Thursday, December 29, 2011

Cardiology JC 12.20.11

Dr. Branch: Reduced Dietary Salt for the Prevention of Cardiovascular Disease

Click here to watch a recording of the presentation.

Dr. Branch presented a recent meta-analysis from the Cochrane Collaboration which reviewed the published evidence that reductions in dietary salt intake impact cardiovascular morbidity and mortality.

The authors found 7 studies meeting their inclusion criteria, which enrolled either normotensives, hypertensives, or mixed populations, with a single study looking at salt reduction in patients with decompensated heart failure.  

Their results were, as they say, "consistent with the belief that salt reduction is beneficial in normotensive and hypertensive people."  However, the size of that beneficial effect was miniscule for most outcomes, and not especially consistent with recent, highly publicized statistical projections of the impact large-scale salt reduction might have on public health.  The authors found a nonsignificant reduction in mortality among normotensives and hypertensives (Relative Risk = 0.67 and 0.97, respectively, both confidence intervals including 1.0).  They found "no strong evidence" of a beneficial effect on cardiovascular morbidity (defined as MI, stroke, CABG, PTCA or death from cardiovascular disease) in normotensives or hypertensives either, although one study had appeared to show a large relative risk reduction in its early phase. 

Reductions in blood pressure consistent with previously published findings were observed - normotensives tended to drop by about 1mmHg in systolic blood pressure, hypertensives and those with heart failure by about 4 mmHg.  One disturbing finding was that in the one trial of heart failure patients, low-salt diets seemed to increase morbidity and mortality.  Strangely, none of the studies reviewed reported any outcomes in terms of health-related quality of life.

Overall, while this report did suggest that some people may experience small falls in blood pressure when voluntary salt consumption is reduced, it does not seem to support the idea that salt consumption is a major public health problem.  Moreover, since most salt consumption comes from processed foods, it seems clear that to the extent that this is a problem (which, again, is not clear) it is probably better resolved by regulators than individual patients and their PCPs.

Dr. Vij: Efficacy of Statins for Primary Prevention in People at Low Cardiovascular Risk

 Click here to watch a recording of the presentation.  Due to technical issues, this recording has no audio.

Dr. Vij presented a meta-analysis published in CMAJ which reviewed the available evidence on the use of statins for primary prevention in people at low cardiovascular risk.  In terms of the inclusion criteria applied to studies, "low-risk" was defined as having a projected 10-year risk of death from a cardiovascular cause or non-fatal MI of less than 20%.  

The main outcomes assessed were the risk of all-cause mortality, MI, stroke, and adverse events related to statin therapy.  The authors also undertook meta-regression analysis to assess whether a number of other variables influenced the association of statins with these outcomes, including the absolute and percent change in serum LDL during the trial.

They reported their results in terms of pooled relative risk (RR), absolute risk reduction (ARR), and number needed to treat (NNT).  Table 3 in the article gives their full report of results, but here are the highlights:
  • The RR of all-cause mortality was 0.9 (95% CI 0.79-1.03), corresponding to an ARR of 0.42%.
  • The RR of MI was 0.68 (95% CI 0.53-0.87), although weirdly the relative risk of fatal MI was not significantly different).  This corresponded to an ARR of 0.46%.
  • The RR of stroke was 0.89, with a confidence interval crossing unity.
  • The RR of revascularization was 0.66 (95% CI 0.55-0.78), corresponding to an ARR of 0.77%. 
  • No adverse event was significantly more common in the statin group, including rhabdomyolysis, cancer, and incident diabetes.
  • Overall, no absolute risk reduction was greater than 0.77%.
Dr. Frohlich pointed out a number of conceptual flaws with this meta-analysis, and expressed skepticism that the results of the individual trials were actually comparable in this way.  He also noted that such undertakings are prone to massive selection bias, since none of these trials enrolled random cohorts of healthy people; everyone in these trials was referred by a doctor who thought that, despite their low risk, they might benefit from a statin.  Dr. Barbant agreed that it is generally best, if possible, to make decisions based on original trial data.

This study provides a good opportunity to consider the mechanics and merits of the NNT.  Many clinicians find this a very intuitive measure of treatment outcome, and it has been shown that both physicians and patients understand outcomes expressed in terms of ARR and NNT better and tend to make more conservative decisions when data are presented in these terms rather than in therms of relative outcome measures such as RR or hazard ratios

However, the NNT has limitations.  First, it always needs to be considered in light of the background risk of the condition.  If this is low, the absolute value of the ARR will always be low, and even a highly efficacious intervention will be associated with a large NNT.  Second, it has to be associated with a period of time, since it changes with duration of therapy and there's no good reason to expect that change to be linear.  For example, if 10 patients are given a lethal drug which has a mean onset of action of 5 minutes and a maximum onset of 10 minutes, the NNT to produce one death at 1 minute is infinite, the NNT at 5 minutes is 2, and the NNT at 10 minutes is 1.  Thus, whenever confronted with an NNT, one's first question should be "for how long?"  When these limitations are taken into account, it's unclear how one could calculate NNT in a study like this one or how one should interpret them, and in this case it's probably best to stick with the relative risk.  For those interested, there is an excellent discussion of NNT here.

Dr. Lahsaeizadeh: Ruling Out Acute MI With Sensitive Troponin Assays

 Click here to watch a recording of the presentation.

Dr. Lahsaeizadeh presented two trials looking at the characteristics of biochemical tests for acute (non-ST elevation) myocardial infarction. 

In the first article, from the JACC, Body et al. tested a novel, fifth-generation, troponin-T assay in a prospective cohort of patients presenting to the ED with chest pain, and then validated their preliminary results in a subsequent prospective study.  They were primarily interested in the ability of this novel test to rule out MI with a single measurement on presentation. 

The second study, by Scharnhorst et al., published in Clinical Chemistry, looked at the characteristics of a conventional, second-generation troponin-I assay when a novel interpretive strategy was applied to the results.  Troponin-I levels were obtained at hour 0, 2, 6 and 12.  The authors defined a positive result as an initial troponin-I level greater than the 99th percentile of results obtained from normal subjects (0.06 micrograms/L) or an increase of more than 30% in serum troponin over 2 hours with absolute concentrations below 0.06 micrograms/L.  The investigators in this study also measured CK and myoglobin.

In both studies, the "gold standard" was considered to be clinical diagnosis of acute MI by a cardiologist, who was blinded to the experimental troponin levels; in the first study, the troponin-T level, and in the second study the troponin-I level at hour 2.

Both studies found very high sensitivities and negative predictive values for their respective diagnostic strategies.  Body et al. reported a sensitivity for acute MI of 100% (95% CI 97.2%-100%) in their initial cohort and 99.8% (99.1%-100%) in their audit, which would have allowed confident exclusion of myocardial infarction in 17.5% of their patients.  Scharnhorst et al. found a sensitivity of 100% (95% CI 86%-100%), and calculated a negative predictive value of 100% (95% CI 95%-100%).  They found no additional diagnostic value in measuring CK or myoglobin.

Dr. Frohlich thought that while the results seemed to be applicable in the appropriate context, it is important to remember what exactly that context is: this strategy enables exclusion of myocardial infarction in the setting of chest pain in patients presenting to the ED.  It might not be applicable, for instance, in the setting of sudden chest pain on an inpatient ward, where onset is witnessed and diagnosis needs to be made before the troponin starts to rise.  Additionally, while myoglobin might not add any diagnostic information in the study setting, in a patient with positive troponins and an unclear onset of symptoms it can be useful in determining the age of an infarct.  Dr. Barbant pointed out that the prevalence of acute MI in both study populations (around 20%) was higher than we see in our ED, and that this means the NPV they calculated is not neatly applicable to our population (the influence of prevalence on predictive value is explained elsewhere).  Dr. Stullman thought the entire discussion highlighted the need for a careful history in diagnosing questionably cardiac chest pain, since one can only make the most of these biochemical results if one knows the precise character and timing of the patient's symptoms.

Overall, these results suggest that the "myo/trop q6 x3" strategy of excluding NSTEMI is unnecessary, and that a significant proportion of patients presenting to the ED with chest pain can be reassured that they are not having a myocardial infarction based on a comparatively brief testing interval using conventional assays.  If Scharnhorst et al.'s results can be validated in a larger cohort, yielding a narrower confidence interval, it seems that many patients could be discharged without inpatient investigation after 2 hour serial testing using currently available troponin assays. The major caveat is that this strategy rules out infarction, not unstable angina, and so (as above,) can only be applied by an intelligent clinician in the appropriate clinical context.

2 comments:

  1. Nick, The sodium reduction meta-analysis didn't seem to be as valuable as you made it seem. You had me to believe that salt doesn't affect CVD based on this meta analysis but this was such a silly meta-analysis to begin with I don't think we can take anything away from it.

    My Summary: They looked at 7 studies where patients were assisted in reducing sodium intake and reductions were measured by USE. (Not sure what that translates into as far as mg or sodium/day) but the authors say it was significant but don't say to what degree. Then they say the reduction in sbp was 1mmhg or so, which we both know is silly. How do they have that sort of accuracy to depict 1mmhg change in BP. Then they extrapolate that to reflect the impact on mortality and measure mortality some distant time out from their last USE sample. I have several questions about the utility of this meta-analysis.

    1) To what extent was the reduction in salt. We know that most salt comes from processed food so did the studies achieve significant enough reductions (at least 3g or 1200mg/Na).
    2) Is the follow up time of 6-36 months long enough to show a change when people are eating 7-12g of salt/day for years
    3) Even the study that followed the group for 12yrs, were their eating habits still the same or did people go back to their previous salt intake, primarily since salt in our food nationally has been creeping up.
    4) How can we trust a study that is saying it can detect minuscule reductions in BP
    5) Were these patients on antihypertensives.

    I could go on and should probably read more into the 7 particular studies but I think the clinical take home point is not substantial enough to bother.
    Should we encourage people to reduce their salt intake. I think trying to focus on salt added at the table is the wrong strategy to begin with. obviously policies have to change our food service but in the mean time we should be advocating healthy diets high in veggies, low in processed foods, meals high in fiber and nutrient density. We should be teaching doctors about healthy diet and activity and be instructing them on how to prescribe and support patients in changing overall eating habits.

    -Dr. Peoples

    ReplyDelete
    Replies
    1. Right. I agree with many of your points. The reason I thought it was an interesting meta-analysis (and I think you already know my feelings on meta-analyses in general) was precisely because it FAILED to show the sort of large effect which would justify all the public hand-wringing and popular concern about added salt which seems to obsess the media, patients, and many doctors. You would think that, if this actually mattered, a meta-analysis of 7 large trials would have showed something interesting; to me, the fact that it didn't means that I can stop bugging my patients about eating salt and, in some cases, reassure them that they can preserve this as a pleasurable domain of eating in the face of other distasteful changes I'm asking them to try, like, say, not drinking 3 liters of Coke every day.

      Delete