I am very pleased to write this commentary about the "Effect of Feedback Signal on Blood Pressure Self-regulation Capability in Individuals With Prehypertension or Stage I Hypertension: A Randomized Controlled Study,"1 which appears in this issue of the Journal of Cardiovascular Nursing. This is an exemplary article investigating the effect of feedback signal on blood pressure (BP) self-regulation. I have 3 major comments focused on the quality of the article, control groups for behavioral studies, and floor effects of hypertension research.

First, this is one of the most well-written articles I have read. The article was clear and concise, and publication guidelines were followed. The authors included a coherent power analysis and followed the Consolidated Standards of Reporting Trials (CONSORT) statement,2 which are readily available on the CONSORT Web site at http://www.consort-statement.org/. These guidelines provide specific information regarding what should be included in randomized clinical trial reports, which are title, abstract, introduction, methods, results, and discussion, and are well worth reviewing as researchers prepare articles about clinical trials. Following these reporting guidelines ensures that study participants are accounted for and enables other researchers to replicate the study. I used this checklist to determine whether these authors followed the guidelines and am pleased that they did. They included each of the applicable criteria in their report. This is critical if we are going to be able to advance nursing science.

Second, having spent a number of years thinking about biofeedback and conducting research on the effects of biofeedback on hypertension, I am able to empathize with the authors about the challenges in this type of work.3 The criterion standard in research requires a control group, but unlike medicinal interventions, where an inactive pill or placebo can be administered, behavioral intervention research does not have a clear control group. It is impossible to blind either group because both participate in the intervention. Even with a wait list control group, researchers cannot be sure whether the effects are caused by the training or by the attention and focus on BP after signing the consent document.

There is no good choice for a control group. In the early 2000s, a group of nurse researchers and I confronted all of the problems that arise in researching behavioral interventions as we performed a meta-analysis of biofeedback for hypertension.4 We quickly ran into the problem of suitable control groups. To address this, we categorized the study groups into biofeedback, active control, and inactive control. The biofeedback therapy group included different biofeedback modalities: thermal, electromyography, electrodermal activity, BP, pulse wave velocity, respiratory sinus dysrhythmia, and heart rate, and may have been accompanied by elements of cognitive therapy and relaxation training. The active control group included those treatments known to reduce BP: relaxation training, cognitive therapy, and home BP monitoring. The inactive control group included waiting list, clinic BP monitoring, and sham biofeedback. In our meta-analysis of 23 articles, biofeedback and active control treatments resulted in a reduction in systolic BP and diastolic BP, but only biofeedback (with related cognitive therapy and relaxation training) showed a significantly greater reduction in systolic BP (6.7 mm Hg) and diastolic BP (4.8 mm Hg) when compared with inactive control treatments. In the study published in this issue of Journal of Cardiovascular Nursing, both treatment groups involved sitting quietly for half hour and thinking about lowering their BPs. It is not surprising that both groups were successful in reducing their BPs.

Third, it is easier to show differences in interventions when the change expected is large. This is similar to needing smaller groups when expected effect sizes are large. The sample selected for this study, including those with prehypertension or stage 1 hypertension, had starting BPs of 140.6/90.7 (experimental) and 143.7/91.2 (control) mm Hg, providing less opportunity for improvement than among those with stage 2 or 3. Had the researchers used patients with stage 2 hypertension or above, they would have had to deal with the confounding effects of antihypertensive medications. Given the generally low levels of BPs among their subjects, the researchers could not expect to see large decreases in BPs. This floor effect has been noted by other researchers as well. In a review of more than 100 randomized controlled trials, Linden and Moseley5 reported that behavioral treatments reduce BP to a modest degree. This reduction was highly consistent across studies and was partially caused by the powerful impact of pretreatment BP levels on subsequent BPs. Specifically, they suggested that 40% of the variance in BP reductions after treatment was accounted for by the initial BP level rather than the treatment.

The problems inherent in biofeedback studies for hypertension that I discussed previously are common across other disorders as well.3 Given these challenges, adherence to the CONSORT reporting guidelines becomes even more important. In conclusion, I congratulate the authors for preparing a well-planned and -executed study, given the challenges of this type of research.

REFERENCES