- Email: [email protected]
Use of an embedded N-of-1 trial to improve adherence and increase information from a clinical study
Contemporary Clinical Trials 26 (2005) 397 – 401 www.elsevier.com/locate/conclintrial
Use of an embedded N-of-1 trial to improve adherence and increase information from a clinical study Andrew L. Avinsa,b,c,T, Stephen Bentb, John M. Neuhausc a
Northern California Kaiser-Permanente Division of Research, Oakland, CA, United States b Department of Medicine, University of California, San Francisco, CA, United States c Department of Epidemiology and Biostatistics, University of California, San Francisco, CA, United States Received 30 March 2004; accepted 4 February 2005
Abstract Withdrawal of participants from randomized trials can occur because of symptoms thought to be related to the study medicine, but the causal relationship between the study medicine and the symptoms is often unclear. Singlepatient trials (bN-of-1 trialsQ), developed to identify optimal therapy for an individual patient in the clinical setting, may provide a means of resolving some of these dilemmas. We describe here the use of an N-of-1 study embedded within a placebo-controlled trial of saw palmetto for a participant who considered withdrawing because he believed the study medication caused an increase in his blood pressure. In this case, the N-of-1 study not only reassured the patient, who decided to remain in the study, but provided potentially useful new information regarding the study medication. Wider use of formal N-of-1 studies may be a valuable tool for improving adherence and determining whether observed side effects are caused by study medication in clinical trials. D 2005 Elsevier Inc. All rights reserved.
1. Background Symptomatic adverse effects are a common reason for participant withdrawal from a clinical trial. Individuals who voluntarily participate in clinical trials may not be willing to tolerate even minor adverse effects unless the perceived benefit of participation is very high. T Corresponding author. Northern California Kaiser-Permanente Division of Research, 2000 Broadway, 3rd Floor, Oakland, CA 94612, United States. Tel.: +1 510 891 3557; fax: +1 510 891 3606. E-mail address: [email protected] (A.L. Avins). 1551-7144/$ - see front matter D 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.cct.2005.02.004
398
A.L. Avins et al. / Contemporary Clinical Trials 26 (2005) 397–401
Often, symptomatic reactions are expected side effects of a study medication and participant withdrawal due to development of these symptoms can be anticipated. Other times, however, the relationship between a side effect and the study medication is unclear. Under these circumstances, withdrawal of an otherwise willing participant is unfortunate, given that the participant’s symptoms may not be due to the intervention. Identifying a method that informs the participant and investigator about the likelihood that a symptom is due to the study medicine (without breaking a double-blind condition) could help improve study participation and may also provide new information that could be useful for study participants, investigators, and institutional review boards in assessing the possible causal relationship between symptoms and study interventions. bSingle-patientQ or bN-of-1Q trials are a well-described method generally used to identify optimal treatment, dosage, or type of medication (e.g., proprietary or generic) for an individual patient in the clinical setting [1]. This method, however, can be readily adapted to the randomized, double-blind, placebo-controlled clinical trial, in order to shed light on the causal relationship between a symptom and a study intervention and, potentially, to improve adherence to the study protocol. We report here the use of such an bembedded N-of-1 trialQ that resulted in continued participant adherence and provided new information about the study medication.
2. Methods The Saw Palmetto Treatment for Enlarged Prostates (STEP) study is an NIH-funded randomized, double-blind, placebo-controlled clinical trial of a standardized extract of the saw palmetto berry for the treatment of symptoms of benign prostatic hyperplasia (BPH). Participants are men 50 years of age or older with at least moderate symptoms of BPH. Participants are randomized to 160 mg twice-daily doses of a saw palmetto extract (Rexall-Sundown, Inc, Boca Raton, FL; Indena USA, Inc., Seattle, WA) or a matching placebo and followed for 1 year for outcomes of BPH symptoms and uroflowmetry. This study is approved by the Institutional Review Boards at the University of California, San Francisco and the Kaiser Foundation Research Institute. One month after randomization, a participant reported to the study team a concern that his blood pressure may have become elevated since he began taking the study medicine and he was considering withdrawing from the study. This participant checks his blood pressure regularly with a home blood pressure monitoring device and reported that his readings had increased since randomization. Hypertension is not a known or expected side effect of saw palmetto extract [2,3]. Because neither the participant nor the study investigators were convinced that the recently elevated blood pressure readings were due to the study medication (which might have been placebo), we offered the participant the option of undergoing an N-of-1 study to examine the association between his study medication and his blood pressure; the participant willingly agreed. The N-of-1 study was conducted in the following manner: the research pharmacist prepared six bottles of 14 capsules; three bottles contained the participant’s assigned study medicine and three bottles contained placebo. The blinding of the STEP study was not broken, and the investigators did not know if the participant’s study medication (used in both the STEP study and the N-of-1 study) was placebo or saw palmetto extract. Unbeknownst to the participant, the six bottles were ordered so that the participant took study medicine for 2 weeks, followed by 3 weeks of placebo,
A.L. Avins et al. / Contemporary Clinical Trials 26 (2005) 397–401
399
followed by 1 week of study medicine. (This order was chosen to provide a continuous period of placebo administration. If the patient’s study medicine was, in fact, saw palmetto, many components of this extract are fat soluble and take a relatively long period to be eliminated from the body.) The participant was instructed to take one capsule twice daily (as he does normally in the STEP study) and to record his blood pressure twice daily using his own blood pressure measuring device. After the six-week N-of-1 substudy, the participant returned his bottles and blood pressure diary to the study clinic. We analyzed the blood pressure data following the approach of Rochon [4]. Using this approach, we fit a regression model to each of the four blood pressure outcomes with treatment indicator (study medicine versus placebo) as the single explanatory variable and assumed that the repeated measures followed a first-order autoregressive model. Because the variability and longitudinal correlation of blood pressure measurements seemed to vary with treatment period, we fit models that allowed separate variance and correlation parameters for each of the three treatment periods. Our analyses yielded estimates of the difference in mean blood pressure values between study medicine and placebo periods and associated 95% confidence intervals for this difference. The autoregressive models were fit using PROC MIXED in SAS [5].
3. Results The participant missed 1 day of medication and 1 day of recording during the N-of-1 substudy. Overall, based on estimates derived from the analytic model, the results did not suggest a strong effect of the study medication on blood pressure (Table 1). The only estimate that achieved conventional levels of statistical significance was the morning systolic blood pressure, which showed a small decrease during the study medicine period compared with known placebo. However, both the morning systolic and evening diastolic blood pressures showed a trend toward some increase while on study medicine. After a full presentation and discussion of the data with the participant, he elected to continue his participation in the study and was pleased with his experience. In addition, while the data from the N-of1 study do not strongly suggest an effect of the participant’s study medicine on blood pressure, the results have prompted the study investigators to plan a more careful examination of the blood pressure data from the trial.
Table 1 Blood pressure response data in N-of-1 substudy Measurement time
Blood pressure outcomes Difference (mm Hg)T
Morning systolic BP Morning diastolic BP Evening systolic BP Evening diastolic BP
7.51 2.06 3.78 3.45
95% confidence interval 1.20 1.21 7.54 0.87
to 16.22 to 5.33 to 0.02 to 7.77
p-value 0.09 0.21 0.05 0.11
T Difference between adjusted mean blood pressures in study phases (positive values denote higher mean blood pressures in study-medicine periods compared to known placebo periods).
400
A.L. Avins et al. / Contemporary Clinical Trials 26 (2005) 397–401
4. Discussion Because of the need for high-quality data and the expense of enrolling and maintaining participants in clinical trials, there is a great imperative to optimize participants’ study adherence. Methods that directly address reasons for non-adherence, particularly those that are very participant-specific, may greatly assist investigators in maintaining participants in clinical trials. N-of-1 trials are not commonly performed in clinical practice for a variety of reasons, including the difficulty and costs of performing the study, obtaining the placebo formulation, and finding resources for analyzing the data [6,7]. None of these practical barriers exist in placebo-controlled clinical trials: participants are already part of a study structure, placebo is readily available from the research pharmacy, and data analysis is integral to all clinical studies. For participants already enrolled in a research study, there is unlikely to be a psychological barrier to participating in another, embedded, clinical trial. In fact, the participant in this study was quite enthusiastic about the N-of-1 trial, identifying this substudy as an opportunity to learn more about his response to the study medicine and increasing his interaction with study staff. In addition to improving adherence, N-of-1 clinical trials have the potential to provide additional information regarding causality of adverse events. In many cases, patients who experience adverse events drop out of a study, and there is limited opportunity to assess whether the study intervention caused the adverse effect. Embedded N-of-1 clinical trials allow for blinded and placebo-controlled dechallenge and rechallenge of the study medication, providing additional information to help investigators determine if the side effect is likely to have been caused by the intervention. This type of information has the potential to improve early detection and closer monitoring for possible side effects. In the STEP study, additional analyses are now planned to more closely examine the effects of saw palmetto extract on blood pressure, since the N-of-1 trial did not exclude the possibility of a small but real increase in this participant’s blood pressure (though, if the participant was randomized to placebo, this issue could become irrelevant). Design of formal embedded N-of-1 trials must be tailored to the specific situation in the clinical trial. For example, highly variable outcomes will require more determinations and some subjective outcomes may require multiple cross-over periods to better establish a causal association. Such trials can also be double-blind (though this was thought to be unnecessary in the present case, since all outcome measurements were taken by the participant at his home). Procedures for the design of N-of-1 studies have been published previously [8–10]. Embedded N-of-1 trials offer an opportunity for helping improve participants’ adherence to clinical protocols and increasing the information gathered from the study, while being fairly easy to implement. Investigators should consider planning the use of embedded N-of-1 trials for addressing relevant issues of adverse effects and participant adherence as part of their study protocols.
Acknowledgements Financial support: supported by a grant from the National Institute of Diabetes, Digestive, and Kidney Diseases (#R01 DK56199) and the National Center for Complementary and Alternative Medicine (#K08 AT001338), National Institutes of Health.
A.L. Avins et al. / Contemporary Clinical Trials 26 (2005) 397–401
401
The authors gratefully acknowledge the contributions of Suzanne Staccone, M.A., Evelyn Badua, M.D., Amy Padula, M.S., Bertina Lee, and Arlene Sakamoto, R.N. for their outstanding assistance and, in particular, to the research participant who so willingly took part in this study.
References [1] Spilker B. Single-patient clinical trials. In: Spilker B, editor. Guide to Clinical Trials. Philadelphia7 Lippencott-Raven; 1996. pp. 277 – 82. [2] Wilt TJ, Ishani A, Stark G, MacDonald R, Lau J, Mulrow C. Saw palmetto extracts for treatment of benign prostatic hyperplasia: a systematic review. Jama 1998;280(18):1604 – 9. [3] Wilt T, Ishani A, Mac Donald R. Serenoa repens for benign prostatic hyperplasia. Cochrane Database Syst Rev 2002;3:CD001423. [4] Rochon J. A statistical model for the bN-of-1Q study. J Clin Epidemiol 1990;43:499 – 508. [5] SAS Institute Inc. SAS statistical software; 2003. Cary, NC. [6] Larson EB, Ellsworth AJ, Oas J. Randomized clinical trials in single patients during a 2-year period. Jama 1993;270(22):2708 – 12. [7] Guyatt GH, Keller JL, Jaeschke R, Rosenbloom D, Adachi JD, Newhouse MT. The N-of-1 randomized controlled trial: clinical usefulness. Our three-year experience. Ann Intern Med 1990;112(4):293 – 9. [8] Guyatt G, Sackett D, Taylor DW, Chong J, Roberts R, Pugsley S. Determining optimal therapy-randomized trials in individual patients. N Engl J Med 1986;314(14):889 – 92. [9] Guyatt G, Sackett D, Adachi J, Roberts R, Chong J, Rosenbloom D, et al. A clinician’s guide for conducting randomized trials in individual patients. Cmaj 1988;139(6):497 – 503. [10] Larson EB. N-of-1 clinical trials: a technique for improving medical therapeutics. West J Med 1990;152(1):52 – 6.
Use of an embedded N-of-1 trial to improve adherence and increase information from a clinical study Andrew L. Avinsa,b,c,T, Stephen Bentb, John M. Neuhausc a
Northern California Kaiser-Permanente Division of Research, Oakland, CA, United States b Department of Medicine, University of California, San Francisco, CA, United States c Department of Epidemiology and Biostatistics, University of California, San Francisco, CA, United States Received 30 March 2004; accepted 4 February 2005
Abstract Withdrawal of participants from randomized trials can occur because of symptoms thought to be related to the study medicine, but the causal relationship between the study medicine and the symptoms is often unclear. Singlepatient trials (bN-of-1 trialsQ), developed to identify optimal therapy for an individual patient in the clinical setting, may provide a means of resolving some of these dilemmas. We describe here the use of an N-of-1 study embedded within a placebo-controlled trial of saw palmetto for a participant who considered withdrawing because he believed the study medication caused an increase in his blood pressure. In this case, the N-of-1 study not only reassured the patient, who decided to remain in the study, but provided potentially useful new information regarding the study medication. Wider use of formal N-of-1 studies may be a valuable tool for improving adherence and determining whether observed side effects are caused by study medication in clinical trials. D 2005 Elsevier Inc. All rights reserved.
1. Background Symptomatic adverse effects are a common reason for participant withdrawal from a clinical trial. Individuals who voluntarily participate in clinical trials may not be willing to tolerate even minor adverse effects unless the perceived benefit of participation is very high. T Corresponding author. Northern California Kaiser-Permanente Division of Research, 2000 Broadway, 3rd Floor, Oakland, CA 94612, United States. Tel.: +1 510 891 3557; fax: +1 510 891 3606. E-mail address: [email protected] (A.L. Avins). 1551-7144/$ - see front matter D 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.cct.2005.02.004
398
A.L. Avins et al. / Contemporary Clinical Trials 26 (2005) 397–401
Often, symptomatic reactions are expected side effects of a study medication and participant withdrawal due to development of these symptoms can be anticipated. Other times, however, the relationship between a side effect and the study medication is unclear. Under these circumstances, withdrawal of an otherwise willing participant is unfortunate, given that the participant’s symptoms may not be due to the intervention. Identifying a method that informs the participant and investigator about the likelihood that a symptom is due to the study medicine (without breaking a double-blind condition) could help improve study participation and may also provide new information that could be useful for study participants, investigators, and institutional review boards in assessing the possible causal relationship between symptoms and study interventions. bSingle-patientQ or bN-of-1Q trials are a well-described method generally used to identify optimal treatment, dosage, or type of medication (e.g., proprietary or generic) for an individual patient in the clinical setting [1]. This method, however, can be readily adapted to the randomized, double-blind, placebo-controlled clinical trial, in order to shed light on the causal relationship between a symptom and a study intervention and, potentially, to improve adherence to the study protocol. We report here the use of such an bembedded N-of-1 trialQ that resulted in continued participant adherence and provided new information about the study medication.
2. Methods The Saw Palmetto Treatment for Enlarged Prostates (STEP) study is an NIH-funded randomized, double-blind, placebo-controlled clinical trial of a standardized extract of the saw palmetto berry for the treatment of symptoms of benign prostatic hyperplasia (BPH). Participants are men 50 years of age or older with at least moderate symptoms of BPH. Participants are randomized to 160 mg twice-daily doses of a saw palmetto extract (Rexall-Sundown, Inc, Boca Raton, FL; Indena USA, Inc., Seattle, WA) or a matching placebo and followed for 1 year for outcomes of BPH symptoms and uroflowmetry. This study is approved by the Institutional Review Boards at the University of California, San Francisco and the Kaiser Foundation Research Institute. One month after randomization, a participant reported to the study team a concern that his blood pressure may have become elevated since he began taking the study medicine and he was considering withdrawing from the study. This participant checks his blood pressure regularly with a home blood pressure monitoring device and reported that his readings had increased since randomization. Hypertension is not a known or expected side effect of saw palmetto extract [2,3]. Because neither the participant nor the study investigators were convinced that the recently elevated blood pressure readings were due to the study medication (which might have been placebo), we offered the participant the option of undergoing an N-of-1 study to examine the association between his study medication and his blood pressure; the participant willingly agreed. The N-of-1 study was conducted in the following manner: the research pharmacist prepared six bottles of 14 capsules; three bottles contained the participant’s assigned study medicine and three bottles contained placebo. The blinding of the STEP study was not broken, and the investigators did not know if the participant’s study medication (used in both the STEP study and the N-of-1 study) was placebo or saw palmetto extract. Unbeknownst to the participant, the six bottles were ordered so that the participant took study medicine for 2 weeks, followed by 3 weeks of placebo,
A.L. Avins et al. / Contemporary Clinical Trials 26 (2005) 397–401
399
followed by 1 week of study medicine. (This order was chosen to provide a continuous period of placebo administration. If the patient’s study medicine was, in fact, saw palmetto, many components of this extract are fat soluble and take a relatively long period to be eliminated from the body.) The participant was instructed to take one capsule twice daily (as he does normally in the STEP study) and to record his blood pressure twice daily using his own blood pressure measuring device. After the six-week N-of-1 substudy, the participant returned his bottles and blood pressure diary to the study clinic. We analyzed the blood pressure data following the approach of Rochon [4]. Using this approach, we fit a regression model to each of the four blood pressure outcomes with treatment indicator (study medicine versus placebo) as the single explanatory variable and assumed that the repeated measures followed a first-order autoregressive model. Because the variability and longitudinal correlation of blood pressure measurements seemed to vary with treatment period, we fit models that allowed separate variance and correlation parameters for each of the three treatment periods. Our analyses yielded estimates of the difference in mean blood pressure values between study medicine and placebo periods and associated 95% confidence intervals for this difference. The autoregressive models were fit using PROC MIXED in SAS [5].
3. Results The participant missed 1 day of medication and 1 day of recording during the N-of-1 substudy. Overall, based on estimates derived from the analytic model, the results did not suggest a strong effect of the study medication on blood pressure (Table 1). The only estimate that achieved conventional levels of statistical significance was the morning systolic blood pressure, which showed a small decrease during the study medicine period compared with known placebo. However, both the morning systolic and evening diastolic blood pressures showed a trend toward some increase while on study medicine. After a full presentation and discussion of the data with the participant, he elected to continue his participation in the study and was pleased with his experience. In addition, while the data from the N-of1 study do not strongly suggest an effect of the participant’s study medicine on blood pressure, the results have prompted the study investigators to plan a more careful examination of the blood pressure data from the trial.
Table 1 Blood pressure response data in N-of-1 substudy Measurement time
Blood pressure outcomes Difference (mm Hg)T
Morning systolic BP Morning diastolic BP Evening systolic BP Evening diastolic BP
7.51 2.06 3.78 3.45
95% confidence interval 1.20 1.21 7.54 0.87
to 16.22 to 5.33 to 0.02 to 7.77
p-value 0.09 0.21 0.05 0.11
T Difference between adjusted mean blood pressures in study phases (positive values denote higher mean blood pressures in study-medicine periods compared to known placebo periods).
400
A.L. Avins et al. / Contemporary Clinical Trials 26 (2005) 397–401
4. Discussion Because of the need for high-quality data and the expense of enrolling and maintaining participants in clinical trials, there is a great imperative to optimize participants’ study adherence. Methods that directly address reasons for non-adherence, particularly those that are very participant-specific, may greatly assist investigators in maintaining participants in clinical trials. N-of-1 trials are not commonly performed in clinical practice for a variety of reasons, including the difficulty and costs of performing the study, obtaining the placebo formulation, and finding resources for analyzing the data [6,7]. None of these practical barriers exist in placebo-controlled clinical trials: participants are already part of a study structure, placebo is readily available from the research pharmacy, and data analysis is integral to all clinical studies. For participants already enrolled in a research study, there is unlikely to be a psychological barrier to participating in another, embedded, clinical trial. In fact, the participant in this study was quite enthusiastic about the N-of-1 trial, identifying this substudy as an opportunity to learn more about his response to the study medicine and increasing his interaction with study staff. In addition to improving adherence, N-of-1 clinical trials have the potential to provide additional information regarding causality of adverse events. In many cases, patients who experience adverse events drop out of a study, and there is limited opportunity to assess whether the study intervention caused the adverse effect. Embedded N-of-1 clinical trials allow for blinded and placebo-controlled dechallenge and rechallenge of the study medication, providing additional information to help investigators determine if the side effect is likely to have been caused by the intervention. This type of information has the potential to improve early detection and closer monitoring for possible side effects. In the STEP study, additional analyses are now planned to more closely examine the effects of saw palmetto extract on blood pressure, since the N-of-1 trial did not exclude the possibility of a small but real increase in this participant’s blood pressure (though, if the participant was randomized to placebo, this issue could become irrelevant). Design of formal embedded N-of-1 trials must be tailored to the specific situation in the clinical trial. For example, highly variable outcomes will require more determinations and some subjective outcomes may require multiple cross-over periods to better establish a causal association. Such trials can also be double-blind (though this was thought to be unnecessary in the present case, since all outcome measurements were taken by the participant at his home). Procedures for the design of N-of-1 studies have been published previously [8–10]. Embedded N-of-1 trials offer an opportunity for helping improve participants’ adherence to clinical protocols and increasing the information gathered from the study, while being fairly easy to implement. Investigators should consider planning the use of embedded N-of-1 trials for addressing relevant issues of adverse effects and participant adherence as part of their study protocols.
Acknowledgements Financial support: supported by a grant from the National Institute of Diabetes, Digestive, and Kidney Diseases (#R01 DK56199) and the National Center for Complementary and Alternative Medicine (#K08 AT001338), National Institutes of Health.
A.L. Avins et al. / Contemporary Clinical Trials 26 (2005) 397–401
401
The authors gratefully acknowledge the contributions of Suzanne Staccone, M.A., Evelyn Badua, M.D., Amy Padula, M.S., Bertina Lee, and Arlene Sakamoto, R.N. for their outstanding assistance and, in particular, to the research participant who so willingly took part in this study.
References [1] Spilker B. Single-patient clinical trials. In: Spilker B, editor. Guide to Clinical Trials. Philadelphia7 Lippencott-Raven; 1996. pp. 277 – 82. [2] Wilt TJ, Ishani A, Stark G, MacDonald R, Lau J, Mulrow C. Saw palmetto extracts for treatment of benign prostatic hyperplasia: a systematic review. Jama 1998;280(18):1604 – 9. [3] Wilt T, Ishani A, Mac Donald R. Serenoa repens for benign prostatic hyperplasia. Cochrane Database Syst Rev 2002;3:CD001423. [4] Rochon J. A statistical model for the bN-of-1Q study. J Clin Epidemiol 1990;43:499 – 508. [5] SAS Institute Inc. SAS statistical software; 2003. Cary, NC. [6] Larson EB, Ellsworth AJ, Oas J. Randomized clinical trials in single patients during a 2-year period. Jama 1993;270(22):2708 – 12. [7] Guyatt GH, Keller JL, Jaeschke R, Rosenbloom D, Adachi JD, Newhouse MT. The N-of-1 randomized controlled trial: clinical usefulness. Our three-year experience. Ann Intern Med 1990;112(4):293 – 9. [8] Guyatt G, Sackett D, Taylor DW, Chong J, Roberts R, Pugsley S. Determining optimal therapy-randomized trials in individual patients. N Engl J Med 1986;314(14):889 – 92. [9] Guyatt G, Sackett D, Adachi J, Roberts R, Chong J, Rosenbloom D, et al. A clinician’s guide for conducting randomized trials in individual patients. Cmaj 1988;139(6):497 – 503. [10] Larson EB. N-of-1 clinical trials: a technique for improving medical therapeutics. West J Med 1990;152(1):52 – 6.