Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Aug 18.
Published in final edited form as: Eur J Cancer Care (Engl). 2019 Feb 4;28(2):e12994. doi: 10.1111/ecc.12994

Modifiable pathways from pain to functional status: Confirmatory baseline results from a randomised trial of African American patients with cancer pain

April Hazard Vallerand 1, Thomas N Templin 1, Susan M Hasenau 1, Angelika Maly 1
PMCID: PMC8373261  NIHMSID: NIHMS1728924  PMID: 30719796

Abstract

Objective:

This study tested a model of cancer-related pain and functional status in African American patients, including beliefs about the ability to control pain as a key determinant of distress and functional status.

Methods:

Baseline data from a randomised clinical trial consisting of clinical and patient-reported outcomes were used. Participants were 228 African American patients experiencing moderate to severe pain within the past 2 weeks. The model comprised four latent constructs: pain, perceived control over pain, pain-related distress and functional status. Confirmatory factor analysis was used to validate the factor structure of the measurement model. Structural equation modelling was used to estimate direct and mediated effects.

Results:

The measurement model fit well (RMSEA = 0.06, SRMR = 0.05) with all loadings significant (p < 0.05). The structural model also fit well (RMSEA = 0.04, SRMR = 0.05). The complex mediated pathway from pain to functional status through perceived control over pain and pain-related distress was strong and significant (specific indirect effect = −0.456, p = 0.004). Mediation by perceived control accounted for a 47% reduction of the effects of pain on functional status.

Conclusion:

If these results hold up longitudinally, interventions to increase perceived control over pain have the potential to improve functional status by decreasing pain-related distress.

Keywords: functional status, pain, pain-related distress, perceived control over pain, structural equation modelling

1 |. INTRODUCTION

Pain occurs frequently in patients with cancer and is one of the most distressing symptoms experienced (National Cancer Institute, 2015). Despite the options of using analgesics and other modalities, pain continues to be problematic in more than 50% of cancer patients (van den Beuken-van Everdingen et al., 2007). Cancer pain is a multidimensional experience comprised of many factors, including distress and perceived control over pain, all of which affect the patient’s functional status (Leung, Pachana, & McLaughlin, 2014; Schaller, Larsson, Lindblad, & Liedberg, 2015; Vallerand, Templin, Hasenau, & Riley-Doucet, 2007; Wells, 1994; Wells & Sandlin, 2012). Although distress is experienced by all patients with cancer at some point during diagnosis and treatment (Hulbert-Willams, Storey, & Wilson, 2015; National Comprehensive Cancer Network, 2015), in a study of ambulatory patients with cancer-related pain, general symptom distress was found to be low, whereas, pain-related distress was high (Vallerand, Templin et al., 2007). In addition, Wells found that pain-related distress was the most important factor explaining the relationship between pain and pain-related interference with function (Wells, Murphy, Wujcik, & Johnson, 2003). While pain-related distress is important and should be assessed in all patients with pain (NCI, 2015), designing interventions to decrease distress is challenging due to the affective nature of the concept. The factors that lead to pain-related distress are more amenable to intervention strategies. Perceived control over pain, a factor that had not been previously considered, was found to have a direct effect on pain-related distress and mediated the effect of beliefs about pain and pain level on distress in ambulatory patients with cancer-related pain (Vallerand, Templin et al., 2007).

Perceived control over pain has been defined as the perception that one has a way of gaining and/or maintaining control over an aversive event, such as pain (Ovayolu, Ovayolu, Aytac, Serce, & Sevinc, 2014). Control does not need to be exercised for it to be effective; it just needs to be perceived (Thompson, 1981). Beliefs about one’s ability to control daily symptoms have been found to be more influential on well-being than beliefs about one’s ability to control the progression of disease (Thompson, 1981; Vallerand & Ferrell, 1995). Clinically, these patients are more willing to go about their daily activities without fear of increased pain because they know what to do if their pain increases. In our previous study (Vallerand, Hasenau, Templin, & Collins-Bohler, 2005), black patients with cancer pain were found to have more pain, higher symptom distress and lower functional status than white patients. After controlling for perceived control over pain, this disparity between the races was diminished. These finding led to this current study which consisted of African American patients with cancer-related pain.

Our previous work developed a model of factors that affect functional status in patients with cancer pain: pain level, beliefs about pain (knowledge, attitudes and barriers), symptom distress and perceived control over pain. (Vallerand, et al, 2005). The relationship between pain level and functional status is mediated by perceived control over pain and symptom distress. Beliefs about pain management have a significant effect on perception of control over pain, and increased perception of control over pain decreases symptom distress (chiefly pain-related distress), ultimately leading to better functional status.

The primary purpose of this study was to test a mediation model of cancer-related pain and functional status in African Americans. Mediation models are concerned with process: “A theoretical premise posits that an intervening variable is an indicative measure of the process through which an independent variable is thought to impact a dependent variable” (Iacobucci, Saldanha, & Deng, 2007, p 139). In the current model, two intervening variables/mediators are proposed. It is hypothesised that patients’ perceived control over pain and pain-related distress mediates the relationship between pain and functional status. If these relationships are supported, interventions targeting perceived control could impact both pain-related distress and functional status.

A secondary purpose of this study was to examine alternative models of the relationship between pain and functional status, leaving out, one at a time, perceived control and pain-related distress. These models support the primary results by showing how the relationship between pain and functional status is explained by incorporating the proposed mediation.

2 |. METHODS

Using baseline data from our larger randomised control trial (R01 CA149432–01A1), testing the Power Over Pain—Coaching (POP-C) intervention, a hypothesised model of cancer-related pain and functional status was used to test the mediating role of perception of control and symptom distress.

2.1 |. Participants and setting

Participants were recruited from the waiting room of a large urban, comprehensive cancer centre. Inclusion criteria consisted of self-selection as African American, diagnosis of cancer, 18 years of age or older, cognitively intact, English speaking and having moderate to severe (≥4 on a 0–10 scale) pain in the previous 2 weeks. Table 1 displays the demographics of the participants (N = 228).

TABLE 1.

Demographic data

Variable n Mean or %
Gender
 Male 90 40%
 Female 138 60%
Age, mean (range) 228 55 (20–87)
Marital status
 Single 112 49%
 Married 51 23%
 Separated 8 4%
 Divorced 35 15%
 Widowed 21 9%
Education
 Grade School 6 3%
 High School 131 58%
 College 79 35%
 Graduate School 9 4%
Primary site of cancer
 Breast 48 21%
 Gastrointestinal 34 15%
 Lung 34 15%
 Genitourinal 31 14%
 Oral 24 11%
Months since diagnosis Mean (range) 221 30 (0 to 193)
Metastasised
 No 139 61%
 Yes 82 36%
Open in a new tab

2.2 |. Procedures

Following IRB approval from the University, patients were approached in the waiting room by the study’s recruiters to determine whether they met the inclusion criteria and were interested in participating in the study. Patients were then assigned a Data Collection Nurse who contacted the patient by phone to arrange a time and date to meet that was convenient for the patient for the first, baseline meeting. Participants were seen in their home, informed consent was obtained, and data were collected by the Data Collection Nurses for the first study visit. Demographic data were measured by a demographic data sheet developed by the investigators. The questionnaires were then completed with the Data Collection Nurse reading the questions to the participant while the participant sat next to the nurse or nearby, whichever was the most comfortable for the participant. The average length of time to complete the questionnaires was 30 min. Data were collected from August 2011 to December 2014.

2.3 |. Measures

The indicators used to define each of the four primary study constructs are described below and in Table 2:

TABLE 2.

Constructs, indicators and bootstrapped estimates of factor loadings (estimates), and 95% confidence intervals

Constructs Predetermined indicators No. Items Estimates Lower Upper
Pain Brief pain inventory-short form [BPI]—Least pain 1 0.80 0.72 0.86
BPI—Average pain 1 0.74 0.62 0.84
BPI—Worst pain 1 0.68 0.56 0.79
BPI—Pain now 1 0.79 0.70 0.86
Patient Pain Questionnaire [PPQ 10] (Pain over the last week)a 1
Patient Pain Questionnaire [PPQ 11] (Pain now) 1 0.79 0.71 0.86
Pain-related distress Distress Thermometer [DT] (Distress in past week) 1 0.80 0.65 0.85
Memorial Symptom Assessment Scale [MSAS] (Bothered by pain) 1 0.78 0.68 0.83
Symptom Distress Scale [SDS 4] (level of pain during past week)b 1
Patient Pain Questionnaire [PPQ 13] (Distress to self) 1 0.62 0.45 0.72
Patient Pain Questionnaire [PPQ 14] (Distress to family) 1 0.52 0.37 0.65
Perceived control over pain Perceived Control Scale [PCS] (Pharmacological control) 8 0.47 0.32 0.58
Survey of Pain Attitudes [SOPA] (Cognitive Control Subscale) 10 0.66 0.54 0.75
Pain Catastrophizing Scale [PCat] (Feelings about your pain) 13 −0.78 −0.87 −0.67
Patient Pain Questionnaire [PPQ 15] (Feel you can control your pain) 1 0.44 0.25 0.58
Functional status Brief Pain Inventory-Short Form [BPI] (Interference subscale) 7 −0.79 −0.88 −0.68
SF-12 [PCS] (Physical composite) 6 0.47 0.30 0.59
SF-12 [MCS] (Mental composite) 6 0.58 0.45 0.67
American Chronic Pain Association-Quality of Life Scale [ACPA-QOL] 1 0.49 0.30 0.63
Open in a new tab

Notes. Indicators without estimates were excluded due to high cross-factor residual correlations.

a

Indicator had a high cross-factor residual correlation with distress (r = 0.23).

b

Indicator had a high cross-factor residual correlation with worst pain (r = 0.30).

2.3.1 |. Pain

Four items rating pain intensity (worst, least and average in the last 2 weeks and now) from the Brief Pain Inventory—Short Form (BPI-SF; Cleeland & Syrjala, 2001) and one item from the Patient Pain Questionnaire (PPQ; Ferrell, Eberts, McCaffery, & Grant, 1991) were used to measure pain intensity. The BPI-SF was developed to assess pain in cancer patients, is short enough to be considered for routine clinical use with cancer patients and has a reliability factor of 0.85 (Ware, Kosinski, & Keller, 1996). The four selected pain items used are numerical rating scales that ask the patient to rate the severity of their pain on a 0–10-point scale. The item from the PPQ asked how much pain you are having now, also on a 0–10-point scale. A second item from the PPQ, asking how much pain they had over the last week was eliminated as an indicator as it was found to have a residual correlation with the latent variable, pain-related distress. The internal consistency reliability score for the pain items in this study was 0.90.

2.3.2 |. Perceived control over pain

As a latent factor, perceived control over pain (PCP) was defined by three scales from existing instruments and one question about perceived control. The three scales were the Survey of Pain Attitudes Control Subscale (Jensen, Turner, Romano, & Lawler, 1994); Pain Catastrophizing Scale (PCat; Sullivan, Bishop, & Pivik, 1995); and the Perceived Control Scale (PCS; Pellino & Ward, 1998). The Control Subscale (CS) from the Survey of Pain Attitudes (SOPA) is 10 items rated on a 5-point scale of (0) very untrue to (4) very true. The CS measures perceptions of control over pain and the ability to influence the amount of pain experienced. The instrument has a reliability of 0.81(Jensen, Karoly, & Chant, 1987). In this study, the reliability was 0.80. The PCat rates the frequency of 13 thoughts or feelings when experiencing pain, on a 5-point scale from (0) not at all to (4) all the time yielding a total score (r = 0.91). In this study, the reliability was 0.92. Items from the PCS reflect perceived control of post-operative pain and perceptions of control over taking pain medication and efficacy of pain relief. Items relating to post-operative pain have been modified, with permission, to relate to cancer pain. Responses are summed and divided by 8 to get an overall control score. In this study, the reliability was 0.68. One item from the PPQ questioning how controlled the patient believed their pain was, on a Likert 0 to 10 scale, was also used. The internal consistency reliability score for the pain items in this study was 0.71.

2.3.3 |. Pain-related distress

The construct pain-related distress was measured using two items from the PPQ (Ferrell et al., 1991), the distress thermometer (DT; Ransom, Jacobsen, & Booth-Jones, 2006), an item from the Symptom Distress Scale (SDS; McCorkle & Young, 1978) and one item from the Memorial Symptom Assessment Scale (MSAS; Chang, Hwang, Feuerman, & Kasimis, 2000). The items from the PPQ asked how distressing the pain was to the participant and how distressing the pain was to their family members. The distress thermometer is a one-item, self-report measure of psychological distress with endpoints labelled 0 for no distress and 10 for extreme distress and was developed by the National Comprehensive Cancer Network (NCCN), with a reported reliability of 0.80 (Ransom et al., 2006). The item from the SDS was eliminated from analysis as, on face validity, it did not correlate to this construct, but was highly correlated to the construct, pain. The MSAS scale was developed to provide multidimensional information about a diverse group of common symptoms and has an established reliability of 0.87 (Webber & Davies, 2011). Only one item from the MSAS, pain distress, was used for this study. Pain distress or bother was rated on a 5-point distress scale of 0 not at all to 4 very much. The internal consistency reliability score for the pain-related distress items in this study was 0.76.

2.3.4 |. Functional status

Functional status, for the purposes of this study, was defined as the ability to participate in activities that are meaningful or important to the patient. Functional status was measured by the seven-item interference subscale of the BPI-SF (Cleeland & Syrjala, 2001), the SF-12 physical composite and mental composite (acute version; Ware et al., 1996) and the Quality of Life scale of the American Chronic Pain Association (ACPA-QOL, 2003). Interference with daily life from pain is measured on the BPI-SF with 0 no interference to 10 interferes completely on ratings of how much pain interferes with mood, walking, other physical activity, work, social activity, relations with others and sleep. The interference subscale has previously been validated for use as a measure of the effect of pain on daily life (Tyler, Jensen, Engel, & Schwartz, 2002; Williams, Smith, & Fehnel, 2006). In the current study, the internal consistency for the seven interference items was 0.93. The SF-12 is a 12-item self-report instrument for health-related quality of life with a composite score for mental function and a composite score for physical function and has reported reliability ranging from 0.76 to 0.89 (Ware et al., 1996). The ACPA-QOL is a 1-item measure of activity level for people with pain. The internal consistency reliability score for the functional status items in this study was 0.82.

2.4 |. Data analysis

2.4.1 |. Analysis strategy

Mediation models can be tested using a series of regression equations (Baron & Kenny, 1986; Hayes & Rockwood, 2017) or with structural equation modelling (SEM). SEM has several advantages including more flexibility in the types of mediation models that can be specified and the use of latent variables that take measurement error into account (Iacobucci et al., 2007). A two-step SEM approach was used to test the hypothesised model (Anderson & Gerbing, 1988). In Step 1, a measurement model was fit using a priori specified indicators (see Table 2). Refinements at this stage were made without reference to the structural model. In Step 2, the hypothesised structural model was tested. The structural model is shown in Figure 1. The structural pathways are labelled a–f. Two pathways are hypothesised to mediate the relationship between pain and functional status. The first includes pain-related distress and consists of the paths labelled e and c. The second includes both perceived control over pain and pain-related distress and consists of the paths labelled a, b and c. The two paths labelled d and f were set to zero in order to simultaneously test the proposed mediational pathways. The fit of this model will provide confirmatory evidence for the proposed mediation model with perceived control as a central mediating construct.

FIGURE 1.

FIGURE 1

Open in a new tab

Structural model showing direct effect pathways (a–f). Pathways d and f were fixed to zero in the hypothesized mediation model and freely estimated in an alternative model

2.4.2 |. Estimation

Maximum likelihood (ML) estimation of the variance–covariance matrix was used to estimate parameters and asymptotic standard error. Bootstrapping was used to estimate the standardised indirect effects. Bootstrapping was also used as an alternative estimation method.

2.4.3 |. Model fit

Fit was evaluated using conventional criteria and established guidelines for model evaluation (Kline, 2016). Two measures of approximate fit were used including the root mean square error of approximation (RMSEA < 0.08) and the standardised root mean residual (SRMR < 0.10). Two measures of exact fit were also calculated: the likelihood ratio chi-square based on maximum likelihood, and the Bollen–Stine bootstrapped chi-square. These statistics test the null hypothesis that the entire set of model residuals are simultaneously zero. Failure to reject this null hypothesis in a model that otherwise fits using conventional criteria indicates the existence of sources of misfit that need further inspection.

2.4.4 |. Measurement model

A four-factor congeneric measurement model was estimated (Joreskog, 1971). The indicators used in the initial measurement model were determined a priori, and each measure was expected to load on a single factor (see Table 2). Modification indices were used to identify sources of model misfit. In order to achieve an acceptable fit, correlations among errors of measurement within a factor were allowed. The correlated error was modelled by allowing a correlation between the otherwise independent error terms. Cross-factor measurement error correlations were not allowed. In order to improve fit in this case, the offending item was omitted from the factor.

2.4.5 |. Power and sample size

The sample size of 228 patients was deemed sufficient to test the proposed model. The power for the RMSEA test of close fit (MacCallum, Browne, & Sugawara, 1996) was determined to be 0.98 using the web utility provided by Preacher and Coffman (2006). In addition, a sample size of 200 was recommended as “a goal” for SEMs with latent variables (Kenny, 2015) with smaller sample sizes for simple models, models without latent variables and models with all loadings fixed. For confirmatory factor analysis, it is required that the number of participants is greater than the number of free parameters that must be estimated. Therefore, we used total scores for scales with multiple items rather than modelling all the items individually.

2.4.6 |. Causal model assumptions

Causal interpretation depends on the assumption of no confounding (MacKinnon & Piriott, 2015). Possible confounding of each of the pathways was considered. Pathways are confounded when a third variable is correlated with exposure and causally associated with outcome. To aid in the specification of possible confounders, we examined the correlation of four covariates—age, education, married and metastasised—with each of the model constructs. These correlations were obtained by adding the covariates to the measurement model. Using these results, each covariate was added to the structural model with a causal pathway to the outcome factor it was most strongly correlated with and that made theoretical sense.

The proposed mediation pathways were tested using bootstrapped standard errors as recommended (Shrout & Bolger, 2002). Because the operational definition of mediation includes a reduction in the direct effect once the mediator is added to the equation (Baron & Kenny, 1986), we also estimated direct and indirect effects of mediators taken one at a time.

3 |. RESULTS

3.1 |. Latent variable measurement model

A four-factor confirmatory factor analysis model for the latent variables was estimated prior to the structural equation model (see Figure 2). The latent constructs were comprised of pain, perceived control over pain, pain-related distress and functional status.

FIGURE 2.

FIGURE 2

Open in a new tab

Four factor confirmatory factor analysis results showing correlations among the factors and standardized factor loadings. All coefficients shown were significant, p < .05

3.1.1 |. Estimation

Examination of the individual indicators prior to fitting the model showed no extreme skewness (>2) or kurtosis (>7); however, Mardia’s coefficient of multivariate kurtosis was significant (p < 0.01) and >5 (Bentler, 2005; Cohen, Cohen, West, & Aiken, 2003). Consequently, bootstrapped standard errors and statistics were used for tests of model fit and parameter estimates in addition to the usual statistics that assume multivariate normality.

3.1.2 |. Model fit

After allowing within factor residual correlations, and removing two indicators with a high cross-factor residual, the measurement model fit well (chi-square = 191.83 [df = 107] p < 0.001; root mean square error of approximation [RMSEA] = 0.059, 95% CI = [0.045, 0.072]; CFI = 0.953; the standardised root mean square residual [SRMR] = 0.05; see Figure 2). All factor loadings were significant (p < 0.05). Additional tests allowing for non-normality were also conducted. The Bollen–Stine bootstrap chi-square of the hypothesis that all residuals are simultaneously zero was rejected (p = 0.015). Bootstrap estimates of model parameters were very close to the ML estimates (see Table 2). No bootstrapped estimate of a standardised parameter was different than 0.02 in absolute value.

3.1.3 |. Convergent validity

The standardised factor loadings are shown in Figure 2 adjacent to paths pointing to indicator variables. Two statistics based on these loadings are useful for assessing convergent validity; the ratio of the factor loadings to their standard errors, and average variance explained (AVE) (Segars, 1997). The standardised loadings ranged from 0.43 (Feel in control) to 0.80 (Least pain), and all were significant (p < 0.05). The AVEs were 0.58, 0.48, 0.36 and 0.36 for pain, perceived control, distress and functional status respectively.

3.1.4 |. Discriminant validity

The correlations among the constructs ranged in absolute value from 0.70 to 0.91. The hypothesis of equal correlations was tested rejected (Chi-square [212.5–191.8 = 20.7; with 5 df]) indicating true systematic differences among factor correlations (Fornell & Larcker, 1981).

The correlations among the indicators in the final CFA are shown in Table 3. The shaded blocks show groups of indicators defining a common construct. Correlations within a block were expected to be larger than correlations across blocks. This pattern was seen for pain, perceived control over pain and pain-related distress. It was less evident for the indicators of functional status. The relatively high cross-block correlations are consistent with a factor model in which the factors are highly correlated.

TABLE 3.

Descriptive statistics and correlation matrix of indicators used to define model constructs—pain, perceived control, pain-related distress and function status, N = 228

Pain Perceived control Distress Function
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1. Pain now (PPQ11) 1.00
2. Worst pain (BPI) 0.52 1.00
3. Pain now (BPI) 0.85 0.51 1.00
4. Ave. pain (BPI) 0.50 0.56 0.56 1.00
5. Least pain (BPI) 0.67 0.51 0.76 0.61 1.00
6. Cog. control (SOPA) −0.37 −0.35 −0.33 −0.37 −0.32 1.00
7. Pharm. control (PCS) −0.31 −0.29 −0.31 −0.29 −0.34 0.45 1.00
8. Catastrophizing (PCat) 0.40 0.40 0.37 0.40 0.38 −0.50 −0.33 1.00
9. Feel in control (PPQ15) −0.28 −0.25 −0.27 −0.25 −0.23 0.36 0.27 −0.32 1.00
10. Distress to family (PPQ14) 0.28 0.36 0.27 0.34 0.32 −0.28 −0.27 0.43 −0.04 1.00
11. Distress to self (PPQ13) 0.36 0.40 0.33 0.33 0.33 −0.36 −0.23 0.41 −0.09 0.54 1.00
12. Bothered by pain (MSAS) 0.49 0.36 0.45 0.37 0.33 −0.39 −0.31 0.55 −0.31 0.37 0.46 1.00
13. Distress in past week (DT) 0.47 0.38 0.43 0.39 0.44 −0.46 −0.28 0.51 −0.28 0.33 0.50 0.63 1.00
14. QOL (ACPA) −0.25 −0.33 −0.24 −0.30 −0.26 0.32 0.35 −0.37 0.11 −0.31 −0.23 −0.29 −0.33 1.00
15. Interference (BPI) 0.54 0.48 0.53 0.57 0.55 −0.37 −0.29 0.48 −0.17 0.44 0.46 0.49 0.51 −0.40 1.00
16. Physical Score (SF12) −0.30 −0.27 −0.27 −0.19 −0.21 0.29 0.25 −0.27 0.15 −0.31 −0.35 −0.37 −0.32 0.41 −0.36 1.00
17. Mental Score (SF12) −0.35 −0.19 −0.29 −0.31 −0.32 0.31 0.21 −0.48 0.25 −0.29 −0.32 −0.40 −0.47 0.29 −0.44 −0.02 1.00
Mean 4.96 7.59 4.75 5.90 3.55 1.72 4.70 2.86 5.68 6.69 7.60 2.85 6.23 4.27 5.72 31.21 40.42
SD 3.41 2.35 3.15 2.34 2.80 0.89 1.18 1.01 3.27 3.46 2.79 1.05 3.07 2.29 2.83 9.18 11.25
Open in a new tab

3.2 |. Structural equation model of proposed mediation pathways

The proposed SEM model with the addition of potential confounders is shown in Figure 3. Because each confounder was significantly associated with only one construct, they were not a source of bias and could have been omitted completely. The model fit well (χ2 [df = 177] = 265.34, RMSEA = 0.05, SRMR = 0.05) with covariates (see Figure 3) and without covariates (χ2 [df = 107] = 190.02, RMSEA = 0.06, SRMR = 0.05). The significant path coefficients are shown in bold italics. As expected, the mediated pathway from pain to functional status through perceived control over pain and pain-related distress was significant (specific standardised indirect effect = −0.456, 95% CI = −0.816 to −0.322, p = 0.004). However, the proposed effect of pain on functional status through pain-related distress alone was not significant.

FIGURE 3.

FIGURE 3

Open in a new tab

Structural equation model results of proposed mediation pathways. The effect of pain on functional status through pain related distress alone was not significant (p > .05). The other three proposed pathways were significant

3.2.1 |. Sub-models and reduction in direct effect

Pain was strongly related to functional status but that was not evident from the modelling results because this pathway (d in Figure 2) was fixed to zero, consistent with the hypothesis of complete mediation. A series of simpler models was fit in order to show how mediation pathways affected the correlation between pain and functional status. Covariates were not included because they appeared to make little difference in model fit or path coefficients. When functional status was regressed on Pain alone, Pain accounted for 64% of the variance in functional status (Figure 4, Panel A). Adding perceived control over pain to the model resulted in a 41% reduction in the direct effect of pain and functional status (Figure 4, Panel B). An alternative to the proposed model was also tested. In this model, no restrictions were put on the estimated pathways; this structural model was just identified. The model is shown in Figure 4, Panel C. In this model, the proposed mediation pathway involving perceived control over pain and pain-related distress was still significant (p = 0.028) but mediation was partial rather than complete. The complex of pathways in this model reduced the direct effect of pain on functional status by 60%. The model was also estimated with covariates and all path coefficients were within 0.01 of those shown.

FIGURE 4.

FIGURE 4

Open in a new tab

Sub-models showing reduction in direct effect due to mediated pathways. (a) Latent variable regression of functional status on pain. (b) 41% reduction in direct effect due to mediation by perceived control. (c) Further reduction in direct effect due to mediation by complex of pathways including perceived control and distress

4 |. DISCUSSION

This study of African Americans with cancer pain tested a model of cancer-related pain and functional status. Previous studies found that distress mediated the relationship between pain and functional status (NCCN, 2015; Tyler et al., 2002). Recognising that interventions to decrease distress are difficult to develop, and factors that influence distress were considered.

The National Comprehensive Cancer Network (NCCN, 2015) recommended that all patients with cancer should be screened for distress at their initial visit, at appropriate intervals and as clinically indicated, especially with changes in disease status. All patients with cancer will experience some amount of distress during their illness (Leung et al., 2014). The NCCN suggested that distress extends along a continuum of feelings of vulnerability to problems that can become disabling, such as pain. As supported by the findings of our previous study (Vallerand, Saunders, & Anthony, 2007), increasing one’s perceived control over pain enables patients with a sense that they can make a difference in their own feelings and situations, decreasing their distress and sense of vulnerability. Clinically, patients who lack perceived control over pain are often distressed by increases in their pain and limit activities to prevent pain. Patients who perceive they can control their pain rarely become distressed and are more willing to participate in activities because they know how to manage their pain if it increases. Several methods have been identified in the literature to improve perceived control over pain (Vallerand, Templin et al., 2007; Vallerand, Templin, Robinson-Lane, & Hasenau, 2018). Therefore, applied to pain, increasing perceived control over pain decreases pain-related distress and improves function or potentially disabling pain.

Lack of clarity in the definitions of pain-related distress and symptom distress has led to confusion and difficulty utilising these concepts (Wells & Ridner, 2008). Symptoms or psychological distress is a broader concept that has been used to measure multiple symptoms at one time (Goodell & Nail, 2005). Intervening in these situations becomes challenging due to the variability of responses. Clark et al. (2003) suggested that pain is comprised of intensity and distress, similar to the measures of pain in the Symptom Distress Scale (McCorkle & Young, 1978). Although symptom distress and pain-related distress are often used interchangeably, one of the major defining characteristics of psychological distress is the inability to cope effectively (Ridner, 2004). When considering pain-related distress, increasing perceived control over pain is synonymous with increasing effective coping. Though pain intensity may not decrease, when perceived control over pain is increased, pain-related distress may decrease. Therefore, intervening on the patient’s perceived control over pain may be a more effective way of decreasing pain-related distress. Using perceived control over pain as a factor affecting distress may be more beneficial in developing interventions to increase functional status then attempting to address distress directly.

4.1 |. Limitations

Causal claims can be supported but not validated with observational data used in this study. Because of this, some researchers (Kline, 2016) have argued that mediation in cross-sectional models like the one tested here should be referred to using the more neutral operational term, indirect effect. The contrasting point of view was well articulated by Hayes and Rockwood (2017).

Inferences about substantive meaning are made not with output from routines built into statistical software, but by researchers who are attempting to make sense of and interpret that output. Inferences are products of our minds, not our mathematics. …. So, we don’t agree that one cannot conduct a mediation analysis with correlational data. … Most any statistical tool can provide some insight into the story you ultimately end up telling with your data

(p. 40).

A causal interpretation assumes the constructs in the proposed model were ordered correctly with perceived control antecedent to pain-related distress. A different ordering or reciprocal effects could yield different results. The model also assumes no unmeasured confounders (James, Mulaik, & Brett, 1982; VanderWeele, 2015). We tested for possible confounding with covariates we observed but it is possible that unobserved covariates could be also affect the paths.

This study presents a new measurement model of the relationship between pain and functional status providing empirical support for the conceptual distinction between pain-related distress and perceived control over pain. The model includes empirically supported constructs measuring pain-related distress and perceived control over pain. The measurement model and the proposed mediation were supported but could still be improved. Interventions focused on improving perceived control over pain have the potential for decreasing pain-related distress and increasing functional status.

ACKNOWLEDGEMENTS

This article is dedicated to the memory of Dr. Stephanie M. Schim, a colleague and friend, instrumental in developing this model, who is dearly missed.

Funding information

Funded by the National Cancer Institute 1 R01 CA149432-01A1.

REFERENCES

  1. American Chronic Pain Association. (2003). Quality of life scale: A measure of function for people with pain. ©2003.http://www.theacpa.org
  2. Anderson JC, & Gerbing DW (1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103, 411–423. 10.1037/0033-2909.103.3.411 [DOI] [Google Scholar]
  3. Baron RM, & Kenny DA (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. 10.1037/0022-3514.51.6.1173 [DOI] [PubMed] [Google Scholar]
  4. Bentler PM (2005). Structural equations program manual. Encicno, CA: Multivariate Software. [Google Scholar]
  5. Chang VT, Hwang SS, Feuerman M, & Kasimis B (2000). The memorial symptom assessment scale short form (MSAS-SF) validity and reliability. Cancer, 89, 1162–1171. 10.1002/1097-0142(20000901)89:5<1162:AID-CNCR26>3.0.CO;2-Y [DOI] [PubMed] [Google Scholar]
  6. Clark WC, Kuhl J, Keohan M, Knotkova H, Winer RT, & Griswold GA (2003). Factor analysis validates the cluster structure of the dendrogram underlying the Multidimensional Affect and Pain Survey (MAPS) and challenges the a priori classification of the descriptors in the McGill Pain Questionnaire (MPQ). Pain, 106, 357–363. 10.1016/j.pain.2003.08.005 [DOI] [PubMed] [Google Scholar]
  7. Cleeland C, & Syrjala KL (2001). How to assess cancer pain. In Turk DC, & Melzack R(Eds.), Handbook of pain assessment (pp. 362–387). New York, NY: The Guilford Press. [Google Scholar]
  8. Cohen J, Cohen P, West SG, & Aiken LS (2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Lawrence Erlbaum Associates Publishers. [Google Scholar]
  9. Ferrell BR, Eberts MT, McCaffery M, & Grant M (1991). Clinical decision making and pain. Cancer Nursing, 14, 289–297. 10.1097/00002820-199112000-00002 [DOI] [PubMed] [Google Scholar]
  10. Fornell C, & Larcker DF (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18, 39–50. 10.1177/002224378101800104 [DOI] [Google Scholar]
  11. Goodell T, & Nail L (2005). Operationalizing symptom distress in adults with cancer: A literature synthesis. Oncology Nursing Forum, 32, E42–E47. 10.1188/05.ONF.E42-E47 [DOI] [PubMed] [Google Scholar]
  12. Hayes AF, & Rockwood NJ (2017). Regression-based statistical mediation and moderation analysis in clinical research: Observations, recommendations, and implementation. Behaviour Research and Therapy, 98, 39–57. 10.1016/j.brat.2016.11.001 [DOI] [PubMed] [Google Scholar]
  13. Hulbert-Willams NJ, Storey L, & Wilson KG (2015). Psychological interventions for patients with cancer: Psychological flexibility and the potential utility of Acceptance and Commitment Therapy. European Journal of Cancer Care, 24, 15–27. 10.1111/ecc.12223 [DOI] [PubMed] [Google Scholar]
  14. Iacobucci D, Saldanha N, & Deng X (2007). A meditation on mediation: Evidence that structural equations models perform better than regressions. Journal of Consumer Psychology, 17(2), 139–153. 10.1016/S1057-7408(07)70020-7 [DOI] [Google Scholar]
  15. James LR, Mulaik SA, & Brett JM (1982). Causal analysis: Assumptions, models, and data. Thousand Oaks, CA: Sage Publishers. [Google Scholar]
  16. Jensen MP, Karoly P, & Chant D (1987). The development and preliminary validation of an instrument to assess patients’ attitudes towards pain. Journal of Psychosomatic Reseach, 31, 393–400. [DOI] [PubMed] [Google Scholar]
  17. Jensen MP, Turner JA, Romano JM, & Lawler BK (1994). Relationship of pain-specific beliefs to chronic pain adjustment. Pain, 57, 301–309. 10.1016/0304-3959(94)90005-1 [DOI] [PubMed] [Google Scholar]
  18. Joreskog KG (1971). Statistical analysis of sets of congeneric tests. Psychometrika, 36, 109–133. 10.1007/BF02291393 [DOI] [Google Scholar]
  19. Kenny D (2015). Measuring model fit. Retrieved from http://davidakenny.net/cm/fit.htm
  20. Kline RB (2016). Principles and practice of structural equation modeling. New York, NY: Guilford Press. [Google Scholar]
  21. Leung J, Pachana NA, & McLaughlin D (2014). Social support and health-related quality of life in women with breast cancer: A longitudinal study. Psychooncology, 23, 1014–1020. 10.1002/pon.3523 [DOI] [PubMed] [Google Scholar]
  22. MacCallum RC, Browne MW, & Sugawara HM (1996). Power analysis and determination of sample size for covariance structure modeling. Psychological Methods, 1, 130–149. 10.1037/1082-989X.1.2.130 [DOI] [Google Scholar]
  23. MacKinnon DP, & Piriott AG (2015). Statistical approaches for enhancing causal interpretation of the M to Y relation in mediation analysis. Personality and Social Psychology Review, 19, 30–43. 10.1177/1088868314542878 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. McCorkle R, & Young K (1978). Development of a symptom distress scale. Cancer Nursing, 1, 373–378. 10.1097/00002820-197810000-00003 [DOI] [PubMed] [Google Scholar]
  25. National Cancer Institute. (2015). National Cancer Institute: PDQ® Pain. Bethesda, MD: National Cancer Institute. Date last modified<07/16/2015>.Retrieved from http://www.cancer.gov/about-cancer/treatment/side-effects/pain/pain-hp-pdq [Google Scholar]
  26. National Comprehensive Cancer Network. (2015). NCCN clinical practice guidelines in oncology (NCCN®): Distress management version 2.2015, 7/15 update. Retrieved from http://wwwnccn.org [DOI] [PMC free article] [PubMed]
  27. Ovayolu O, Ovayolu N, Aytac S, Serce S, & Sevinc A (2014). Pain in cancer patients: Pain assessment by patients and family caregivers and problems experienced by caregivers. Supportive Care in Cancer, 23, 1857–1864. 10.1007/s00520-014-2540-5 [DOI] [PubMed] [Google Scholar]
  28. Pellino TA, & Ward SE (1998). Perceived control mediates the relationship between pain severity and patient satisfaction. Journal of Pain & Symptom Management, 15, 110–116. 10.1016/S0885-3924(97)00255-8 [DOI] [PubMed] [Google Scholar]
  29. Preacher KJ, & Coffman DL (2006). Computing power and minimum sample size for RMSEA [Computer software]. Retrieved fromhttp://quantpsy.org/
  30. Ransom S, Jacobsen PB, & Booth-Jones M (2006). Validation of the distress thermometer with bone marrow transplant patients. Psycho-Oncolology, 15, 604–612. 10.1002/pon.993 [DOI] [PubMed] [Google Scholar]
  31. Ridner SH (2004). Psychological distress: Concept analysis. Journal of Advance Nursing, 45, 536–545. 10.1046/j.1365-2648.2003.02938.x [DOI] [PubMed] [Google Scholar]
  32. Schaller A, Larsson B, Lindblad M, & Liedberg GM (2015). Experiences of pain: A longitudinal, qualitative study of patients with head and neck cancer recently treated with radiotherapy. Pain Management Nursing, 16, 336–345. 10.1016/j.pmn.2014.08.010 [DOI] [PubMed] [Google Scholar]
  33. Segars AH (1997). Assessing the unidimensionality of measurement: A paradigm and illustration within the context of information systems research. Omega, 25(1), 107–121. 10.1016/S0305-0483(96)00051-5 [DOI] [Google Scholar]
  34. Shrout PE, & Bolger N (2002). Mediation in experimental and nonexperimental studies: New procedures and recommendations. Psychological Methods, 7(4), 422–445. 10.1037/1082-989X.7.4.422 [DOI] [PubMed] [Google Scholar]
  35. Sullivan ML, Bishop S, & Pivik J (1995). The pain catastrophizing scale: Development and validation. Psychological Assessment, 7, 524–532. 10.1037/1040-3590.7.4.524 [DOI] [Google Scholar]
  36. Thompson SC (1981). Will it hurt less if I can control it? A complex answer to a simple question. Psychological Bulletin, 90, 89–101. 10.1037/0033-2909.90.1.89 [DOI] [PubMed] [Google Scholar]
  37. Tyler EJ, Jensen MP, Engel JM, & Schwartz L (2002). The reliability and validity of pain interference measures in persons with cerebral palsy. Archives of Physical Medicine & Rehabilitation, 83, 236–239. 10.1053/apmr.2002.27466 [DOI] [PubMed] [Google Scholar]
  38. Vallerand AH, & Ferrell BR (1995). Issues of control in patients with cancer pain. Western Journal of Nursing Research, 17, 467–481. 10.1177/019394599501700502 [DOI] [PubMed] [Google Scholar]
  39. Vallerand AH, Hasenau SM, Templin T, & Collins-Bohler D (2005). Disparities between Black and White patients with cancer pain: The effect of perception of control over pain. Pain Medicine, 6, 242–250. 10.1111/j.1526-4637.2005.05038.x [DOI] [PubMed] [Google Scholar]
  40. Vallerand AH, Saunders MM, & Anthony M (2007). Perceptions of control over pain by patients with cancer and their caregivers. Pain Management Nursing, 8(2), 55–63. 10.1016/j.pmn.2007.02.001 [DOI] [PubMed] [Google Scholar]
  41. Vallerand AH, Templin T, Hasenau SM, & Riley-Doucet C (2007). Factors that affect functional status in patients with cancer-related pain. Pain, 132, 82–90. 10.1016/j.pain.2007.01.029 [DOI] [PubMed] [Google Scholar]
  42. Vallerand AH, Templin T, Robinson-Lane SG, & Hasenau SM (2018). Improving functional status in African Americans with cancer pain: A randomized control study. Oncology Nursing Forum, 45, 260–272. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. van den Beuken-van Everdingen MHJ, de Rijke JM, Kessels AG, Schouten HC, van Kleef M, & Patijn J (2007). Prevalence of pain in patients with cancer: A systematic review of the past 40 years. Annuals of Oncology, 18, 1437–1449. 10.1093/annonc/mdm056 [DOI] [PubMed] [Google Scholar]
  44. VanderWeele T (2015). Explanation in causal inference: Methods for mediation and interaction. Oxford, UK: Oxford University Press. [Google Scholar]
  45. Ware J, Kosinski M, & Keller SD (1996). A 12-Item Short-Form Health Survey: Construction of scales and preliminary tests of reliability and validity. Medical Care, 34, 220–233. 10.1097/00005650-199603000-00003 [DOI] [PubMed] [Google Scholar]
  46. Webber K, & Davies AN (2011). Validity of the memorial symptom assessment scale-short form psychological subscales in advanced cancer patients. Journal of Pain & Symptom Management, 42, 761–767. 10.1016/j.jpainsymman.2011.02.007 [DOI] [PubMed] [Google Scholar]
  47. Wells N (1994). Perceived control over pain: Relation to distress and disability. Research in Nursing & Health, 17, 295–302. 10.1002/nur.4770170408 [DOI] [PubMed] [Google Scholar]
  48. Wells N, Murphy B, Wujcik D, & Johnson R (2003). Pain-related distress and interference with daily life of ambulatory patients with cancer with pain. Oncology Nursing Forum, 30, 977–986. 10.1188/03.ONF.977-986 [DOI] [PubMed] [Google Scholar]
  49. Wells N, & Ridner SH (2008). Examining pain-related distress in relation to pain intensity and psychological distress. Research in Nursing & Health, 31, 52–62. 10.1002/nur.20262 [DOI] [PubMed] [Google Scholar]
  50. Wells NL, & Sandlin V (2012). Expectations of pain and accompanying symptoms during cancer treatment. Current Pain & Headache Reports, 16, 292–299. 10.1007/s11916-012-0272-0 [DOI] [PubMed] [Google Scholar]
  51. Williams VS, Smith MY, & Fehnel SE (2006). The validity and utility of the BPI interference measures for evaluating the impact of osteoarthritic pain. Journal of Pain & Symptom Management, 31, 48–57. 10.1016/j.jpainsymman.2005.06.008 [DOI] [PubMed] [Google Scholar]

RESOURCES