Yoga Therapy as an Intervention to Improve Patient-Reported Outcomes Among Adults After Treatment for Cancer: Preliminary Findings From a Trial Using Single-Subject Experimental Design

Abstract

Background:

Cancer is a chronic condition associated with a substantial symptom burden, which can impair recovery after treatment. Investigating interventions with potential to improve self-reported disease and/or treatment effects—known as patient-reported outcomes (PROs)—is paramount to inform cancer care. The objective of this study was to evaluate the effects of a yoga therapy (YT) intervention on key PROs (ie, cancer-related fatigue, anxiety, cognitive function, depression, stress, quality of life [QoL]) among adults after treatment for cancer.

Methods:

Data from 20 adults (M_age = 55.74 years, 85% female; M_{time since diagnosis} = 2.83 years) who had completed treatment for cancer were analyzed for this study. In this single-subject exploratory experimental study, the YT intervention comprised a 1:1 YT session (ie, 1 participant with 1 yoga therapist) followed by 6 weekly small (ie, 2–3 participants) group YT sessions. Group sessions were facilitated by the same yoga therapist who delivered participants’ 1:1 session to ensure an in-depth personalized approach. PROs were assessed before (ie, pre-intervention) and after the 1:1 YT session (ie, during the intervention), as well as after the last group YT session (ie, post-intervention). Hierarchical linear modeling was used to analyze the data.

Results:

Participants showed improvements in cancer-related fatigue, state anxiety, trait anxiety, perceived cognitive impairments, impacts of perceived cognitive impairments on QoL, and 1 dimension of QoL (ie, functional wellbeing) over time. Notably, cancer-related fatigue and state anxiety increased immediately after the 1:1 session, but showed greater improvements over time afterward (ie, during the intervention phase). No changes were observed for the remaining PROs.

Conclusion:

Although results require confirmation in future trials, this study highlights the importance of continuing to investigate YT as an intervention to enhance important PROs (ie, cancer-related fatigue and state anxiety) after treatment for cancer. More research is needed to identify additional beneficial effects and factors that influence participants’ responses to 1:1 and group YT (ie, moderators and mediators).

Registration number:

ISRCTN64763228

Date of registration:

December 12, 2021. This trial was registered retrospectively.

URL of trial registry record:

https://www.isrctn.com/ISRCTN64763228

Published Protocol:

Brunet, J., Wurz, A., Hussien, J., Pitman, A., Conte, E., Ennis, J. K., . . . & Seely, D. (2022). Exploring the Effects of Yoga Therapy on Heart Rate Variability and Patient-Reported Outcomes After Cancer Treatment: A Study Protocol. Integrative Cancer Therapies, 21, 15347354221075576.

Background

Cancer and Symptom Burden

In 2020, there were 19.29 million new cancer cases worldwide. ¹ With more than half of people diagnosed with cancer expected to survive 10 or more years, ² patient-centered research and care has become more prominent because people with cancer experience multiple symptoms that impair function and recovery.^3,4 Many people describe unrelenting cancer-related fatigue, heightened feelings of anxiety, stress, and depression, persistent and upsetting cognitive complaints, and worsened quality of life (QoL) following cancer treatment.^5,6 Importantly, symptoms may persist for weeks, months, or years after treatment completion.^7,8 Left untreated, symptoms can increase risk of morbidity and mortality, lead people to abandon important (and possibly curative) therapies, limit vocational activity, and impair cancer recovery. ⁹ Accordingly, there has been growing investigation into interventions that can help to reduce the consequences of cancer and its treatment on people, and thus improve outcomes, people’s experiences, and quality of survival.^{6,10 -12}

Yoga Interventions

Recognizing that symptoms do not occur in isolation and that people typically experience multiple symptoms throughout their cancer trajectory, growing attention has been paid to understanding the role of yoga because it has been shown to be feasible, acceptable, and beneficial to those with varying symptom severity, symptom interference, or symptom distress. ¹³ Yoga is a mind-body practice including meditation and relaxation techniques, breath practices, and physical postures. ¹⁴ Among adults diagnosed with cancer, the benefits of different styles of yoga from different lineages (eg, Hatha, Iyengar, Vinyasa, Ashtanga, Sivananda, and Restorative) throughout diagnosis, treatment, survivorship, and palliative care are clear. Systematic reviews and meta-analyses demonstrate that yoga can reduce cancer-related fatigue, depression, stress, and anxiety, and enhance a range of physical, psychosocial, and cognitive outcomes, including different dimensions of QoL among adults with a history of cancer.^{13,15 -20} These results are promising and highlight the importance of conducting more research investigating yoga interventions that can provide safe, acceptable, and effective therapeutic options for adults after treatment for cancer.

Yoga Therapy (YT) as a Focus

YT is a therapeutic approach that uses yoga principles and practices adapted to people with varied health challenges. YT is delivered by a yoga therapist who has completed 800+ hours of in-depth training and clinical practice, beyond their yoga instructor/teacher training (https://www.iayt.org/page/ContemporaryDefiniti). YT may be therapeutic for people with a history of cancer given the complexity of the disease and associated treatments. Further, YT may be ideally suited to support people after cancer treatment completion, as it is non-prescriptive, inclusive, personalized, modulated, and centers on people’s agency and eventual self-regulation. Anecdotally, YT can foster people’s ability to navigate the burden of cancer through promoting self-management skills, practices of self-compassion and self-awareness, and other yogic practices (eg, breath, movement and meditation), which can reduce allostatic load, downregulate the nervous system, and enhance wellbeing. YT may also increase people’s self-regulation capacity through lifestyle and nervous system education, helping them better understand their lived experiences (eg, thoughts, experiences, memories, and emotions) and learn how to work compassionately with shock, anxiety, and fear to effectively manage their emotions and rest well amid life’s demands. The latter is a fundamental value of YT that differentiates it from yoga. Despite the potential of YT and application of YT in practice, research is limited, making it difficult to draw a conclusion on whether YT specifically can improve outcomes, people’s experiences, and quality of survival; this highlights the need for further research in this area. Indeed, additional research on YT could support efforts to increase dissemination and implementation of YT interventions designed to reduce the burden of cancer.

Current Study

A single-subject experimental exploratory study was conducted to assess the effects of YT on heart rate variability (HRV; primary outcome) and patient-reported outcomes (PROs; secondary outcomes) among adults who had completed treatment for cancer. PROs are indicators that come directly from people themselves, and thus reflect how they interpret their experiences and the conditions not observable by others. ²¹ They include symptoms, functional health, wellbeing, psychological issues, QoL, among other outcomes that can be reported by people without interpretation by a clinician. To capture these, PRO measures (PROMs), also called self-report measures, have been developed with different focuses, including cancer-related fatigue, anxiety, depression, stress, cognitive function, and QoL, which have been shown to be predictive and prognostic indicators of survival;^22,23 hence, they are increasingly used to inform patient-centered care, clinical decision making, and health policy or reimbursement decisions. ²⁴ PROMs related to these constructs were administered repeatedly during the study to gain insight into changes over time in general and cancer-specific PROs. In this manuscript, results for secondary outcomes are reported to advance the knowledge base on the effects of YT on PROs; a manuscript presenting results for the primary outcome of HRV is forthcoming. The specific aim was to assess the effects of YT on PROMs of cancer-related fatigue, anxiety, depression, stress, cognitive function, and QoL. Drawing on existing empirical work cited above, it was hypothesized that YT would improve cancer-related fatigue, anxiety, depression, stress, cognitive function, and QoL.

Methods

Study Design

The study design and rationale have been previously reported. ²⁵ Briefly, a single-subject experimental design was adopted because it is a rigorous and efficient method of establishing preliminary data on the likely causal effects of an intervention while controlling for multiple threats to internal validity such as maturation (ie, simple passage of time), repeated assessment, and regression to the mean.^26,27 This design allows for the assessment of intra-individual changes over time in response to an intervention (ie, YT). As such, in lieu of a control condition, participants’ baseline assessments serve as their control condition. Participants completed a 90-minute 1:1 YT session (ie, 1 participant with 1 yoga therapist; intervention phase 1), followed by 6 weekly 60-minute group YT sessions (ie, 2-3 participants per group with the same yoga therapist who performed their 1:1 YT session, using a consistent physical and psychosocial spiritual (kosha) framework; intervention phase 2). Throughout the study, participants completed 7 assessments: 3 baseline assessments (ie, 6 weeks, 3 weeks, and immediately before intervention phase 1), 2 assessments between intervention phases (ie, immediately and in the week following phase 1), and 2 assessments after completing intervention phase 2 (ie, 1 week and 6 weeks after, which corresponds to 6 weeks and 12 weeks after beginning the YT intervention). See Figure 1 in the published protocol ²⁵ for a diagram depicting the flow of participants throughout the YT intervention and assessments.

Figure 1.

Flow diagram reflecting the flow of participants in the study. ^a= indicates same participant.

The study was carried out in accordance with the principles of the Declaration of Helskinki, and was approved by the ethics committees at the University of Ottawa (no.: H01-17-04), as well as at the Canadian College of Naturopathic Medicine (no.: CCNMREB016). The study was registered retrospectively in a publicly accessible registry for clinical trials (https://www.isrctn.com/ISRCTN64763228). Written informed consent was obtained from all participants to participate prior to data collection. This manuscript is reported following Consolidated Standards Of Reporting Trials (CONSORT) extension for reporting N-of-1 trials (CENT) 2015 standards. ²⁸ As the trial protocol has been published, ²⁵ only the methodological information pertaining to this study are reported herein.

Setting and Participants

The current study was conducted at the Ottawa Integrative Cancer Centre (OICC; now known [and here forth referred to] as the Centre for Health Innovation; CHI), Ontario (Canada). Adults were recruited from May 2017 to September 2019 via referral from naturopathic doctors and CHI staff, and via self-referral. To participate, individuals had to: (1) be ≥18 years of age, (2) have received a cancer diagnosis when ≥18 years of age, (3) have completed conventional treatment for cancer (chemotherapy, radiotherapy, surgery) ≥3 months prior (estimated time to allow for transition from acute survivorship into extended survivorship where the support from the health system is reduced and patients’ priorities shift to living with the impact of cancer) and have no scheduled treatments for cancer for the next 5 months (given possible effects on treatment-related experiences and outcomes), (4) not be taking cardiac medication, (5) self-report no YT practice since their cancer diagnosis, (6) be cleared for physical activity as indicated by physical activity readiness questions (https://csep.ca/2021/01/20/pre-screening-for-physical-activity/), and (7) be willing and able to provide informed consent in English. No financial compensation was offered or provided to participants for taking part in this study.

Sample Size

The plan was originally to enroll 30 participants. A moderate effect size (Cohen’s d = 0.50) was expected for standardized mean differences in the primary outcome of HRV. Power analysis was conducted to detect a medium effect size (0.50) with a statistical power of 80% power at the 5% level of significance. Results indicated a required sample size of 25 participants. Accounting for 20% attrition, the intention was to enroll 30 participants. However, recruitment stopped after 25 participants had enrolled due to restrictions put in place at the onset of the COVID-19 pandemic, and did not resume for logistical reasons (ie, limited financial resources, slow recruitment, and difficulties planning group sessions). Nevertheless, sample sizes for single-subject studies are generally small (ie, average N = 3.64, range = 1-13) as a function of the repeated assessments, which results in a larger number of observations than other study designs. Moreover, authors have shown that a sample size of 3 offers sufficient statistical power to detect pre-to-post intervention effects using hierarchical linear modeling (HLM),^29,30 and HLM approaches have been validated with small sample sizes (eg, N = 4-8 ³¹ ). Thus, it was determined that there would be adequate statistical power as the analyses for this study involved HLM.

YT Intervention

The YT intervention is detailed elsewhere following the CheckList stAndardising the Reporting of Interventions for Yoga (CLARIFY) guidelines. ³² Briefly, the intervention was developed by an expert yoga therapist (AP) who integrated lineage (eg, Hatha), post-lineage (eg, somatic movement), and modality focused (eg, restorative) practices within the intervention with the goal of helping participants regulate their autonomic nervous system functioning. Participants completed a 90-minute 1:1 YT session followed by 6 weekly 60-minute group sessions at the CHI over a 7-week period. Groups sessions were delivered to 2–3 participants by the same yoga therapist who delivered their 1:1 session to ensure an in-depth personalized approach; see the published protocol ²⁵ for a description of the yoga therapists’ background and specific training given. Sessions were delivered by 1 of 2 yoga teachers who were training with AP to become yoga therapists or by AP herself. The 1:1 YT session provided empathetic listening, co-assessment of participants’ concerns, yogic autonomic nervous system education, and yoga practices specific to participants’ needs and responses. The focus of the group YT sessions derived from the information gleaned from participants’ 1:1 sessions and included a group check-in, more yogic autonomic nervous system education, gentle Hatha poses, lateral somatic movements, breath awareness practices (including possible breath exhalation extension practices), full body relaxation, and a check-out for participants to share their responses to the practice. Yoga sessions were delivered in English using an established modifiable protocol continually tailored to meet participants’ needs and abilities, as learned in the 1:1 sessions, or as they arose and/or changed during the intervention. Participants were encouraged to modify practices themselves, ask for help, rest anytime they wanted, and/or consider imagining the practices (rather than completing them physically) depending on how they were feeling. Soft background music (ie, “Bija” by Todd Norian) was played during the sessions. Of note, participants also received care as usual from conventional healthcare and practitioners at the CHI.

Measures

PROs were prespecified secondary outcomes that were assessed with the use of questionnaires with evidence of score reliability and validity, as detailed in the published protocol. ²⁵ At each timepoint, the following hardcopy questionnaires were completed at the CHI: Functional Assessment of Cancer Therapy-Fatigue Scale (FACT-F ³³ ), State-Trait Anxiety Inventory (STAI ³⁴ ), Functional Assessment of Cancer Therapy-Cognitive Function version 3 (FACT-Cog ³⁵ ), Center for Epidemiologic Studies Depression short-form Scale, ³⁶ Perceived Stress Scale, ³⁷ and Functional Assessment of Cancer Therapy General version 4 (FACT-G ³⁸ ). PROMs were scored and analyzed as recommended by the authors of the measures, except for the FACT-F where data were scored such that higher scores reflect higher levels of cancer-related fatigue. The baseline questionnaire also included questions on sociodemographic, behavioral, and medical characteristics to describe the sample. See Table 2 in the published protocol ²⁵ for the assessment schedule.

Table 2.

Results for 2-Level Hierarchical Level Models.

Outcomes	Fixed effects				Random effects
	Estimate (SE)	95% CI	df	P	σ2	τ00 ID	τ11 ID.Time	ρ01 ID	ICC
State anxiety a
Intercept	22.61 (2.69)	17.29, 27.93	58.0	<.001	48.58	64.29	-	-	0.57
Time	−3.75 (0.72)	−5.17, −2.34	107.0	<.001
Phase	8.83 (2.45)	3.99, 13.68	107.01	<.001
Time-by-phase	3.51 (1.39)	0.77, 6.26	107.26	.01
Trait anxiety a
Intercept	35.42 (2.63)	30.20, 40.64	22.53	<.001	11.04	115.35	0.49	0.29	0.91
Time	−1.06 (0.38)	−1.80, −0.31	88.66	.01
Phase	1.34 (1.17)	−0.97, −3.65	90.03	.25
Time-by-phase	−0.28 (0.67)	−1.60, 1.04	94.08	.68
Depression b
Intercept	7.38 (1.51)	4.39, 10.38	41.62	<.001	7.26	28.71	-	-	0.80
Time	−0.13 (0.43)	−0.97, 0.72	91.97	.77
Phase	−0.12 (1.08)	−2.26, 2.02	91.99	.91
Time-by-phase	−0.44 (0.62)	−1.67, 0.79	92.09	.48
Cancer-related fatigue b
Intercept	35.93 (2.23)	31.52, 40.35	42.71	<.001	16.72	60.25	1.81	0.25	0.79
Time	−1.59 (0.71)	−3.00, −0.17	88.94	.03
Phase	3.88 (1.64)	0.62, 7.13	74.10	.02
Time-by-phase	2.56 (0.96)	0.66, 4.45	77.06	.01
QoL: Physical wellbeing b
Intercept	25.46 (0.77)	23.94, 26.98	55.55	<.001	2.47	6.02	-	-	0.71
Time	0.28 (0.25)	−0.21, 0.77	92.00	.27
Phase	−0.52 (0.63)	−1.77, 0.73	92.04	.41
Time-by-phase	−0.37 (0.36)	−1.09, 0.36	92.19	.32
QoL: Social wellbeing b
Intercept	23.84 (1.02)	21.83, 25.85	45.10	<.001	3.57	12.29	-	-	0.78
Time	−0.10 (0.30)	−0.70, 0.49	92.01	.73
Phase	−0.18 (0.76)	−1.68, 1.32	92.04	.81
Time-by-phase	0.79 (0.44)	−0.08, 1.65	92.15	.08
QoL: Emotional wellbeing b
Intercept	19.38 (0.93)	17.54, 21.22	49.99	<.001	3.31	9.50	-	-	0.74
Time	0.55 (0.29)	−0.02, 1.12	91.88	.06
Phase	−0.27 (0.73)	−1.71, 1.18	91.91	.72
Time-by-phase	−0.10 (0.42)	−0.93, 0.74	92.04	.82
QoL: Functional wellbeing b
Intercept	22.18 (1.22)	19.77, 24.59	42.08	<.001	4.76	18.45	-	-	0.79
Time	0.75 (0.35)	0.07, 1.43	91.96	.03
Phase	−0.45 (0.88)	−2.18, 1.29	91.99	.61
Time-by-phase	−0.79 (0.50)	−1.79, 0.21	92.08	.12
Stress b
Intercept	14.33 (1.86)	10.63, 18.02	48.02	<.001	13.29	38.42	1.04	0.24	0.75
Time	−0.76 (0.62)	−1.99, 0.47	91.11	.22
Phase	−1.28 (1.46)	−4.18, 1.62	74.43	.38
Time-by-phase	0.03 (0.85)	−1.66, 1.72	77.54	.97
QoL: Overall b
Intercept	90.87 (2.78)	85.35, 96.39	40.31	<.001	23.72	99.56	-	-	0.81
Time	1.48 (0.77)	−0.05, 3.00	91.98	.06
Phase	−1.41 (1.96)	−5.28, 2.46	92.01	.47
Time-by-phase	−0.47 (1.13)	−2.70, 1.77	92.09	.68
Perceived cognitive impairments b
Intercept	55.44 (3.46)	48.58, 62.31	34.88	<.001	30.18	169.36	-	-	0.85
Time	2.16 (0.87)	0.44, 3.88	92.05	.02
Phase	−1.61 (2.21)	−5.98, 2.76	92.06	.47
Time-by-phase	−0.83 (1.27)	−3.34, 1.69	92.13	.52
Comments from others b
Intercept	15.50 (0.22)	15.07, 15.93	108.00	<.001	0.34	0.17	-	-	0.33
Time	−0.08 (0.09)	−0.26, 0.11	92.21	.42
Phase	0.15 (0.23)	−0.32, 0.61	92.35	.53
Time-by-phase	0.16 (0.14)	−0.11, 0.43	92.87	.24
Perceived cognitive abilities b
Intercept	18.75 (1.56)	15.67, 21.83	50.93	<.001	9.41	26.48	-	-	0.74
Time	0.05 (0.49)	−0.91, 1.01	92.05	.92
Phase	0.27 (1.23)	−2.18, 2.71	92.08	.83
Time-by-phase	1.02 (0.71)	−0.38, 2.43	92.21	.15
Impact of perceived cognitive impairments on QoL b
Intercept	13.08 (0.97)	11.16, 15.01	52.01	<.001	3.86	9.80	0.46	−0.63	0.77
Time	0.73 (0.35)	0.04, 1.41	88.10	.04
Phase	−0.73 (0.79)	−2.29, 0.83	74.14	.36
Time-by-phase	0.11 (0.46)	−0.79, 1.02	77.57	.81

Notes. ^aN = 19. ^bN = 20. Effect interpretations are as follows: Intercept: average levels upon starting the YT intervention, Phase: immediate shift in level between baseline and intervention phases, Time: linear slope, Time-by-phase: change in linear slope between baseline and intervention phases. Some cells are empty due to low/no variance. Abbreviations: estimate = unstandardized regression coefficient; SE = standard error; df = degrees of freedom; P = P-value; σ² = within-person variance; τ_{00 ID} = between-person variance; τ_{11 ID.Time} = random-slope variance (ie, between-participant variance in slopes); ρ_{01 ID} = slope-intercept correlation (ie, how random intercepts and random slopes are related); ICC = intraclass correlation coefficient. *P < .05.

Public and Patient Involvement

No members of the public (including patients) were involved in the design, conduct, and interpretation of this study.

Statistical Analysis

Descriptive statistics were used to summarize baseline characteristics of the sample. Whilst methods of data analysis in single subject studies have been the topic of ongoing debates, recent developments suggest inferential statistics in the form of HLM—also called multilevel modeling—can be useful in analyzing data of single-subject studies involving multiple participants.^39,40 Thus, HLM was used to assess the effects of the YT intervention on PROs.

After checking the tenability of regression assumptions, a 2-level HLM approach suggested by Van den Noorgate and Onghena ³⁹ was used to analyze each PRO. HLM is ideally suited to analyze the data because it takes the hierarchical nature of the data into account (ie, measurements are nested within participants), and it allows for analysis of multiple participants in an analytical model instead of interpreting each participant’s data individually. Accordingly, it allows for a more meaningful assessment of the average effects of an intervention. ³¹ In addition, HLM allows for the estimation of between-person variability, an important consideration because the magnitude of intervention effects may vary from person to person.

After determining that HLMs were appropriate based on P-values < .05 for unconditional models, predictor variables were added. These included fixed effects for time, phase, and time-by-phase. The time variable allowed for estimating the overall rate of change during the study, the phase variable allowed for testing if there was a significant change immediately upon YT exposure (ie, right after the 1:1 YT session), and the time-by-phase interaction variable allowed for estimating whether the amount and direction of change differed upon between the 2 study phases (ie, baseline and YT intervention). ⁴¹ Importantly, time was centered at the start of the intervention phase (thus, the intercepts are interpreted as average levels of PROs at the start of the YT intervention), and the models assumed that the overall influence of time on the outcome was linear. All models initially included random effects to estimate within- and between-person variances; however, as the study hypotheses related to the fixed effects (not the random-effects structure), variance components were removed from the models if they were not supported by the data. ⁴² Within each model, the intraclass correlation coefficient reflects the proportion of the total variance that is due to between-person differences. Full information REML (Restricted Maximum Likelihood) estimation was used as it can provide reliable estimates with missing data in small samples. ⁴³ As such, missing data were not imputed; all data from the 20 participants who initiated the YT intervention were included in the analysis. Additionally, because observations collected closer in time are likely to be more correlated than observations collected further apart in time, a first-order autoregressive error structure was specified to account for autocorrelation. ⁴⁴ Sensitivity analyses were conducted to examine the degree to which estimates were affected by different modeling decisions (eg, allowing the Level-1 error covariance matrix to vary across participants ⁴⁵ and exclusion of 2 cases with different patterns); no discrepancies were observed. Descriptive and main data analyses were completed using SPSS and R packages, respectively, and a significance level of α = .05 was used.

Results

Sample and Descriptive Statistics

Figure 1 shows the CENT flow diagram of participants from the screening process onward (based on the CENT 2015 statement). In total, 120 adults were referred or self-referred to participate in this study, and of these, 25 were successfully contacted, met eligibility criteria, provided written informed consent, and completed the first baseline assessment. Five participants dropped out after completing the first (n = 4) or second (n = 1) assessment prior to commencing the intervention; their data were excluded from all analyses. Reasons for dropout included: dissatisfaction with the assessment protocol (n = 1), too busy (n = 2), self-consciousness when wearing the Hexoskin shirt (n = 1), or complications following an (unrelated) eye cataract surgery (n = 1). Of the remaining 20 participants who commenced the intervention, all completed the 1:1 YT session and 19 completed all YT group sessions; 1 discontinued the intervention after completing 5 YT group sessions (no reason provided for the missed session). The intervention completion rate (defined as the number of intervention sessions attended by participants) was therefore 99%. The same participant who discontinued the intervention did not complete the last 2 assessments (no reason provided); however, their data were included in the analyses, yielding an analytical sample size of 20. The dropout rate (defined as the number of participants who did not return to finish the last assessment) was 5%.

Participants flowed through the study protocol in 8 different waves (ie, 4 in 2017, 2 in 2018, and 2 in 2019). Five waves included 3 participants in the group YT sessions, while the other 3 included 2 participants due to dropouts. Of note, 1 participant joined another wave’s group sessions to serve as a third participant once they had completed all study-related procedures (ie, intervention and assessments). Table 1 provides a description of the analyzed sample at baseline.

Table 1.

Sociodemographic and Clinical Characteristics of the Analytical Sample (N = 20).

Characteristic	M	SD
Age (years) a	55.74	8.62
Time since cancer diagnosis (years) b	2.83	3.45
Time since last cancer treatment (years) b	1.50	2.08
	N	%
Sex (female)	17	85.0
Civil (relationship) status (married/common law)	13	65.0
Education (completed university/college/graduate school) b	15	78.95
Annual household income (>$100 000 CAD) c	8	47.06
Employment status (part/full-time worker)	10	50.0
Ethnicity (White [Caucasian])	17	85.0
Health issues
Heart disease risk factors (eg, high blood pressure, high cholesterol)	6	30.0
Mood disorder (eg, anxiety)	3	15.0
Cancer type
Bladder	1	5.0
Blood	2	10.0
Brain	1	5.0
Breast	11	55.0
Colorectal	1	5.0
Pancreatic	1	5.0
Prostate	1	5.0
Other (eg, eye melanoma and esophageal)	2	10.0
Cancer stage d
I	4	26.67
II	4	26.67
III	4	26.67
IV	1	6.67
Not applicable or not reported	7	13.33
Medical treatment received for cancer
Surgery	16	80.0
Chemotherapy	14	70.0
Radiotherapy	15	75.0
Hormonal therapy	4	20.0

Notes. ^an = 18, ^bn = 19, ^cn = 17, ^dn = 15.

Estimated Effects

Supplemental File Figures 1–8 show the visualization of predicted values from fixed intercepts/slopes (shaded areas display 95% confidence intervals [CI] of the estimates). Table 2 reports the HLM results. The intercepts reflect the model estimated average levels of PROs at the start of the YT intervention, and the fixed effects can be thought of as the effects for a typical participant. To aid in interpretation, the intercept for cancer-related fatigue suggests a typical participant showed a level of 35.93 (from a total possible score of 0–52) upon starting the YT intervention. The significant negative fixed effect of time for cancer-related fatigue demonstrates that levels for a typical participant decreased by 1.59 (3.06%) on a scale of 0–52 at each time point across the study. The significant positive fixed effect of phase for cancer-related fatigue shows that levels for a typical participant increased by 3.88 (7.46%) on a scale of 0–52 immediately after the 1:1 YT session. Finally, the significant positive effect of time-by-phase for cancer-related fatigue indicates a change in the amount, but not direction, of change wherein a typical participant demonstrated a larger decrease in levels over time once the YT intervention was introduced.

As for remaining PROs, the models for state and trait anxiety showed significant negative time effects. On a scale of 20–80, state anxiety levels linearly decreased by 3.75 (6.25%) and trait anxiety levels linearly decreased by 1.06 (1.77%) at each time point across the study. Positive phase and time-by-phase interaction effects were also observed for state anxiety. The phase effect indicates that levels increased by 8.83 (14.72%) on a scale of 20–80 immediately after the 1:1 YT session. The time-by-phase interaction effect indicates a larger decrease in state anxiety levels over time once the YT intervention was introduced.

The models for the Perceived Cognitive Impairments and Impact of Perceived Cognitive Impairments on QoL subscales of the FACT-Cog and the Functional Wellbeing subscale of the FACT-G showed evidence of linear change during the study, but none showed evidence of significant change upon YT exposure nor a significant time-by-phase interaction effect (meaning that the amount or direction of change did not differ across phases). The time effects shows that Perceived Cognitive Impairments subscale scores increased linearly by 2.16 (3.0%) at each time point on a scale of 0–72 (indicative of fewer cognitive problems over time), Impact of Perceived Cognitive Impairments on QoL subscale scores increased linearly by 0.73 (4.5%) at each time point on a scale of 0–16 (indicative of better QoL over time), and Functional Wellbeing subscale scores increased linearly by 0.75 (2.68%) at each time point on a scale of 0–28 (indicative of better functional wellbeing over time).

There was no evidence of linear change over time, immediate YT exposure effect, nor a change in the slope (ie, amount or direction) between phases for depression, 3 FACT-G subscales (ie, Physical Wellbeing, Social Wellbeing, and Emotional Wellbeing), 2 FACT-Cog subscales (ie, Comments From Others and Perceived Cognitive Abilities), FACT-Cog total score, and stress.

Discussion

Several studies involving large cohorts of adults diagnosed with cancer have shown that conventional treatments negatively impact PROs.^{3 -9} This study evaluated the effects of a YT intervention that combined individualized 1:1 YT followed by group YT. Whilst developed to improve autonomic nervous system functioning, it was hypothesized that YT would also improve a range of PROs (ie, cancer-related fatigue, anxiety, cognitive function, depression, stress, and QoL)—the focus of this manuscript. Though results will have to be confirmed in future trials, improvements in cancer-related fatigue, anxiety, perceived cognitive impairments, impact of perceived cognitive impairments on QoL, and functional wellbeing were observed over time. However, the most important finding is that cancer-related fatigue and state anxiety, but no other PRO, showed greater improvements during the group YT intervention phase compared to before, despite levels increasing immediately following the 1:1 YT session. These results suggest 6 weekly small (ie, 2-3 participants) group YT sessions may produce notable benefits for these specific symptoms.

Published studies and systematic reviews have demonstrated the beneficial effects of mind-body practices (including yoga) to alleviate various symptoms after treatment for cancer, including cancer-related fatigue and anxiety.^{13,15 -20,46,47} The significant findings of the present study complement this evidence by demonstrating that a 7-week YT intervention designed to be gentle, self-modulated, and focused on nervous system regulation can reduce cancer-related fatigue and state anxiety. Understanding mechanisms by which the intervention facilitated decreases in cancer-related fatigue and state anxiety over time is crucial to inform how future YT interventions should be designed and delivered to maximize effectiveness. As YT is a mind-body practice that can enhance the interaction between the mind and body, possible mechanisms should be investigated from multiple perspectives.^48,49 On the one hand, YT may reduce cancer-related fatigue and state anxiety via physiological mechanisms. Indeed, based on evidence suggesting that yoga and other mind-body practices including meditation, Qigong, and mindfulness directly impact physiological parameters,^{50 -54} and evidence of associations between physiological parameters and cancer-related fatigue and anxiety, these mechanisms may include autonomic nervous system activation and homeostasis maintenance, improvements in heart rate variability measures, anti-inflammatory effects, neural pathways, neurogenesis, neuroplasticity, and/or improved pulmonary function. As well, YT may reduce cancer-related fatigue and state anxiety by fostering a calm and pleasant state, promoting focus on and acceptance of physical sensations, feelings, and thoughts without judgment, enhancing emotional self-regulation abilities, and allowing oneself to live in the present moment. As such, measures should be diverse to assess a range of potential physiological and psychological mechanisms of symptom improvement.

Moreover, as the YT intervention featured small group sessions, similar to prior studies offering yoga to people with cancer,^13,55,56 social mechanisms should be investigated. Speculatively, safe sharing and shared understanding, peer-to-peer interactions, and opportunities for building personal relationships with other participants may promote the continuous relief of cancer-related fatigue and state anxiety. This perspective aligns with sociocultural frameworks that emphasize the importance of the mechanisms, processes, and bi-directional exchanges related to the individual and contextual relations in the study of experiences ⁵⁷ and research demonstrating that people diagnosed with cancer value participating in yoga with others. ⁵⁸ Based on previous research, ⁵⁹ the group may have engendered a forum to connect with peers, learn from others’ experiences, receive peer support (eg, encouragement, companionship, and informational), and share disease experiences before and/or after the sessions. In turn, the group may have become an arena in which participants helped each other understand their cancer-related fatigue and anxiety and develop personalized coping strategies based on shared information. The growing trend to offer group yoga to effectively use limited resources speaks to the need for further research regarding the role of social processes (eg, peer-to-peer interactions, opportunities for building personal relationships with other participants, socioemotional context, group atmosphere, group engagement, and communication patterns and interactions between participants) in reducing cancer-related fatigue and anxiety to maximize opportunities for positive experiences when people participate in group YT interventions after treatment for cancer.

Although it was posited that YT would also improve trait anxiety, stress, depression, cognitive function, and QoL, the a priori aim of demonstrating effects of the YT intervention on these PROs was not met as there were no associated significant phase or time-by-phase effects, though some aspects of cognitive function and QoL improved during the study (possibly due to non-specific factors such as contact with a research assistant and natural improvements that occur with increasing time since treatment). The null effects observed stand in contrast to previous studies showing positive effects from yoga on these outcomes.^{15,60 -62} Future research needs to explore the reasons for this. Whilst it is plausible that the YT intervention simply did not affect these outcomes (as perhaps they are less closely linked to autonomic nervous system functioning than cancer-related fatigue and state anxiety) and that YT, based on its nature, could be developed to target specific PROs, there are other reasons that could explain the current findings. In this sample, average levels were already good for most PROs prior to the YT intervention, leaving little room for improvement (ie, ceiling effect). Another possible explanation is that participants may have attempted other interventions (eg, pharmaceutical treatments, psychotherapy, behavioral treatments, and mind-body therapies) to enhance their cancer recovery, and therefore shown no further symptom response as a result of the YT intervention. Indeed, any reported symptoms for these PROs may be residual (ie, symptoms that linger despite an adequate dose and duration of an intervention). To investigate this possibility, it may prove helpful to add a measure of other intervention participation as a possible adjustment factor.

Additionally, at present, there remains a lack of understanding with regards to for whom and how YT might be successful after treatment for cancer. PROs are complex and can be influenced by a range of interacting factors (eg, physical, lifestyle, psychological, and environmental), which vary from person to person. Accordingly, individualization was central in this study. For example, the small group YT sessions worked with the information gleaned from each participant’s 1:1 session with further sessions continually tailored to meet participants’ needs and abilities. Nevertheless, people with cancer are a heterogeneous group with distinct patterns of symptoms and the groups were not created based on participants’ presentation of symptoms. Perhaps creating groups based on the symptoms participants present during the 1:1 YT would allow for an even higher level of individualization specifically targeted to participants’ presentation of symptoms (eg, depression), which may yield larger effects. Alternatively, the null effects observed may be due to the dosage (ie, frequency, duration, and length) and/or timing of the YT intervention evaluated in this study. It is possible that it was not intensive or long enough and/or it did not start early enough to improve specific PROs. Considering that some yoga interventions for adults diagnosed with cancer showing improvements in anxiety, stress, depression, cognitive function, and QoL have offered multiple sessions per week over a longer period of time, ¹³ the introduction of a YT intervention with ≥2 sessions per week for ≥ 6 weeks may result in improvements. Last, the timing of the intervention and participants’ expectations on the efficacy of the sessions to directly improve their trait anxiety, stress, depression, cognitive function, and QoL may have directly affected these outcomes in this study. Therefore, it would be worthwhile to explore the impact of intervention timing and participants’ perceptions of sessions and their relevance on PROs.

Finally, despite extensive research investigating the effects of yoga interventions for persons diagnosed with cancer,^13,55,56 there is limited information on the effects of an initial or single yoga (or YT) session (see Mackenzie et al ⁶³ for exception). Indeed, outcomes are often assessed prior to and after interventions comprised of multiple sessions in quantitative studies, or participants are prompted to discuss their experiences after completing a multisession intervention in qualitative studies.^13,55,56 Results consistently show that people experience more favorable outcomes after practicing yoga for a period of time,^13,55,56 though it is possible that “null” or “negative” findings are not published (see Cramer et al ⁶⁴ for exception) nor are “adverse effects.” ⁶⁵ However, examining outcomes and experiences after weeks of practice and often days (or weeks) following the final session obscures how people respond immediately upon starting the intervention. Participants’ initial responses may differ from their responses after weeks of practice; yet this paradoxical effect is seldom examined. A strength of the current study is that the effects of YT on PROs were investigated both immediately after the (first) 1:1 YT session and across the entire YT intervention (including group YT sessions). Compared to the results observed across the entire YT intervention phase wherein cancer-related fatigue and state anxiety decreased, levels increased immediately after the 1:1 YT session. Whilst this may not be related to the YT practice itself, but rather the result of having to complete assessments before and after a 90-minute YT session, this contradictory effect could be due to “relaxation-induced anxiety” ⁶⁶ or suggest that there is a rebound effect. This rebound hypothesis is consistent with the Opponent-Process Theory, ⁶⁷ which posits that the hedonic response to a stimulus can be considered through 2 oppositive processes. The a process reflects the immediate unconditioned hedonic response to a stimulus, whereas the b process is a delayed and slower response that occurs once the stimulus is terminated and the a process has ended. ⁶⁸ The a and b processes have opposite hedonic qualities, such that the b process will return a person to their hedonic homeostasis or beyond. ⁶⁸ Another explanation for this contradictory effect might be due to abreaction (ie, therapeutic process of bringing forgotten or lived experiences [eg, thoughts, experiences, memories, and emotions] from the unconscious into consciousness or to present moment awareness). Anecdotally, this phenomenon is not uncommon, especially among people who have had traumatic experiences and wish to “get on with life.” Plausibly, participants may have focused on their inner selves in a clear (Vidya) and honest (Satya) manner for the first time, enabling better awareness of internal feelings and sensations right now. In turn, this may have left them feeling fatigued and anxious immediately after the session before they had a chance to work through their inner feelings and sensations in the subsequent sessions. YT works to create the conditions to help participants tune in to their inner selves with compassion. Accordingly, participants may not have been more anxious after their 1:1 YT session, but more aware of their nervous system (dis)regulation; in essence, more aware of their embodied anxiety. The same may be true for fatigue. These explanations are, of course, speculative, but warrant continued careful examination.

Nevertheless, if these hypotheses were to hold true, these results exemplify the relevance of disentangling the acute and chronic effects of YT (and yoga) in future research, as well as the need to actively and systematically assess effects frequently, especially in the beginning. Relatedly, the results underscore the importance of mechanistic research that focuses on understanding how and under what circumstances 1:1 YT yields different effects, giving attention to the different components (eg, empathetic listening, co-assessment of participants’ concerns [including persistent side effects], yogic education with regard to the regulation of the nervous system, and yoga practices) and other plausible mechanisms. Also, the inflection point (ie, when decreases start appearing) could be further examined to provide indication of how long programs should be at a minimum. Last, from a practical standpoint, the results highlight the importance of qualified yoga therapists who conduct a check-out for participants to share their responses to the practice.

Strengths and Limitations

There are strengths and limitations in the design and conduct of this study for researchers and stakeholders to consider. Key strengths are the broad, pragmatic eligibility criteria, which better reflects the “real-world” and individuals who might seek YT interventions, as well as the use of validated PRO measures and high adherence rates. Another strength is the study design used (ie, single-subject experimental design), which reduces important threats to internal validity (eg, maturation, multiple testing, and statistical regression) in an efficient manner.^26,27 However, an important limitation is that the sample comprised few men, persons of non-White race, and persons with lower socioeconomic status. Relatedly, the sample was a volunteer sample; as such “volunteer bias” or the Hawthorne effect may be present. Indeed, an enthusiastic and enlightened sample of volunteers (independent of the intervention) willing to commit to the intervention and 7 assessment timepoints might be expected to seek supportive care and have better outcomes. Second, the study design presents some threats to internal validity (eg, time-related factors such as history and testing effects). ²⁶ Moreover, without a no-YT intervention control group within a randomized controlled trial, the possibility that the YT interventions effects were observed due to other reasons (eg, participants overstating change taking place as they were hoping for positive changes) cannot be ruled out. Third, the conclusions are based on group-level statistical analyses and do not mean that the effect was the same for every participant in the sample. Fourth, based on visual inspection of baseline data (data not reported, but available upon request to the corresponding author), there were few notable differences between the analyzed sample and dropouts in sociodemographic, behavioral, and medical characteristics or in PROs at baseline. Specifically, dropouts did not include men, tended to be more often married/common law than completers, and had slightly higher trait anxiety, overall QoL, and social wellbeing than completers. Consequently, results may not generalize well beyond this sample of participants. Fifth, engagement in other therapeutic modalities (eg, naturopathy and acupuncture) that could affect PROs may be an important factor to consider when interpreting the results from this study; however, these data were not collected from participants. Finally, no long-term follow-up was included, thus, maintenance data cannot be reported. Addressing these limitations in the future by including larger and more diverse samples, assessments over a longer period, and a comparison group is necessary to provide convincing data to support the purported positive effects of continued YT on cancer-related fatigue and state anxiety, and in turn provide justification for the resources needed for large-scale implementation.

Implications and Future Research

Pending future studies, the findings demonstrate that a YT intervention that combined individualized 1:1 YT followed by group YT can alleviate 2 common and debilitating symptoms reported after treatment for cancer—that is, cancer-related fatigue and state anxiety. Nevertheless, in its current form, it could not be linked to improvements in other PROs. Future studies may benefit from: (1) tailoring YT interventions to target specific PROs relevant to each participant, (2) testing different dosages, intervention timing, and combinations of delivery modes (ie, changing the 1:1 to group ratio), and (3) exploring underlying mechanisms of change (ie, mediating factors). The latter were not examined, yet they are a critical piece moving forward to identify the “active ingredients” or strategies that are particularly effective. Thus, intrapersonal (ie, internal factors such as motivation, self-efficacy, perceived control, and degree of empowerment), interactional (ie, participants’ effective interaction with the environment [including the YT therapist and other participants], and behavioral (ie, actions participants take to influence the outcomes through their participation] should be considered, both as process and outcome variables to learn more about how/why YT can (or cannot) improve specific PROs.

Relatedly, future studies with larger sample sizes could provide the opportunity to investigate factors that influence response to YT (ie, moderating factors) to answer questions such as: For which people and under which conditions does the intervention have an effect/stronger effect? Last, this intervention was delivered face-to-face, which makes it less/not accessible for some people due to geographic barriers. Considering alternative ways to offer the intervention, such as via videoconferencing platforms that have seen immense growth, and examining feasibility, acceptability, and comparable/differential effects is warranted. Relatedly, for adults who struggle to participate in structured programs (face-to-face YT or otherwise), it is important to consider the implications of including motivational reinforcements and behavioral supports to help them engage more fully, hopefully resulting in better overall outcomes. Collectively, this type of research will contribute to optimized YT interventions with discernable effects on key PROs in this cohort.

Conclusions

The scope of cancer care research and practice has evolved over the years with growing efforts to develop and evaluate interventions to improve PROs to curtail healthcare costs and meet the complex needs of people diagnosed with cancer. This study is among the first to examine the effects of a YT intervention on PROs among adults who have completed treatment for cancer. The results exemplify the potential relevance of YT. Nevertheless, before any definite conclusions are made, future studies are needed to replicate the current results in larger, more diverse samples. Moreover, as depressive symptoms, multiple aspects of QoL (ie, physical wellbeing, social wellbeing, and emotional wellbeing), stress, and cognitive functioning did not improve with the YT intervention, possible explanations for the null findings should be investigated. Finally, in an effort to better understand whether future YT interventions yield benefits, additional PROs should be assessed and mixed-methods adopted to better understand factors underlying intervention effects or lack thereof.