The validity and utility of the M.D. Anderson Symptom Inventory in patients with prostate cancer: evidence from the Symptom Outcomes and Practice Patterns (SOAPP) data from the Eastern Cooperative Oncology Group

Abstract

Background

The M. D. Anderson Symptom Inventory (MDASI) is a psychometrically validated patient-reported outcome measure that assesses the severity and impact of multiple symptoms related to cancer and its treatment and has the potential to guide treatment specific to prostate cancer patients. Although the original MDASI validation study encompassed various cancer types, the instrument’s psychometric properties have not been examined in a large, homogeneous sample of patients with prostate cancer.

Patients and Methods

This study involved secondary analysis of data from the nationwide, multicenter Eastern Cooperative Oncology Group (ECOG) Symptom Outcomes and Practice Patterns (SOAPP) study, which enrolled patients from 38 ECOG-affiliated institutions, including 6 academic centers and 32 community clinics. Data were used to establish the psychometric properties of the MDASI in a subsample of 320 patients with prostate cancer. The instrument was administered twice, approximately 1 month apart.

Results

The MDASI demonstrated good internal consistency and test-retest reliability (with Cronbach alphas of ≥ 0.84 and intraclass correlations of ≥ 0.76 for all subscales), strong ability to discriminate between clinically different patient groups (by performance status, tumor response, and disease stage), and high sensitivity in detecting symptom change (with respect to patient-reported quality of life between the baseline and 1-month follow-up visits).

Conclusion

The MDASI is a valid, reliable, and sensitive symptom-assessment instrument that can be utilized with confidence in descriptive and clinical studies of symptom status in patients with prostate cancer.

Introduction

The prostate is the most common cancer site in older men,¹ and although the prognosis for prostate cancer has steadily improved,² patients are faced with treatment-related toxicities that can substantially impede treatment adherence and success. Patients also experience disease and treatment-related sequelae that can have a negative effect on disease outcomes and quality of life.^{1, 3} In light of these observations, recent studies^{1, 4} and reviews^{2, 3, 5} have underscored the need to examine patient-reported outcomes in adult oncology practice because they represent the most direct measure of the patient’s experience with disease and treatment and serve as an essential guide to informed clinical decision-making.⁴ Optimal treatment outcomes for patients with prostate cancer thus depend upon rigorous and accurate symptom assessment.

A variety of symptom assessment instruments have been used both clinically and in studies of patients with prostate cancer to assess the subjective patient experience both during and after therapy. Some of the most frequently used prostate-specific questionnaires are the Functional Assessment of Cancer Therapy-Prostate (FACT-P)⁶; the European Organisation for Research and Treatment of Cancer prostate-cancer-specific quality of life (QOL) questionnaire (EORTC QLQ-PR25)⁷; the UCLA Prostate Cancer Index (UCLA-PCI)⁸; the Expanded Prostate Cancer Index Composite (EPIC-26),⁹ and the Prostate Cancer-Specific QOL Instrument (PROSQOLI).¹⁰ Although these and other similar instruments are of benefit in capturing some prostate-specific symptoms, most were developed within the conceptual framework of health-related quality of life, rather than of symptom assessment. Additionally, instruments that are specific to a particular disease or cancer type may not include symptoms that are widely applicable across different cancers. In 2011, the National Cancer Institute (NCI) Symptom Management and Quality of Life Steering Committee underscored the need to identify a standard set of core patient-reported symptoms to be assessed in NCI-sponsored clinical trials in patients with cancer.¹¹ In response to this call, recent studies of symptom prevalence and severity in patients receiving active cancer treatment demonstrated that a discrete set of symptoms is frequently common across cancer types, and may serve as the foundation for defining a “core” symptom set that is recommended for consideration across cancer clinical trials, particularly among patients with advanced or metastatic cancers.^3–5 Consequently, the ability of an instrument to capture these core critical symptoms⁴ and thus to facilitate a comparison of their prevalence and severity across multiple cancers is a significant asset in epidemiological studies and clinical trials that include patients with different cancer types.

The M. D. Anderson Symptom Inventory (MDASI) is an instrument with a core module that assesses the severity of common cancer-related and treatment-related symptoms and their interference with functioning across different cancers.¹² The MDASI is comprehensive yet brief, thus minimizing patient burden, has a readily understood numeric scale that can be adapted for telephone or digital forms of administration, and is thus highly suited to assess major symptoms of concern for oncology patients. Although the MDASI’s psychometric properties were previously established in a diverse sample of oncology outpatients,¹² they have not been evaluated in a large, homogeneous sample of patients with prostate cancer. Thus, the objective of the current study was to establish the reliability, validity, and sensitivity of the MDASI using data from the Eastern Cooperative Oncology Group (ECOG) Symptom Outcomes and Practice Patterns (SOAPP) study, a nationwide, multicenter project that included a substantial sample of patients with prostate cancer.

Patients and Methods

Study Participants

From March 3, 2006 to May 19, 2008, outpatients with breast, lung, colorectal or prostate cancer were enrolled from ECOG-affiliated institutions, including 6 academic centers and 32 community clinics. This report focuses on the patients who were enrolled with prostate cancer. The sample size for this analysis was a product of accrual of the overall study. Patients were enrolled at any point in the trajectory of their care. Patients receiving treatment at academic centers were enrolled from disease site-specific clinics; community oncology patients were enrolled from general oncology clinics. To minimize selection bias, a prespecified algorithm for patient recruitment was approved and utilized by the coordinating center.

Eligible patients were 18 years or older; recipients of care at an ECOG-affiliated institution; willing to complete the follow-up survey, and cognitively competent to complete study surveys. Written informed consent was provided by all participants and the study protocol was approved by the Institutional Review Board at each participating site.

Procedures

Patients were recruited when they checked in for their appointment. Basic clinical and demographic data, including cancer treatment history and current therapies, was obtained prior to the clinician visit. Patients completed the MDASI via paper-and-pencil administration, and global quality of life (QOL) items at this initial (baseline) visit and at a follow-up visit 28–35 days later.

Each patient was asked to read the instructions at the beginning of each questionnaire and to complete all items in terms of his or her experience during the prior 24 hours. Reasons for incomplete forms were documented on an Assessment Compliance Form. Patients who could not complete the follow-up questionnaire because of acute illness were given the option to mail the forms to the treating clinic by day 42.

Survey Measures

M. D. Anderson Symptom Inventory

Symptom data was obtained via the MDASI, which contains 13 core symptom severity items and 6 interference items. Symptoms (pain, fatigue/tiredness, nausea, disturbed sleep, distress, shortness of breath, difficulty remembering, lack of appetite, drowsiness, dry mouth, sadness, vomiting, and numbness/tingling) were rated at their worst in the previous 24 hours on a 0–10 scale, with 0 representing “not present” and 10 representing “as bad as you can imagine.” On the basis of recent literature reviews, 6 additional symptoms (diarrhea, constipation, mouth sores, skin rash, hair loss, and coughing) deemed to be potentially important were included.^{4, 5} Patients also rated the degree to which symptoms interfered with various aspects of life during the past 24 hours. Each interference item (general activity, mood, work [including both work outside the home and housework], relations with other people, walking ability, and enjoyment of life) was rated on a 0–10 scale, with 0 representing “did not interfere” and 10 representing “interfered completely.”¹²

Scoring the MDASI

MDASI ratings can be averaged into subscale scores: mean core (13 core symptom items only), mean core plus any additional symptom items, and mean interference (6 interference items only). A more-sensitive characterization of symptoms may use a subset of the most-severe symptoms reported by that group. The interference items can further be separated into mean activity-related interference (WAW: work, general activity, and walking ability) and mean mood-related interference (REM: relations with people, enjoyment of life, and mood).¹² Our previous studies have shown that a rating of 5 or greater (on a 0–10 numeric rating scale) for pain and for fatigue indicates a moderate to severe symptom that significantly impairs daily functioning.¹³ This cut point was used to determine moderate to severe symptom levels.

Global Quality of Life

A global QOL item asked patients to rate their QOL as excellent, good, fair, poor, or very poor. At follow up, patients were also asked how they would characterize the change in their global QOL since their last visit: much better, better, the same, worse, or much worse. Using global QOL as an anchor, we described the smallest symptom score difference that represents the minimal clinically significant difference or minimally important difference [MID]¹⁴ of concern to patients.

Statistical Analysis

For this study, we analyzed data from the subset of patients with prostate cancer. Analyses were conducted using STATA statistical software.¹⁵ Descriptive statistics were computed for symptoms and subscales. Statistical significance was set using a 2-tailed alpha level of .05.

Reliability

Internal consistency reliability of baseline and follow up values was estimated by computing Cronbach alpha values for 4 MDASI subscales: the core symptom subscale (13 MDASI symptom items), the interference subscale (6 interference items), and the WAW and REM interference subscales. The criterion for good internal-consistency reliability requires a Cronbach alpha value of 0.70 or higher.¹⁶ To evaluate test-retest reliability, we calculated the intraclass correlations of the 4 subscales between the baseline and follow-up MDASI administrations, using 2-way analysis of variance.

Construct Validity and Known-Group Validity

Construct validity of the MDASI was assessed by performing factor analysis of baseline data to determine the underlying constructs that the 13 core MDASI items measure. Known-group validity was assessed by evaluating the MDASI’s ability to distinguish between groups known to be clinically different with respect to ECOG performance status, tumor response, and cancer stage. To test the statistical significance of the difference between mean symptom and interference scores for comparison groups (e.g. patients with poor vs. good ECOG performance status), we conducted two independent sample t-tests. Differences in mean symptom scores between groups were considered clinically meaningful if ≥ 0.5 standard deviation (SD).^{17, 18}

Criterion Validity

To evaluate the MDASI’s criterion validity, we used area under the receiver operating characteristics (ROC) curve to demonstrate the relationship between baseline MDASI subscale scores and QOL ratings (categorized as excellent or good versus fair, poor, or very poor). An ROC area of 1.0 represents a perfect test; that of 0.5 represents a worthless test. A rough guide for classifying the accuracy of a diagnostic test is the traditional academic point system: .90 to 1.00 = excellent (A), .80 to .90 = good (B), .70 to .80 = fair (C), .60 to .70 = poor (D), less than .60 = fail (F).¹⁹

Sensitivity

At follow up, patients indicated whether their global QOL had changed since the baseline visit. Responses were assigned a variable: much better = 1, better = 2, nearly the same = 3, worse = 4 or much worse = 5. This variable was used to evaluate the sensitivity of the MDASI. We first conducted 1-sample T tests to examine whether the mean change in the subscale scores between baseline and follow-up was significantly different from 0 in each of the QOL change groups. Next, we conducted 2 independent sample t-tests to test the significance of the difference in mean changes of subscale scores between comparison groups (eg, patients who rated their overall QOL as much better/better versus those who rated it as nearly the same). Change scores for the MDASI subscales and individual items were considered clinically meaningful at 0.5 SD or higher.¹⁷ We developed bar graphs to illustrate the relationship between changes in symptom severity and symptom interference and change in patient-reported QOL.

Results

Patient Demographics and Clinical Characteristics

Of the 3123 patients recruited for the SOAPP study, 320 had prostate cancer. Demographic and clinical characteristics of these patients are summarized in Table 1. The sample was predominantly white, non-Hispanic men (83%) with a median age of 71.9 years. Most patients had good ECOG performance status; 21% had no evidence of disease, and 14% exhibited complete response to therapy. A relatively large percentage of patients (44%) had metastatic disease only; 10% had both locoregional and metastatic disease, and 23% had progressive disease. Approximately 57% had previously undergone chemotherapy, immunotherapy, or hormonal therapy, and 52% had previous radiation therapy.

Table 1.

Patient Demographic and Disease Characteristics at Baseline (N = 320)

Variables	n	%
Age
Median	71.9
Range	38.6–93.1
Sex
Male	320	100.0
Race
Unknown	8	2.5
White	265	82.8
Black	43	13.4
Asian	1	0.3
Native American	1	0.3
Multiracial	2	0.6
Ethnicity
Unknown	38	11.9
Hispanic or Latino	34	10.6
Non-Hispanic	248	77.5
ECOG Performance Status
Unknown	2	0.6
0	161	50.3
1	127	39.7
2	22	6.9
3	7	2.2
4	1	0.3
Current Status of Disease
Unknown	1	0.3
Complete response	45	14.1
Partial response	26	8.1
Stable disease	174	54.4
Progressive disease	74	23.1
Current Stage of Disease
Unknown	1	0.3
No evidence of disease	68	21.3
Local/regional	77	24.1
Metastatic	142	44.4
Local/regional and metastatic	32	10.0
Weight Loss in Previous 6 Months
Unknown	4	1.3
< 5%	286	89.4
5% to < 10%	22	6.9
10% to < 20%	6	1.9
≥ 20%	2	0.6
Total Number of Metastatic Sites
0	144	45.0
1	129	40.3
2	32	10.0
3	9	2.8
4 or more	6	1.8
Prior Chemo/Immuno/Hormonal Therapy	181	56.6
Prior Radiation Therapy	167	52.2
Currently Receiving Cancer Treatment	220	68.8
Type of Current Therapy
No current therapy	100	31.3
Unknown	1	0.3
Adjuvant therapy	42	13.1
Neoadjuvant therapy	17	5.3
Therapy for recurrent/non-metastatic disease	15	4.7
Therapy for metastatic disease	145	45.3
Current Treatment for Cancer
Chemotherapy	75	23.4
Immunotherapy	4	1.2
Hormonal therapy	171	53.4
Radiation therapy	22	6.9

Abbreviation: ECOG = Eastern Cooperative Oncology Group.

Symptom Severity at Initial and Follow-up Assessments

At baseline, mean scores for the core and interference subscales were 1.60 and 1.95, respectively (Table 2). The percentages of patients who rated each symptom as moderate to severe at both assessment time points are also presented. In order of severity, the most-severe symptoms reported by patients were fatigue, disturbed sleep, drowsiness, pain, dry mouth, and difficulty remembering. Nausea and vomiting were the least-severe symptoms reported. Moderate to severe levels¹³ of fatigue and pain were reported by nearly 36% and 18% of patients, respectively.

Table 2.

Prevalence, Severity and Percent Missing of Individual MDASI Symptoms and Subscales at Baseline and Follow-Up (N = 320)

Baseline Assessment
Symptom	Mean	SD	% Moderate to Severea	% No Symptoma	% Missing/Unknownb
MDASI Core Symptoms
Fatigue	3.14	2.89	35.5	27.4	4.1
Disturbed sleep	2.38	2.80	25.6	41.3	0.9
Drowsiness	2.14	2.45	18.6	38.8	0.9
Pain	1.80	2.71	17.5	55.9	1.6
Dry mouth	1.69	2.45	15.9	54.7	3.4
Difficulty remembering	1.68	2.22	13.6	44.8	0.9
Shortness of breath	1.65	2.49	16.1	56.5	0.9
Numbness/tingling	1.62	2.43	15.8	58.0	0.9
Distress	1.50	2.33	13.5	55.0	0.6
Sadness	1.42	2.40	14.5	61.3	0.6
Lack of appetite	1.14	2.28	11.1	69.3	1.3
Nausea	0.56	1.55	4.8	82.0	1.3
Vomiting	0.16	0.92	1.6	95.0	0.6
Additional Symptoms
Constipation	1.28	2.39	12.4	65.4	1.6
Coughing	1.04	1.98	8.5	65.1	0.6
Hair loss	1.02	2.47	10.7	77.7	0.6
Diarrhea	0.98	1.98	8.8	70.4	0.9
Rash/pruritus	0.57	1.71	4.7	82.7	0.6
Sore Mouth	0.29	1.29	3.2	92.4	1.3
MDASI Subscale	Mean	SD	LCL	UCL	% with Missing Itemsb
Core	1.60	1.52	1.43	1.77	8.5
Interference	1.95	2.38	1.69	2.21	1.6
WAW	2.30	2.74	2.00	2.60	0.6
REM	1.60	2.25	1.35	1.85	1.3
Follow-Up Assessment
Symptom	Mean	SD	% Moderate to Severea	% No Symptoma	% Missing/Unknownb
MDASI Core Symptoms
Fatigue	3.06	2.66	31.1	24.8	10.6
Disturbed sleep	2.36	2.66	21.7	34.6	10.6
Drowsiness	2.24	2.53	22.0	36.9	10.3
Distress	1.78	2.51	18.3	50.2	9.7
Difficulty remembering	1.75	2.08	11.8	36.6	10.3
Pain	1.74	2.52	17.0	51.6	9.7
Numbness/tingling	1.72	2.53	16.7	51.0	10.0
Shortness of breath	1.65	2.42	14.9	52.9	9.7
Dry mouth	1.65	2.39	15.3	52.1	10.0
Sadness	1.55	2.42	13.9	56.4	10.3
Lack of appetite	1.11	2.15	9.4	66.4	10.6
Nausea	0.43	1.24	2.4	81.5	10.6
Vomiting	0.16	0.87	1.4	92.7	10.6
Additional Symptoms
Constipation	1.34	2.28	10.4	60.6	10.3
Coughing	1.06	1.77	6.6	61.7	10.3
Diarrhea	1.03	1.86	6.9	65.0	9.7
Hair loss	1.00	2.33	10.1	75.3	10.3
Rash/pruritus	0.61	1.68	4.2	80.5	10.3
Sore mouth	0.34	1.16	2.4	88.2	9.7
MDASI Subscale	Mean	SD	LCL	UCL	% with Missing Itemsb
Core	1.64	1.49	1.47	1.81	6.9
Interference	2.07	2.38	1.80	2.34	1.7
WAW	2.43	2.79	2.11	2.75	0.3
REM	1.71	2.24	1.45	1.97	1.4

Abbreviations: MDASI = M. D. Anderson Symptom Inventory; SD = standard deviation; LCL = 95% lower confidence limit; UCL = 95% upper confidence limit; WAW = activity-related interference items (work, general activity, and walking ability); REM = mood-related interference items (relations with people, enjoyment of life, and mood).

^aPercentage of patients who answered the question

^bPercentage missing from all 320 patients

At follow up, mean scores for the core and interference subscales were 1.64 and 2.07, respectively. Fatigue, disturbed sleep, drowsiness, being distressed, difficulty remembering, and pain registered as the most-severe symptoms. About 31% and 17% of patients had moderate to severe fatigue and pain, respectively, at follow up.

Psychometric Validation of the MDASI

Internal Consistency Reliability

The MDASI subscales showed good internal consistency reliability. Baseline Cronbach coefficient alpha values were 0.88 for the core subscale (13 core symptom items) and 0.92 for the interference subscale (all 6 interference items). At follow up, Cronbach coefficient alpha values were 0.89 and 0.93 for the core and interference subscales, respectively.

Test-Retest Reliability

The intraclass correlations of the MDASI core and interference subscales administered approximately 1 month apart were indicative of good test-retest reliability. Values for each of the subscales (core, interference, WAW, and REM) were ≥ 0.79.

Construct Validity and Known-Group Validity

Results from factor analysis indicated that factor loadings of the core MDASI items were distributed across 2 factors. Nausea, vomiting, and lack of appetite appeared to load on the same underlying construct (gastrointestinal symptoms), and the remaining items loaded onto the separate underlying construct of general symptoms. These results are consistent with those obtained for the original MDASI validation in a heterogeneous sample of cancer patients.¹²

To evaluate the MDASI’s ability to distinguish between groups known to be clinically different (known-group validity), we compared MDASI symptom and interference subscale scores with ECOG performance status scores, tumor response, and current disease stage. Patients with poor ECOG performance status reported significantly higher scores for both MDASI subscales (Table 3). Patients with complete response reported significantly lower MDASI subscale scores than did patients with disease progression. Patients with no evidence of disease reported significantly lower MDASI subscale scores than did patients with either locoregional/metastatic or metastatic cancer. Effect sizes between these groupings for the MDASI subscales were clinically significant (0.5 or higher).

Table 3.

Known-Group Validity of the MDASI: Comparison of Baseline Symptom and Interference Subscale Scores by ECOG Performance Status, Tumor Response, and Disease Stage

MDASI Subscale: Mean Score for 13 Core Symptoms
Variable		Patients (n)	Mean	SD	Diff	Effect Sizea
ECOG Performance Status	Good (0)	159	1.02	1.10
	Poor (≥1)	157	2.16	1.59	1.14	0.83
Tumor Response	Complete responseb	44	1.20	1.20
	Partial response	26	1.24	1.07	0.04	0.03
	Stable disease	173	1.43	1.39	0.23	0.17
	Progressive disease	74	2.39	1.83	1.19	0.73
Current Stage of Disease	NEDc	67	1.22	1.31
	Locoregional	76	0.91	0.88	−0.31	−0.28
	Metastatic	142	2.05	1.70	0.83	0.52
	Locoregional/metastatic	32	2.02	1.52	0.80	0.58
MDASI Subscale: Mean Score for 6 Interference Items
Variable		Patients (n)	Mean	SD	Diff	Effect Sizea
ECOG Performance Status	Good (0)	159	1.01	1.64
	Poor (≥ 1)	157	2.87	2.60	1.86	0.86
Tumor Response	Complete responseb	44	1.42	2.13
	Partial response	26	1.24	1.56	−0.18	−0.09
	Stable disease	173	1.79	2.24	0.37	0.17
	Progressive disease	74	2.92	2.80	1.50	0.58
Current Stage of Disease	NEDc	67	1.38	2.00
	Locoregional	76	1.08	1.70	−0.30	−0.16
	Metastatic	142	2.47	2.59	1.09	0.45
	Locoregional/metastatic	32	2.84	2.62	1.46	0.66

Abbreviations: MDASI = M. D. Anderson Symptom Inventory; ECOG = Eastern Cooperative Oncology Group; SD = standard deviation; NED = no evidence of disease.

^aEffect size = Diff/ (Pooled SD), Pooled SD=sqrt [((n1−1)*Var1+(n2−1)*Var2)/(n1+n2−2)]. Var = SD^2

^bComplete response = reference group for tumor response

^cNED = reference group for disease stage

Comparing good performance status = 0 vs. poor performance status ≥ 1, P < .001 for both core and interference subscales; comparing complete response vs. progressive disease, P < .001 for core subscale, and P = .003 for interference subscale; comparing NED vs. advanced disease (metastatic or locoregional/metastatic), P < .001 for core subscale, P = .001 for interference subscale, using 2 independent-sample t tests

Criterion Validity

Table 4 shows the area under the ROC curve for MDASI subscales. All MDASI subscales demonstrated good accuracy, with values of approximately 0.77 when compared with dichotomized QOL (excellent or good versus fair, poor, or very poor).

Table 4.

Criterion Validity of the MDASI: Area Under the ROC Curve for MDASI Subscales Compared with QOL at Baseline (N = 318)

MDASI Subscale	Area Under the ROC Curve (95%CI)a
Mean core symptom items	0.769 (0.716, 0.822)
Mean interference items	0.797 (0.745, 0.848)
Mean WAW	0.781 (0.729, 0.834)
Mean REM	0.792 (0.740, 0.844)

Abbreviations: MDASI = M.D. Anderson Symptom Inventory; ROC = receiver operating characteristics; QOL = quality of life; WAW = activity-related interference items (work, general activity, and walking ability); REM = mood-related interference items (relations with people, enjoyment of life, and mood), CI=confidence interval.

^aArea under the ROC curve was calculated using a logistic regression model with QOL as the dependent variable, dichotomized as excellent/good vs. fair/poor/very poor.

Sensitivity

Figure 1A shows that when patient-reported QOL remained nearly the same from baseline to follow up, scores on MDASI symptom items changed only slightly during the same period. Conversely, when patients reported worse or much-worse change in QOL, noticeably large changes occurred in scores of the MDASI symptom items. Figure 1B demonstrates that an analogous pattern was observed for the MDASI symptom interference items.

Figure 1. Sensitivity of the MDASI.

(A) Change in Severity for 19 Symptoms (13 MDASI Core plus 6 Additional Symptoms) by Patient-Reported Change in QOL*

(B) Change in Severity for 6 MDASI Interference Items by Patient-Reported Change in QOL*

*Of 320 patients, 31 had missing values for all items at follow up

Abbreviations: MDASI = M. D. Anderson Symptom Inventory; QOL = Quality of Life

Table 5 shows similar observations for the MDASI subscales. Change scores for MDASI subscales between baseline and follow-up indicate that patients with worse or much-worse QOL experienced significant increases in symptom severity and interference. Conversely, significant decreases were observed for patients with better or much-better QOL at follow-up, whereas changes in the MDASI subscales for patients whose QOL remained nearly the same were not always significant.

Table 5.

Sensitivity of the MDASI: Change in MDASI Subscales Compared with Changes in QOL, from Baseline to Follow-Up^*

	Core		REM		WAW
Patient’s Overall QOL Change During Follow-Upa	Mean Changeb	SD	Mean Changeb	SD	Mean Changeb	SD
Much Better/Better (n = 83)	−0.24	1.10	−0.49	1.81	−0.58	2.27
Nearly the Same (n = 183)	0.14	1.00	0.24	1.68	0.32	2.14
Worse/Much Worse (n = 17)	1.06	1.65	1.90	2.68	1.94	2.38

^*Of 320 patients, 31 had missing values for all items at follow up (13 MDASI core plus 6 additional symptoms, and 6 interference items)

Abbreviations: MDASI = M. D. Anderson Symptom Inventory; QOL = quality of life; WAW = activity-related interference items (work, general activity, and walking ability); REM = mood-related interference items (relations with people, enjoyment of life, and mood); SD = standard deviation.

^aChange in overall QOL at follow-up was assessed in this format: Compared to your previous visit, would you say your QOL is: 1 = much better, 2 = better,3 = nearly the same, 4 = worse, 5 = much worse.

^bChange in MDASI subscale score = MDASI subscale score at follow up – MDASI subscale score at baseline.

P < .05 for all pairwise comparisons of mean change in subscale scores between QOL change groups (better/much better, worse/much worse) with QOL nearly the same group (reference group), using 2 independent-sample t tests. Comparing the mean change in subscale scores with 0 using 1-sample t test: P < .05 for all three subscales for much better/better and worse/much worse groups. For nearly the same group, P < .05 for WAW score change, and P > .05 for core and REM score change.

For patients whose overall QOL worsened, the mean increases in MDASI subscale scores were 1.06 for the core symptom subscale, 1.94 for the mean activity-related interference subscale, and 1.90 for the mean mood-related interference subscale (Table 5). Similarly, for patients whose overall QOL improved, reductions in scores were 0.24, 0.58, and 0.49 points, respectively, for the mean core symptom, activity-related, and mood-related subscales. For the 5 most severe symptoms and for patients whose overall QOL worsened, fatigue showed the largest increase at 2.15, followed by distress at 1.63, disturbed sleep at 1.02, drowsiness at 1.12, and hair loss at 0.76.

Discussion

Although the MDASI’s psychometric properties have previously been established in a diverse sample of oncology outpatients, the objective of the current study was to evaluate its psychometric properties in a large, national, multicenter cohort of patients with prostate cancer. An important feature of our study was that the MDASI was administered to a sufficient sample of prostate cancer patients at 2 time points—at baseline and at their follow-up visit. This allowed us to examine its sensitivity to change in patients’ perception of QOL, as well as to evaluate the stability of or change in patients’ symptom reporting between the 2 visits. Our results confirmed the relative invariance in the order of the most-severe patient-reported symptoms across the 2 assessments.

The MDASI’s core symptom and interference subscales demonstrated acceptable internal consistency reliability; high test-retest reliability, and strong ability to discriminate between clinically different patient groups (by ECOG performance status, tumor response to therapy, and disease stage). For example, patients with poor ECOG performance status reported clinically significant higher scores for the MDASI symptom and interference subscales; those with complete response reported significantly lower MDASI subscale scores than did those with disease progression, and those with no evidence of disease reported significantly lower MDASI subscale scores than did those with both locoregional/metastatic and metastatic cancer. Our study reports the effect sizes between these groupings for the MDASI subscales, almost all of which were clinically significant (0.5 or higher). Additionally, our results indicated that the MDASI’s subscales were sensitive to changes in patient-reported QOL, as demonstrated by statistically and clinically significant change in symptom severity and interference from baseline to follow-up.

With the exception of 2 symptom items (i.e. constipation and diarrhea), the MDASI core module contains the symptoms that have been recommended for consideration in studies of advanced or metastatic cancers.⁴ Thus, one of its main advantages over prostate cancer-specific scales is that it not only captures key symptoms that have a meaningful impact on the patient experience, but also allows for a comparative evaluation of symptom prevalence and severity across cancer types. This is of significant benefit in clinical trials and epidemiological studies that include patients with different cancers. An additional, critical asset of the MDASI is that, unlike several prostate cancer-specific assessment tools, it was developed specifically with the objective of capturing patients’ symptom burden rather than health-related quality of life.

Several context-specific modules of the MDASI, such as the MDASI-Brain Tumor,²⁰ the MDASI-Lung Cancer,²¹ the MDASI-Multiple Myeloma,²² and others have previously been developed and validated; however, this study was not a module validation. The 6 additional items included and explored in our study were based on recommendations from the literature, rather than from qualitative patient interviews as suggested in the Food and Drug Administration (FDA) guidelines for developing patient-reported outcomes. These items did not show high prevalence or severity in our patient sample. Future studies in patients with prostate cancer might include cognitive debriefing and qualitative interviews to further explore these items or identify other disease-specific items for inclusion in a MDASI prostate cancer module. While the core MDASI is quite sensitive to common symptoms in patients with cancer, its use in clinical trials would greatly benefit from the addition of a set of prostate-specific symptom questions developed and psychometrically assessed in this manner to form a new MDASI prostate-specific module.

Conclusion

Recent commendations for incorporating patient-reported outcomes in adult oncology emphasize the measurement of multiple symptoms that are common across different cancers in order to improve quality of patient care,^{3, 4, 23} obtain better insight into the patients’ subjective experience of both disease and treatment,⁴ and facilitate informed decision making.⁴ As an instrument that assesses the severity of multiple, common cancer-related and treatment-related symptoms and their interference with functioning across cancer types, the MDASI adequately meets this recommendation. In this nationwide, multicenter study, results from the psychometric analysis of the MDASI administered twice to a sample of 320 prostate cancer patients establish that the MDASI has excellent reliability, is valid, and has high sensitivity to detect change in prostate patients’ quality of life. As symptoms are prevalent and can be severe in prostate cancer patients, a measure shown to perform well in this patient population is likely to be useful to investigators. Additionally, due to its brevity, simple 0–10 assessment scale (which is favored by the FDA, as evidenced by the recent approval of ruxolitinib using the Myelofibrosis Symptom Assessment Form²⁴), and ease of administration, the MDASI is a practical tool for use in the outpatient setting, and it makes the longitudinal assessment of patients’ symptoms across a treatment trajectory more feasible.

Our results thus support the use of the MDASI in patients with prostate cancer, and extend the generalizability of its psychometric properties beyond its original validation in a heterogeneous sample of cancer patients at a single tertiary cancer center.

Clinical Practice Points.

• The M.D. Anderson Symptom Inventory (MDASI) is a psychometrically validated patient-reported outcome measure that assesses the severity and impact of multiple symptoms related to cancer and its treatment. The MDASI’s core module can help oncologists obtain insight into prostate cancer patients’ major symptoms of concern, as well as their interference with daily functioning.

• This study established the psychometric properties and utility of the MDASI in a large multicenter study of outpatients with prostate cancer. Psychometric analysis of the MDASI administered twice to a sample of 320 prostate cancer patients from the Eastern Cooperative Oncology Group demonstrated that the MDASI has excellent reliability, is valid, and has high sensitivity to detect changes in prostate patients’ quality of life.

• Due to its brevity, and simple 0–10 assessment scale, the MDASI is feasible to administer for longitudinal assessment of prostate cancer patients’ symptoms in the outpatient setting. The MDASI can be used with confidence in descriptive and clinical studies of patients with prostate cancer and is general enough to allow for comparisons of the care of prostate cancer patients to other cancer patient groups.