The Diuretic Duel: Still 2 Options?
Medscape: Why did you feel that it would be useful to carry out this study of thiazide diuretic use in hypertensive patients treated in Ontario? How did you predict it would add to the evidence already accumulated in the chlorthalidone vs HCTZ debate?
Dr. Dhalla: Hypertension is an extremely common, serious condition, and diuretics are frequently used to treat it. In North America, probably the 2 most commonly prescribed diuretics are HCTZ and chlorthalidone. Many, if not most, patients with hypertension will eventually end up on a diuretic at some point, and so it would be useful to know whether one of those 2 drugs is better than the other.
I think that most physicians take the view that ultimately, this kind of a question can only be definitively settled by a large, well-designed, and well-executed randomized controlled trial. Unfortunately, we were not in a position to carry out such a trial. However, in Ontario, we have access to a great deal of administrative data, and so we were able to compare safety and effectiveness outcomes in individuals who were prescribed HCTZ with safety and effectiveness outcomes with individuals who were prescribed chlorthalidone. In my view, it is nowhere near as good as a randomized controlled trial would be, but it is a "second-best" kind of study design, if you will.
Medscape: Could you tell us more about the origins of the data you used?
Dr. Dhalla: Routine information is collected on every hospitalization in Canada, and we have access to the information pertaining to all of those hospitalizations in Ontario. Similarly, we have some information pertaining to every physician visit. We also have information regarding every medication that was paid for by the public. Everyone aged 65 years or older has public coverage; therefore, virtually every drug prescribed to someone in that age group is paid for through the public purse, and we have access to those data. These routinely collected administrative data allow researchers such as us to do studies here that are not possible in other jurisdictions.
Medscape: You found that in Ontario, between 1993 and 2010, 643,529 patients aged 65 years or older were prescribed HCTZ compared with 11,389 prescribed chlorthalidone. The relative frequency with which the 2 drugs are prescribed seems similar to that in the United States.
Dr. Dhalla: As far as I'm aware, yes, I think the situation is similar in Canada and the United States.
Medscape: Are Canadian national guidelines for hypertension treatment similar to the Seventh Report of the Joint National Committee (JNC7): that a thiazide diuretic should be first-line therapy in the treatment of uncomplicated hypertension?
Dr. Dhalla: The Canadian guidelines here are less prescriptive. They suggest a range of different options for first-line treatment, and diuretics are among those options. They don't say that thiazides are preferred.
Medscape: To overcome differences between treatment groups, you used high-dimensional propensity-score matching and matched each chlorthalidone recipient with up to 2 HCTZ recipients. Could you explain more about that, and why you didn't just study the rate of events in both populations?
Dr. Dhalla: In recent years, researchers have learned to use more sophisticated analytic techniques when we are doing observational studies to minimize the risk for bias. One way of doing so is to try to mimic a randomized controlled trial. So we matched the HCTZ patients with chlorthalidone patients on certain criteria, including age, sex, and what year the treatment was started.
We also used a large number of other baseline variables -- for example, prescription drug claims, physician services fee codes, physician services diagnosis codes, hospital diagnosis codes, hospital procedure codes, and emergency department diagnosis codes -- to construct a propensity score, which is basically the probability that a patient will be treated with one drug or the other. We then matched patients on the propensity score, and in so doing, created a cohort where the HCTZ patients look very similar to the chlorthalidone patients.
The big difference between an observational study such as ours and a randomized controlled trial is that we are only able to produce balance on the observed baseline characteristics, whereas in a randomized controlled trial you get balance on both the observed characteristics and baseline characteristics that have not been measured. This is one of the main reasons why randomized controlled trials are so powerful. No matter how sophisticated an observational study is, you cannot be confident that you have achieved balance on the unmeasured confounders.
Medscape: What would an unmeasured confounder be in this case?
Dr. Dhalla: It could be something simple, such as the skill of the physician. It may be the case that physicians who are more skillful are more likely to prescribe 1 of the 2 drugs. It may also be the case that patients cared for by more skillful physicians are likely to live longer.
Medscape: You noted in your article that people taking chlorthalidone were likely to also be taking a beta-blocker and less likely to be treated with an angiotensin-converting enzyme (ACE) inhibitor or an angiotensin receptor blocker, possibly because a chlorthalidone/atenolol combination is available in the public formulary in Ontario. Thus, a physician's decisions about what drug to prescribed may not be completely based on personal choice?
Dr. Dhalla: If you were going to design a randomized controlled trial comparing these 2 drugs, you would be very careful about the dose, about medications that were started at the same time, and about monitoring. We obviously could not do that kind of study. So we decided to compare safety and effectiveness outcomes in these 2 groups of patients, basically looking at the 2 drugs as they are usually prescribed. And when you look at how they are usually do this in Ontario, they are actually prescribed in quite different ways. Chlorthalidone is typically prescribed in a much higher dose, and it is much more likely to be prescribed with atenolol. This is because a product on the public formulary here has both active ingredients in 1 pill.
We adjusted for that, and we also did a series of post hoc analyses to address the dose issue, but our main objective was to compare the effectiveness outcomes in these 2 drugs as they are usually prescribed. This is in contrast to comparing safety and effectiveness outcomes in these 2 drugs as an expert in hypertension might prescribe them.
Medscape: Were you surprised to see patients aged 66 years or older being prescribed a diuretic plus a beta-blocker?
Dr. Dhalla: No, we weren't. I think beta-blockers in particular are becoming less popular in older people, and I certainly would not use one as a first-line treatment in older adults. But we looked at data from as far back as the 1990s, and beta-blockers were more commonly prescribed then.
Medscape: The main result of your study, that there appeared to be no difference in cardiovascular events or mortality between chlorthalidone and HCTZ, was surprising, however. What was your reaction?
Dr. Dhalla: We were hoping that chlorthalidone would be associated with better outcomes than HCTZ. Although that wouldn't have settled this question once and for all, it would have provided more evidentiary support for physicians who are inclined to prescribe chlorthalidone over HCTZ. Instead, we were unable to find a difference in effectiveness. It is worth noting, however, that despite our study being so large, it was not large enough to exclude a small yet clinically significant difference between the 2 drugs.
On the safety side, we found that patients treated with chlorthalidone were considerably more likely to end up in hospital with hypokalemia or hyponatremia. That is why in the conclusion of our article, we do not really provide a recommendation to physicians that they should prescribe either chlorthalidone or HCTZ preferentially. What we are saying is that ultimately to settle this question, we would need a very large randomized trial, and in the absence of such a trial, physicians who prescribe HCTZ can probably feel comfortable continuing to prescribe HCTZ. Similarly, physicians who prefer chlorthalidone can probably feel comfortable continuing to prescribe chlorthalidone. And physicians who prescribe either drug should be mindful of the risk for electrolyte abnormalities, particularly immediately after starting the drug and with any illness that predisposes to fluid and electrolyte imbalance.
Medscape: You mentioned in the article that you found, as others have previously reported, that chlorthalidone was being prescribed at higher doses than HCTZ. Was that why you found the greater risk for electrolyte abnormalities with chlorthalidone?
Dr. Dhalla: That is a great question. I would speculate that the reason chlorthalidone was prescribed at a higher dose was simply because of the way it is available on the formulary. There is a 50-mg tablet of chlorthalidone available on the public formulary, but the 25-mg formulation is available only in combination with atenolol. Obviously, if physicians want to prescribe 25 mg of chlorthalidone per day without atenolol, patients can just cut the 50-mg tablet or have the pharmacist cut it for them. Similarly, a physician could prescribe 12.5 mg daily and the tablet could be cut into quarters, although that is not always so easy for older adults to do.
But the availability of the 50-mg formulation may be driving physicians to prescribe a substantially higher dose than they would with HCTZ. That may be part of the reason why we saw more electrolyte abnormalities with chlorthalidone, although in the post hoc dose-specific analyses, we still saw an increased risk for hypokalemia and hyponatremia with chlorthalidone at pretty much all doses. So there may be other reasons that are independent of the dose. For example, the fact that chlorthalidone has a longer half-life may contribute.
Medscape: You mentioned the post hoc analyses. Could you tell us more about these?
Dr. Dhalla: The main post hoc analyses were the ones that compared a particular starting dose of HCTZ with a particular starting dose of chlorthalidone: for example, 25 mg of chlorthalidone with 50 mg of HCTZ. All of the results in the post hoc analyses were consistent with the results in the main analysis.
Medscape: Why do you think a real-world-practice study produced different results from previous analyses and meta-analyses?
Dr. Dhalla: I don't think our results are that different from some of the previous analyses, such as the meta-analysis by Bruce Psaty and his group. But you are right that our results don't draw the same conclusion as some of the other studies that have been published. I would speculate that it may be that there really isn't a big difference between the 2 drugs, especially as they are prescribed in routine practice.
Medscape: The results of such studies as the recent retrospective analysis of MRFIT data have been quoted in support of the superiority of chlorthalidone, but in the Discussion section of your article, you point out that important differences between the 2 patients groups in MRFIT led you and your colleagues to conclude that the results are not as compelling as previously maintained.
Dr. Dhalla: I wouldn't view the MRFIT results as conclusive, for a number of reasons. First of all, the comparison of HCTZ vs chlorthalidone was post hoc. Also, the data are from so long ago that I don't think they reflect contemporary practice.
Medscape: Presumably a randomized, controlled clinical trial is unlikely to happen.
Dr. Dhalla: Over the past few years, we have seen more and more focus on comparative effectiveness studies, where 2 active treatments are compared. Some of these studies are observational in nature, such as ours, and others are randomized controlled trials.
There are some challenges in terms of doing a large randomized controlled trial of chlorthalidone vs HCTZ. The first challenge is that both of these drugs are off patent, and so it is unlikely that a pharmaceutical company would have a serious interest spending the large amount of money necessary to do the clinical trial. But I hope that one day, the National Institutes of Health or another public research funder will sponsor a trial comparing the 2 drugs.
Medscape: In the meantime, would there be a better option to give another diuretic, or not to give a diuretic at all? Thiazide diuretics, such as chlorthalidone or HCTZ, may have more adverse effects in elderly persons. They may exacerbate hyperuricemia, glucose intolerance, and dyslipidemia; they may promote orthostatic hypotension; and they may exacerbate age-related physiologic changes. Would a different diuretic -- for example, indapamide, which is associated with a lower incidence of serious hypokalemia and hyperglycemia -- be a better option in older patients, such as the ones in your study?
Or why give them a thiazide diuretic at all? They are said to be overused in elderly patients. The UK National Institute for Health and Clinical Excellence (NICE) hypertension guidelines recommend a calcium-channel blocker as first-line treatment in patients aged 55 years or older, with a renin/angiotensin system blocker second, and a diuretic only as third-line therapy.
Dr. Dhalla: I think you can make the argument that elderly patients with hypertension potentially should be started on a drug that is associated with a lower risk for complications, and many physicians would indeed start with an ACE inhibitor or a calcium-channel blocker. However, many people will still end up on diuretics, because many require multiple drugs for hypertension and many experience side effects of the medications that they are initially prescribed. So the fact is that diuretics are an important part of our therapeutic armamentarium, and ideally we would know which of the 2 drugs is better.
Medscape: How would you summarize what we know now?
Dr. Dhalla: What we did was a real-world study looking at safety and effectiveness outcomes in patients who were prescribed HCTZ and chlorthalidone in everyday practice. What we were hoping to show is that patients prescribed chlorthalidone did better than patients who were prescribed HCTZ. That would lend support to the hypothesis that chlorthalidone was superior and might have resulted in a shift away from HCTZ to chlorthalidone. But the reason you do the study is that you don't really know the answer to the question until you do it.
On the basis of our findings, I can't recommend that physicians who prescribe HCTZ should shift away from the drug. In fact, I would take the neutral view: On the basis of the results of our study, physicians who are prescribing HCTZ should probably feel comfortable continuing to prescribe HCTZ, and similarly, physicians who are currently favoring chlorthalidone should feel comfortable continuing to prescribe chlorthalidone. Not every study changes practice, and I think ours falls into that category.
The Interview
Medscape: Why did you feel that it would be useful to carry out this study of thiazide diuretic use in hypertensive patients treated in Ontario? How did you predict it would add to the evidence already accumulated in the chlorthalidone vs HCTZ debate?
Dr. Dhalla: Hypertension is an extremely common, serious condition, and diuretics are frequently used to treat it. In North America, probably the 2 most commonly prescribed diuretics are HCTZ and chlorthalidone. Many, if not most, patients with hypertension will eventually end up on a diuretic at some point, and so it would be useful to know whether one of those 2 drugs is better than the other.
I think that most physicians take the view that ultimately, this kind of a question can only be definitively settled by a large, well-designed, and well-executed randomized controlled trial. Unfortunately, we were not in a position to carry out such a trial. However, in Ontario, we have access to a great deal of administrative data, and so we were able to compare safety and effectiveness outcomes in individuals who were prescribed HCTZ with safety and effectiveness outcomes with individuals who were prescribed chlorthalidone. In my view, it is nowhere near as good as a randomized controlled trial would be, but it is a "second-best" kind of study design, if you will.
Medscape: Could you tell us more about the origins of the data you used?
Dr. Dhalla: Routine information is collected on every hospitalization in Canada, and we have access to the information pertaining to all of those hospitalizations in Ontario. Similarly, we have some information pertaining to every physician visit. We also have information regarding every medication that was paid for by the public. Everyone aged 65 years or older has public coverage; therefore, virtually every drug prescribed to someone in that age group is paid for through the public purse, and we have access to those data. These routinely collected administrative data allow researchers such as us to do studies here that are not possible in other jurisdictions.
Medscape: You found that in Ontario, between 1993 and 2010, 643,529 patients aged 65 years or older were prescribed HCTZ compared with 11,389 prescribed chlorthalidone. The relative frequency with which the 2 drugs are prescribed seems similar to that in the United States.
Dr. Dhalla: As far as I'm aware, yes, I think the situation is similar in Canada and the United States.
Medscape: Are Canadian national guidelines for hypertension treatment similar to the Seventh Report of the Joint National Committee (JNC7): that a thiazide diuretic should be first-line therapy in the treatment of uncomplicated hypertension?
Dr. Dhalla: The Canadian guidelines here are less prescriptive. They suggest a range of different options for first-line treatment, and diuretics are among those options. They don't say that thiazides are preferred.
Medscape: To overcome differences between treatment groups, you used high-dimensional propensity-score matching and matched each chlorthalidone recipient with up to 2 HCTZ recipients. Could you explain more about that, and why you didn't just study the rate of events in both populations?
Dr. Dhalla: In recent years, researchers have learned to use more sophisticated analytic techniques when we are doing observational studies to minimize the risk for bias. One way of doing so is to try to mimic a randomized controlled trial. So we matched the HCTZ patients with chlorthalidone patients on certain criteria, including age, sex, and what year the treatment was started.
We also used a large number of other baseline variables -- for example, prescription drug claims, physician services fee codes, physician services diagnosis codes, hospital diagnosis codes, hospital procedure codes, and emergency department diagnosis codes -- to construct a propensity score, which is basically the probability that a patient will be treated with one drug or the other. We then matched patients on the propensity score, and in so doing, created a cohort where the HCTZ patients look very similar to the chlorthalidone patients.
The big difference between an observational study such as ours and a randomized controlled trial is that we are only able to produce balance on the observed baseline characteristics, whereas in a randomized controlled trial you get balance on both the observed characteristics and baseline characteristics that have not been measured. This is one of the main reasons why randomized controlled trials are so powerful. No matter how sophisticated an observational study is, you cannot be confident that you have achieved balance on the unmeasured confounders.
Medscape: What would an unmeasured confounder be in this case?
Dr. Dhalla: It could be something simple, such as the skill of the physician. It may be the case that physicians who are more skillful are more likely to prescribe 1 of the 2 drugs. It may also be the case that patients cared for by more skillful physicians are likely to live longer.
Medscape: You noted in your article that people taking chlorthalidone were likely to also be taking a beta-blocker and less likely to be treated with an angiotensin-converting enzyme (ACE) inhibitor or an angiotensin receptor blocker, possibly because a chlorthalidone/atenolol combination is available in the public formulary in Ontario. Thus, a physician's decisions about what drug to prescribed may not be completely based on personal choice?
Dr. Dhalla: If you were going to design a randomized controlled trial comparing these 2 drugs, you would be very careful about the dose, about medications that were started at the same time, and about monitoring. We obviously could not do that kind of study. So we decided to compare safety and effectiveness outcomes in these 2 groups of patients, basically looking at the 2 drugs as they are usually prescribed. And when you look at how they are usually do this in Ontario, they are actually prescribed in quite different ways. Chlorthalidone is typically prescribed in a much higher dose, and it is much more likely to be prescribed with atenolol. This is because a product on the public formulary here has both active ingredients in 1 pill.
We adjusted for that, and we also did a series of post hoc analyses to address the dose issue, but our main objective was to compare the effectiveness outcomes in these 2 drugs as they are usually prescribed. This is in contrast to comparing safety and effectiveness outcomes in these 2 drugs as an expert in hypertension might prescribe them.
Medscape: Were you surprised to see patients aged 66 years or older being prescribed a diuretic plus a beta-blocker?
Dr. Dhalla: No, we weren't. I think beta-blockers in particular are becoming less popular in older people, and I certainly would not use one as a first-line treatment in older adults. But we looked at data from as far back as the 1990s, and beta-blockers were more commonly prescribed then.
Medscape: The main result of your study, that there appeared to be no difference in cardiovascular events or mortality between chlorthalidone and HCTZ, was surprising, however. What was your reaction?
Dr. Dhalla: We were hoping that chlorthalidone would be associated with better outcomes than HCTZ. Although that wouldn't have settled this question once and for all, it would have provided more evidentiary support for physicians who are inclined to prescribe chlorthalidone over HCTZ. Instead, we were unable to find a difference in effectiveness. It is worth noting, however, that despite our study being so large, it was not large enough to exclude a small yet clinically significant difference between the 2 drugs.
On the safety side, we found that patients treated with chlorthalidone were considerably more likely to end up in hospital with hypokalemia or hyponatremia. That is why in the conclusion of our article, we do not really provide a recommendation to physicians that they should prescribe either chlorthalidone or HCTZ preferentially. What we are saying is that ultimately to settle this question, we would need a very large randomized trial, and in the absence of such a trial, physicians who prescribe HCTZ can probably feel comfortable continuing to prescribe HCTZ. Similarly, physicians who prefer chlorthalidone can probably feel comfortable continuing to prescribe chlorthalidone. And physicians who prescribe either drug should be mindful of the risk for electrolyte abnormalities, particularly immediately after starting the drug and with any illness that predisposes to fluid and electrolyte imbalance.
Medscape: You mentioned in the article that you found, as others have previously reported, that chlorthalidone was being prescribed at higher doses than HCTZ. Was that why you found the greater risk for electrolyte abnormalities with chlorthalidone?
Dr. Dhalla: That is a great question. I would speculate that the reason chlorthalidone was prescribed at a higher dose was simply because of the way it is available on the formulary. There is a 50-mg tablet of chlorthalidone available on the public formulary, but the 25-mg formulation is available only in combination with atenolol. Obviously, if physicians want to prescribe 25 mg of chlorthalidone per day without atenolol, patients can just cut the 50-mg tablet or have the pharmacist cut it for them. Similarly, a physician could prescribe 12.5 mg daily and the tablet could be cut into quarters, although that is not always so easy for older adults to do.
But the availability of the 50-mg formulation may be driving physicians to prescribe a substantially higher dose than they would with HCTZ. That may be part of the reason why we saw more electrolyte abnormalities with chlorthalidone, although in the post hoc dose-specific analyses, we still saw an increased risk for hypokalemia and hyponatremia with chlorthalidone at pretty much all doses. So there may be other reasons that are independent of the dose. For example, the fact that chlorthalidone has a longer half-life may contribute.
Medscape: You mentioned the post hoc analyses. Could you tell us more about these?
Dr. Dhalla: The main post hoc analyses were the ones that compared a particular starting dose of HCTZ with a particular starting dose of chlorthalidone: for example, 25 mg of chlorthalidone with 50 mg of HCTZ. All of the results in the post hoc analyses were consistent with the results in the main analysis.
Medscape: Why do you think a real-world-practice study produced different results from previous analyses and meta-analyses?
Dr. Dhalla: I don't think our results are that different from some of the previous analyses, such as the meta-analysis by Bruce Psaty and his group. But you are right that our results don't draw the same conclusion as some of the other studies that have been published. I would speculate that it may be that there really isn't a big difference between the 2 drugs, especially as they are prescribed in routine practice.
Medscape: The results of such studies as the recent retrospective analysis of MRFIT data have been quoted in support of the superiority of chlorthalidone, but in the Discussion section of your article, you point out that important differences between the 2 patients groups in MRFIT led you and your colleagues to conclude that the results are not as compelling as previously maintained.
Dr. Dhalla: I wouldn't view the MRFIT results as conclusive, for a number of reasons. First of all, the comparison of HCTZ vs chlorthalidone was post hoc. Also, the data are from so long ago that I don't think they reflect contemporary practice.
Medscape: Presumably a randomized, controlled clinical trial is unlikely to happen.
Dr. Dhalla: Over the past few years, we have seen more and more focus on comparative effectiveness studies, where 2 active treatments are compared. Some of these studies are observational in nature, such as ours, and others are randomized controlled trials.
There are some challenges in terms of doing a large randomized controlled trial of chlorthalidone vs HCTZ. The first challenge is that both of these drugs are off patent, and so it is unlikely that a pharmaceutical company would have a serious interest spending the large amount of money necessary to do the clinical trial. But I hope that one day, the National Institutes of Health or another public research funder will sponsor a trial comparing the 2 drugs.
Medscape: In the meantime, would there be a better option to give another diuretic, or not to give a diuretic at all? Thiazide diuretics, such as chlorthalidone or HCTZ, may have more adverse effects in elderly persons. They may exacerbate hyperuricemia, glucose intolerance, and dyslipidemia; they may promote orthostatic hypotension; and they may exacerbate age-related physiologic changes. Would a different diuretic -- for example, indapamide, which is associated with a lower incidence of serious hypokalemia and hyperglycemia -- be a better option in older patients, such as the ones in your study?
Or why give them a thiazide diuretic at all? They are said to be overused in elderly patients. The UK National Institute for Health and Clinical Excellence (NICE) hypertension guidelines recommend a calcium-channel blocker as first-line treatment in patients aged 55 years or older, with a renin/angiotensin system blocker second, and a diuretic only as third-line therapy.
Dr. Dhalla: I think you can make the argument that elderly patients with hypertension potentially should be started on a drug that is associated with a lower risk for complications, and many physicians would indeed start with an ACE inhibitor or a calcium-channel blocker. However, many people will still end up on diuretics, because many require multiple drugs for hypertension and many experience side effects of the medications that they are initially prescribed. So the fact is that diuretics are an important part of our therapeutic armamentarium, and ideally we would know which of the 2 drugs is better.
Medscape: How would you summarize what we know now?
Dr. Dhalla: What we did was a real-world study looking at safety and effectiveness outcomes in patients who were prescribed HCTZ and chlorthalidone in everyday practice. What we were hoping to show is that patients prescribed chlorthalidone did better than patients who were prescribed HCTZ. That would lend support to the hypothesis that chlorthalidone was superior and might have resulted in a shift away from HCTZ to chlorthalidone. But the reason you do the study is that you don't really know the answer to the question until you do it.
On the basis of our findings, I can't recommend that physicians who prescribe HCTZ should shift away from the drug. In fact, I would take the neutral view: On the basis of the results of our study, physicians who are prescribing HCTZ should probably feel comfortable continuing to prescribe HCTZ, and similarly, physicians who are currently favoring chlorthalidone should feel comfortable continuing to prescribe chlorthalidone. Not every study changes practice, and I think ours falls into that category.