What's the Risk?

Nearly six months into the pandemic, we ought to know the important risk factors for serious illness or death from Covid-19. Whether because of poor record-keeping, lack of international cooperation, or sloppy statistical analysis, the information until now has been limited. 

Not just limited; the claims about risk factors to date have been quite misleading. I complained in my blog post in early May that the rates of certain conditions in patients dying of Covid-19 were actually no different from the rates of those same conditions in older people in general. For example, I noted that one study reported that the rate of high blood pressure in patients with severe cases of Covid-19 was 56.6 percent—but failed to comment that the rate of high blood pressure in the elderly population is 60 percent. Far from indicating that high blood pressure increases the likelihood of severe Covid-19 in older adults, this finding suggests that high blood pressure confers no extra risk or maybe is even protective. 

The only consistently observed risk factor for both severe illness and death has been older age, with age greater than 80 representing very high risk. Now, thanks to the existence of widespread, compatible electronic medical records in the British National Health Service (NHS), we have some useful data.

The study, published early on line in Nature Reviews, compares British patients who died of Covid-19 with all other British patients who are cared for in a group practice that used the necessary software (approximately one-third of the population). By using “a secure analytics platform inside the data centre of major electronic health records vendors, running across the full live linked pseudonymised electronic health records,” and after excluding people under age 18 and those with less than a year's worth of data, the investigators were able to collect health information on over 17 million individuals, including just under 11,000 with Covid-19 related deaths. The results confirm age as the single most potent risk factor, with a small number of other major risks.

To capture the most striking findings, I extracted data from the chart listing the hazard ratios (HRs) and 95% confidence intervals (CI) for Covid-19 death (Table 2) and present 3 separate tables: one highlighting very high-risk characteristics, one highlighting high-risk characteristics, and one showing characteristics associated with no or minimally increased risk. For added emphasis, I highlighted hazard ratios of greater than 3 in red and hazard ratios between 2 and 3 in blue.

Characteristics Associated with Covid-19 Deaths

The important difference between these charts and previous attempts at quantifying risk is that the hazard ratios reported have been age and sex-adjusted and have been further adjusted for other potential confounders along with age and sex.  

The conclusion from this analysis is that when we consider the age- and sex-adjusted hazard ratios, there are only four very high-risk conditions: old age, a hematologic malignancy diagnosed within the previous year, severe kidney impairment, and organ transplantation. Within the old age categorization, the hazard ratio goes from 8.62 to 38.29 as the age increases from 70-79 to 80 and older (the reference group is people aged 50-60). The only medical condition that comes close to this magnitude is organ transplantation, with a hazard ratio of 6. 

Another handful of conditions are in what I have classified as high-risk: obesity with a BMI of 40 or greater, poorly controlled diabetes, stroke or dementia, a hematologic malignancy diagnosed between 1 and 5 years earlier, liver disease, and other forms of immunosuppression. Incidentally, four out of six of these conditions drop out if we look at the “fully adjusted” column.

It’s worth noting some of the chronic conditions that were not associated with increased risk. High blood pressure, as I had previously suggested by comparing the rate in the very ill Covid-19 patients with the rate in the general older population, does not appear to be a risk factor. Mild to moderate asthma, defined in this study as someone with asthma who did not use oral steroids within the previous year, is also not a risk factor.

It’s also worth noting that the CDC has issued its own guidance about risk factors for Covid-19. Their conclusions, while overlapping with the NHS data, differ in important ways. Most important, their methodology differs. The CDC, lacking a domestic large, comprehensive data base, is forced to draw on evidence from small case series, cohort studies, and some meta-analyses, as well as a much earlier preliminary report from the NHS. The new NHS data, rigorously obtained and meticulously analyzed, should be seen as the gold standard.

Down, Down, Down

         The most important recent development affecting older patients, without a doubt, is the issuance of new guidelines for the diagnosis and treatment of high blood pressure. The Wall Street Journal proclaimed “Nearly Half of US Adults Have High Blood Pressure Under New Guidelines,” and venerable health columnist Gina Kolata of the NY Times wrote: “The nation’s leading heart experts on Monday issued new guidelines for high blood pressure that mean tens of millions more Americans will meet the criteria for the condition, and will need to change their lifestyles or take medicines to treat it.” The report, which took me a while to track down and read since most of the references are to summaries of the report or commentaries on the report but not to the actual document, is 175 pages. It was issued or endorsed by eleven organizations, whose initials actually form part of the name of the report: its full name is the “2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPCNMA/PCNA Guidelines for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults.” The subtitle is: “A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines.” Now that I’ve had a chance to digest this—what should we make of it?

         I remember the day in 1991 when JAMA published the results of the “SHEP” trial, a study of the treatment of isolated systolic hypertension in the elderly that upended what I’d been taught in medical school. Gone was the belief that high blood pressure was good for older people because they needed greater force to push blood through their stiff arteries; suddenly, there was compelling evidence that high blood pressure was dangerous—it caused strokes and heart attacks. Keeping the systolic blood pressure below 160 reduced the risk of stroke by 36 percent and the risk of a cardiac event by 27 percent. Overnight, the practice of medicine changed.

         To be sure, there were skeptics. Wouldn’t lowering the blood pressure cause older people to become dizzy and to faint? Just how far were we supposed to lower blood pressure anyway? But to me and many others, the choice was clear. Stroke was one of the worst fates that could befall older patients: sometimes it killed them, but more often, it left them impaired, often profoundly, significantly diminishing the quality of life.

         But I also remember a book by physician and historian of science, Jeremy Greene, published in 2008 and called Prescribing by Numbers: Drugs and the Definition of Disease. Greene argued persuasively that the pharmaceutical industry has a vested interest in lowering the cut-off for treatment of a variety of chronic conditions, such as hypertension, high cholesterol, and diabetes. Defining “pre-diabetes” as a real condition warranting treatment or repeatedly dropping the threshold for treating cholesterol with statins had clear implications for the drug companies: they would sell more pills. More volume, more revenue, more profits. So are the new guidelines just another instance of “diagnosis creep,” a way for Pharma to make more money?

         It’s important to realize that what tipped the scales for the guideline writers, what’s different now from the last time they wrote a guideline, three years ago, is the 2015 “SPRINT” study (Systolic Blood Pressure Intervention Trial), a randomized trial of high risk patients over 50 to either a systolic blood pressure target of 140 or 120. The study, which was reported in the New England Journal of Medicine, found a 25 percent reduction in cardiovascular events (such as stroke or heart attacks) when physicians tried to reduce the blood pressure to the lower value.

         As Gilbert Welch of the Dartmouth Institute, co-author of Overdiagnosed: Making People Sick in the Pursuit of Health (2012) and Less Medicine, More Health (2016), commented, that 25 percent sounds pretty good, but it’s the relative risk reduction. In fact, the rate of cardiac events fell from 8 percent in one group to 6 percent in the other—an absolute difference of only 2 percent. Or, looked at differently, 92 percent of people in one group and 94 percent of people in the other group did just fine. Suddenly, the impressive results don’t look quite so impressive. Moreover, all these patients were already at increased risk of cardiovascular disease because of other factors such as cigarette smoking or diabetes. Presumably, had the same study been carried out in people of average or lower risk, the benefits would have been correspondingly smaller.

         So if the new guidelines are based predominantly on the findings from one study, and the one study isn’t quite as compelling as it seemed at first glance, what should we make of the new recommendations? Should we really treat everyone with a systolic blood pressure over 120 who has at least a 10 percent risk of cardiac disease in the next decade? 

         The answer, I think, is yes, but with caveats. First, the guideline writers are at great pains to insist that blood pressure be measured the way it was measured in the SPRINT trial—in a way that it’s almost never measured in the doctor’s office, namely after sitting quietly for five minutes and averaged over three readings. They also advocate use of home monitoring to confirm (or refute) the diagnosis of high blood pressure, as well as to guide medication adjustment once treatment is started. Next, physicians need to be prudent about what medications to use if they’re going to embark on pharmacologic treatment. We have ample numbers of cheap, effective medicines that have been around for years, such as diuretics and beta blockers, and these should be tried before pricier medicines. Finally, both physicians and patients need to be vigilant about medication side effects, which means doctors need to tell their patients what to look for, patients need to report their symptoms, and doctors need to change course if symptoms develop. In addition, non-pharmacologic treatment should be attempted, including exercise, weight loss, and a low salt diet. If we follow all these steps, we can be confident we will be changing how we deal with blood pressure because it’s what’s best for patients, not because it’s what’s best for the pharmaceutical industry.

Brave New World of Genetic Testing

In a provocative piece in the NY Times last week, science writer Gina Kolata suggests that the long term care insurance industry may be in a “death spiral.” The culprit, she argues, is genetic testing, which got a boost last month when the FDA approved testing by the company 23andMe. Previously best known for providing genealogical information to those who send in a saliva sample and a $99 fee, the personal genomics company is now authorized to provide information about genetic risk factors as well—for only an additional $26.

One of the ten conditions about which companies may offer information is Alzheimer’s disease. And perhaps the best established genetic risk factor for late-onset Alzheimer’s disease (sometimes called LOAD) is apoE. A gene that codes for a protein involved in cholesterol metabolism, apoE comes in three varieties, prosaically named apoE2, apoE3, and apoE4. Everyone has two copies of the apoE gene, so there are six possible genotypes, of which the most common are E2/E2, E3/E3, E4/E4, E2/E3, and E3/E4. The majority of people (63 percent, in one study of the distribution of the allele in 9 different populations) are E3/E3.

But while there is no way to definitively predict who will develop Alzheimer’s disease, fully 40 percent of people who develop LOAD are among the 25-30 percent of people who carry the e4 variant of the gene. And roughly half the residents of nursing homes have Alzheimer’s disease. As a result, according to spokesmen from the industry, a growing number of people are seeking testing from 23andMe to see if they have the e4 gene. If they have it, they buy long term care insurance. If they don’t, they take their chances.

If this trend continues—and as of 2017, 2 million people have obtained the direct to consumer genetic analyses---the pool of people buying long term care insurance could be heavily weighted towards those who actually will develop the disease. The health insurance industry, however, as consumers are perhaps finally coming to understand in the ongoing Obamacare wars, depends on pooled risk; it only works if the people who buy insurance include some people who will get sick and others who won’t. 

The reason the direct-to-consumer marketing of genetic information potentially spells doom for the long term care industry is that Americans are protected by the Genetic Information Nondiscrimination Privacy Act, which prevents insurers from requiring gene tests or using the results of genetic testing in coverage decisions. That means that ordinary people, without any physician input, can find out if they are at high risk, they can make decisions about buying coverage based on that assessment—and the insurance companies are powerless to intervene. If most of the people who are destined to get Alzheimer’s disease end up with insurance, the insurance company will end up paying out a lot more than they originally anticipated—leading to enormous increases in the rates or bankruptcy of the industry.

Now there are plenty of other reasons that the long term care insurance industry may collapse, and a good number of pre-existing reasons why it’s a poorly designed program. The amount of money it actually provides people is rarely enough to cover their actual costs, whether of home or institutional care. There are many barriers in the way of people using their benefits—for example, many policies require three months of disability before they kick in, which may be three months too long, especially if they are only going to be in a nursing home for three months. And Medicaid is currently available as a back up to pay for institutional care, provided people have “spent down” their personal savings, so the value of long term care insurance derives from its ability to shelter assets.

Analyzing the value of long term care insurance is a conversation for another blog post. But the point I want to make today is that before you rush out and get tested for Apo E, you should be aware of the limited predictive value of the test. While the likelihood of getting LOAD if you are one of the 2.6 percent of the population who have two copies of E4 is 91 percent, the likelihood of getting LOAD if you are one of the 22 percent of the population with one copy of the E4 allele falls to 47 percent. And if you are one of the 76 percent of the population with no E4 alleles, you still have a 20 percent of getting Alzheimer’s. Because the poor “negative predictive value”—because even with a negative test, you have a substantial risk of getting the disorder—physicians and organizations such as the Alzheimer’s Association have for years recommended against routine Apo E screening.

Interestingly, Apo E determination has been available to Europeans as a direct to consumer test for years. A study to find out what the short- and long-term psychological consequences were for patients with a positive test. Though the study only looked at people who requested testing, it found that there were no significant adverse consequences of getting bad news. So fear that you will become anxious or depressed is probably not grounds for resisting the impulse to be tested for Apo E4. But you should remember that it’s just a risk factor—and one over which you have no control—and plenty of people who test negative will still develop Alzheimer’s disease.