Pattern of declining blood pressure across replicate population surveys of the WHO MONICA project, mid-1980s to mid-1990s, and the role of medication

Blood pressure measurement

Investigators followed standard protocols.3 8 In every population, with ethical approval, they targeted 200 or more participants from each 10 year age and sex group, from age 35 to 64, for a random sample survey at the beginning of the MONICA decade, replicated independently at its end.1 2 Blood pressure was measured in the right arm by using a random zero or traditional mercury sphygmomanometer after seating the participant for five minutes. Systolic and diastolic blood pressures (phase 5) were recorded twice to the nearest 2 mm Hg and the averages obtained.8 Data were forwarded to the MONICA data centre in Helsinki for analysis and quality assessment.3 8 The original database is available to MONICA investigators; detailed tabulations are in the public domain.5

Statistical methods

We extracted published information on numbers of participants and response rates,9 use of antihypertensive medication,5 scores assigned in the quality assessment of the blood pressure data,3 and summary statistics on numbers, blood pressure means, standard deviations, and previously defined low, middle, and high readings to be used for tracking change—the 20th, 50th and 80th centiles.5 Results for the age range 35-64 were standardised to the world standard population.1 4 10 We added selected analyses from the MONICA database.

Subtraction of initial from final survey values means that a negative difference indicates a fall. We used differences obtained between surveys in individual populations to produce simple mean pooled values. The number of populations, generally 38 or 19, was used to calculate the standard errors of these pooled means.

Our hypothesis was that hypertension control would selectively depress the top end of the population bell curve, resulting in pooled results showing differential falls in the 80th centile compared with the 20th centile; the mean compared with the 50th centile; and the mean compared with the average change in these three centiles, so we obtained these differences by subtraction. Null hypotheses of no treatment (but a mass population) effect would be disproved by significant values for these differences. In order to test the robustness of the results obtained from pooling all 38 populations together, we then compared the findings in contrasting subgroupings of these populations.

Validity of analyses

We are confident that our analyses are valid14 and alternatives, analysing those on and off medication separately are not.15 The WHO MONICA project is unique in the range and quality of its data, and in publishing detailed quality assessments along with its data books.1^–5 There are no equivalent databases in which to replicate our analyses. However, by basing our main analyses on published summary statistics5 and publishing our spreadsheets (see bmj.com) we have left a transparent audit trail by which others, if they wish, can rework them. We gave equal weight to all populations in pooling results. Our tests of robustness (table 5), imply that different weightings would produce similar answers.

Unfortunately for public health, we are unable to say what caused the blood pressure decline in MONICA. Other repeat surveys have found population blood pressures falling. Detailed discussion is beyond the scope of this paper, although doubtless claims will now be made for different nostrums. MONICA cannot give answers outside what it measured. Obesity, a known contributor, moved in the wrong direction.1 Possibilities, apart from measurement errors,3 16 include cohort effects,17 dietary factors (including consumption of alcohol, sodium, potassium and other electrolytes), and other lifestyle and environmental factors.18

Use of medication

During the MONICA study, the use of antihypertensive medication in MONICA populations increased by a rounded average of 0.5%, from 10.8% to 11.4% (see tables in bmj.com), but drugs in use changed. For 11.4% medication to be the exclusive cause of an average population-wide fall of 2.26 mm Hg between surveys, the blood pressure fall of people taking medication would need to be an improbable 19.8 mm Hg. Individual data show that medication was concentrated among high readings: 25.5% usage above the 80th centile and 3.4% below the 20th.

Change in medication across populations and change in blood pressure were not correlated. In 21 market economy populations where blood pressure declined, use increased in 11 and decreased in 10, implying no simple association. Better blood pressure control in people taking medication could result from underlying population trends in blood pressure—the goal posts remaining stationary but the ground moving under them. This possibility is welcome, but potentially embarrassing for those publishing and interpreting analyses of repeated surveys studying the “rule of halves,” the percentage of people with hypertension in the population identified, treated and controlled.19^–21 In populations where blood pressure declined overall, with all else unchanged, the resulting improved percentages might be credited mistakenly to clinical vigilance and improved medication.

Conclusions

Our current findings, questioning the contribution of treatment, contrast with earlier MONICA collaborative results that showed unexpectedly strong links between population trends in medication for coronary care, and declining case fatality and mortality.22 This paradox results from our dispassionate analyses of standardised data across 38 different populations. But MONICA can report only on its own decade, now over. Medication may contribute more than spontaneous population change after the mid-1990s, although similar multicentre databases for analysis may not be available to show it. Data from the mid-1980s to mid-1990s from the WHO MONICA project may still conceal surprises in other areas.