Does Hypertension Cause Cardiovascular Disease?

And, does “Mendelian Randomization” help answer this question?

It is not uncommon for press reports of newly published studies in medicine, the social sciences, or economics, to describe a finding of a correlation between two variables as causative, that is, one variable causes the other. Often, the authors of the study (or the journal editors) do nothing to correct this logical fallacy, perhaps because causation is more sensational than correlation, and gets more clicks and more reads.

A recent example (2023-6-23) is a news article directed to health care professionals, entitled “Alcohol reduces rheumatoid arthritis symptoms”. The headline suggests that drinking alcohol will improve the illness. The authors of the referenced research (Alfredsson et al 2023¹), case-control study which found an association, muddied the waters with “…favorable effects of alcohol consumption in RA…” implying causality. ²

The gold standard for inferring causation (causality) in an observed relationship is a randomized controlled trial (double-blinded, where feasible). Unfortunately, doing RCTs is often not possible, or would be too time-consuming or expensive, or poses ethical problems. In these situations, other means of inferring causation might be tried.

For infectious diseases, Robert Koch in 1890 developed a set of four postulates although it was subsequently found that they could not be applied for viral pathogens or asymptomatic carrier states. For conditions other than infections, the Bradford Hill criteria, described in 1965, can be useful if correctly applied.

More recently, a statistical technique called instrumental variable estimation (IV) is being used to infer causation. This requires choosing an additional variable, often called the “instrument” which satisfies certain requirements. The advent of relatively inexpensive gene sequencing and the proliferation of genome-wide association studies (GWAS) has led to researchers choosing instrument variables (typically single nucleotide polymorphisms or SNPs) where germline genetic variation can be applied to estimate a causal effect of a given exposure on the disease being studied. This technique, termed “Mendelian randomization” (MR) has additional conditions that need to be met, including that of random mating (ie, choosing one’s mate regardless of any physical, social, or genetic preference).

In the past couple of decades, there has been an explosion of MR studies, especially ones using large numbers of genetic variants. Because some of these variants are less than ideally suitable, a number of algorithmic modifications to the basic MR approach are being tried, leading to increasing confusion for consumers of research literature, and perhaps for the researchers also. For example, a study exploring whether nonalcoholic fatty liver disease is causally related to pancreatic cancer (King et al 2023³) applied four different MR methods; another on allergic diseases and cardiovascular disorders (Wang et al 2023⁴) compared six MR methods.

Overall, though, “the eagerness with which MR has been adopted has, in some cases, outstripped caution about its use” (Mukamal et al 2020⁵). These authors suggest that it be treated just like any other form of observational epidemiology, and “urge the elimination of randomization or causality in reports of its use”.

A common condition, hypertension (HT), has been recognized since the 1960s or so as having a strong correlation with cardiovascular disease (CVD) and death. Indeed, since the risks increased in a graded fashion with higher blood pressure, and since an early study (Freis 1974⁶) demonstrated a marked reduction in stroke incidence and death in men treated with 3 antihypertensive medications (hydrochlorothiazide plus reserpine plus hydralazine), it was suggested that there is a causal relationship. Hypertension has come to be known as the “silent killer”, clearly indicating a belief that it causes CVD.

In the case of hemorrhagic stroke, this is likely true. And there is a plausible physiologic mechanism. Even as early as 1868 it was suspected that the cerebral micro-aneurysms found in hemorrhagic stroke patients were caused by sustained hypertension and animal models in rats and rabbits have confirmed this (Kido et al 1978⁷; Lee & Berry 1978⁸). The study by Freis reported on in 1974 emphasized that “effective long-term control of hypertension markedly reduces the incidence of stroke in hypertensive patients, particularly with respect to hemorrhagic stroke.”

What about other cardiovascular diseases, such as atherosclerosis? Here, we do not seem to have good explanatory models of how HT might increase the formation of atheromatous lesions in blood vessels. On the other hand, it is quite easy to explain how something that restricts blood flow to the kidney or to the brain, such as an atheromatous plaque, will result in elevations of blood pressure, in the system’s attempt to maintain adequate blood flow to those vital organs. Indeed, animal models of HT are typically based on restricting blood circulation to a kidney (“Goldblatt hypertension”; see, eg Morgan 2003⁹), demonstrating a reverse causation.

In spite of this, the research literature overwhelmingly concludes that HT causes CVD, given the many clinical trials of a variety of medications that lower blood pressure which also reduce CVD outcomes such as stroke, angina, cardiac infarcts, and mortality. And there is plenty of evidence that systolic blood pressure (SBP), diastolic blood pressure (DBP), and pulse pressure (ie, SBP – DBP) are correlated with CVD, including coronary disease and stroke; these correlations remain strong even after accounting for the effects of age, gender, and other factors such as diet and exercise.

What if the improvements in CVD events and mortality induced by medications were due to medication effects other than lowering blood pressure (medication pleiotropy)? While searching for specific effects of these drugs that play a role in reducing CVD is complicated by the fact that several different classes of drugs are involved¹⁰, an often overlooked issue is that all drugs have effects, both desired and adverse, which stress the organism, and stress in general can be beneficial by increasing healthy longevity, in addition to the specific beneficial effects of certain stresses. For example, moderate amounts of exercise are healthy, while very little or no exercise, as well as excessive exercise, can be harmful. For further examples, see my essay ”Don’t stress about stress!”.

Physiologic processes often manifest a triphasic hormetic response, ie a U or J-shaped dose-response relationship. This can play havoc with traditional statistical methods, leading to searches for better algorithms (Cedergreen 2005¹¹) to tease out hormetic processes. Blood pressure is one such hormetic process, and it is likely that blood pressure lowering medications have hormetic effects also. This may complicate the process for inferring causation in the relationship between HT and CVD, and may help to explain why attempts continue to be made, including the use of MR. While some of these studies conclude that there is support for inferred causation (eg, Ehret et al 2011¹²; Wan et al 2021¹³; Clarke et al 2023¹⁴) other authors challenge these claims (Bowden et al 2015¹⁵) based on methodologic issues; even while obtaining results against causation (using their own methodology), however, they continue to express the belief that blood pressure is causally related to coronary heart disease risk, calling it a well established notion, and referring to a meta-analysis (Law et al 2009¹⁶) which they describe as “definitive analyses of the randomized trial evidence on the effectiveness of blood pressure-lowering treatments”.

Unfortunately, though, the Law et al meta-analysis, while including hundreds of drug trials, had a number of limitations. For example, the reduction in blood pressure had not been recorded in most trials for people with a history of CHD, and so the average reduction in BP had to be estimated. Another was that effect sizes were “standardized” to reductions of SBP by 10 mm Hg and DBP by 5 mm Hg, while the median reduction actually reported in 27 trials was only 6 and 3 mm Hg, respectively. Outcomes were recorded regardless of whether participants took their allocated tablets. Changes in BP were averages for the treated group and for the control group, and similarly, CVD outcomes were comparisons between treatment and control groups.

Is this adequate when attempting to assess the effect of blood pressure reduction on CVD outcome? When comparing treatment to control group, one is essentially comparing the effect of taking the medication, and not necessarily the effect of the blood pressure lowering. If blood pressure lowering actually had a causative effect on reducing CVD risk, shouldn’t the study show that, at least within the treatment group, greater BP reductions had better CVD outcomes (either based on a single cutoff, or quartiles, or linear regression)? I personally have looked at dozens of these drug trials, and I have so far not seen one that reported outcomes in this way. I suspect it’s not reported because the actual data do not support the hypothesis, and possibly the trial sponsors (almost always pharmaceutical companies) would prefer that you not know this.

Interestingly, the meta-analysis authors suggest that any BP reduction will be beneficial, whatever the original blood pressure, and that therefore “there is then little or no gain in measuring a person’s blood pressure—a conclusion that will undoubtedly stimulate discussion since it is at variance with a 100 years of medical practice”. Of course, one would arrive at the same conclusion, that there is little need to measure BP if the improvement in CVD outcomes were due to medication effects unrelated to BP lowering!

Also interesting is the finding that, while “data from before the 2000s indicate that the majority of incident cardiovascular disease (CVD) events occur among US adults with systolic and diastolic blood pressure (SBP/DBP) ≥140/90 mmHg”, in the modern era, the majority of incident CVD events occur in those with blood pressures below those cutoffs (Tajeu et al 2017¹⁷). Their study was a pooled analysis of 3 US cohort studies, involving almost 32,000 study participants without prior CVD. Overall, 78.3% had BP < 140/90 mm Hg; this group had new onset CVD at a rate of 8.0 per 1000. For the 21.7% with BP ≥ 140/90 the rate was 18.1 per 1000 individuals. Surprising, to me at least, was the finding that the 38% of the lower BP group who were taking antihypertensive medications had more than double the rate of new onset CVD compared to those not on meds! There was also a trend for those in the higher BP group taking meds to have a higher rate of new onset CVD, although this trend was significant only for females (for any CVD outcome); for strokes, nonsmokers taking meds had a significantly higher rate than those not on meds, while for coronary heart disease, significance was only found for those with diabetes. Could we take this to mean that the medications caused the CVD?

A parting word on how the insistence on a causative role for blood pressure reduction in reducing CVD (an insistence which certainly leads to profits for the pharmaceutical industry when they can convince physicians to subscribe to this idea) can lead to harm. In my own clinical experience with elderly patients (who almost always have been prescribed antihypertensives) suggests that the blood pressure lowering from these medications causes cognitive impairment. Consider, for example, orthostatic hypotension, that is, a drop in blood pressure caused by sitting up from lying down, or standing up. People experiencing this side effect of antihypertensive medication have dizziness, faintness, and may even lose consciousness and fall. Losing consciousness is, of course, on the extreme end of a spectrum of cognitive impairment. But lesser impairments may be simply ascribed to aging or to dementia. Unfortunately, the parts of the brain affected most by low blood pressure are the frontal and temporal lobes, leading to what is called frontal lobe syndrome whose symptoms include disinhibition, anger, excitement, apathy, depression, and difficulty in understanding others’ points of view. And of course, the person experiencing brain dysfunction has little insight, because when your brain is not working it won’t be able to tell you that it’s not working!

Many outpatients do not take their medication as prescribed, either forgetting or because of side effects or lack of efficacy. If their doctor is unaware, he or she may simply up the dosage or prescribe additional antihypertensives when blood pressure targets are not met. If the patient is now admitted for any reason, or even moves to an assisted living facility, they will be given their prescribed medication by staff, without consideration of what they were actually taking at home. This can result in a huge and sudden increase in side effects, including cognitive impairment and even falling. I learned in medical school that around 25% of seniors who fall and break a hip never walk again; of this group, half will be dead within 2 years.

In my work providing medicolegal expertise to lawyers in cases involving competence, I have several times encountered situations like the above; when the person’s doctor followed my recommendation to reduce or discontinue the person’s antihypertensive medication, their cognitive impairment improved or disappeared!

In conclusion, I believe that hypertension has not been demonstrated to be a cause of cardiovascular disease or mortality, except for hemorrhagic stroke; the beneficial effects on health and mortality from taking antihypertensive medication is partially due to pleiotropic effects; the promotion of hypertension as a “silent killer” is excessive but profitable for the pharmaceutical industry, and finally, overprescribing of antihypertensives especially in seniors can have serious and even tragic consequences.

On reflection, then, perhaps hypertension is a “silent killer”: it kills when it results in overprescription to seniors, and in the scenario above, falling, hip fracture, immobility, and death; silent when the players don’t talk about it either because they are unaware or, like in the tobacco or fossil fuel industries, the lure of profits overcomes ethical concerns.

1. Alfredsson L, Klareskog L, Hedström AK. Disease Activity and Health-Related Quality of Life Among Patients With Rheumatoid Arthritis With Different Alcohol Consumption Habits. Arthritis Rheumatol. 2023;75:872-878. PMID 36629408 ↩
2. Of course, the opposite (that is, minimizing or even denying causality) can occur, for example studies showing beneficial effects of treatments which might eat into the profits of the pharmaceutical industry. See this essay. ↩
3. King SD, Veliginti S, Brouwers MCGJ et al. Genetic susceptibility to nonalcoholic fatty liver disease and risk for pancreatic cancer: Mendelian randomization. Cancer Epidemiol Biomarkers Prev. 2023EPI-23. PMID 37351909 ↩
4. Wang S, Liu H, Yang P et al. Exploring the genetic association of allergic diseases with cardiovascular diseases: a bidirectional Mendelian randomization study. Front Immunol. 2023;14:1175890. PMID 37334359 ↩
5. Mukamal KJ, Stampfer MJ, Rimm EB. Genetic instrumental variable analysis: time to call mendelian randomization what it is. The example of alcohol and cardiovascular disease. Eur J Epidemiol. 2020;35:93-97. PMID 31761964 ↩
6. Freis ED. The Veterans Administration Cooperative Study on Antihypertensive Agents. Implications for Stroke Prevention. Stroke. 1974;5:76-77. Link ↩
7. Kido DK, Gomez DG, Santos-Buch CA, Caston TV, Potts DG. Microradiographic study of cerebral and ocular aneurysms in hypertensive rabbits. Neuroradiology. 1978;15:21-26. PMID 643169 ↩
8. Lee J, Berry CL. Cerebral micro-aneurysm formation in the hypertensive rat. J Pathol. 1978;124:7-11. PMID 722374 ↩
9. Morgan T. Renin, angiotensin, sodium and organ damage. Hypertens Res. 2003;26:349-354. PMID 12887125 ↩
10. The fact that different medication classes are cause improvements in CVD is usually taken to mean that the effect must be due to what they have in common, ie they all lower blood pressure. But read on… ↩
11. Cedergreen N, Ritz C, Streibig JC. Improved empirical models describing hormesis. Environ Toxicol Chem. 2005;24:3166-3172. PMID 16445100 ↩
12. International CFBPG-WAS, Ehret GB, Munroe PB et al. Genetic variants in novel pathways influence blood pressure and cardiovascular disease risk. Nature. 2011;478:103-109. PMID 21909115 ↩
13. Wan EYF, Fung WT, Schooling CM et al. Blood Pressure and Risk of Cardiovascular Disease in UK Biobank: A Mendelian Randomization Study. Hypertension. 2021;77:367-375. PMID 33390054 ↩
14. Clarke R, Wright N, Walters R et al. Genetically Predicted Differences in Systolic Blood Pressure and Risk of Cardiovascular and Noncardiovascular Diseases: A Mendelian Randomization Study in Chinese Adults. Hypertension. 2023;80:566-576. PMID 36601918 ↩
15. Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol. 2015;44:512-525. PMID 26050253 ↩
16. Law MR, Morris JK, Wald NJ. Use of blood pressure lowering drugs in the prevention of cardiovascular disease: meta-analysis of 147 randomised trials in the context of expectations from prospective epidemiological studies. BMJ. 2009;338:b1665. PMID 19454737 ↩
17. Tajeu GS, Booth JN, Colantonio LD et al. Incident Cardiovascular Disease Among Adults With Blood Pressure <140/90 mm Hg. Circulation. 2017;136:798-812. PMID 28634217 ↩

• ← Poisonings, not overdoses
• What is a council meeting? What is their purpose? →