Here we go again. The same month that it approved Entresto (look here), the US Food and Drug Administration approved a new PCSK9 inhibitor cholesterol lowering agent, alirocumab, immediately marketed as the pricy Praluent by Sanofi and Regeneron, and heralded by a blast of media hype. Yet the evidence that this drug benefits patients is lacking, and critical review of the one big published randomized controlled trial of it raises many concerns.
Media Hype
Let us first consider the media hype. The TIME coverage started with this headline,
The New York Times article by Andrew Pollack quoted Katherine Wilemon, founder and president of the FH foundation, an advocacy group for patients with familial hypercholesterolemia, who have very high cholesterol values and increased risk of heart and vascular disease,
The Washington Post article started with,
USA Today reported,
Another NY Times article by Gina Kolata directly described the drug as
Ms Colata also quoted a cardiologist who characterized the drug again as a "game changer."
To be fair, note that while the WaPo article, NYT article by Pollack and the USA Today article provided hype attributed to "doctors," or identified individuals, they also quoted some people who were very skeptical about the drug. However, in most of the media coverage, the positivity seemed to be more prominent and extreme than the skepticism.
The High Price
In general, the media coverage noted that the "breakthrough," "blockbuster," "powerful" new drug would not come cheap. Praluent would cost about $14,600 a year. Naturally, those selling it saw this as a bargain. For example, Andrew Pollack wrote in the NYT,
Gina Kolata went farther,
She justified this by comparing the cost to apheresis, a radical procedure to treat high cholesterol. She did not discuss whether it had any evidence of clinical benefit. Yet,
Presumably, Dr Soffer was the one who had recommended the apheresis treatment.
Note that at best, the company that sells this drug can justify the price only in terms of potential, not actual value or results.
No Evidence for Clinical Benefit
Praluent, generic name alirocumb, is certainly a breakthrough in that it seeks to lower cholesterol through a novel mechanism. The drug is a biologic, a monoclonal antibody that inhibits the enzyme PCSK9.
Yet a close reading of the one large published randomized controlled trial of alirocumab(1) belays the hype beyond that. The study by Robinson et al was a double blind randomized controlled trial of alirocumab injections every 2 weeks versus placebo. The protocol called for patients to be treated for 78 weeks, and followed for 8 more weeks, a bit more than one and one-half years.
Loss to Follow Up and Missing Data
The study enrolled 1553 patients in the alirocumab group, and 788 in the placebo group. However, many patients did not complete the study: a total of 437 (28.1%) in the alirocumab group, and 193 (24.5% in the placebo group). Reasons for noncompletion were adverse events (113, 7.2% alirocumab vs 47, 6.0% placebo); "nonadherent" to treatment (60, 3.9% vs 38, 4.7%), and "other reason," (264, 17.0% vs 108, 13.7%).
So the drop out rate was fairly high. It was particularly troubling that the reasons for most of the drop outs were vague "other reaons." I could not find a clarification of this term in the main article or supplemental materials.
Furthermore, it was not clear how the investigators intended to collect data from patients after they dropped out, and how complete data collection about clinical events was for patients who dropped out. (Note that for patients that dropped out, the investigators simply imputed, that is estimated cholesterol values, but did not necessarily measure them. So even this measure was "potential.")
Drop outs and missing data are classically problematic because patients may drop out after suffering events that could be counted as study outcomes. The rate of these events could differ according to treatment group. If patients who dropped out of the alirocumab group had more adverse events than those who dropped out of the placebo group, and these events were not recorded, the high drop out rate could have concealed important harms of the drug.
Thus it is quite possible that the study by Robinson et al undercounted adverse events due to aliromucab.
Multiple Study Sites
The study enrolled patients at a remarkable number of sites, 320 in 27 countries, so that the average number of patients enrolled per site was only seven. It seems improbable that a study involving so many investigators and centers, most of whom must have devoted little of their time and effort to this particular study, would have adequate quality control. I could not find a discussion of implementation quality control in the published article.
Thus it is possible that poor quality of study implementation, which could have affected enrollment and data collection, may have challenged the validity of the Robinson et al study.
Lack of Generalizability in the Patient Population
The complete list of exclusion criteria, only appearing in the supplementary material, was extensive. Patients with many common problems were supposed to be excluded, and the definition of the some exclusion criteria were vague and subjective.
Common conditions leading to exclusion were:
- Recent heart and cardiovascular problems, i.e., "(within 3 months prior to the screening visit [Week -3] or between screening and randomization visits) MI, unstable angina leading to hospitalization, uncontrolled cardiac arrhythmia, CABG, PCI, carotid surgery or stenting, cerebrovascular accident, transient ischemic attack, endovascular procedure or surgical intervention for peripheral vascular disease."
- "Planned to undergo scheduled PCI, CABG, carotid or peripheral revascularization during the study"
- Severe congestive heart failure, i.e., "New York Heart Association Class III or IV heart failure within the past 12 months"
- Poorly controlled hypertension, i.e., "Systolic blood pressure >180 mmHg or diastolic blood pressure >110 mmHg at screening visit or randomization visit."
- "History of hemorrhagic stroke."
- "History of active optic nerve disease."
- Use of systemic corticosteroids, other than for replacement
- "History of cancer within the past 5 years, except for adequately treated basal cell skin cancer, squamous cell skin cancer, or in situ cervical cancer."
- "History of HIV positivity."
- "Positive test for Hepatitis B surface antigen and/or Hepatitis C antibody (confirmed by reflexive testing)."
- Kidney dysfunction, specifically, "eGFR <30 nbsp="" p="">- Poorly controlled diabetes, specifically, HbA1c >10%.
- Abnormal liver enzymes, specifically, ALT or AST > x ULN
Vaguely described exclusions were:
Also,
Thus the study would have excluded patients with a variety of common conditions, and may have excluded many other patients based on rather poorly defined decisions by individual investigators. Since patients in clinical practice commonly have common conditions, the generalizability of the results of this study to many practices and patients was not clear.
No Evidence of Clinical Benefit
Patients should not be subject to treatments whose benefits do not clearly outweigh their harms. The Robinson et al article focused on reductions in measured cholesterol, particularly LDL cholesterol. The new drug certainly did seem to clearl reduce cholesterol, particularly LDL cholesterol. However, these are only the results of laboratory tests.
Although high cholesterol and high LDL cholesterol indicate increased risk of future cardiac events, many patients with abnormal values do not have such events. Having a high cholesterol or LDL cholesterol does not directly cause symptoms, or dysfunction. Thus simply lowering cholesterol does not immediately or directly benefit patients. Furthermore, other drug have been shown to lower cholesterol, but ultimately they accomplished this without ever being shown to benefit patients, e.g., by preventing heart attacks, strokes, or premature death.
However, cholesterol values are considered intermediate or surrogate variables. They are not directly related to what happens to patients, who they feel or function, whether they get new diseases, or when they die. So only showing that the new drug lowers cholesterol does not prove clinical patient benefit.
Although the published trial did attempt to record cardiovascular events, it did not find that the drug prevented them. The small difference in total cardiovascular events affecting patients given alirocumab (4.6%) versus placebo (5.1%) did not reach statistical significance, that is, could well have been due to chance alone.
Furthermore, while elevated cholesterol is a chronic problem, and the problems with which it is correlated occur over the long run, the study ran for less than 2 years. It could not measure the effects of the new drug beyond that.
So the clinical benefit of the drug was not evident in this trial.
On the other hand, the drug was not without its own risks. More patients who received aliromucab left the study due to adverse events (7.2%) thand did those who got placebo (5.8%), as noted above. Also, as noted above, it was possible that adverse events affecting dropouts were not fully recorded. Given that there were higher rates of dropouts due to non adherence and "other" reasons among patients who received alirocumab, the study might still have missed important adverse effects of the new drug.
So the study did not prove that the new drug has any clinical benefits, showed it does have clinical harms, and could still have easily underestimated its harms. So it certainly did not show it had benefits that outweighed its harms.
Summary and Conclusions
The NEJM study was accompanied by an editorial by Stone and Lloyd-Jones(2) which documented that drugs previously shown to lower cholesterol were never proved to do any good for patients, and concluded,
After an FDA advisory committee recommended approval of aliromucab in June, 2015, John Mandrola entitled a Medscape article,
His reasons included lack of proof of clinical benefits, and concerns that harms may have been missed but mainly because of its inability to detect long-term outcomes.
Again, the current media articles also noted the concerns raised by Dr Mandrola and the NEJM editorial These concerns, however, did not dissuade the FDA from approving aliromucab. These concerns did not apparently affect the pricing of Praluent. These concerns will likely not deter the drug manufacturers from continuing an aggressive marketing campaign. Whether these concerns will deter physicians from prescribing, or patients from asking for these drugs is unknown, but unlikely.
And I have not seen anything published so far that addressed how the problem with dropouts and missing data may have lead to further underestimation of aliromucab's harms, the multiplicity of study sites may have lead to quality control problems further challenging the study's validity, and the extensive exclusion criteria may have reduced the study's generalizability.
So here we go again. Another new drug is put on the market accompanied by a mighty hoopla, yet in the absence of clear data that it does more good for patients than harm.
As we said last year about valsartan-sacubitril, also just (July, 2015) put on the market as Entresto, at a high price and with lots of hype,...
All the enthusiasm about this drug may be premature, and does not appear to be evidence-based. That clinical research sponsored by organizations that sell health care goods and services may be manipulated to make the sponsors' products look better than they really are is now an old story. We have seen multiple instances in which drugs and devices turned out to be less efficacious and/or more dangerous than originally advertised. Excess enthusiasm about such new innovations may drive up costs, and worse, hurt patients. Physicians, other health care professionals, and those concerned about health policy ought to be much more skeptical about every new instance of a purportedly wondrous innovation.
Evidence-based medicine rigorously applied suggests that individual health care and health policy decisions should be driven by the best available evidence, mostly from clinical research, about the benefits and harms of tests, treatments, programs, and so on, in the context of what outcomes matter to patients. The skepticism EBM should engender could lead to health care that is more about patients and their outcomes, and less about ideology, hype, and hucksterism.
How high must our health care costs go, and how many unproven treatments must eventually be exposed as such before we learn that lesson?
ADDENDUM (6 August, 2015) - Fixed minor errors: misspelling of Ms Kolata's name fixed, misstatement re apharesis fixed, erroneous reference to second approved PCSK9 inhibitor removed.
ADDENDUM (9 August, 2015) - Note that his post was republished on the Naked Capitalism blog.
References
1. Robinson JG, Farnier M, Krempf M et al. Efficacy and safety of alirocumab in reduincg lipids and cardiovascular events. N Engl J Med 2015; 372: 1489-1499. Link here.
2. Stone NJ, Lloyd-Jones DM. Lowering LDL cholestero is good, but how in whom? N Engl J Med 2015; 372: 1564-5. Link here.
Media Hype
Let us first consider the media hype. The TIME coverage started with this headline,
This New FDA-Approved Cholesterol Drug is a Game Changer
The New York Times article by Andrew Pollack quoted Katherine Wilemon, founder and president of the FH foundation, an advocacy group for patients with familial hypercholesterolemia, who have very high cholesterol values and increased risk of heart and vascular disease,
It represents a new era of hope for us.
The Washington Post article started with,
The Food and Drug Administration on Friday approved the first in a new class of cholesterol-busting drugs that many doctors believe will trigger a breakthrough in reducing the incidence of strokes and heart attacks, which kill hundreds of thousands of Americans each year.
USA Today reported,
The drugs are predicted to be blockbusters many times over, adding billions of dollars to prescription drug costs, said Steve Miller, senior vice president and chief medical officer at Express Scripts, a leading pharmacy benefit manager.
Another NY Times article by Gina Kolata directly described the drug as
powerful almost beyond belief.
Ms Colata also quoted a cardiologist who characterized the drug again as a "game changer."
To be fair, note that while the WaPo article, NYT article by Pollack and the USA Today article provided hype attributed to "doctors," or identified individuals, they also quoted some people who were very skeptical about the drug. However, in most of the media coverage, the positivity seemed to be more prominent and extreme than the skepticism.
The High Price
In general, the media coverage noted that the "breakthrough," "blockbuster," "powerful" new drug would not come cheap. Praluent would cost about $14,600 a year. Naturally, those selling it saw this as a bargain. For example, Andrew Pollack wrote in the NYT,
Sanofi and Regeneron Pharmaceuticals, which developed the product, said the price was justified by the potential benefits to patients and savings to the health care system that the drug would provide by preventing heart attacks and strokes — though the ability of the drug to do that has not been proved.
'We came to a price that is reflective of value, not what the market will bear,' said Elias Zerhouni, head of research and development at Sanofi, who said his own brother had suffered three heart attacks and needed new options to control cholesterol.
Gina Kolata went farther,
The $14,600 yearly price of the drug, which is injected under the skin once every two weeks, is a stunner. Yet for some patients, that might actually be a bargain.
She justified this by comparing the cost to apheresis, a radical procedure to treat high cholesterol. She did not discuss whether it had any evidence of clinical benefit. Yet,
'Cost is in the eye of the beholder,' said Dr. Daniel Soffer, [Mr. DeRuchie’s cardiologist at the University of Pennsylvania.
Presumably, Dr Soffer was the one who had recommended the apheresis treatment.
Note that at best, the company that sells this drug can justify the price only in terms of potential, not actual value or results.
No Evidence for Clinical Benefit
Praluent, generic name alirocumb, is certainly a breakthrough in that it seeks to lower cholesterol through a novel mechanism. The drug is a biologic, a monoclonal antibody that inhibits the enzyme PCSK9.
Yet a close reading of the one large published randomized controlled trial of alirocumab(1) belays the hype beyond that. The study by Robinson et al was a double blind randomized controlled trial of alirocumab injections every 2 weeks versus placebo. The protocol called for patients to be treated for 78 weeks, and followed for 8 more weeks, a bit more than one and one-half years.
Loss to Follow Up and Missing Data
The study enrolled 1553 patients in the alirocumab group, and 788 in the placebo group. However, many patients did not complete the study: a total of 437 (28.1%) in the alirocumab group, and 193 (24.5% in the placebo group). Reasons for noncompletion were adverse events (113, 7.2% alirocumab vs 47, 6.0% placebo); "nonadherent" to treatment (60, 3.9% vs 38, 4.7%), and "other reason," (264, 17.0% vs 108, 13.7%).
So the drop out rate was fairly high. It was particularly troubling that the reasons for most of the drop outs were vague "other reaons." I could not find a clarification of this term in the main article or supplemental materials.
Furthermore, it was not clear how the investigators intended to collect data from patients after they dropped out, and how complete data collection about clinical events was for patients who dropped out. (Note that for patients that dropped out, the investigators simply imputed, that is estimated cholesterol values, but did not necessarily measure them. So even this measure was "potential.")
Drop outs and missing data are classically problematic because patients may drop out after suffering events that could be counted as study outcomes. The rate of these events could differ according to treatment group. If patients who dropped out of the alirocumab group had more adverse events than those who dropped out of the placebo group, and these events were not recorded, the high drop out rate could have concealed important harms of the drug.
Thus it is quite possible that the study by Robinson et al undercounted adverse events due to aliromucab.
Multiple Study Sites
The study enrolled patients at a remarkable number of sites, 320 in 27 countries, so that the average number of patients enrolled per site was only seven. It seems improbable that a study involving so many investigators and centers, most of whom must have devoted little of their time and effort to this particular study, would have adequate quality control. I could not find a discussion of implementation quality control in the published article.
Thus it is possible that poor quality of study implementation, which could have affected enrollment and data collection, may have challenged the validity of the Robinson et al study.
Lack of Generalizability in the Patient Population
The complete list of exclusion criteria, only appearing in the supplementary material, was extensive. Patients with many common problems were supposed to be excluded, and the definition of the some exclusion criteria were vague and subjective.
Common conditions leading to exclusion were:
- Recent heart and cardiovascular problems, i.e., "(within 3 months prior to the screening visit [Week -3] or between screening and randomization visits) MI, unstable angina leading to hospitalization, uncontrolled cardiac arrhythmia, CABG, PCI, carotid surgery or stenting, cerebrovascular accident, transient ischemic attack, endovascular procedure or surgical intervention for peripheral vascular disease."
- "Planned to undergo scheduled PCI, CABG, carotid or peripheral revascularization during the study"
- Severe congestive heart failure, i.e., "New York Heart Association Class III or IV heart failure within the past 12 months"
- Poorly controlled hypertension, i.e., "Systolic blood pressure >180 mmHg or diastolic blood pressure >110 mmHg at screening visit or randomization visit."
- "History of hemorrhagic stroke."
- "History of active optic nerve disease."
- Use of systemic corticosteroids, other than for replacement
- "History of cancer within the past 5 years, except for adequately treated basal cell skin cancer, squamous cell skin cancer, or in situ cervical cancer."
- "History of HIV positivity."
- "Positive test for Hepatitis B surface antigen and/or Hepatitis C antibody (confirmed by reflexive testing)."
- Kidney dysfunction, specifically, "eGFR <30 nbsp="" p="">- Poorly controlled diabetes, specifically, HbA1c >10%.
- Abnormal liver enzymes, specifically, ALT or AST > x ULN
Vaguely described exclusions were:
E 25. Conditions/situations such as:
A) Any clinically significant abnormality identified at the time of screening that in the
judgment of the Investigator or any sub-Investigator would preclude safe completion
of the study or constrain endpoints assessment such as major systemic diseases,
patients with short life expectancy.
B) Patients considered by the Investigator or any sub-Investigator as inappropriate
for this study for any reason, e.g.:
i) Those deemed unable to meet specific protocol requirements, such as scheduled
visits.
ii) Those deemed unable to administer or tolerate long-term injections as per the
patient or the investigator.
Also,
iv) Presence of any other conditions (eg, geographic, social….) actual or anticipated,
that the Investigator feels would restrict or limit the patient’s participation for the
duration of the study.
Thus the study would have excluded patients with a variety of common conditions, and may have excluded many other patients based on rather poorly defined decisions by individual investigators. Since patients in clinical practice commonly have common conditions, the generalizability of the results of this study to many practices and patients was not clear.
No Evidence of Clinical Benefit
Patients should not be subject to treatments whose benefits do not clearly outweigh their harms. The Robinson et al article focused on reductions in measured cholesterol, particularly LDL cholesterol. The new drug certainly did seem to clearl reduce cholesterol, particularly LDL cholesterol. However, these are only the results of laboratory tests.
Although high cholesterol and high LDL cholesterol indicate increased risk of future cardiac events, many patients with abnormal values do not have such events. Having a high cholesterol or LDL cholesterol does not directly cause symptoms, or dysfunction. Thus simply lowering cholesterol does not immediately or directly benefit patients. Furthermore, other drug have been shown to lower cholesterol, but ultimately they accomplished this without ever being shown to benefit patients, e.g., by preventing heart attacks, strokes, or premature death.
However, cholesterol values are considered intermediate or surrogate variables. They are not directly related to what happens to patients, who they feel or function, whether they get new diseases, or when they die. So only showing that the new drug lowers cholesterol does not prove clinical patient benefit.
Although the published trial did attempt to record cardiovascular events, it did not find that the drug prevented them. The small difference in total cardiovascular events affecting patients given alirocumab (4.6%) versus placebo (5.1%) did not reach statistical significance, that is, could well have been due to chance alone.
Furthermore, while elevated cholesterol is a chronic problem, and the problems with which it is correlated occur over the long run, the study ran for less than 2 years. It could not measure the effects of the new drug beyond that.
So the clinical benefit of the drug was not evident in this trial.
On the other hand, the drug was not without its own risks. More patients who received aliromucab left the study due to adverse events (7.2%) thand did those who got placebo (5.8%), as noted above. Also, as noted above, it was possible that adverse events affecting dropouts were not fully recorded. Given that there were higher rates of dropouts due to non adherence and "other" reasons among patients who received alirocumab, the study might still have missed important adverse effects of the new drug.
So the study did not prove that the new drug has any clinical benefits, showed it does have clinical harms, and could still have easily underestimated its harms. So it certainly did not show it had benefits that outweighed its harms.
Summary and Conclusions
The NEJM study was accompanied by an editorial by Stone and Lloyd-Jones(2) which documented that drugs previously shown to lower cholesterol were never proved to do any good for patients, and concluded,
it would be premature to endorse these drugs for widespread use before the ongoing randomized trials, appropriately powered for primary end-point analysis and safety assessment, are available.
After an FDA advisory committee recommended approval of aliromucab in June, 2015, John Mandrola entitled a Medscape article,
Dear FDA: Resist the Urge on PCSK9 Drugs
His reasons included lack of proof of clinical benefits, and concerns that harms may have been missed but mainly because of its inability to detect long-term outcomes.
Again, the current media articles also noted the concerns raised by Dr Mandrola and the NEJM editorial These concerns, however, did not dissuade the FDA from approving aliromucab. These concerns did not apparently affect the pricing of Praluent. These concerns will likely not deter the drug manufacturers from continuing an aggressive marketing campaign. Whether these concerns will deter physicians from prescribing, or patients from asking for these drugs is unknown, but unlikely.
And I have not seen anything published so far that addressed how the problem with dropouts and missing data may have lead to further underestimation of aliromucab's harms, the multiplicity of study sites may have lead to quality control problems further challenging the study's validity, and the extensive exclusion criteria may have reduced the study's generalizability.
So here we go again. Another new drug is put on the market accompanied by a mighty hoopla, yet in the absence of clear data that it does more good for patients than harm.
As we said last year about valsartan-sacubitril, also just (July, 2015) put on the market as Entresto, at a high price and with lots of hype,...
All the enthusiasm about this drug may be premature, and does not appear to be evidence-based. That clinical research sponsored by organizations that sell health care goods and services may be manipulated to make the sponsors' products look better than they really are is now an old story. We have seen multiple instances in which drugs and devices turned out to be less efficacious and/or more dangerous than originally advertised. Excess enthusiasm about such new innovations may drive up costs, and worse, hurt patients. Physicians, other health care professionals, and those concerned about health policy ought to be much more skeptical about every new instance of a purportedly wondrous innovation.
Evidence-based medicine rigorously applied suggests that individual health care and health policy decisions should be driven by the best available evidence, mostly from clinical research, about the benefits and harms of tests, treatments, programs, and so on, in the context of what outcomes matter to patients. The skepticism EBM should engender could lead to health care that is more about patients and their outcomes, and less about ideology, hype, and hucksterism.
How high must our health care costs go, and how many unproven treatments must eventually be exposed as such before we learn that lesson?
ADDENDUM (6 August, 2015) - Fixed minor errors: misspelling of Ms Kolata's name fixed, misstatement re apharesis fixed, erroneous reference to second approved PCSK9 inhibitor removed.
ADDENDUM (9 August, 2015) - Note that his post was republished on the Naked Capitalism blog.
References
1. Robinson JG, Farnier M, Krempf M et al. Efficacy and safety of alirocumab in reduincg lipids and cardiovascular events. N Engl J Med 2015; 372: 1489-1499. Link here.
2. Stone NJ, Lloyd-Jones DM. Lowering LDL cholestero is good, but how in whom? N Engl J Med 2015; 372: 1564-5. Link here.
