Subgroup effects of occupational therapy-based intervention for people with advanced cancer

Abstract

Background: Many people with advanced cancer have decreased ability to perform activities of daily living (ADL). We recently performed a randomized, controlled trial (RCT) assessing the efficacy of an occupational therapy-based program, the ’Cancer Home-Life Intervention’ in people with advanced cancer (N = 242) and found no overall effects on ADL ability. However, heterogeneity of treatment effect may disguise subgroup differences.

Objective: To investigate whether subgroups of people with advanced cancer gain positive effects from the ‘Cancer Home-Life Intervention’ on ADL ability.

Material and method: An exploratory subgroup analysis including 191 participants from a RCT. The outcome was ADL motor ability measured by the Assessment of Motor and Process Skills (AMPS). Subgroups were defined by age, gender, years of education, type of primary tumor, functional level, and activity problems.

Results: The ‘Cancer Home-Life Intervention’ had no statistically significant effect in the six subgroups. Modifying effects of age (0.30 [95% CI: −0.05 to 0.64]) and gender (0.23 [95% CI: −0.11 to 0.57]) were not found.

Conclusion: There were no subgroup effects of the ‘Cancer Home-Life Intervention’on ADL motor ability. Some indications suggest greater effects for those aged below 69 years; however, this result should be interpreted with caution.

Keywords:

• Activities of daily living
• neoplasms
• occupational therapy
• randomized controlled trial
• moderator variables

Introduction

Although people with advanced cancer are living with an incurable disease, they still need and value to be engaged in activities of daily living (ADL) [1,2]. ADL encompass the ability to perform personal ADL (PADL) related to self-care (e.g. eating and dressing) and instrumental ADL (IADL) related to home maintenance (cooking, housework, and shopping) [3]. Studies show that between 43%–74% of those with advanced cancer have decreased ability to perform ADL [2,4,5]. Problems with ADL may contribute to loss of independence [3] and may also have unfavorable consequences for individuals’ health-related quality of life (HRQoL) [6]. One of the aims of occupational therapy-based (OT-based) interventions is to support people to perform their ADL [7]. Yet, little is known about efficacious OT-based interventions for people with advanced cancer [8–10]. We therefore performed a full-scale randomized, controlled trial (RCT) [11] to assess the efficacy of a tailored, adaptive, OT-based program, the ‘Cancer Home-Life Intervention’, compared with usual care received by people with advanced cancer living at home (N = 242).

The ‘Cancer Home-Life Intervention’ showed no effect compared with usual care with respect to ADL motor ability, which was the primary outcome, measured by means of the Assessment of Motor and Process Skills (AMPS) [11]. However, since the trial included a heterogeneous population, the effect may vary in subgroups of people with advanced cancer [12]. To our knowledge, only two RCTs have investigated subgroup effects of OT-based interventions [13,14]. Gitlin et al. conducted a subgroup analysis of an RCT to investigate the effectiveness of an OT-based intervention in community-living older adults and found effects on ADL in three subgroups according to age, gender, and level of education [13]. The applied OT-based intervention [13] had five OT sessions involving education, problem-solving, home modification, and energy conservation training. Sheffield et al. also conducted a subgroup analysis in their RCT to examine the effectiveness of an OT-based intervention similar to that of Gitlin et al. in people aged 65 and older. However, they found no subgroup effects on ADL [14].

Identifying subgroups with advanced cancer that may benefit from the ‘Cancer Home-Life Intervention’ can help clinical practitioners and OT researchers target those who would benefit most from the intervention. The purpose of the present study was therefore to explore whether we could identify subgroups of people with advanced cancer who gained positive effects from the ‘Cancer Home-Life Intervention’ with respect to ADL motor ability.

Methods

This study was an exploratory subgroup analysis of data from a rater-blinded, two-armed RCT. The study protocol and the main results from the RCT have been published previously [11,15].

The RCT evaluating the efficacy of the ‘Cancer Home-Life Intervention’

In total, 242 adults (≥18 years) who lived at home were recruited from two Danish hospitals from February 2015 to October 2016 [10]. The sample in the present study comprised 191 participants who had completed the AMPS at baseline (T1) and at 12-weeks of follow-up (T2). They had been diagnosed with advanced cancer and had a functional level of 1–2 on the World Health Organization (WHO) performance status (PS). WHO PS1 describes that the person is ambulatory and able to manage daily activities of a lighter nature, but is restricted in performing physically demanding activities. WHO PS2 denotes people who are ambulatory and able to manage self-care, but are restricted in performing work activities [16]. An additional inclusion criterion was that the participants lived at home or in sheldered accomodation. Patients were excluded if they lived in a nursing home or hospice, had cognitive impairments, or lacked the proficiency in Danish required for them to complete questionnaires and participate in interviews.

The ADL ability of the included participants was measured at T1 and T2 by eight blinded data-collection occupational therapists. After T1, participants were randomized to the intervention group or the control group in a 1:1 ratio. The randomization was stratified by hospital (Aarhus University Hospital (AUH), Odense University Hospital (OUH) and palliative care unit at OUH.

The intervention group was given the ‘Cancer Home-Life Intervention’ as supplement to usual care. The ‘Cancer Home-Life Intervention’ is an adaptive OT-based program targeted towards the participants prioritized everyday activities, including ADL. It is delivered by an occupational therapist (OT) and has six intervention components: (1) a mandatory initial interview that addresses everyday activity problems and needs. Based on this interview, one or more of the following components are selected by the OT and the participant: (2) prioritization of resources, energy, and activities; (3) adaptation of activities; (4) adaptation of posture and seating positioning; (5) provision of assistive technology; and (6) modification of the physical home environment. The delivery of the intervention taskes place during 1–3 home visits and 1–3 follow-up telephone contacts. The purpose of the follow-up telephone contacts are to support the participants’ use of the adaptive strategies learned in the selected components and to resolve new problems which might arise in-between the home visits. Since the ‘Cancer Home-Life Intervention’ is tailored, the combination of components 2–6, number of home visit and telephone contacts vary in each participant. See the results and the delivery of the ‘Cancer Home-Life Intervention’ in our published paper [11].

Outcome

ADL motor ability was measured by the AMPS. The AMPS is a standardized, observation-based, occupational therapy instrument that reflects two domains: ADL motor ability and ADL process ability [3]. ADL motor ability measures are adjusted for rater severity, ADL task challenge, and skill item difficulty. Higher positive measures represent better ADL ability. ADL motor ability evaluates the amount of physical effort, clumsiness, and/or fatigue a person demonstrates during ADL task performance. A change of ≥0.3 logits indicates a clinically relevant change [3]. The AMPS is found to be a valid and reliable instrument for people with advanced cancer [17,18] and has also demonstrated good responsiveness in people with other diseases [3].

Subgroups

Our study protocol outlined four subgroups to be included in a subgroup analysis [15] – age, gender, primary tumor, and WHO PS – as research suggests that these factors are associated with ADL or physical functioning [19–23]. These factors may also be associated with a different treatment response to the ‘Cancer Home-Life Intervention’. Two additional variables were chosen after the data collection was completed: having activity problem(s) and years of education. This post-rationalisation was done for two reasons. First, 28% of the participants in the RCT [11] reported not having any activity problems. This may indicate that these individuals had less need for an OT-based intervention and were thus less likely to benefit from the intervention. Second, we included educational level, since Gitlin et al. found that those with less education benefited more from their OT-based intervention on ADL than did those with a longer education [13].

The subgroups were categorized as follows: gender (men versus women), age (<69 years versus ≥69 years, as 69 years was the median value), primary tumor (lung, head and neck, gynaecological, prostate, breast, gastrointestinal, bladder, and other), functional level (WHO PS1 versus WHO PS2), activity problems (yes versus no), and years of education (≤10 years, 11–12 years, and ≥13 years).

Age and gender were registered using a study-specific questionnaire. Hospital nurses at two hospitals collected information about the participants’ primary tumor from the responsible oncologist and rated the participants’ WHO PS. Activity problems were assessed by the data-collection occupational therapists who used the Individually Prioritised Problems Assessment (IPPA) instrument. The IPPA is a self-report interview-based instrument where a person prioritizes up to seven activity problems to be solved [24–26]. Information on educational level was collected using the study-specific questionnaire. All six variables were collected either before or at T1.

Statistical analysis

ADL motor ability at T1 and T2 are presented with mean values and standard deviation (SD) in the intervention group and in the control group, divided into six subgroups. Between-group mean changes were calculated for each subgroup with 95% confidence intervals (95% CI). A multiple regression analysis [27] was performed with ADL motor ability at T2 as the dependent variable, and treatment (intervention versus control), baseline ADL motor ability, gender, age, years of education, and activity problems as predictors. Regression coefficients for each predictor were estimated with 95% CI, and p values were reported. Interaction terms were then added to the regression analysis between the treatment and the selected modifiers (age and gender). Functional level and primary tumor were not included in the regression analysis for two reasons. Firstly, there was collinearity between baseline ADL motor and functional level in the model [28]. Secondly, the size of the primary tumor subgroups were too small to be included in the model. Only age and gender were selected in the interaction test since these characteristics seem to be the most important modifiers [13,29]. Furthermore, research consistently shows significant age and gender ADL differences [19,22,23]. A statistically significant interaction was tested by a Wald test [27]. Model assumptions were investigated by QQ plot and histogram. A p value of 0.05 was considered statistically significant, and all analyzes were performed using Stata 14.

Ethics

The Danish Regional Committees on Health Research Ethics considered approval not to be required (S-20122000-96). The Danish Data Protection Agency (FN 215-57-0008) approved the study. All included participants provided oral and written informed consent and were informed that participation in the study was voluntary and that they could withdraw at any time. The trial was registered at www.controlled-trials.com/ClinicalTrials.gov NCT02356627.

Results

Table 1 shows the T1 mean ADL motor ability and T2 mean ADL motor ability in the intervention group and the control group, stratified for each subgroup. The between-group mean change for the subgroups is also shown in Table 1. In the six subgroups, the mean ADL motor ability changes from T1 to T2 for both treatment groups were approximately normally distributed. Overall, the between-group mean change in the six subgroups was small, all confidence intervals included 0 and the estimates were generally below the threshold of a clinically relevant difference (≥0.3 logits) [3] (see Table 1). For instance, in those aged <69 years, the T1 ADL motor ability was 1.19 logits in the intervention group and 1.33 logits in the control group. During the 12-week follow-up, ADL motor ability decreased to 1.12 logits in the intervention group and 1.08 logits in the control group. The intervention group therefore decreased less than the control group (0.18 logits [95% CI: −0.09 to 0.45]), although this was not statistically significant, as 0 was included in the 95% CI (see Table 1). Overall, the results in Table 1 did neither favor the intervention group nor the control group. Eight subgroups favor the intervention group (positive measures) while 11 subgroups favor the control group (negative measures). Most of the measures were, however, far from being clinical relevant.

Table 1. Mean ADL motor ability in the intervention group andcontrol group and subgroup differences from baseline to 12-week follow-up.

	T1 mean ADL motor ability (SD)		T2 mean ADL motor ability (SD)
Subgroups	Intervention group	Control group	Intervention group	Control group	Between-group mean change^a (95% CI)
Treatment (n = 191)
Intervention group (n = 97)	1.16 (0.58)	–	1.01 (0.69)	–	–
Control group (n = 94)	–	1.21 (0.58)	–	1.10 (0.61)
Age (n = 191)
<69 (n = 95)	1.19 (0.56)	1.33 (0.57)	1.12 (0.52)	1.08 (0.75)	0.18 (−0.09 to 0.45)
≥69 (n = 96)	1.13 (0.59)	1.08 (0.58)	0.91 (0.82)	1.11 (0.44)	−0.25 (−0.53 to 0.03)
Gender (n = 191)
Women (102)	1.05 (0.64)	1.17 (0.61)	1.15 (0.40)	1.14 (0.61)	0.13 (−0.10. to 0.37)
Men (n = 89)	1.26 (0.49)	1.26 (0.54)	0.88 (0.87)	1.04 (0.62)	−0.16 (−0.46 to 0.15)
Education (n = 190)
≤10 years (n = 50)	1.15 (0.68)	1.09 (0.58)	0.95 (0.94)	1.16 (0.42)	−0.27 (−0.73 to 0.19)
11-12 years (n = 52)	1.21 (0.42)	1.16 (0.53)	0.94 (0.69)	0.99 (0.61)	−0.10 (−0.48 to 0.28)
≥13 years (n = 88)	1.12 (0.62)	1.30 (0.59)	1.12 (0.47)	1.11 (0.70)	0.19 (−0.07 to 0.44)
Primary tumor (n = 190)
Lung (n = 36)	1.21 (0.44)	1.08 (0.61)	1.05 (0.60)	0.85 (0.78)	0.07 (−0.49 to 0.64)
Head and Neck (n = 11)	0.90 (0.56)	1.06 (0.94)	0.25 (1.14)	0.82 (1.16)	−0.41 (−1.90 to 1.08)
Gynecological (n = 9)	NA	1.44 (0.32)	NA	1.41 (0.35)	NA
Prostate (n = 25)	1.29 (0.41)	1.34 (0.45)	1.10 (0.48)	1.17 (0.68)	−0.02 (−0.48 to 0.42)
Breast (n = 32)	0.91 (0.41)	1.03 (0.65)	1.03 (0.33)	1.08 (0.43)	0.07 (−0.31 to 0.44)
Gastrointestinal (n = 61)	1.32 (0.59)	1.44 (0.51)	1.07 (0.81)	1.30 (0.41)	−0.11 (−0.47 to 0.26)
Bladder (n = 12)	0.64 (1.00)	1.01 (0.34)	0.98 (0.50)	0.97 (0.39)	0.38 (−0.41 to 1.18)
Other (n = 4)	0.95 (0.79)	NA	1.15 (0.53)	NA	NA
Functional level (n = 190)
PS1 (n = 139)	1.28 (0.44)	1.32 (0.51)	1.12 (0.66)	1.13 (0.61)	0.03 (−0.20 to 0.24)
PS2 (n = 51)	0.88 (0.75)	0.83 (0.64)	0.78 (0.74)	0.97 (0.62)	−0.24 (−0.65 to 0.18)
Activity problems (n = 191)
No (n = 57)	1.47 (0.38)	1.45 (0.43)	1.11 (0.74)	1.20 (0.44)	−0.11 (−0.42 to 0.19)
Yes (n = 134)	1.02 (0.59)	1.11 (0.61)	0.97 (0.67)	1.05 (0.67)	0.01 (−0.20 to 0.24)

ADL: Activities of Daily Living.

T1: Baseline.

T2: 12-week follow-up.

^a A change of ≥0.3 logits in ADL motor ability indicates a clinically relevant change.

The adjusted regression analysis with interaction terms is shown in Table 2. All model assumptions were fulfilled. The statistically significant predictors of T2 ADL motor ability were baseline ADL motor ability (0.48 logits [95% CI: 0.32 to 0.64], p = <0.000) and gender (−0.24 logits [95% CI: −0.42 to −0.06], p = 0.01). Gender and age did not statistically significantly modify the effects of the treatment on T2 ADL motor ability (see Table 2). We did, however, observe a tendency for interaction effect by age group, as the 95% CI tended to be in the positive direction with 0.64 logits in the upper end (0.30 logits [95% CI: −0.05 to 0.64], p = 0.09).

Table 2. Linear regression model investigating predictors for ADL motor ability and interaction of age and gender.

Predictors	Adjusted mean ADL motor ability at T2^a^,^b [95% CI]	p values	Interaction with treament group^a^,^b [95% CI]	p of Interaction
Treatment
Control group (n = 94)	Ref.
Intervention group (n = 97)	−0.02 [−0.20 to 0.15]	0.82
Baseline ADL motor ability	0.48 [0.32 to 0.64]	<0.000
Age		0.30 [−0.05 to 0.64]		0.09
Age <69 (n = 95)	Ref.
Age ≥69 (n = 96)	−0.03 [−0.21 to 0.15]	0.76
Gender		0.23 [−0.11 to 0.57]		0.19
Women (n = 101)	Ref
Men (n = 89)	−0.24 [−0.42 to −0.06]	0.01
Education
≤10 years (n = 50)	Ref
11–12 years (n = 52)	−0.08 [−0.32 to 0.16]	0.54
≥13 years (n = 88)	−0.03 [−0.25 to 0.19]	0.78
Activity problems
No (n = 57)	Ref.
Yes (n = 134)	−0.01 [−0.21 to 0.20]	0.96

T2: 12-week follow-up.

^a Multiple regression adjusted for age, gender, education, perceived activity problems, center, and baseline ADL motor ability.

^b A change of ≥0.3 logits in ADL motor ability indicates a clinically relevant change.

Discussion

The aim of the present exploratory study was to identify subgroups with advanced cancer who may have gained positive effects from the ‘Cancer Home-Life Intervention’ regarding ADL motor ability. Overall, we found that the intervention had no effects in any of the investigated subgroups. Men compared with women and younger participants (<69 years) compared with older participants (≥69 years) did not respond differently in ADL motor ability after the intervention. We did, however, observe some indications of greater effects of the ‘Cancer Home-Life Intervention’ amongst younger participants.

Previous studies investigating subgroup effects of an OT-based intervention have focused on older adults living in the community [13,14]. Only the study by Gitlin et al. found subgroup effects on ADL. They showed that their OT-based intervention had greater effects on ADL for individuals aged ≥80 [13]. This is contrary to our findings of greater effects for those aged <69 years. The two study populations are, however, not immediately comparable since people with advanced cancer live with a life-threatening illness and often experience fluctuating symptoms and problems [20]. Furthermore, Gitlin et al. included individuals only aged ≥70 years, and thus had no younger group as a frame of reference [13].

We performed two post-hoc analyzes to investigate the robustness of our results regarding the possible influence of age on ADL motor ability. First, we wanted to challenge the indication of greater effects for the younger group by setting an even lower age cutoff (>60 years versus ≤60 years). Second, when dichotomizing age, statistical power is lost, and therefore we entered age as a continuous covariate. These adjustments to the age variable was entered in the multiple regression analysis described in the Method section. Lowering the age cutoff still showed greater effects in the younger group even though the result still was not statistically significant (0.08 logits [95% CI: −0.36 to 0.52], p = 0.71). Furthermore, entering age as a continuous covariate in the regression model did not lead to statistically significantly results (−0.02 logits [95% CI: −0.04 to 0.03], p = 0.10). Even though age did not moderate the effect of the ‘Cancer Home-Life Intervention’ statistically, participants aged <69 years seemed to benefit more than those aged ≥69 years. This result may be partly explained by the fact that the younger group have greater unmet needs [30] and may therefore be more receptive to applying the components of the intervention into their daily life at home. However, our trial may be underpowered to detect statistically significant subgroup effects or interactions, and the results should therefore be interpreted cautiously.

Methodological considerations

With respect to general recommendations for conducting subgroup analysis [31], our study has several strengths. Four subgroups were chosen in the design phase of the RCT with all variables being collected before or at the time of randomization. We investigated effects on the primary trial outcome, ADL motor ability, as research shows that people with advanced cancer demonstrate increased clumsiness or physical effort or inefficiency during ADL task performance [5]. In the present study, we studied only one outcome, the primary trial outcome, as recommended [31]. The AMPS data were collected by trained and calibrated occupational therapists, and data quality was validated during the entire study period by the Center for Innovative OT Solutions.

The study has two main limitations: (1) Even though the intervention group and the control group were balanced in terms of gender, baseline ADL motor ability, education, center and activity problems they may not be comparable in the analysis regarding other characteristics because the stratification of participants into subgroups was made after the randomization (see Table 2). This means that our indication of greater effect in participants aged <69 years may be attributable to other causes than age alone. For example, the two groups were not balanced for tumor diagnoses which may be a factor that influences the results. Our sample size calculation was not powered to detect a clinically significant interaction, wherefore the estimates are less likely to reach statistical significance [1,15].

The ‘Cancer Home-Life Intervention’ showed no effects in the main RCT [11] or in the present exploratory subgroup study. We need more knowledge about why the ‘Cancer Home-Life Intervention’ was not effective. The Medical Research Council recommends performing process evaluation to assess the fidelity and implementation of interventions, and which mechanisms of impact are at stake when trying to produce change through an intervention [32]. In line with these recommendations, we are currently conducting a process evaluation of the RCT. Furthermore, only 36 out of 121 intervention participants received more than a single home visit and only 46 received more than a single follow-up telephone contact [11]. The lack of effect may be explained by the low intervention intensity and the short time of delivery (three weeks). However, to our knowledge the minimum intervention intensity required to instill a change in ADL ability is not yet established. Thus, more research is needed that investigates the potential dose-response relationship of an OT-based intervention in people with advanced cancer and what might be the optimal dose. In addition, it may also be relevant to investigate the most efficacious delivery format of an OT-based intervention (telephone-delivered and in-person contacts or telephone-delivered alone).

Implications for occupational therapy practice and research

The purpose of subgroup analysis in an RCT is to identify groups of people that may respond differentially to an intervention [12].

• Our results are mainly useful to OT researchers as they show the importance of being aware of the possible influence of age on ADL ability in an OT-based intervention like the ‘Cancer Home-Life Intervention’.

• Future studies should consider stratifying for age in their randomization scheme.

• This kind of knowledge is useful when planning future RCT studies that investigate the effectiveness of OT-based interventions.

• The present OT-based intervention is the largest RCT so far performed in people with advanced cancer, [33,34] and indicates the potential usefulness of conducting subgroup analyzes of future OT-based interventions.

Conclusion

This is the first study to investigate subgroup effects of an OT-based intervention in people with advanced cancer. The ‘Cancer-Home Life Intervention’ had no statistically significant effects in subgroups defined by gender, age, years of education, type of primary tumor, extent of functional limitations, or having activity problems or not. No modifying effects of age and gender were found. There was a tendency towards greater effects in participants aged <69 years, but this result should be interpreted with caution and viewed as hypothesis-generating rather than be used to set guidelines for practice.

Acknowledgements

We wish to thank all the participants who took part in the study, the OTs, and statistician Professor Jakob Hjelmborg, University Southern Denmark, for statistical consultancy on the analyzes.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The trial was supported financially by TrygFonden, the Danish Cancer Society (R53-A2783), the University of Southern Denmark, the Danish Association of Occupational Therapists (FF 2 14 – 3), and the Region of Southern Denmark (15/23775).