Weak Evidence Can Hurt Important Ideas

The Study of the Week explores an observational analysis of deprescribing blood pressure meds in older adults. But the study falls short--and it could hurt the cause

JAMA-IM has published an observational study that sought causal effects from stopping blood-pressure-lowering drugs in older adults who were in long-term care.

The UCSF-led team asked a specific question: did deprescribing of BP-lowering meds lessen cognitive decline? They didn’t write their question that specifically; instead they wrote that their purpose was to

investigate the association of deprescribing antihypertensive medication with changes in cognitive function in nursing home residents.

I begin with this semantic point about causal language because it ends up being one of my two main criticisms of the analysis.

First Some Background

Older patients often ask me why I want them to come in for yearly visits. My answer is that the main reason clinicians should see older patients is to consider removing therapies—like BP-lowering tablets.

I often tell anyone who will listen that the purpose of BP-lowering tablets (or any preventive meds) is to take them in middle age with the goal of becoming elderly.

Once elderly, these drugs can cause more harm than benefit.

In prior eras, middle aged people who developed severe high blood pressure had strokes, heart attacks and kidney failure, which prevented them from becoming elderly. Treating high blood pressure is obviously a way to extend life.

But older patients are not the same as younger patients. The chance of harm from medicine raises with age. This occurs because older patients have more drug-drug interactions, decreased drug clearance, and less robust reflex mechanisms to maintain blood pressure when standing up.

Another nugget I tell older patients is that there may be more risk from low blood pressure than high blood pressure.

Deprescribing, therefore, is an important concept in the care of older adults.

In the matter of blood pressure control, an older adult has—by definition—achieved success. They are old. Now, our goal is to reduce the chance of harm from over-zealous BP control. While you don’t want older adults to have a brain bleed from super-high blood pressure, you don’t need to target intense control.

Intensive vs standard BP control trials, such as SPRINT, enrolled younger ambulatory patients. This data does not apply to older adults with multiple medical conditions.

The Study

The team studied older adults (mean age 77 years) who were in a VA long-term care facility. They started with a sample size of more than 45,000 adults. After excluding those with significant high blood pressure or heart failure and various other reasons, they ended up with 12,000 study subjects.

About 1000 patients had BP-meds stopped. These were the treatment arm. Nearly 11,000 did not have a change in BP meds. These were the control arm. Pause and note that a) these were quite uneven groups, and b) clinicians, not randomization, chose to stop or not stop meds. In addition, there is this copy/paste:

Residents in the control group had higher baseline BP, lower median ADL difficulty counts, and were taking fewer antihypertensive medications compared to the deprescribing group.

They then measured cognitive function on a 4-level ordinal scale. 1 – intact, 2 – mildly impaired, 3 – moderately impaired and 4 – severely impaired.

The authors knew their groups were neither randomized nor similar at baseline. So they made adjustments. They even gave these adjustments the name “target trial emulation,” which, is a way to mimic what a randomized controlled trial would do. (I will discuss this in my comments.)

The main result was positive: Compared to their cognitive function scale (CFS) at the beginning of the follow-up, 12.0% of the residents had a worsened CFS (higher score), while 7.7% had an improved CFS (lower score) by the end of the follow-up. And 10.8% of the residents in the deprescribing group and 12.1% in the stable user group had a worsened CFS score.

In an analysis looking only at those who remained in their category (deprescribed or continued), those in the deprescribed group had a 16% lowered odds of cognitive function worsening per 12-week period. (OR 0.84 95% CI 0.72-0.98).

The authors concluded that this was novel evidence regarding the association of deprescribing BP-lowering meds. They wrote that residents who were deprescribed these meds had slower cognitive decline compared to those who had medicines continued.

Comments

I applaud efforts to study older adults. I strongly believe the care of older people with multiple medical conditions in the US is terrible. One of the reasons for this is that once beneficial medications are not stopped.

Studies like this attempt to persuade caregivers to deprescribe preventive medications with empirical data. I believe in evidence-based medical practice, but persuasion-with-data requires extreme care. You can lose trust when the evidence doesn’t pass muster. We saw this phenomenon during the pandemic. Authorities would propose policy A or B but the supporting evidence was terrible. The end-result was loss of trust in evidence.

---

I am afraid this study falls short of its persuasion goal.

Here are two of many reasons:

It’s a retrospective look back at two non-random groups. Adults who had meds continued (control arm) were different than those who had meds stopped. The authors made statistical adjustments and even set out a specific time to start looking, but, ultimately, there was surely confounding.

Confounding occurs because you can only adjust for factors in the medical record. But clinicians make decision (like deprescribing or not) based on oodles of factors.

And here, to the authors credit, they offer a measure of confounding (the E-score), and it indicates a substantial chance that differences in the groups drove the cognitive score differences, not deprescribing.

The second problem with this study is the primary outcome measure.

Cognition is a tough thing to measure. And it was barely different. The main result differed by less than 2 percentage points. Plus, when you look at eFigure 7, there were tiny differences in mean blood pressure changes.

It’s, therefore, implausible to suggest that a few mmHg difference in blood pressure would drive a change in cognition. More likely, the difference was noise.

Conclusion

In obvious cases, where meds have caused harm, deprescribing in older adults approaches the level of empathy and compassion—that is, evidence is not needed. It’s common sense.

Many older adults haven’t yet incurred harm from preventive meds. While I believe most of these patients could be safely taken off preventive meds, if you want evidence, it must be stronger than confounded observational studies.

To me, this is a classic case for non-inferiority randomized trials. Adults admitted to nursing facilities or those in clinic could be randomized to a deprescribing team or basic care. Then you measure something more obvious—such as mortality or hospital readmissions. No difference, or noninferiority, is a win.

Weak evidence does more to hurt a cause than support it. That’s why I worry a lot about studies like this.