On 2016 Dec 11, Martin Mayer commented:

Cut the fat: Putting the risks of hypertriglyceridemia into context

A brief response to “Nonfasting mild-to-moderate hypertriglyceridemia and risk of acute pancreatitis”

In their article, Pedersen and colleagues present findings from their prospective cohort study on hypertriglyceridemia and its association with both acute pancreatitis and myocardial infarction.¹ With a median follow-up of 6.7 years (interquartile range, 4.0 to 9.4 years) among 116,550 "white individuals of Danish descent from the Danish general population"^1(p1835) selected randomly from two similar prospective studies (the Copenhagen City Heart Study and the Copenhagen General Population Study), this is a sizable study with respectable follow-up, even if generalizability of the findings might be at least somewhat limited. They rightly note “there is no consensus on a clear threshold above which triglycerides are associated with acute pancreatitis,”^1(p1835) and others have highlighted important issues with the evidence base.² Pedersen and colleagues also cite a review³ on triglycerides and cardiovascular disease, but here too the evidence is not entirely clear; the review only concludes evidence “is increasing”^3(p633) and recommends high-intensity statin therapy. The review also considers the future potential of add-on triglyceride-lowering therapy for those already on a statin, pointing to two ongoing trials of ω-3 fatty acids (REDUCE-IT and STRENGTH). However, the currently-available evidence - particularly that with patient-relevant outcomes - does not support such a strategy for ω-3 fatty acids or other agents that can substantially lower triglycerides (such as fibrates and niacin).^2,4,5

Even if their study reflects an underlying truth, Pedersen and colleagues unfortunately demonstrate a relative inattention to absolute risks and the implications thereof. They devote a small amount of text to absolute risks and report absolute numbers in the figures, but they repeatedly state their findings show “high risk” for acute pancreatitis, a perspective seemingly driven by the magnitude of the hazard ratios (HRs). In their concluding statements, they even remark: “Mild-to-moderate hypertriglyceridemia at 177 mg/dL (2 mmol/L) and above is associated with high risk of acute pancreatitis in the general population, with HRs higher than for myocardial infarction.”^1(p1841)

When caring for individual patients, relative metrics such as HRs are most useful when appropriately applied to corresponding baseline absolute risks. Conversely, disproportionate focus on relative metrics or failure to adequately contextualize relative metrics with corresponding absolute risks is considerably less informative and can contribute to a distorted sense of reality. Even if one accepts research findings as being likely reflective of an underlying truth, one must always carefully appraise absolute risks to gain a finer appreciation of the quantitative implications of the research findings. This practice is still useful even if one finds weaknesses in methodology, as one can simply consider the estimates increasingly uncertain in a manner qualitatively proportional to the weaknesses in methodology. A tool customized for this study is available here (TinyURL: http://tinyurl.com/JAMAIMhypertrigcalctool).

According to their own data, comparing the lowest triglyceride level group (<89 mg/dL or <1 mmol/L) to the highest triglyceride level group (≥443 mg/dL or ≥5 mmol/L), one finds an absolute risk difference (ARD) for acute pancreatitis of 0.93% over 10 years if using the absolute numbers reported in Figure 1 to estimate absolute risks, and an ARD of 2.05% over 10 years (95% confidence interval [CI], 0.73% to 4.99%) if using the absolute risk in the lowest triglyceride level group and the multivariable-adjusted HR estimate for the highest triglyceride level group (HR 8.7; 95% CI, 3.7 to 20). Repeating this for myocardial infarction, one finds an ARD of 5.6% over 10 years or an ARD of 5.08% (95% CI, 3.00% to 7.73%) over 10 years. This demonstrates at least one reason why it is important to put relative metrics into context: Although the HRs for acute pancreatitis may be “higher than for myocardial infarction”,^1(p1841) the absolute risks and absolute risk differences are higher for myocardial infarction. Additionally, it is more informative to provide risk estimates in absolute terms than in relative terms. Indeed, as aforementioned, absolute risks give better insight into what research might mean for a patient if one accepts the findings as being reflective of an underlying truth. Unfortunately, The New York Times' coverage of the study exacerbates the issue, with the only attempt to contextualize the relative metrics being a quote from one of the study’s authors. (Such mishandling of evidence is not uncommon in the media, but that is not the focus of this commentary. Including The New York Times’ coverage is not meant to single them out as uniquely bad or good in this regard; it simply serves as an example.) It is ultimately a disservice to say the risk of pancreatitis was 770% higher in patients with triglycerides ≥443 mg/dL (≥5 mmol/L) compared to patients with triglycerides <89 mg/dL (<1 mmol/L) without contextualizing such a metric with absolute risks. More technically, and as discussed in the tool, HRs are also not quite the same as relative risks.

Lastly, while management was not a focus Pedersen and colleagues’ article, sensible lifestyle changes should be emphasized wherever poor lifestyle factors exist. As for interventions beyond lifestyle changes, a medication that can reduce cardiovascular risk – such as a statin – might be instituted after shared decision-making concerning a person’s cardiovascular risk estimate; importantly, however, a person’s cardiovascular risk estimate is not dependent on triglyceride levels, and pharmaceutical intervention targeted at lowering triglycerides per se is not clearly supported by currently-available evidence examining cardiovascular, pancreatic, or other patient-relevant outcomes.  

References

(1) Pedersen SB, Langsted A, Nordestgaard BG. Nonfasting mild-to-moderate hypertriglyceridemia and risk of acute pancreatitis. JAMA Intern Med. 2016 Dec 1;176(12):1834-1842. doi: 10.1001/jamainternmed.2016.6875.

(2) Lederle FA, Bloomfield HE. Drug treatment of asymptomatic hypertriglyceridemia to prevent pancreatitis: where is the evidence? Ann Intern Med. 2012 Nov 6;157(9):662-664. doi: 10.7326/0003-4819-157-9-201211060-00011.

(3) Nordestgaard BG, Varbo A. Triglycerides and cardiovascular disease. Lancet. 2014;384(9943):626-635.

(4) Rizos EC, Ntzani EE, Bika E, Kostapanos MS, Elisaf MS. Association between omega-3 fatty acid supplementation and risk of major cardiovascular disease events: a systematic review and meta-analysis. JAMA. 2012 Sep 12;308(10):1024-1033. doi: 10.1001/2012.jama.11374.

(5) Keene D, Price C, Shun-Shin MJ, Francis DP. Effect on cardiovascular risk of high density lipoprotein targeted drug treatments niacin, fibrates, and CETP inhibitors: meta-analysis of randomised controlled trials including 117,411 patients. BMJ. 2014 Jul 18;349:g4379. doi: 10.1136/bmj.g4379. (Note about this reference: Although the title implies focus on HDL as a therapeutic target, this study nevertheless provides meaningful insight into whether there is any cardiovascular or mortality benefit from adding either niacin or a fibrate to statin therapy, and both these agents can substantially lower triglycerides.)

On 2016 Dec 11, Martin Mayer commented:

Cut the fat: Putting the risks of hypertriglyceridemia into context

A brief response to “Nonfasting mild-to-moderate hypertriglyceridemia and risk of acute pancreatitis”

In their article, Pedersen and colleagues present findings from their prospective cohort study on hypertriglyceridemia and its association with both acute pancreatitis and myocardial infarction.¹ With a median follow-up of 6.7 years (interquartile range, 4.0 to 9.4 years) among 116,550 "white individuals of Danish descent from the Danish general population"^1(p1835) selected randomly from two similar prospective studies (the Copenhagen City Heart Study and the Copenhagen General Population Study), this is a sizable study with respectable follow-up, even if generalizability of the findings might be at least somewhat limited. They rightly note “there is no consensus on a clear threshold above which triglycerides are associated with acute pancreatitis,”^1(p1835) and others have highlighted important issues with the evidence base.² Pedersen and colleagues also cite a review³ on triglycerides and cardiovascular disease, but here too the evidence is not entirely clear; the review only concludes evidence “is increasing”^3(p633) and recommends high-intensity statin therapy. The review also considers the future potential of add-on triglyceride-lowering therapy for those already on a statin, pointing to two ongoing trials of ω-3 fatty acids (REDUCE-IT and STRENGTH). However, the currently-available evidence - particularly that with patient-relevant outcomes - does not support such a strategy for ω-3 fatty acids or other agents that can substantially lower triglycerides (such as fibrates and niacin).^2,4,5

Even if their study reflects an underlying truth, Pedersen and colleagues unfortunately demonstrate a relative inattention to absolute risks and the implications thereof. They devote a small amount of text to absolute risks and report absolute numbers in the figures, but they repeatedly state their findings show “high risk” for acute pancreatitis, a perspective seemingly driven by the magnitude of the hazard ratios (HRs). In their concluding statements, they even remark: “Mild-to-moderate hypertriglyceridemia at 177 mg/dL (2 mmol/L) and above is associated with high risk of acute pancreatitis in the general population, with HRs higher than for myocardial infarction.”^1(p1841)

When caring for individual patients, relative metrics such as HRs are most useful when appropriately applied to corresponding baseline absolute risks. Conversely, disproportionate focus on relative metrics or failure to adequately contextualize relative metrics with corresponding absolute risks is considerably less informative and can contribute to a distorted sense of reality. Even if one accepts research findings as being likely reflective of an underlying truth, one must always carefully appraise absolute risks to gain a finer appreciation of the quantitative implications of the research findings. This practice is still useful even if one finds weaknesses in methodology, as one can simply consider the estimates increasingly uncertain in a manner qualitatively proportional to the weaknesses in methodology. A tool customized for this study is available here (TinyURL: http://tinyurl.com/JAMAIMhypertrigcalctool).

According to their own data, comparing the lowest triglyceride level group (<89 mg/dL or <1 mmol/L) to the highest triglyceride level group (≥443 mg/dL or ≥5 mmol/L), one finds an absolute risk difference (ARD) for acute pancreatitis of 0.93% over 10 years if using the absolute numbers reported in Figure 1 to estimate absolute risks, and an ARD of 2.05% over 10 years (95% confidence interval [CI], 0.73% to 4.99%) if using the absolute risk in the lowest triglyceride level group and the multivariable-adjusted HR estimate for the highest triglyceride level group (HR 8.7; 95% CI, 3.7 to 20). Repeating this for myocardial infarction, one finds an ARD of 5.6% over 10 years or an ARD of 5.08% (95% CI, 3.00% to 7.73%) over 10 years. This demonstrates at least one reason why it is important to put relative metrics into context: Although the HRs for acute pancreatitis may be “higher than for myocardial infarction”,^1(p1841) the absolute risks and absolute risk differences are higher for myocardial infarction. Additionally, it is more informative to provide risk estimates in absolute terms than in relative terms. Indeed, as aforementioned, absolute risks give better insight into what research might mean for a patient if one accepts the findings as being reflective of an underlying truth. Unfortunately, The New York Times' coverage of the study exacerbates the issue, with the only attempt to contextualize the relative metrics being a quote from one of the study’s authors. (Such mishandling of evidence is not uncommon in the media, but that is not the focus of this commentary. Including The New York Times’ coverage is not meant to single them out as uniquely bad or good in this regard; it simply serves as an example.) It is ultimately a disservice to say the risk of pancreatitis was 770% higher in patients with triglycerides ≥443 mg/dL (≥5 mmol/L) compared to patients with triglycerides <89 mg/dL (<1 mmol/L) without contextualizing such a metric with absolute risks. More technically, and as discussed in the tool, HRs are also not quite the same as relative risks.

Lastly, while management was not a focus Pedersen and colleagues’ article, sensible lifestyle changes should be emphasized wherever poor lifestyle factors exist. As for interventions beyond lifestyle changes, a medication that can reduce cardiovascular risk – such as a statin – might be instituted after shared decision-making concerning a person’s cardiovascular risk estimate; importantly, however, a person’s cardiovascular risk estimate is not dependent on triglyceride levels, and pharmaceutical intervention targeted at lowering triglycerides per se is not clearly supported by currently-available evidence examining cardiovascular, pancreatic, or other patient-relevant outcomes.  

References

(1) Pedersen SB, Langsted A, Nordestgaard BG. Nonfasting mild-to-moderate hypertriglyceridemia and risk of acute pancreatitis. JAMA Intern Med. 2016 Dec 1;176(12):1834-1842. doi: 10.1001/jamainternmed.2016.6875.

(2) Lederle FA, Bloomfield HE. Drug treatment of asymptomatic hypertriglyceridemia to prevent pancreatitis: where is the evidence? Ann Intern Med. 2012 Nov 6;157(9):662-664. doi: 10.7326/0003-4819-157-9-201211060-00011.

(3) Nordestgaard BG, Varbo A. Triglycerides and cardiovascular disease. Lancet. 2014;384(9943):626-635.

(4) Rizos EC, Ntzani EE, Bika E, Kostapanos MS, Elisaf MS. Association between omega-3 fatty acid supplementation and risk of major cardiovascular disease events: a systematic review and meta-analysis. JAMA. 2012 Sep 12;308(10):1024-1033. doi: 10.1001/2012.jama.11374.

(5) Keene D, Price C, Shun-Shin MJ, Francis DP. Effect on cardiovascular risk of high density lipoprotein targeted drug treatments niacin, fibrates, and CETP inhibitors: meta-analysis of randomised controlled trials including 117,411 patients. BMJ. 2014 Jul 18;349:g4379. doi: 10.1136/bmj.g4379. (Note about this reference: Although the title implies focus on HDL as a therapeutic target, this study nevertheless provides meaningful insight into whether there is any cardiovascular or mortality benefit from adding either niacin or a fibrate to statin therapy, and both these agents can substantially lower triglycerides.)