Measurement of Irritability in Cancer Patients

Abstract

Background:

Irritability is common among people who are physically ill, but a physical underpinning of irritability is not assessed by existing measures. A measure that assesses multidimensionality of irritability can help nurses and clinicians provide better care for people with cancer and, thus, reduce a risk for developing depression during cancer treatment.

Objectives:

We pilot tested a new measure, The Irritability Scale–initial version (TISi), for assessing irritability of cancer patients on three dimensions: physical, affective, and behavioral.

Methods:

We conducted thee pilot studies to develop the 35-item TISi on a 5-point Likert scale. TISi was tested in 48 early-stage, nonmetastasized breast cancer patients at baseline (before) and three months (during chemotherapy). Of these patients, 62.5% received neoadjuvant and 37.5% received adjuvant chemotherapy, but none received hormonal treatment before or during the study. Measures of other correlates, including depression, anxiety, symptom distress, and social disconnectedness were also administered, and biomarkers of hsCRP, TNF-α, IL-6, and BDNF were obtained from blood draws at both assessments.

Results:

TISi has a high internal consistency (Cronbach’s α = .97), satisfactory test–retest reliability (retest r = .69; ICC = .86), and moderate correlation with other constructs over time (r ≈ .40-.70). Its physical subscale significantly correlated with hsCRP (r = .32, p = .025) at baseline and TNF-α (r = .44, p = .002) at three months. A confirmatory factor analysis yields three factor loadings that are in line with conceptualization of the subscales.

Discussion:

The findings support psychometric properties of TISi and its application for assessing cancer patients’ irritability in multiple dimensions. Further investigation using a large study sample is necessary for improving construct and criterion validity and reducing item redundancy.

Irritability has become a growing area of interest in child and adolescent research in recent years and was found to predict depression and suicidal behaviors in this population (Benarous et al., 2019; Eyre et al., 2019). Studies of irritability in the adult population are few, but suggest that irritability is more common than other affects (e.g., guilt) in depression and present in half of major depression cases of adult Americans (Balbuena, Bowen, Baetz, & Marwaha, 2016; Fava et al., 2010). What has been rarely noted in research is that irritability is common in adults with chronic diseases. Literature has shown that elevated irritability exists in adult patients with not only neurological disorders, such as Parkinson’s disease (Vescovelli, Sarti, & Ruini, 2019) and brain injury (Yang, Huang, Lin, Tsai, & Hua 2013), but also physical diseases such as Huntington’s disease (van Duijn et al., 2014), prostate cancer (Sharpley et al., 2018), and chronic obstructive pulmonary disease (Blinderman, Homel, Billings, Tennstedt, & Portenoy, 2009). Irritability is a risk factor for depression and predictor of poor quality of life in adult patients (Blinderman et al., 2009; Orri, Perret, Turecki, & Geoffroy, 2018), but has not received as much research attention as depression and anxiety.

A challenge to irritability research is that the conceptualization of irritability is not as clear as anxiety or depression. Irritability was viewed as a spontaneous bodily sensation and an intermediate between the body and higher cognition (i.e., sense or sensibility) in the 18th century (Giglioni, 2008). In the 1980s, Snaith and Taylor (1985) described irritability as susceptibility to bodily stimuli and a “feeling state characterized by reduced control over temper.” Later, Craig, Hietanen, Markova, and Berrios (2008) defined irritability as a mood state that “predisposes towards certain emotions (e.g., anger), certain cognitions (e.g., hostile appraisals), and certain actions (e.g., aggression).” This conceptualization of irritability as a mood state with a behavioral tendency is influential and accepted nowadays, but shifted away from a bodily connection. However, more recently, psychoneuroimmunology theory has stipulated interactions among psychological process, the central nervous system and immune function, and particularly an influence of immune system on brain activity through the hypothalamic-pituitary-adrenal axis and the sympathetic nervous system—the so-called “immune-brain loop” (Vedhara & Irwin, 2005). Evidence has shown an association between irritability and immunological inflammation (Blacklaws, Gardner, & Usher, 2011; Jyonouchi, Geng, & Davidow, 2014; Kraus, Schäfer, Faller, Csef, & Scheurlen, 2003; Meyer et al., 2010; Russo et al., 2005), suggesting that irritable mood is interrelated with immunological and nervous activities. A measure that includes a physical aspect—in addition to affective and behavioral aspects—of irritability is particularly important for people with immunological dysregulation. Such a comprehensive assessment of irritability is imperative for accurately evaluating irritability and developing effective interventions to reduce the risk of depression.

Evidence from our previous study (Zhang, Gary, & Zhu, 2015) supports the notion of multidimensionality of irritability. Previously, we conducted a mixed-method study to understand depressive symptoms of African-American cancer patients, because the literature has shown that depressed African Americans report predominantly physical symptoms but are emotionally guarded (Ghods et al., 2008); this can result in difficulty detecting depression in cancer patients due to a similarity of physical symptoms between cancer and depression. In that study, we interviewed and compared depressed and nondepressed African-American cancer patients, while including depressed White cancer patients as a comparison group. As reported elsewhere (Zhang et al., 2015), the study findings showed that the depressed African-American cancer patients reported irritability more frequently than nondepressed African-American cancer patients (p < .05), but reported sad feelings and fears less frequently than depressed White cancer patients, and physical discomfort more frequently than all the other patient subjects (p < .05). In the self-report, they associated irritable feelings with physical discomfort such as pain or lack of sleep. They did not differ in sociodemographic (gender, education, income, marital status, and employment), diagnostic, and treatment (surgery, chemotherapy, or radiation) variables from others, except that they were younger (p < .05) and sicker, with significantly more diseases, prescribed medications, and pain than the nondepressed African Americans (p < .01). These findings suggested a possibility that underlying physical condition affected the experience of depressed African-American cancer patients and their acknowledgement of “irritability” in self-report (Akinyemi et al., 2018; Zhang et al., 2015); hence, a physical dimension in irritability.

Measuring irritability in adults with chronic conditions is difficult without a valid and reliable instrument. There are only a few measures of irritability, including some published in recent years, but they do not assess irritability alone or only measure a unidimensional affect of irritability. For example, the 21-item Irritability Questionnaire (IRQ) (Craig et al., 2008), the 14-item Born-Steiner Irritability Scale developed for females (Born, Koren, Lin, & Steiner, 2008), and Snaith, Constantopoulos, Jardine, and McGuffin’s (1978) 18-item Irritability, Depression, Anxiety (IDA) Scale were developed for adult populations, but they assess a mixture of irritability, anger, and other emotions. The 7-item Affective Reactivity Index assesses irritability, but this measure is designed for children only (Stringaris et al., 2012). Two other measures, the Sheehan Irritability Scale (SIS) (Khan et al., 2016) and the Brief Irritability Test (BITe) (Holtzman, O’Connor, Barata, & Stewart, 2015) assess irritability in adults; however, they conceptualized irritability as a unidimensional affect. The SIS was developed from 61 patients from the United States and 525 patients from India with major depressive disorder. It measures irritability on seven mood items (irritability, frustration, edginess/impatience/overreaction, moodiness, anger with self, anger with others, and temper) without considering a physical aspect of irritability. The BITe was developed from 887 undergraduate students and 229 patients with chronic pain through an online survey, which tested 65 items, including 21 items from the IRQ and 14 items from the Born-Steiner Irritability Scale. Five items were identified and retained in BITe, but they assessed irritable mood (i.e., feeling irritable, grumpy, snappish, bothering, and nervous) alone. An undercurrent of irritability due to pain and its contribution to overall irritable mood were not considered nor measured in BITe. These limitations of existing measures, either lacking clarity or having a limited scope in assessing irritability, can lead to an inaccurate assessment of irritability and undercut the application of these measures to adults with chronic diseases.

It should be noted that irritability is associated with, but different from, anxiety (Eckhardt, Norlander, & Deffenbacher, 2004). Craig and colleagues viewed irritability as a more transitory mood state and anxiety as a more persistent state or being associated with a personality trait. Anger was considered an emotion directed toward a particular object, while irritability was viewed as a nonspecific mood state (Craig et al., 2008). Irritability is not the same as symptom distress, which reflects unpleasant feelings about diseases (McCorkle & Young, 1978). These constructs can be used to assess the discriminant validity of a new measure of irritability.

The need for a measure to include multidimensionality of irritability in the physically ill population is apparent and particularly urgent for people diagnosed with cancer. Cancer affects millions of Americans and has become a chronic condition as medical technology advances (American Cancer Society, 2019). Depression is common in cancer patients, but difficult to detect, partly because depression and cancer share similar physical symptoms (e.g., fatigue, loss of sleep or appetite) while depressive feelings (e.g., sadness) are underreported in cancer patients (Pasquini & Biondi, 2007). Evidence shows that depression predicts cancer death (Pinquart & Duberstein, 2010). An accurate assessment of irritability can help nurses and other clinicians to better identify at-risk cancer patients and provide needed care in a timely manner, so that these patients will have a reduced chance of developing depression and a better chance of survival. Therefore, we developed a new measure of the multidimensionality of irritability in cancer patients. We conducted a series of exploratory studies and tested a hypothesis that irritability exhibits in physical, affective, and behavioral dimensions among cancer patients.

Methods

The research project was conducted in a U.S. midwestern city and involved three pilot studies that contributed to the development of a new measure of irritability. The three studies included an initial study that generated a pool of statement items, a pilot test to verify wording of selected statement items and scale ratings, and a repeated measures test of the selected item-statements by cancer patients undergoing chemotherapy. All participants of these studies provided written consent following institutional review board approvals.

Study 1: Item Generation

The above mentioned mixed-method study (Zhang et al., 2015) was conducted between 2006 and 2009 to identify depressive symptoms among White versus African-American cancer patients. Seventy-four non-Hispanic cancer patients (34 depressed and 23 nondepressed African Americans, and 17 depressed Whites) diagnosed for early-stage breast or prostate cancer within the past three years and had completed cancer treatments at least six months prior were interviewed. Three professionals (two African Americans and one White) independently conducted the interviews. They were licensed psychiatric nurses or social workers with advanced degrees and matched with study participants by racial/ethnic status to ensure cultural sensitivity. The interviewers used an interview guide to solicit the recall of experience during cancer diagnosis and treatment. The interviews were audiotaped and transcribed. Two coders who were experienced in qualitative research conducted qualitative data analysis of the transcripts using Atlas 9.0 software. The details of the study methodology are published elsewhere (Zhang et al., 2015), but additional information on the item-generation process was not published and is presented in the following text:

The qualitative data analysis identified a number of depression-related feelings and generated codes according to words or phrases used, including fear, worry, anxiety, anger, irritability, sadness, a loss of interest or pleasure, suicidal, hopelessness, and others. Under the code of irritability, three themes emerged from narratives:

1. a physical feeling of irritability (e.g., “It is a deep pain that makes you unable to sleep and makes you irritable”), when physical suffering becomes unbearable;

2. an emotional feeling of irritability (e.g., “I am irritated by things that usually do not bother me”), reflecting awareness of irritable mood; and

3. a reactive feeling of irritability (e.g., “I would snap at people,” “I didn’t want to be bothered”), in which an impulse of emotional outburst and a tendency to social withdrawal for avoiding emotional triggers coexisted.

These narratives reflected a coherent experience of irritability in physical, affective, and behavioral channels, suggesting three manifesting aspects of the same underlying construct of irritability. Patients’ words and phrases that described irritability were reviewed, and a number of statements were identified as proxy item-statements.

Study 2: Verifying Verbatim and Rating Scale

Ninety-three item-statements that describe irritability and other depressive feelings were abstracted from interview transcripts and pooled together. A research team including the principal investigator and staff checked item wording and decided to assess the acceptability of three alternative scale ratings, including one 4-point Likert scale (true, somewhat true, somewhat false, and false) and two 5-point Likert scales (Scale 1: 1 = frequently and 5 = never. Scale 2: 1 = strongly agree and 5 = strongly disagree). A new questionnaire containing the 93 items and an instruction sheet was mailed to nine consenting cancer patients for testing the appropriateness of wording and rating scales. These cancer patients were randomly selected from depressed African-American participants in Study 1 and included five men and four women aged 49 to 68 years (M = 57.4). Based on their responses, the item-statements were modified verbatim and a 5-point Likert scale (never to frequently) was chosen.

Study 3: Testing an Initial Version of The Irritability Scale

Study 1 produced a finding that depressed African Americans reported irritability significantly more frequently than nondepressed African Americans (p < .05), but not significantly different from depressed Whites, perhaps due to similar mental experiences or a smaller sample of Whites. Given this observation, we decided to narrow the focus of the questionnaire assessment to irritability. Accordingly, the items relevant to irritability were retained with some modification according to participants’ responses, and 10 other items were rewritten to better mirror irritability experiences. A new measure, “The Irritability Scale-initial version” (TISi), with 35 item-statements containing physical, mood, and behavioral subdomains, was developed using a 5-point Likert scale (1 = never, 2 = rarely or 1-2 days, 3 = occasionally or 3-4 days, 4 = often or 5-6 days, and 5 = frequently or 7 days). The 10-item physical subscales included statements that our subjects used to attribute irritability to pain, fatigue, insomnia, and physical discomfort (e.g., “I’m cranky because I am in pain,” “I’m not sleeping, which makes me irritable”). The 11-item mood subscale included statements about feelings of irritability such as feeling irritated, sensitive, and grouchy. The 14-item behavioral subscale included statements about tendencies to outburst and social withdrawal (e.g., “I snap at people,” “I just want to be left alone”). All items asked about personal experience in the past seven days to minimize recall errors. Some items expressed essentially the same meaning but were stated in different words. We deliberately kept this redundancy for further testing of the words to best describe irritability in cancer patients.

We used a repeated measures design to test TISi at baseline prior to chemotherapy (T₁) and three months later during chemotherapy (T₂). The chemotherapy is considered an agent of immune suppression that incurs adverse physical response, hence more irritability (Franzen, Buysse, Rabinovitz, Pollock, & Lotrich, 2010). The repeated measures help determine whether the item-statements measure irritability consistently despite its fluctuation. The following measures were also administered at both assessments for evaluating TISi’s construct validity.

The Hamilton Rating Scale for Depression (HRSD) is an interviewer-administered rating scale of depressive symptoms and severity. It has well-established validity (r = .65-.90) and reliability (r = .65-.90; Cronbach’s α = .48–.92) (American Psychiatric Association [APA], 2000). Its 21-item scale, with a total score of 0 to 68, was used in this study.

Hospital Depression and Anxiety Scale (HADS) is commonly used to detect anxiety and depression in people with health problems. It has high internal consistency (Cronbach’s α = 0.90), reliability (r = .92) and validity (r > .70). It contains two subscales with seven items each to assess depression and anxiety, respectively (APA, 2000).

Irritability Questionnaire (IQR) has 21 items and assesses cognitive and emotional aspects and behavioral expression of irritability, with a total score of 0 to 63. It contains two rating scales of the frequency (how often) and the amount (how much) of irritability and was used to validate TISi. The scale has satisfactory psychometric properties (Cronbach’s α = .58–.90; reliability r = .77-.86) (Craig et al., 2008).

McCorkle and Young Symptom Distress Scale (SDS) assesses the level of distress for 11 cancer symptoms (nausea, appetite, insomnia, pain, fatigue, bowel pattern, concentration, appearance, outlook, breathing, and cough) on a 5-point Likert scale with satisfactory reliability (α = .82) (McCorkle & Young, 1978). This measure is used to validate TISi’s physical subscale.

The serum concentrations of hsCRP as an indicator of chronic inflammation and human brain-derived neurotrophic factor (BDNF) that is associated with chronic pain, along with IL-6 and TNF-α, were measured for validating TISi’s physical subscale. IL-6 and tumor necrosis factor (TNF)-α levels were chosen among other proinflammatory cytokines because they were found significantly more elevated in depressed subjects than in control subjects in a meta-analysis of the literature (Dowlati et al., 2010).

The UCLA Loneliness Scale assesses distributions of social disconnectedness and perceived isolation. It has high internal consistency (α = .89-.94), reliability (r = .73), and construct validity. Its 20 items have a total score of 20 to 80 and were used to validate TISi’s behavioral subscale (Russell, 1996).

The participants completed TISi one more time at home within two weeks of a TISi assessment and returned it by mail for a preliminary evaluation of test–retest reliability. The pilot test was conducted in 2016 using a sample of newly diagnosed early-stage (I, II, and III), nonmetastasized, adult breast cancer patients because they usually receive chemotherapy pre- or postsurgery. Patients were eligible if they were 21 years of age or older, received a nonmetastisized breast cancer diagnosis within 12 months, had been asked to take chemotherapy, and were experiencing distress. We used the hospital tumor registry to identify potentially eligible patients and then mailed the patients a letter to inform them about the study, followed up with a phone call. Trained research staff screened the patients on the telephone, using the Distress Thermometer (DT), a screening tool recommended by the National Comprehensive Cancer Network (NCCN). Evidence supports a cutoff of DT for clinically significant distress at “4” (Donovan, Grassi, McGinty, & Jacobsen, 2014) or a clinically elevated level of distress at “3” (Cutillo et al., 2017) in cancer patients. The patients scoring “2” or more on the DT were eligible for this study, because this score indicates the presence of distress and a potential of increasing irritability during chemotherapy, while excluding patients whose mental states are unlikely affected by cancer. Patients who received or were receiving hormonal therapy or had preexisting psychotic disorders (e.g., schizophrenia, bipolar disorder) were identified from medical charts and excluded. The eligible and consenting participants were invited for participation and interviewed in a private hospital room for 60 minutes at T₁ and T₂ during their hospital visits. A clinical nurse drew a 12-ml blood sample from participants at the beginning of each interview, and the sample was later stored and analyzed in a hospital lab using standard procedures. Demographic, medical, and cancer treatment information was obtained from the participants and verified through a medical chart review. The participants received $60.00 for completing the study.

Data analysis.

We analyzed the data to evaluate internal consistency, test–retest reliability, and construct validity of TISi. Descriptive statistics for demographic and medical treatment variables were computed. The standardized Cronbach’s α test was performed for the total and the subscales of TISi. Due to the nonnormal nature of the data, Spearman correlation coefficients were calculated to assess test–retest reliability of the total and the subscales of TISi. For comparison, the intraclass correlation coefficients (ICCs) were also computed in the test–retest analysis of the total and the subscales of TISi. We then used the Spearman and Pearson correlation analyses to correlate TISi with measures of other correlates (e.g., HRSD, HADS) or biomarkers to assess construct validity. All analyses were performed for both timepoints. A confirmatory factor analysis was performed on TISi at T₁ to explore underlying dimensions, using the oblimin extraction method allowing for three factors. SAS version 9.4, Stata 15.1, and Mplus 8.1 software were used to perform various segments of the statistical analysis. Criterion for statistical significance was maintained at a .05 level.

Results

The sample consisted of 20 White (41.7%) and 28 African American (58.3%) non-Hispanic early-stage breast cancer patients. The ages ranged from 27 to 76 years, with a mean of 55.2 years (median = 56.5). Approximately 37.5% of the sample had completed a bachelor’s or graduate degree. Nearly one half (47.9%) were married. Most (60.4%) continued working, and over 64% had annual incomes over $25,000. With respect to medical characteristics, the majority (60.4%) had a stage II cancer and all patients received chemotherapy for three months, with 62.5% receiving chemotherapy prior to surgery or radiotherapy and 37.5% receiving chemotherapy after surgery. Preexisting diagnoses of depression, anxiety, and panic disorders were reported by 13 patients (27%). The sample had an average of 3.4 diseases (median = 3) and a Charlson Comorbidity Index mean score of 0.52 (median = 0) (see Table 1).

Table 1.

Demographics and Medical Characteristics (n=48)

Variables	n		%
Race
White	20		41.7
African American	28		58.3
Education
High school or less	10		20.9
Some college/associate degree	20		41.6
Bachelor’s degree or higher	18		37.5
Marital status
Currently married	23		47.9
Single/Widowed/Separated/Divorced	25		52.1
Employed
Full time	17		35.4
Part time	12		25.0
Unemployed	19		39.6
Household annual gross income
< $25,000	12		25.0
$25,000 - $49,999	11		22.9
$50,000 - $100,000	14		29.2
$100,000+	6		12.5
Missing	5		10.4
Cancer stage
I	13		27.1
II	29		60.4
III	6		12.5
Received surgery	18		37.5
Chemotherapy
Neoadjuvant	30		62.5
Adjuvant	18		37.5
Pre-existing mental disorders
Panic disorder	1		2.1
Anxiety only	5		10.4
Depression only	3		6.3
Depression/Anxiety	4		8.3
	Mean	Median (min., max.)	Std. Dev.
Age	55.2	56.5 (27, 76)	12.7
Charlson Comorbidity Index	0.52	0.0 (0, 5.0)	0.97
Number of diseases	3.4	3.0 (0, 11)	3.0
Distress Thermometer baseline	4.8	5.0 (0, 10)	2.4

A confirmatory factor analysis (CFA) was conducted on TISi at T₁ using Mplus 8.1. The factor loadings generally supported the assignment of the items to the 3 subscales of TISi if a cut-point value of 0.4 is used to retain items (Table 2); however, there is no consensus regarding an absolute cut-point value.

Table 2.

Confirmatory Factor Analysis

	Oblimin Rotated Factor Loadings
	Time 1
The Irritability Scale (initial version) items:	1	2	3
Please indicate how you felt in the past week (7 days) with regard to the following statements:
Physical Dimension
1. My body hurts so bad that it makes me feel irritable	0.809	−0.099	0.031
2. I’m not sleeping, which makes me irritable	0.328	0.034	0.374
3. I’m in pain, which makes me irritable	0.723	0.150	0.062
4. I’m cranky because I am in pain	0.697	0.246	0.046
5. I feel so tired and drained in my body that I get irritated easily	0.438	0.450	−0.272
6. I’m in a lot of pain, which makes me crabby and mean	0.386	0.324	0.339
7. I feel so wore out that I can’t do anything anymore	0.492	0.252	0.174
8. Since I have been sick, I do not have patience	0.002	0.610	0.221
9. I am drained because of cancer	0.483	0.431	−0.034
10. I am easily agitated because my body hurts	0.664	0.281	0.207
Mood Dimension
11. I am getting bothered by things that usually do not bother me	−0.043	0.816	−0.093
12. Every little thing can get on my nerves	0.000	0.928	−0.114
13. I find myself getting irritated with everything	0.146	0.908	−0.112
14. I’m sensitive and touchy	0.268	0.649	−0.041
15. I am sometimes grouchy, jumpy, or mean	0.115	0.750	0.101
16. I don’t feel like myself	0.533	0.346	0.120
17. I easily get irritated	0.234	0.743	0.041
18. I am very irritable	0.215	0.738	0.048
19. I have mood swings	0.243	0.418	0.339
20. I am on edge	0.034	0.726	0.065
21. I feel anger inward at myself	0.216	0.474	−0.054
Behavioural Dimension
22. I am very short with people	−0.087	0.829	0.080
23. I become short and stoic in response to others	−0.060	0.765	0.167
24. I am not as nice or considerate with people as I used to be	0.066	0.761	0.116
25. I am short-tempered	−0.003	0.512	0.453
26. When people keep bothering me, I get snappish	0.056	0.421	0.452
27. I snap at people	−0.062	0.525	0.466
28. I isolate myself because I do not want to hurt anyone’s feelings	0.233	−0.251	0.889
29. I stay away from people because I want to be by myself	0.117	−0.071	0.948
30. I now tend to be more on my own	0.180	−0.181	0.977
31. Most times, I’d rather be alone	−0.054	0.083	0.922
32. I just want to be left alone	−0.179	0.351	0.794
33. I really do not want to talk to anybody	−0.239	0.408	0.762
34. I do not want to be bothered	−0.118	0.319	0.710
35. I feel alone and isolated	0.387	0.202	0.550

A 5-Likert scale (1-5) is used:
Never	Rarely	Occasionally	Often	Frequently
(0 day)	(1-2 days)	(3-4 days)	(5-6 days)	(7 days)

Cronbach’s α analysis indicated a strong internal consistency among 35 items. At T₁, α values were .97 overall, .91 on physical, .95 on mood, and .96 on behavioral subscale. At T₂, α values were .97 overall, .92 on physical, .93 on mood, and .94 on behavioral subscale. There was little difference in α values over time. Further evaluating correlations among three subscales, we performed a Spearman correlation test that revealed r = .76 at T₁ and .86 at T₂ for the physical and mood subscales, r = .74 at T₁ and .82 at T₂ for the mood and behavioral subscales, and r = .65 at T₁ and .70 at T₂ for the physical and behavioral subscales. All correlations had p-values < .001.

Spearman correlation analyses of test–retest reliability provided statistically significant (p < .001) correlation coefficients for TISi (r = .69) and its subscales (physical, r = .61; mood, r = .76; behavioral, r = .67). Intraclass correlation coefficients (ICCs) were high: .86 overall; physical, .81; mood, .80; and behavioral, .77. All ICC p-values were < .001.

Table 3 presents Spearman correlations of TISi or its subscales with the measures of other constructs. Correlations of TISi and subscales with HRSD, HADS, IRQ, SDS, and the UCLA Loneliness Scale were significant but moderate, and coefficients fell within a range of .40 to .70, except those for HADS Depression and HADS Anxiety (.25-.40).

Table 3.

Spearman Correlation for TISi and Psychosocial Correlates

	Time 1				Time 2
Correlates	Physical	Mood	Behavior	Total	Physical	Mood	Behavior	Total
HRSD	***0.63	***0.67	***0.51	***0.66	***0.65	***0.52	***0.46	***0.60
HADS Depression	***0.49	*0.29	**0.38	***0.45	*0.34	**0.38	*0.32	**0.40
HADS Anxiety	***0.52	***0.45	***0.47	***0.54	0.25	**0.40	**0.37	**0.38
IRQ Often	***0.48	***0.53	***0.60	***0.59	***0.52	***0.57	***0.55	***0.61
IRQ How Much	***0.50	***0.53	***0.60	***0.61	***0.61	***0.64	***0.64	***0.70
UCLA Loneliness	***0.56	**0.46	***0.55	***0.55	*0.32	**0.41	***0.54	***0.47
SDS	***0.48	**0.38	**0.43	***0.48	***0.50	***0.51	**0.38	***0.52

Note.

^*p < .05

^**p < .01

^***p < .001.

Pearson correlation—a common test of biomarker variables in the literature—was computed for biomarkers and TISi (see Table 4). Biomarkers significantly correlated with the TISi physical subscale at T₁ (hsCRP: r = .32, p = 0.025) and T₂ (TNF-α: r = .44, p = 0.002); with TISi mood subscale at T₁ (BDNF: r = .33, p = 0.024) and total TISi at T₂ (TNF-α: r = .31, p = 0.03). Biomarkers did not correlate with total TISi at T₁ or other subscales.

Table 4.

Pearson Correlation Between TISi and Biomarkers

	Time 1				Time 2
Correlates	Physical	Mood	Behavior	Total	Physical	Mood	Behavior	Total
IL-6	0.02	−0.60	−0.09	−0.06	0.19	0.07	0.04	0.10
TNF-α	0.14	0.06	0.16	0.14	**0.44	0.17	0.25	*0.31
hsCRP	*0.32	0.24	0.07	0.20	0.18	0.09	0.03	0.10
BDNF	0.18	*0.33	0.15	0.24	−0.07	−0.01	−0.08	−0.06

Note. Neither biomarkers had a significant correlation with any other study measures, except one between IL-6 and SDS at baseline (r = 0.40, p = 0.005)

^*p < .05

^**p < .01.

Discussion

The findings suggest that TISi has satisfactory internal consistency, test–retest reliability, and validity for assessing irritability in cancer patients. The high Cronbach’s α (.97; for all subscales > .90 at T₁ and T₂) suggests that item-statements of each subscale assess the same underlying construct and the three subscales are highly interrelated, assessing different aspects of the same underlying construct. The results of Spearman correlation analysis further support this inference, showing significant correlations among the three subscales. Correlations between mood and physical or behavioral subscales are stronger than that between physical and behavioral subscales, and this pattern held steady over time. These findings are important because they show that irritability manifests in a physical dimension that interrelates with affective and behavioral dimensions in cancer patients, supporting psychoneuroimmunology theory and its related evidence. While the relationships among the three dimensions have yet to be investigated, this emerging evidence urges a new conceptualization of irritability for developing better assessment and management strategies to benefit people with chronic diseases.

TISi has a moderate but satisfactory test–retest reliability; the ICC was higher. The three subscales yielded a similar test–retest reliability r to that of the total TISi. These findings suggest that TISi and its subscales are testing the underlying construct, i.e., irritability, with a good consistency.

Findings also suggest a satisfactory construct validity of TISi. TISi has a moderate correlation with IRQ. This may be explained by the fact that IRQ actually assesses a mixture of irritable and angry feelings rather than irritability alone, and it does not include a physical component of irritability. TISi also has a moderate correlation with HRSD and HADS Depression, and a low-to-moderate correlation with HADS Anxiety. Further, its physical subscale—which measures irritability due to physical symptoms—correlated moderately at time one and two with SDS, which measures general distress due to physical symptoms; and its behavioral subscale correlated moderately with the UCLA Loneliness Scale, which assesses loneliness and isolation rather than a tendency of social withdrawal. The moderate correlation was observed not only across multiple measures, but also across TISi’s three subscales on these measures, and such correlational relations remained largely unchanged over time. Therefore, TISi has shown an ability to differentiate multidimensional irritability of its measure from other related but different constructs, including depression, anxiety, symptom distress, and social disconnectedness. We were unable to find a well-validated measure of irritability at the time for assessing convergent validity. However, we observed a significant increase in TISi total score at T₂ (mean difference from T₁= −9.5, p = .005), suggesting that TISi was responding to a change of irritability following the administration of chemotherapy.

Notably, we found a significant correlation between TISi physical subscale and biomarkers; hsCRP is an indicator of low-grade, chronic inflammation, while TNF-α signals systemic inflammation and is part of acute-phase response to inflammation. The fact that both correlated with TISi physical subscale at T₁ or T₂, without significantly correlating with other study measures (e.g., HRSD, IRQ), supports our assumption that irritability has a physiological basis in immunosuppression. We noticed inconsistency in such correlations and insignificant correlations between TISi physical subscale and IL-6 or BDNF. This may be explained by these biomarkers’ unique roles in the immune-brain loop in this sample or by the small sample size. The observed correlation with biomarkers needs to be replicated. It is possible that chronic inflammation contributes to depression through mediation of irritable mood, but this speculation about inflammation, irritability, and depression will need to be tested in the future.

Regarding structural validity, a factor analysis provided preliminary supporting evidence. It showed three factor loadings that are largely in line with our conceptualization of the three domains of irritability. Three items (#22-24) in the behavioral subscale had better loading on the mood subscale, but the observed factor loadings on the 35 items may not be stable due to a small sample and are subject to change in a large sample. It should be noted that minimum sample size requirements for factor analysis found in the literature may range from 100 to 1000 subjects, or a multiple of 3 to 20 times the number of items, but the values are not based on empirical evidence. TISi appears to have good content validity considering its clear relevance to irritability in wording, breadth of symptoms covered, and a firm grounding of the item-statements in qualitative interviews. However, a high Cronbach’s α suggests item redundancy (Streiner, 2003). Since the study sample was too small to obtain a valid result from the factor analysis, our ability to identify and remove less relevant item-statements is limited. Using the alpha value for item deletion in this sample can result in a loss of criterion validity (Raykov, 2008). We will need to perform the item-response analysis in a larger sample to reduce item redundancy. Therefore, we recommend that the total score of TISi be used in the assessment of irritability for the time being.

We were unable to test criterion validity of TISi, and this is a major study limitation. Ideally, a comparison between those suffering from irritability and normal controls will strengthen the validity test. Because there is no gold standard for detecting irritability in adult cancer patients, the best way to classify participants by irritability would be through in-depth interviews. Future investigation will benefit from using comparison groups, repeated measures, personal interviews, and a large study sample for fully evaluating TISi’s validity, sensitivity to change over time, and predictive validity for depression or anxiety.

Another limitation concerns the study sample. Our study oversampled African Americans and caution is warranted for a potential effect of race on study results. The study sample contains female patients and one type of cancer (breast) only; hence, the findings may vary when male cancer patients and other types of cancers are taken into consideration. The study sample has a relatively high education status and household income. Whether these socioeconomic factors affected study findings remains to be seen. Moreover, we used TISi to assess chemotherapy-induced irritability. Irritability in cancer patients who do not undergo chemotherapy or those with other chronic diseases may vary. TISi’s ability to detect irritability across diverse populations has yet to be assessed. Nonetheless, we note that TISi’s items were generated from cancer survivors and showed stability before and after receiving chemotherapy. Therefore, TISi may be usable and deserves to be tested in patients with other diseases due to their resemblance of cancer as a chronic condition and TISi’s stable psychometric properties.

Conclusion

TISi in its current version is preliminary, but useful. Its total score can be used to inform the level of irritability in cancer patients. Further testing of TISi in a larger, more representative sample is necessary for improving its psychometric properties. The dimensions of irritability and their dynamic relations need to be theorized based on evidence, tested, and understood, so that we can use TISi to better detect irritability and develop appropriate interventions for improving mood and quality of life in cancer patients, and hopefully, in people with other health problems as well.