Good morning friends! The SPRINT trial, our topic of conversation today, is a little older than most of the stuff we’ve been presenting. But 2015 isn’t that long ago, and the buzz surrounding the Systolic Blood Pressure Intervention Trial certainly hasn’t died down much since its being published. I’ve turned it over in my head about 100 times, and there’s so much to unpack that I felt like I could use some help. Fortunately, Josh Uy, Penn geriatrician extraordinaire, has already taken the time to catalogue his always-useful insights. Josh’s summary is excellent but a bit bullet-pointy, and at times I have (with his permission) expanded it to read a little more easily and be more in-line with what you’re used to seeing on IM HEAT. So if something is poorly written or seems not to make sense, odds are that’s a “me” problem and not a “Josh” problem.
Warning: Long piece, about 3500 words. Maybe save for a weekend long read, but man is it worth reading. Josh has great insight. His approach to prioritizing this intervention in the context of the whole patient, towards the end of the piece, is one of my favorite frameworks, and I’m excited to share it with you. Read and enjoy! I know I did.
Lower Is Better, Except For All The Times It Isn’t
In the 2015 NEJM article, the SPRINT investigators describe their efforts to “[compare] the benefit of treatment of systolic BP to a target of less than 120 with treatment to a target of less than 140.” Is lowering BP from 140 to 120 systolic more beneficial than harmful, more harmful than beneficial or both beneficial and harmful in a population that represents both men and women, older than 75 and younger than 75, black and white, with and without previous heart disease, and across a range of baseline blood pressures? From the Sprint Trial Protocol, “The goal is to assemble a representative population of older patients for whom intensive BP lowering is reasonable to consider from a medical perspective. This goal is motivated by the perspective that there may be some older persons with advanced frailty and/or multiple comorbid conditions whose health is so poor that it would not be reasonable to attempt to treat SBP as intensively as needed to control SBP to less than 120 mm Hg.”
In attempting to define this population, the investigators have a long list of inclusion and exclusion criteria. To be included, all patients had to be at least 50 years old with a baseline BP of 130-180. Patients had to have either known symptomatic atherosclerotic cardiovascular disease (previous ACS, MI, or revascularization, or known stenosis > 50% anywhere), known subclinical ASCVD (by coronary artery calcium scoring, ABI < 0.9, or LVH), or be at elevated risk (defined as CKD with GFR 20-59 by MDRD, Framingham risk score > 15%, or age > 75 alone).
Patients were excluded for not being on meds they should already be on (e.g. BB after MI, ACE if CKD), but could be reassessed for inclusion after starting these meds. Patients with BP > 180 or not meeting specific thresholds on a given number of meds were excluded for severe hypertension. Known secondary hypertension was grounds for exclusion, as was orthostasis defined as a BP of < 110 on standing. Most interestingly, patients with a history of diabetes were excluded, as were patients with stroke. (“This is likely because blood pressure targets have been well-studied for diabetic patients previously in the ACCORD BP, and BP goals after stroke are kind of a whole thing.” – Bill).
Also excluded were patients who for a variety of reasons were likely too frail to benefit from the long-term effects of antihypertensive therapy: patients with dementia, nursing home residents, malignancy requiring treatment in the past two years, GFR < 20, or EF < 35%. (“There was also a broad consideration for exclusion based on history of poor compliance, which on the one hand seems arbitrary, and on the other is a very good reason that a primary care doctor might choose not to treat to a lower BP target. Minimizing pill burden might be more important. They also excluded pregnant patients — there aren’t a lot of pregnant 50 year olds, but you can be sure there were none in the trial.” – Bill).
On the whole, this patient population represents the very healthy old. Not only are they healthy, but probably represent a population of the most stable older adults. They can’t have had recurrent hospitalizations or come from a nursing homes. They have to be medically stable enough and without active symptoms to allow them to even consider being part of a medical research study. Patients with active symptoms or unstable medical issues are unlikely to have the energy or desire to focus on being part of a research study. Research subjects likely are the most adherent and invested patient population. Patients with multimorbidity and frailty were effectively excluded. Notably, polypharmacy was not assessed.
In this trial of “standard” versus “intensive” BP control, the standard treatment BP goal was 135-139 systolic. If <130 at a single visit or <135 at two visits, then medications were decreased. For intensive treatment, the BP goal was <120. If >120, physicians were to titrate/add therapy until BP is less than 120 and see the patient monthly in follow up. Titration was continued until patients were at goal, or until a clinical decision was made that therapy should not be increased further. This means that caution was exercised and it was not get to <120 at all costs. It is unclear what factors these clinical decisions included.
To get to goal, investigators used pretty much every antihypertensive you can think of to get blood pressure under control, from ARBs to clonidine, but interestingly capped the regimens at 5 medications. BPs were checked monthly for 3 months, then every 3 months thereafter (decreased back to monthly after any adjustment). The methods of measurement themselves were interesting, and not necessarily representative of what we do in clinical practice — with the patient was sitting still for 5 minutes, BP was measured 3 times with an automated cuff and averaged. Given that standard protocol in my office is to walk the patient back from the waiting room and rely on one manual BP from the medical assistant unless I repeat it, this is a significant departure from the norm.
The primary outcome was a composite of MI, acute coronary syndrome without MI, stroke, acute decompensated heart failure, and death from cardiovascular causes. Secondary outcomes included each of the above individual events, call cause mortality, and a variety of renal outcomes including new microalbuminuria, GFR reduction, new hemodialysis, or renal transplant.
Important adverse events included “serious adverse events” (SAE), defined as “Fatal or life threatening, resulted in clinically significant or persistent disability that required or prolonged a hospitalization or that were judged by the investigator to represent a clinically significant hazard or harm to the participant that might require medical or surgical intervention to prevent one of the other events listed above.” Effectively, any event threatening or causing disability or death. In addition to these, injurious falls, syncope, and “any unexpected event [for which] the investigator believes that the SPRINT intervention caused the event or contributed to the immediate cause of the event.” A broad net, to be sure, but nonetheless some harms weren’t counted: these “expected events” included “electrolyte abnormalities (hyponatremia or hypokalemia are known to be associated with diuretics; hyperkalemia and short-term decline in GFR with RAAS blockers, hyperkalemia with potassium-sparing drugs; as well as bradycardia with beta blockers and calcium channel blockers).” These events however could qualify as serious adverse events (SAEs) if they meet criteria.
You will recall that this study was stopped early because there was felt to be overwhelming evidence of benefit in the intensive control arm. Here’s why:
Were those “Exciting!” marks sarcastic? That’s really up to you, isn’t it? Here are the benefits (RRR = relative risk reduction):
Over 3.2 years (median follow up at end of study):
- NNT 61 to prevent 1 primary outcome (RRR 25%)
- NNT 125 to prevent 1 case of acute decompensated CHF (RRR 38%)
- NNT 172 to prevent 1 case of CV death (RRR 43%)
- NNT 90 to prevent 1 all cause deaths (RR 22%)
NNT extrapolated to 1 year based on observed event rates:
- NNT 185 to prevent 1 primary outcome (RRR 25%)
- NNT 385 to prevent 1 case of acute decompensated CHF (RRR 38%)
- NNT 555 to prevent 1 case of CV death (RRR 43%)
- NNT 270 to prevent 1 all cause death (RR 22%)
(“I love the focus on the NNT here; seeing a 22% all cause mortality reduction is very impressive, but it’s important when looking at the balance of risks and benefits to realize just how many people you need to include to make 22% amount to one actual outcome prevented.” – Bill).
Remember here that the primary outcome separated at 1 year while mortality benefits separated at 2 years, so read the extrapolated one-year NNT above with care.
And what did it take to get people to goal? In the intensive group 30% were on 2 meds, 32% on 3 and 24% on 4 or more. In the standard group, 31% were on 1 meds, 33% on 2 meds, 17% on 3 meds, 7% on 4 or more. As you would imagine, more meds led to an excess of harms in the intervention group:
Over 3.2 years
- Serious adverse event HR 1.04 in intensive vs treatment group over 3.2 years (P = 0.25, NS) — no increased harms by composite outcome.
However, looking at individual components:
- Hypotension NNH 71 over 3.2 years (Increased 70%)
- Syncope NNH 91 over 3.2 years (Increased 44%)
- Electrolyte abnormalities NNH 100 over 3.2 years (increased 38%)
- AKI NNH 56 over 3.2 years (increased 71%)
So the bird’s eye view seems to indicate that known adverse effects of antihypertensives were more unsurprisingly more common, but did not translate to a significant increase in overall serious adverse events. But let’s spend some time with table 3:
The harms are very puzzling. There’s no real transparency of how the harms were arrived at, and the total serious adverse events is much greater than the sum of all the events listed below. Anyone who can figure out how to make the events listed below the “Serious adverse event” headline add up to the reported figure of 1793, please tell me. The assumption is that the listed adverse events are only some of the adverse events in the total, so let’s break that out: in the intensive arm, there are 1793 adverse events. Adding up the listed serious adverse events plus ER visits leads to a sum of 1140, giving us 653 unclassified events. For the standard arm there are 1736. Adding up the listed adverse events + ER visits leads to a sum of 870, or 866 unclassified events. This is an excess of almost 25% of unclassified harms in the standard therapy group, and is clearly the sole factor responsible for SAEs not being significantly different! What’s missing?
Bill Steps In
OK, this is Bill, and this was so interesting that I had to find out the answer. Let’s start by turning to page 32 of your supplementary appendix. I’ll wait. A few events show up here that did not show up explicitly in table 3. Cardiac events like MI, CHF, and tachyarrhythmia occurred more frequently in the standard therapy group than the intervention group. With small numbers of observed events, none of these is likely to be significant on its own, but as a composite you could imagine that cardiac adverse events might be significantly different between the two. But fortunately you don’t have to imagine, because that’s the upshot of the damn study. We already knew that cardiac events were less common! That was your headline!
This seems to be frank double-counting. By taking events included in your primary outcome and reporting them as serious adverse events, you are using your treatment effect to mask the potential harms of therapy. With this lens, the SAE composite is more of a balance of benefits vs harms than it is reporting actual harms of therapy. What about SAE just likely to be treatment related — are 1140 vs 870 adverse events that are possibly treatment related statistically significant? It’s certainly worth wondering. We are very possibly underestimating our harms by simply pointing to the nonsignificant top-line comparison.
Back to Josh: Conclusions and Critiques
So by treating 1000 patients with a BP goal of 120 over 3.2 years
- You would prevent 16 primary outcome events
- Including: 8 events of CHF, 6 CV deaths, 12 deaths not included in primary events
- You would cause 22 excess serious adverse events
- You would case 14 cases of hypotension, 11 cases of syncope, 10 cases of electrolyte
- abnormalities and 19 cases of AKI.
As above, the math doesn’t work out.
About the patient population: These patients are truly old, but also very stable healthy older adults due to the exclusion of frailty and multimorbidity.
About the intervention: This was a very cautious intervention with adding just one med at a time and having multiple braking points.
About the outcomes: The benefits are real and are as impressive as any preventive therapy we have other than maybe BB/ACE in systolic CHF or coumadin for afib, AND it decreased all cause mortality. Harms are there but the use of composite outcomes obscure the true rate of harmful events. There was no NNH reported in the paper itself even though the numbers are right there. (“There is concern for double counting benefits of therapy as excess harms in the standard treament group.” – Bill.) Unclear why they chose the composite outcome for harms and what went into it. For example why measure falls in a non frail patient population when risk factors are likely excluded? Why include orthostasis when it is already known that treating hypertension decreases orthostasis? How about fatigue?I believe a true sense of the harms is going to come from subsequent observational studies.
And most importantly, about the application: This is a value sensitive decision because the therapeutic window is small. RRR are huge, but the absolute risk reduction on an individual level is small. When breaking out individual adverse events, serious harms are more common than the benefits and minor harms are likely even more common. The chance of hurting someone is, strictly speaking, greater than helping them. This is a preventative intervention which implies the patient is stable enough to concentrate on preventive therapies (See Josh’s hierarchy of multimorbidity at the end). Preventive treatments are also more effective over longer timeframes and meaningless over short time frames. The placement of this intervention in a broader clinical context is missing (maybe beyond the scope of this article) including patient centered communication. SPRINT goals should not be a performance measure for all patients.
Unanswered questions: Should we persist if someone is actively symptomatic from treatment? Or has an event? If more than 1 medication is needed to get to goal, is the benefit still there? How about 2, 3, 4, 5?
Definitely Bad Applications (My imaginary conversation with a cardiologist)
So a BP of 120 is better than a BP of 140!
Me: Don’t be absurd, we treat patients not numbers. The characteristics of the patient (presence of frailty, multimorbidity, prognosis and values) will determine what’s best. Treatment of hypertension has to be individualized. There is no ideal number taken out of patient context.
Well all patients benefit from a BP of <120 so why individualize?
Cardiologist continues: In every prespecified subgroup analyzed, there was benefit: male vs female, older than 75 vs younger than 75, black vs white, etc.
Me: Yes but the very large and important population of frail patients and multimorbid patients were excluded and polypharmacy not assessed
Okay, for all STABLE patients, it’s worth doing whatever it takes to get to a BP of <120
Me: A goal of 120 systolic in the trial does not equate to a quality indicator goal to achieve a BP of less than 120 in 100% of patients. In fact the benefit was seen even though most people did not get to <120. Second, on average only 1 medication was added. It’s not clear that the benefit would still be there if the patient needs 7 classes of BP meds or BP meds contribute to a polypharmacy burden.
If my healthy, stable, patient resists setting a goal of 120, I need to convince them to agree.
Me: This is not the decision to treat pneumonia with an antibiotic or lasix for decompensated CHF. This isn’t even anticoagulation for a high risk atrial fibrillation patient or wanting to treat a patient with an A1C of 15 or a systolic BP of 220. This is a preventive treatment in patients already well controlled that takes years for benefits and has harms. This is worth approaching as a shared decision with pros and cons.
But a relative risk reduction of 43% for CV death is a big deal!
Me: It is and it isn’t. The NNT over 1 year is absurd. Over 10 years, yes, the NNT is likely in the 20’s. Over 3 years the benefit is 0.6%. So yeah.
The older the patient, the more beneficial the tight control. Don’t be so academic, it’s okay to generalize!
Cardiologist continues: As older adults get older and become at higher risk for CV events, their BP should be more tightly controlled because the benefit will be even greater (for an increasing absolute risk, a fixed relative risk reduction leads to a lower NNT in theory). Me: As adults get older and higher risk they often also become more frail and meet the exclusion criteria for the study. Frailty leads to beneficial interventions being harmful. Also value systems change and patients may not want to sacrifice how they feel now for a potential benefit years down the road.
Feeling dizzy is worth preventing death. Clearly the benefits outweigh the risks.
Me: Not to many patients. Value assumptions of interventions cannot be assumed to match the values of patients. Feeling worse each day may not be worth preventing death. Many patients get to the point where they want to feel as good as possible each day and living longer isn’t the main goal.
Let’s spend some more time on patient values here. From the accompanying editorial: “Rates of some serious adverse events, including hypotension and acute kidney injury or failure, were higher in the intensive-treatment group than in the standard-treatment group, but these higher rates appear unlikely to outweigh the benefits overall.” “Many patients will accept small inconveniences, such as having to get up slowly to avoid symptomatic hypotension or taking an additional pill, to preserve their health.”
My comment is that value assumptions in geriatrics that have quality of life implications is dangerous. Using a hard age cutoff like JNC 8 (age>60) or a BP goal (120 vs 140) will miss having a patient centered approach to management.
Definitely Good Applications
- If a patient is tolerating current treatment and their BP is 120, don’t de escalate
- If a patient is stable, tolerating current treatment and their BP is 140 systolic, consider adding a medication after discussion about pros and cons. Consider making it a recommendation. Do it very judiciously and carefully. (Add one medication and call it a day.)
- Involve shared decision and focus on goals of care (not numbers like age or systolic BP)
But don’t take my word for it. Per the SPRINT PI on a youtube video (https://www.youtube.com/watch?v=EPxheTCswAM):
Q: 80 year old on 2 meds and a systolic of 136, patient is otherwise healthy and stable. What would you do?
A: “It’s a matter of having a discourse. If they are asymptomatic, tolerating their treatment well, careful, judicious reduction in blood pressure, if the patient agrees, is probably a smart, wise course based on SPRINT.
2nd comment: “It’s not that 100% were below 120. We have to distinguish between performance indicators… and what we were doing which was to get to a goal of 120. That’s a big difference.”
To recap, per the study protocol:
The goal is to assemble a representative population of older patients for whom intensive BP lowering is reasonable to consider from a medical perspective. This goal is motivated by the perspective that there may be some older persons with advanced frailty and/or multiple comorbid conditions whose health is so poor that it would not be reasonable to attempt to treat SBP as intensively as needed to control SBP to less than 120 mm Hg.
So Where Does This Fit Into Preventative Care
Josh’s hierarchy of multimorbidity:
- Active symptoms/illnesses
a. Pain, shortness of breath (and whatever the patient cares most about)
b. Acute illness (pneumonia, CHF, hip fracture)
- Frailty syndromes
a. Falls, weight loss, delirium, functional decline, dementia
- Secondary prevention/chronic disease management
a. CHF, hypertension, COPD, depression
b. Osteoporosis, Hyperlipidemia
- Primary prevention
a. Cancer screening
This treatment falls under priority #3a. The data is useful and we should practice with it in mind, but it’s incredibly important that 120 does not become our default decision.