Dr David Barlow is Consultant in Genitourinary Medicine at St Thomas’ and Guy’s Hospitals, London. He has been the lead author for the chapter on gonorrhoea in the last three editions of the Oxford Textbook of Medicine. Between 1986 and 1993, at St Thomas’, he ran the largest linked HIV sero-survey in the United Kingdom. The third edition of his book Sexually Transmitted Infections- The Facts, Oxford University Press, with original cartoons by the late Geoffrey Dickinson, was published in March 2011.
There is something slightly uncomfortable about authoring a book whose cover proclaims: “XXX – The Facts”, with a sub-heading “All the information you need, straight from the experts”. Such is the house style of the OUP for its medical ‘Facts’ series, currently some 35 strong, but going forth and multiplying as you read.
Anyway, it got me thinking about how, in my specialty, when ‘facts’ become ‘figures’, caution is called for. I had an interest in heterosexual transmission of HIV in the 1980s and 1990s which put me in conflict with the official number-crunchers and I’m afraid I’m still suspicious when presented with totals. At the final proof stage of my ‘Facts’ book, I checked the Health Protection Agency’s website for the numbers of UK STIs reported for 2008. Unmentionable diseases including syphilis, gonorrhoea, warts and herpes remained as I had written. Total chlamydial infections, however, had changed from 126,882 (accessed July 2010) to 217,570 (accessed January 2011). A small adjustment might be reasonable. But 70%? This was a DB Type 4 numerical error.
DB’s numerical errors: Types 1-5
• Type 1 Somebody has a vested interest: “If we tell these clap-doctors that laboratory culture of the gonococcus is only 70% sensitive, they’ll shut their lab’ and buy our ‘totally sensitive’ NAAT.”
• Type 2 The totals may be correct but are misleading (1): “It is Government/Department (of Health) policy to pretend that there is a rapidly increasing HIV epidemic in heterosexuals who are transmitting within the UK.”
• Type 3 The totals may be correct but are misleading (2): There is a genuine, probably innocent, misinterpretation of the figures (see horseradish sauce, below)
• Type 4 The totals may be correct but are misleading (3): The explanation is perfectly reasonable and logical, but the calculation is opaque/we are keeping it to ourselves/forgot to tell you/have you read the small print?
• Type 5 The totals are incorrect: Woops!
At the beginning of June, I awoke to BBC headlines about a doubling of UK-acquired HIV between 2001 and 2010. This drew me to the HPA’s website where I found a press release (June 6): ‘Last year there were 3,800 people diagnosed with HIV who acquired the infection in the UK, not aboard [sic], and this number has doubled over the past decade.’ From the same site: ‘… HIV diagnoses among heterosexuals who most likely acquired in the UK have risen in recent years from 210 in 1999 to 1,150 in 2010’. I shall return to these later but if you really have nothing better to do, why not see whether you can confirm the figures quoted above by accessing the HPA’s ‘New HIV Diagnosis,’ Table 5 here. And your next task (5 marks) is to re-word the press release…
Exactly thirty years ago, on 5th June 1981, the sleuths at the Centers for Disease Control published their crafty bit of epidemiology entitled ‘Pneumocystis pneumonia – Los Angeles’. The CDC had picked up an increase, from the West Coast, in requests for pentamidine. This was the drug used to treat PCP, a rare lung infection found in renal transplant patients whose immunity had been weakened (deliberately) to reduce rejection.
These new cases were different. The men were immuno-compromised but none were undergoing transplantation and all were gay. Thus were HIV and AIDS (although not so named for a year or two) introduced to an awe-struck, and soon fear-stricken, public.
Britain had its first AIDS case in 1981 and in August 1982 the Communicable Disease Surveillance Centre (the UK’s CDC) published the first of their monthly updates in the Communicable Disease Report, the CDR. The risk categories were divided into homosexual, haemophiliac, blood transfusion, intravenous drug users and heterosexuals [without other risk]. It was with these heterosexual cases that the distinction between ‘the truth’ and ‘the whole truth’ became lost during late 1986.
The May 1986 CDR tables broke down the heterosexual AIDS cases into: 3 with USA/Caribbean connection, 3 simply ‘heterosexual contact’ (of whom two “…had recently returned from Uganda and Mozambique.”), and 12 associated with sub-Saharan Africa. In October this connection became a footnote: “associated with sub-Saharan Africa” and by November, the categories had become: ‘contact UK’ and ‘contact abroad’. The December, separate, HIV figures were reported, without footnote, simply as ‘heterosexuals’ ( Type 2 numerical error ). Africa had disappeared from the tables.
By one of those coincidences loved by cynics and conspiracy theorists, the UK-wide leaflet drop about AIDS occurred in January 1987, the very next month, to be followed, in February, by the ’_Don’t die of ignorance_’ campaign. The national press then published increasingly doom-laden descriptions, largely unchallenged, of the burgeoning UK AIDS epidemic in heterosexuals.
What actually mattered was the number of cases being transmitted in Great Britain. Was the disease spreading? What was the risk from a bonk?
The change in wording of the heterosexual categories in the late 1980s allowed speculation that the ‘infected abroad’ category was largely made up of British nationals who had gone overseas and returned with HIV/AIDS. This was the CDR’s interpretation when they gave advice to travellers in 1991 ( Type 3 numerical error ).
We published an alternative view in the Lancet (CDR did not print correspondence, commentary or criticism) and the CDSC, unusually given the chance to reply in the same edition, graciously and politely acknowledged our figures from St Thomas’ but said that they were not representative. Neither my first nor last experience as an outlier.
Have you ever made horseradish sauce? Epidemiologists and cookery-writers run similar risks. Counting and cooking need to be in their respective repertoires but, for both, the craft improves with hands-on experience: contact with patients, or trying out the recipe. If your cookbook doesn’t mention wearing goggles with the wind behind you while you grate this vicious root (and most don’t), the author has never made the sauce. Epidemiologists don’t need the formula for horseradish peroxidase either, but they may miss an open goal if they don’t see patients.
Four other hospitals in or near London (I confess to prompting) reported that most of their (no other risk) HIV-positive heterosexuals were, like ours, from Africa, (Outliers 5, Regression Lines 0). It was not until later in the 1990s that the CDSC accepted the UK heterosexual HIV/AIDS epidemic to be largely imported, with little evidence of significant transmission between heterosexuals from, or in, this country.
So, how did you get on with the HPA’s table 5? You found the 210 for 1999 easily enough, I’m sure. But the 1,150 (and 3,800) for 2010? Well, a helpful person in the HPA’s epidemiology section told me they reached this figure by extrapolating the, as yet, uncategorized (‘not reported’ – penultimate row Table 5) cases in the same proportion as the different categories where the region of infection was actually known ( Type 4 ).
“But you didn’t apply that correction to the 210 in 1999”.
“Ah, no. We didn’t!” ( Type 5 ).
And, finally, the Type 1 numerical error? Specificity is also important in diagnostic tests (the 55 year-old Granny who went to her GP for a smear test, was screened for chlamydia, and came out with gonorrhoea. Yes truly!). The Nucleic Acid Amplification Tests for gonorrhoea may give you false positives.
Why didn’t you tell me about this before, Mother?
So, am I advising less sex?
What, and put myself out of a job? Give over!
Barlow D (2004) HIV/AIDS in ethnic minorities in the United Kingdom.
In Ethnicity and HIV: prevention and care in Europe and the USA, Eds, Erwin, Smith and Peters. 21-46
Barlow D, Daker-White G and Band B (1997) Assortative mixing in a heterosexual clinic population – a limiting factor in HIV spread? AIDS; 11:1039-44