and on Peter Armitage himself, primarily, but not exclusively, with reference to his time in the Unit.

2 Early history of the MRC statistical research unit

2.1 Political background

Prime Minister Herbert Henry Asquith’s Liberal Government, 1908–1916, passed, in 1911, the National Insurance Act. This act was shepherded through Parliament by David Lloyd George, the Chancellor of the Exchequer, and Winston Churchill who was initially President of the Board of Trade and subsequently Home Secretary during this time. Churchill told the House of Commons that ‘There is no proposal in the field of politics that I care about more than this great insurance scheme’. The act was a component of the Asquith government’s social reforms and was enabled by the ‘People’s Budget’ of 1909 which introduced taxes on the wealthy to allow the creation of new social welfare programs.

The provisions of Part I of the Insurance Act specified that every worker earning over £160 per year had to pay an insurance tax of four pence per week. To this, the employer was required to add three pence and the state added two pence. Lloyd George advocated for the Act using the phrase ‘nine-pence for four-pence’. These monies provided sick leave benefits of ten shillings per week for thirteen weeks and five shillings per week for the next 13 weeks. Also provided was free treatment for tuberculosis with eligibility for treatment by a ‘panel doctor’. Similar provisions for unemployment insurance were introduced in Part II of the Act.

Sir Robert Morant chaired the implementation of the National Insurance Act and, as part of this, established a Medical Research Fund based on setting aside one penny for each insured person in the UK. To administer this fund, the Medical Research Committee was formed, which subsequently, in 1920, was granted a Royal Charter and became the MRC.

2.2 The scheme of research

As highlighted by Landsborough Thomson in his history of the first 50 years of the MRC (Thomson, 1973), the research focus of the Medical Research Committee was very broad. This is indicated by its first ‘Scheme of Research’, submitted in November 1913 for ministerial approval prior to its implementation. This specified the ‘Type of Research’, which was specified very broadly as ‘the extension of medical knowledge’ with the actual field of research being not limited in any way, and a ‘Method of Conducting the Research’. The latter provided for four ‘departments’, the descriptions of which can be summarized as:

1. Investigators of highest class in permanent employ of scheme, devoting whole time to research.

2. Skilled investigators in permanent or temporary employ of scheme engaged in procuring material clinically or otherwise.

3. Individual investigators not in employ of scheme helped with money or otherwise for researches coordinated with research under the scheme.

4. Statistical Department: mainly consist of persons in the permanent employment of the scheme—statistical investigations useful either as a preliminary to research or confirmatory of its results.

Notably, therefore, from the very beginning the MRC was to have a Statistical Department.

2.3 Early directors

John Brownlee (V. T. Farewell & Johnson, 2014) had studied mathematics and natural philosophy at Glasgow University, graduating with first class honours in 1889. He then went on to qualify in medicine in 1894, subsequently receiving an MD in 1897 with a thesis on scarlet fever. The following year, he sat examinations and received a Diploma in Public Health from the University of Cambridge. These later education qualifications reflected his employment in the fever hospitals of Glasgow and 18 months as a Medical Officer of Health for Guernsey. From 1904 onwards, Brownlee’s publications became increasingly quantitative, facilitated no doubt by his initial training in mathematics, and made use of statistical methods associated with Karl Pearson. It does not appear that Brownlee ever attended lectures by Pearson but his interest in ‘Pearsonian methods’ led ultimately to eight publications in statistical journals.

In 1914, Brownlee was therefore well positioned to be appointed as the first Director of the newly formed MRC Statistical Department. He clearly appreciated the potential of statistical methods to contribute to medical research. While Brownlee continued to publish research papers as Director, the work of the Statistical Department was largely given over to the needs of the War Office, particularly to handling the administrative work needed to compile records about the sick and wounded. An initial staff of 100 persons employed for this purpose by the Statistical Department grew to 225 by 1919. Responsibilities in this area continued until February 2021 with the records being used to answer questions about individual cases at the request of various government departments. Notably, half of the inquiries related to the claims of soldiers which could not be supported by their military records but 90 percent of these could be by the records of the Statistical Department.

As indicated by Peter Armitage, and consistent with discussion of this point in V. T. Farewell and Johnson (2014), Brownlee was not a good communicator. Major Greenwood (V. Farewell & Johnson, 2014, 2016a, 2016b), ‘the key that unlocked Brownlee’s mind’, noted that ‘if one were in a hurry to get from him specific information upon a matter of fact—say, the method by which he had fitted a curve—to find that, after 10 minutes’ conversation, one had learned a great deal about Highland place names, the organization of the Kirk in the 18th century and the history of alchemy, but nothing at all about the curve, might be irritating’.

Because of Brownlee’s lack of communication skills, it is not surprising that the MRC would be looking for alternative input on statistical matters. Sir Walter Fletcher, a noted biochemist, was appointed First Secretary, i.e. Head, of the Medical Research Committee in 1914, holding the post for 20 years. During WW I, Fletcher worked with Major Greenwood, who had studied medicine, qualifying in 1904, but who had quickly turned to research at the London Hospital. Greenwood, while initially funded to work in physiology, pursued statistical studies with Karl Pearson in 1904–1905 and increasingly published papers with statistical objectives and methods reflecting his studies with Pearson. In 1908, he was appointed as the first Director of the Statistical Department at the London Hospital, likely the first department of medical statistics in the modern sense. It was from this post that he moved to be the resident statistician in the Lister Institute in 1911. After war work, Greenwood became a senior statistical officer at the Ministry of Health, although working at the National Institute of Medical Research. It was thus natural for Fletcher to turn to Greenwood, after Pearson declined the post, to chair the newly formed Industrial Health Statistics Committee in 1921 which in 1924 was transformed to the more general MRC Statistical Committee.

Brownlee was a member of the MRC Statistical Committee but it was nevertheless somewhat problematic for the MRC to have two foci of statistical activity. While Brownlee and Greenwood worked alongside each other in their respective roles, and Greenwood was appreciative of Brownlee’s abilities, the situation naturally resolved itself when Brownlee died in 1927 and the activities of the two groups were combined under Greenwood’s leadership. Shortly after Greenwood was appointed to the first Chair of Epidemiology and Vital Statistics at the London School of Hygiene and Tropical Medicine (LSHTM), the MRC Statistical Department was moved to LSHTM as part of Greenwood’s department.

Greenwood remained as Director of the MRC Statistical Department until his retirement in 1946. Greenwood carried a heavy administrative burden while at LSHTM, in terms of the School, the MRC and more broadly, for example as President of the Royal Statistical Society (RSS) in 1934. Greenwood’s publications were varied in terms of medical topics but largely applied in their use of statistical methods. He also began early to write biographical material on medical workers. Nevertheless, his name is still seen today in reference to the Greenwood formula for the variance of the Kaplan–Meier estimate of a survivorship function and, more surprisingly in terms of current application, in the ‘Greenwood statistic’, a measure of clustering of events in time and space. This latter statistic was used in a 2007 publication (Riley et al., 2007) to show that there is an importance to the order where genes are placed on a chromosome, particularly in relation to function. His last major publication while Director of the MRC Statistical Department, published in the 1946 volume of the Journal of the Royal Statistical Society, was on the statistical study of infectious diseases.

While Greenwood was not a major contributor to statistical methods research, he was very aware of the need for researchers with good training and experience in mathematics and statistical methods. This is evidenced by the high regard in which he held Leon Isserlis when seeking, unsuccessfully to employ him at LSHTM (V. Farewell et al., 2006), and his successful hiring of J Oscar Irwin who had worked with both Pearson and Fisher. He was also a great encourager of his staff. It is not surprising that under Greenwood, two individuals, with limited mathematical or statistical training, both completed a DSc in Medical Statistics under his supervision and subsequently wrote what is arguably the first modern British textbook on medical statistics (V. Farewell & Johnson, 2012a; Woods & Russell, 1931). And, of course, Austin Bradford Hill also wrote what is perhaps the best known early book on medical statistics (V. Farewell & Johnson, 2012b; Hill, 1937) while working in Greenwood’s department.

When his wife, Rosa, died in 1945, it is reported that Greenwood lost interest in many aspects of life and while he maintained an office at LSHTM, it appears that he largely associated only with other retirees. His last paper was submitted on the day of his death, which occurred while he was attending a scientific meeting on cancer research.

When Greenwood retired, Austin Bradford Hill was appointed as Professor of Medical Statistics at LSHTM and honorary Director of the MRC Statistical Department which was renamed as the MRC Statistical Research Unit. It was to this Unit that Peter Armitage was recruited in 1947.

3 Peter Armitage

Peter Armitage was born on June 15, 1924 in Huddersfield, Yorkshire. His father, Harry, was a well-regarded headmaster who had studied mathematics at Leeds University. His father had wide-ranging interests and his mother Florence shared many of these, especially music. Growing up in Huddersfield, Peter himself developed wide-ranging interests including sports, English Literature and Language, and music.

Peter had an innovative and inspiring maths teacher, Leslie Horsfall, who had previously worked as a statistician in the cotton industry. In his senior school years, studying for the Higher School Certificate, Peter focused on mathematics and physics. In December 1941, he wrote a Cambridge scholarship exam and was offered a place at Cambridge. In general, scholarship holders would wait until the following September to begin their studies but, because this was war-time and, perhaps, because of an increased sense of urgency given the December, 1941 attack on Pearl Harbour, Peter was encouraged to begin his studies immediately. He therefore left home for Cambridge in January 1941 where he studied for five terms before being recruited for war work after he sat his second year exams, the so-called Preliminary Examination, in 1943.

Peter had taken an uninspiring statistics course in his first year. However, C.P. (later Lord) Snow, visited Cambridge to conduct interviews and advised Peter to take a more advanced course from J. O. Irwin who was temporarily working in Cambridge. This course, although largely theoretical, provided the basis for Peter’s move to work with the Ministry of Supply in London in August 1943.

3.1 Statistical Advisory Service

Peter was employed in the Statistical Advisory Service (SR17) of the Ministry of Supply and was, to his delight, placed in the research section with about half a dozen colleagues. The Section was led by George Barnard who had pre-war industrial experience. He was aided in leading the group by Robin Plackett. Joining at the same time as Peter, also from Cambridge, was Dennis Lindley. Under the guidance of George Barnard, Peter’s work focused on the topic of sequential analysis based on the thinking that an efficient sample inspection scheme might proceed in stages, with an early decision to accept or reject if the sample results were sufficiently good or bad. Work on this topic was aided by receiving, from the USA, classified reports in which Abraham Wald described a general theory of sequential analysis.

At the end of the war, SR17 staff either left for normal jobs or were moved to other sites. Peter moved to a new statistics group at the National Physical Laboratory (NPL) in Teddington, along with Robin Plackett and Dennis Lindley. However, later, in 1946, he returned to Cambridge for a year to finish his degree. During this last year in Cambridge, Peter was able to take statistics courses taught by Maurice Bartlett and Harold Jeffreys. There was no shortage of contact with subsequently leading statisticians in Peter Armitage’s university studies and war-time work!

3.2 To the MRC

After completing his degree, Peter was offered and accepted a regular post at the NPL. However, Peter’s boss at the NPL was Edgar Fieller. Fieller had a chance encounter, at a suburban rail station, with the epidemiologist, Donald Reid, who worked at the LSHTM. Reid indicated that his Professor, Austin Bradford Hill, was looking for a young mathematical statistician and Fieller identified Peter as such. An invitation to interview followed, and Peter Armitage moved to the MRC Statistical Research Unit in December 1947. He was appointed to work with J. Oscar Irwin with whom he worked on problems in bioassay, completing a PhD related to the dynamics of a bacterial population subject to mutation. As Peter has indicated (Armitage, 2024), he was also encouraged by Hill to get involved in other research including sequential methods for clinical trials, building on his war-time work in SR17.

3.3 Peter Armitage and medical statistics

Peter Armitage’s move to the MRC Statistical Unit marked the beginning of his remarkable career in medical statistics. In 1990, The Editors of the journal Statistics in Medicine, wrote in an editorial (McPherson et al., 1990) for a special issue prepared in honour of Peter Armitage, ‘Anyone who delves in earnest into medical statistics will, almost invariably, encounter some facet of the work of Peter Armitage. His areas of interest are broad and cover nearly every aspect of statistics applied to medicine’.

Building on his war-time work on sequential analysis, Peter Armitage provided the basis for its application to medical research in a series of papers, the work forming the basis for the two editions of his book Sequential Medical Trials in 1960 and 1975. His wide-ranging publications on other topics included, for example, two papers on multistage models of carcinogenesis, published with Richard Doll in 1954 and 1957 (Armitage & Doll, 1954, 1957). Nick Day (1990) wrote that ‘The multistage model of carcinogenesis proposed by Armitage and Doll …has provided a conceptual framework of great power for unifying the diverse observations, epidemiological and experimental, on the evolution over time of risk for cancer’. Another illustrative example is his work on cross-over trials leading to the publication of two papers (Armitage & Hills, 1982; Hills & Armitage, 1979) which provided a carefully balanced assessment and investigation of this trial methodology.

Peter Armitage’s teaching at LSHTM formed the basis for the first edition of his book Statistical Methods in Medical Research. Later editions were published in 1987 and 1994 with Berry and in 2002 with Berry and Matthews. This book ‘displays his lucid style of communication and has become a classic …’. (McPherson et al., 1990).

When Austin Bradford Hill retired in 1961, Peter Armitage was appointed Professor of Medical Statistics at LSHTM. He remained there until his appointment to the Chair in Biomathematics (later Applied Statistics) at the University of Oxford. Following his formal retirement from the Oxford post in 1990, Peter was invited to give a variety of conference talks and did some work related to trials in the pharmaceutical industry. He also undertook considerable work on Data Monitoring Committees, especially for AIDS-related trials, motivating two publications on the topic of such committees (Armitage, 1991, 1999).

Even more time consuming, was his acceptance in 1995 of the joint editorship of Wiley’s Encyclopedia of Biostatistics, with Ted Colton, the first edition, with seven volumes, being published in 1998. This work had over 800 contributors and 1,200 articles and a second, 8 volume edition, was published in 2004. At 80 years of age, Peter Armitage, again, retired from active work in medical statistics bringing to the close a remarkable professional career of which the ‘early’ 1990 tribute by the Editors of Statistics in Medicine had become even more appropriate.

Finally, it must be noted that Peter Armitage’s influence on the field of medical statistics has been made even greater, throughout his career, by the personal support he gave to younger researchers, those whom he employed as well as many others whom he supported in a variety of ways.

4 Conclusion

The early history of the MRC reveals foresight regarding the central role that statistical thinking should play in medical research. The Statistical Department established as part of the MRC’s initial research structure, and its successors, have demonstrated the continued importance of this role. Peter Armitage’s career is a prime example of the particular value of combining original methodological research in statistical methods and collaborative medical research, a combination celebrated in the MRC Biostatistics Unit’s establishment of the Armitage Lecture Series. It is very gratifying therefore that the Journal of the Royal Statistical Society, Series A is now publishing Peter Armitage’s recollections, initially presented 20 years ago, of his introduction into the field of medical statistics through his employment in the MRC Statistical Research Unit.

Funding

This work did not receive external funding.

Data availability

There are no new data associated with this work.

References

Author notes

Conflict of interest: None declared.

© The Royal Statistical Society 2024. All rights reserved. For commercial re-use, please contact reprints@oup.com for reprints and translation rights for reprints. All other permissions can be obtained through our RightsLink service via the Permissions link on the article page on our site—for further information please contact journals.permissions@oup.com.