Salivary cortisol sampling compliance: comparison of patients and healthy volunteers

https://doi.org/10.1016/S0306-4530(03)00093-3Get rights and content

Abstract

Objective: Problems of compliance with in vivo data collection and treatment protocols have been identified. This study investigated compliance with salivary cortisol sampling in a 7-day protocol. Impact of non-compliance on cortisol data was evaluated. Methods: Female fibromyalgia patients were matched with healthy female volunteers and randomized to Aware or Unaware conditions regarding objective monitoring of their sampling compliance. The protocol entailed collecting five saliva samples at prescribed times on each of 7 consecutive days. Participants self-reported time of each sample, and electronic monitor caps provided an objective date and time stamp of each sample. Results: Objective compliance among participants unaware of monitoring was 71%, though their self-reported compliance was 93%. Aware participants’ objective compliance was 90% which was consistent with self-reported compliance of 93%. Within-subject comparison of early morning rise and day slope of cortisol for compliant and non-compliant samples found significant differences with non-compliant samples resulting in flatter slopes. Patients were somewhat more compliant than healthy volunteers. Slight decrements in compliance were found for the afternoon sample (1600 h) and for the last 3 days of sampling. Compliance did not differ on weekdays versus weekends. Conclusions: Self-report of compliance in a salivary cortisol sampling protocol substantially overestimates actual compliance in the absence of objective monitoring. Non-compliance with the sampling protocol results in cortisol data that significantly differs from compliant data. Awareness of electronic monitoring of sampling results in satisfactory compliance.

Introduction

Across the past 20 years there has been a prominent movement toward collection of data in the natural environment. While in vivo measurement has distinct advantages for some research questions, it also has the disadvantage of losing the control afforded by the laboratory or clinic. To the degree that patients must actively initiate or participate in measurement outside of the laboratory, factors such as patient motivation or ability can affect the quality of the data. Lack of compliance with the measurement protocol is potentially a significant cause of missing data and of data of questionable validity. For example, the problem has been found to be substantial in investigations of medication dosing compliance where missed doses and inappropriate multiple dosing are common and probably account for many reported treatment failures (Burney et al., 1996, Vanhove et al., 1996). Moreover, the evidence clearly suggests that patients attempt to appear compliant with their medication by generating self-reports of their behavior that inflate actual compliance (Rand et al., 1992, Spector et al., 1986). Similar problems have been observed in studies in which patients report glucose levels (Mazze et al., 1984) and blood pressure readings (Nordmann et al., 2000). Such dissembling places the health provider at a disadvantage in making treatment decisions and provides a researcher with data that are collected at a time deviating from the sampling protocol, thus potentially introducing significant error variance. In the absence of objective information on protocol compliance, unexpected treatment outcomes or research results are difficult to interpret with confidence.

Psychoendocrinology studies often involve collection of data in the field. With the advent of salivary assays, these studies have increased due to their non-invasive nature and the ease of sample collection by research participants (Kirschbaum et al., 1990). Typically, participants are given instruction in the procedures for collection of saliva, are provided with salivettes, and are directed to take samples over one or more days according to a sampling protocol. Participants self-report the date and time each sample was taken. This self-reported information is accepted as veridical by researchers. However, some recent evidence suggests that, as in medication studies, patients’ reports of their compliance with protocol may not be entirely accurate. Kudielka found in a 1-day protocol that 26% of participants collected at least one of six saliva samples outside of the designated sampling times, and the non-compliance significantly affected the cortisol profiles obtained (Kudielka et al., 2003). Sampling of hormones, such as cortisol, that vary quickly over time require accuracy of sampling time to yield valid conclusions (Wehr et al., 2001). Especially in the morning, when cortisol rises rapidly after awakening and then quickly falls, deviations in sampling times can dramatically impact the slopes obtained (Kudielka et al., 2003, Pruessner et al., 1997).

It is reasonable to assume that participant characteristics may impact on level of compliance with a protocol. We hypothesized that patients with an illness would be more motivated in a research protocol to provide valid data in order to facilitate research on their disease than would healthy control individuals who are volunteering for less personal, more altruistic reasons. Thus, we designed this study to compare compliance between a sample of fibromyalgia (FM) patients and a matched, healthy control sample. FM is a chronic pain disorder of unknown etiology (Wolfe et al., 1990) and limited treatment options (Rossy et al., 1999), thus patients express a high degree of enthusiasm for research on the condition. We hypothesized that compliance would be greater in our FM patients than in our controls.

We also assumed that participants who know that they are being monitored for sampling compliance would provide more compliant data than participants who were unaware that they were being monitored. Thus, we randomly assigned participants into groups who were either aware or unaware of objective sampling compliance monitoring. The unaware participants would provide an estimate of the compliance behavior in a typical research protocol in which self-report of sampling times in an “honor system” is provided by participants.

Finally, we wanted to examine the pattern of sampling compliance for different times of day, on weekdays versus weekends, and as the sampling burden increased over the course of 7 days. Compliance might vary systematically on these time dimensions as has been found in treatment studies (Cramer et al., 1990).

The following specific hypotheses will be explored. Hypothesis 1: self-reported times of sample collection will not be consistent with objective evidence of sample times. Hypothesis 2: participant awareness of objective monitoring of sampling times will produce greater compliance with the sampling protocol. Hypothesis 3: patients will provide more compliant samples than healthy control participants. Hypothesis 4: cortisol profiles will be significantly impacted by non-compliance (a) for the early morning rise and (b) for the morning to evening slope.

Secondary analyses will explore the following questions: (1) Will compliance degrade over the 7 days of sampling? (2) Will compliance vary by time-of-day of sampling? (3) Is compliance different on the weekends (Friday–Sunday) versus weekdays?

Section snippets

Participants

Women with FM and healthy controls were recruited through newspaper announcements and fliers in an academic hospital. The stated purpose of the study was to compare adrenal hormone levels of women with FM with matched, healthy women. Interested women were screened on the telephone for eligibility criteria including age (18–70), no significant problems with sight or hearing, fluent in English, not pregnant, no history of Schmidt’s or Cushing’s syndrome, no history of adrenal tumors or renal

Participant characteristics

All 66 participants completed the protocol, however one participant’s data (FM-Aware condition) were not analyzed because she discarded her eDEM™ cap. Table 1 displays the demographic comparisons for FM patients versus Healthy Controls and for Aware versus Unaware groups. As expected, there was a difference in employment (χ2(2,N=65)=3.81, p=0.15), disability status (Fisher’s exact test (N=65), p<0.01), and physical (t(63)=11.33, p<0.001) and mental (t(63)=4.60, p<0.001) health status on the

Discussion

This study was designed to investigate whether self-reported compliance in a typical in vivo 7-day salivary cortisol sampling protocol is accurate, and the degree to which non-compliance affects the cortisol data obtained. As expected, self-reported compliance was high across all participants and did not differ between patients and controls. However, objective measurement of compliance yielded more sobering results. Half of our participants were naïve, thereby representing the conditions of a

Acknowledgments

We would like to thank Arthur Stone and Steven Grossman for their assistance in the data analysis, and Jaclyn Gutlieber and Steven Choi for assistance with research materials preparation. This research was supported in part by the Applied Behavioral Medicine Research Institute, Stony Brook University, New York.

References (22)

  • Cited by (0)

    View full text