20 Journal of Pain and Symptom Management Vol. 35 No. 1 January 2008

Original Article

Use of a Single-Item Screening Tool to DetectClinically Significant Fatigue, Pain, Distress,and Anorexia in Ambulatory Cancer PracticeZeeshan Butt, PhD, Lynne I. Wagner, PhD, Jennifer L. Beaumont, MS,Judith A. Paice, PhD, RN, Amy H. Peterman, PhD, Dan Shevrin, MD,Jamie H. Von Roenn, MD, George Carro, RPh, MS, Joshua L. Straus, MD,J. Cameron Muir, MD, and David Cella, PhDCenter on Outcomes, Research and Education (CORE) (Z.B., L.I.W., J.L.B., D.C.) and Kellogg

Cancer Care Center (G.C.), Evanston Northwestern Healthcare, Evanston, Illinois, and Department of

Psychiatry and Behavioral Sciences (Z.B., L.I.W., J.L.S., D.C.), Institute for Healthcare Studies (L.I.W.,

D.C.), and Division of Hematology/Oncology (J.A.P., D.S., J.H.V.R.), Northwestern University, Feinberg

School of Medicine, Chicago, Illinois; Department of Psychology (A.H.P.), University of North Carolina,

Charlotte, North Carolina; and Capital Hospice (J.C.M.), Falls Church, Virginia, USA

AbstractFatigue, pain, distress, and anorexia are four commonly encountered symptoms in cancer. Toevaluate the usefulness of a single-item screening for these symptoms, 597 ambulatoryoutpatients with solid tumors were administered a self-report screening instrument within thefirst 12 weeks of chemotherapy. Patients rated the severity of each symptom on a 0e10 scale, atits worst over the past three days, with higher ratings associated with higher symptom levels.From this sample, 148 patients also completed a more comprehensive assessment of thesesymptoms. Two criteria were used to determine optimal cut-off scores on the screening items:1) the sensitivity and specificity of each screening item to predict clinical cases usingreceiver-operating characteristics analysis and 2) the proportion of patients at each screeningscore who reported that some relief of the target symptom would significantly improve their life.Optimal cut-off scores ranged from 4 to 6 depending on the target symptom (area under the curverange¼ 0.68e0.88). Use of single-item screening instruments for fatigue, pain, distress, andanorexia may assist routine clinical assessment in ambulatory oncology practice. In turn, suchassessments may improve identification of those at risk of morbidity and decreased quality of lifedue to excess symptom burden. J Pain Symptom Manage 2008;35:20e30. � 2008 U.S.Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved.

Key WordsScreening, assessment, clinical significance, patient-reported outcomes

Portions of these findings were presented at annualmeetingsof the AmericanSociety of ClinicalOncology.

Funding was provided by Ortho Biotech ClinicalAffairs, LLC.

Address correspondence to: Zeeshan Butt, PhD, Centeron Outcomes, Research, and Education (CORE),

� 2008 U.S. Cancer Pain Relief CommitteePublished by Elsevier Inc. All rights reserved.

Evanston Northwestern Healthcare, 1001 UniversityPlace, Suite 100, Evanston, IL 60201, USA. E-mail:z-butt@northwestern.edu

Accepted for publication: February 28, 2007.

0885-3924/08/$esee front matterdoi:10.1016/j.jpainsymman.2007.02.040

Vol. 35 No. 1 January 2008 21Single-Item Screening

IntroductionDue to recent medical advances, the life

expectancy of people with cancer has beenincreasing. However, cancer treatments areoften associated with increased symptom bur-den. Given this trend, treatment interventionsthat effectively manage cancer-related symp-toms and improve the quality of life of cancersurvivors are needed. Among the most fre-quently reported symptoms that people withcancer experience due to their illness and asso-ciated treatments are fatigue and pain.1e4

Emotional distress and anorexia/cachexia, al-though less frequently reported, are neverthe-less associated with significant morbidity.5e7

Each of these cancer-related symptoms hasalso been associated with functional impair-ment and reduced quality of life.2,8,9 It hasbeen suggested that these symptoms are under-recognized and consequently are under-treated.4,10 If these symptoms could beeffectively identified and managed, the overallsymptom burden of cancer would be dramati-cally reduced, and quality of life improved.11

FatigueCancer-related fatigue is the most prevalent

symptom associated with the disease.2,12 Fa-tigue affects patients’ performance in multipleareas, such as physical, emotional, and cogni-tive function.3 Depending on the patient sam-ple and methodology used, the prevalence offatigue in cancer patients has been estimatedto be between 60% and 90%.8,13 In advancedcancer and among long-term cancer survivors,fatigue is a commonly experienced symptomafter chemotherapy or radiotherapy.

PainPain is another common problem experi-

enced by cancer patients, with prevalence esti-mates of up to 70% over the course of patients’illness.14,15 Variability in pain intensity existsbased on the type and stage of cancer. Patientswith primary or metastatic bone tumors or cer-vical cancer report the highest prevalence ofpain, with 85% of patients reporting signifi-cant pain.16 Most frequently, cancer pain is re-lated to tumor involvement, specifically tumorinvasion of bones, nerves, or organs. Surgery,chemotherapy, radiation, and diagnostic tests

may also lead to pain syndromes.16 Cancerpain may not be adequately managed,4,17 andearly and efficient identification may help toimprove management.

Emotional DistressEmotional distress is a normal response to

receiving a cancer diagnosis, discovering a re-currence of cancer, or finding that cancertreatment has failed. Cancer patients demon-strate a characteristic emotional responsethat may include initial shock and disbelief,followed by anxiety and depression, sleepand appetite disturbances, and impairmentsin daily functioning.18,19 The nature and inten-sity of emotional distress among individualswith cancer is largely dependent upon wherethe patient is in the course of illness, as psycho-logical distress may increase with illness sever-ity.20 Recent reviews examining psychologicalinterventions with cancer patients have sup-ported the efficacy of treatment in improvingemotional and functional adjustment andin reducing cancer-related symptoms andtreatment side effects.21,22

AnorexiaCancer-related anorexia predisposes to ca-

chexia, which is among the most documentedcauses of death, making a significant contribu-tion to approximately 20% of cancer deaths.23

It has been estimated that 30%e87% of cancerpatients develop anorexia/cachexia,24,25

which may be related to metabolic abnormali-ties due to tumor by-products and cytokine re-lease.26,27 It may be useful to validate a briefscreening instrument to identify patients withsignificant appetite loss.

Given that comprehensive symptom assess-ments can be time consuming and/or burden-some to patients and staff, it would be useful toestablish clinically significant cut-offs on a briefscreening tool that would identify patients whoshould undergo a more thorough assessment.For example, Kirsh et al.28 evaluated a single-item screening tool for fatigue, and similar re-search has been conducted in efforts to screenfor significant cancer-related pain,29 distress,30

and appetite loss.31 However, we are aware ofno reports that simultaneously screen for mul-tiple symptoms in the same population usingmore than one method of evaluating the

22 Vol. 35 No. 1 January 2008Butt et al.

empirical cut-off score. To that end, the pur-pose of the present study was to evaluate theusefulness of single-item screenings to detectclinically significant fatigue, pain, distress,and anorexia among cancer patients activelyreceiving therapy. Unlike much previous re-search, we used two criteria to evaluate optimalcut-off scores on the screening items: 1) thesensitivity and specificity of each screeningitem to predict clinical cases using receiver-op-erating characteristics (ROC) analysis, and 2)the proportion of patients at each screeningscore who reported that some relief of the tar-get symptom would significantly improve theirlives.

MethodsPatient Eligibility

Patients were recruited from two outpatientoncology clinics affiliated with the Robert H.Lurie Comprehensive Cancer Centers: North-western Medical Faculty Foundation (Chicago,IL) and the Kellogg Cancer Care Center (Evan-ston, IL). The study was approved by the institu-tional review boards of each institution.Eligibility requirements included a diagnosisof a solid tumor, including lymphoma (at anystage), and an Eastern Cooperative OncologyGroup (ECOG) Performance Status Rating32

of 0e3 (max range 0e5). Higher ECOG scoresrepresent increasing disability; a score of ‘‘0’’is used to characterize patients that are fully ac-tive, while a score of ‘‘3’’ describes patientswho are capable of only limited self-care andconfined to a bed or chair for more than 50%waking hours. Patients were also requiredto be within 12 weeks of commencing chemo-therapy or hormone or biologic treatment.Patients were invited to participate in a morecomprehensive assessment of their symptomson the basis of their response to the one-itemscreening questions. For the purposes of invit-ing patients to complete the more comprehen-sive assessment, the score threshold of $4 on0e10 scales was selected a priori as a conserva-tive lower bound for the identification of clini-cally significant cases. As per the NationalComprehensive Cancer Network (NCCN)Guidelines, a score of 4 or more on such screen-ing instruments signifies at least moderatelysevere symptoms.33e35

Screening QuestionsEligible patients completed the screening

questions which were asked in a similar formatacross the four domains of interest (Fig. 1 de-picts the study schema). The questions them-selves were based on NCCN Clinical PracticeGuidelines for Supportive Care. For example,the fatigue screening question was worded:‘‘On a 0e10 scale where 0 means no fatigueand 10 means the worst fatigue imaginable,how would you rate your fatigue at its worstover the past 3 days?’’ (The pain question wasphrased similarly using ‘‘pain,’’ the distressquestion was phrased using ‘‘sadness or dis-tress,’’ and the anorexia question used thephrase ‘‘appetite loss.’’) Patients were thenasked to circle a number between 0 and 10,with both anchors labeled, to indicate theirmaximum symptom level. As a follow-up ques-tion, patients were asked if some alleviation ofthe symptom would improve their quality oflife. Using fatigue as an example: ‘‘Wouldsome relief of your fatigue significantly im-prove your life?’’ Patients responded eitherby circling ‘‘yes’’ or ‘‘no’’ or by indicating thatthey did not have the symptom in question.Responses to the latter question were used toevaluate the cut scores for the 0e10 symptomscreen.

Comprehensive Assessment BatteryPatients who endorsed a score of 4 or higher

on any of the screening items were invited tocomplete a battery of symptom and health-related quality-of-life instruments after com-pleting informed consent. The empiricalselection of cut scores for the screening itemswas based on analysis of this sample. All patientswho completed the comprehensive battery pro-vided basic sociodemographic information,disease/treatment information, and a batteryof self-report instruments. This battery includedthe Functional Assessment of Cancer TherapydGeneral (FACT-G), a general health-relatedquality-of-life instrument,2 and several symptomscales.

Specifically, patients completed the FACT-Fatigue subscale,9,36 the Brief Pain Inventory(BPI),37 the Hospital Anxiety and DepressionScale (HADS),38 and the Functional Assess-ment of Anorexia/Cachexia Therapy (FAACT)5

anorexia subscale. The in-depth symptom

Vol. 35 No. 1 January 2008 23Single-Item Screening

Screening Instrument: 4 domains

(n = 597)

Symptom ≥ 4 on 0-10 scale

(n = 404, on at least one domain)

Self-Report Battery

FACT-GDisease/Treatment InfoTreatment SatisfactionFACIT-F fatigue subscaleBPIHADSFAACT anorexia subscale

Semi-structured Interview(results not reported here)

Characteristics of medicalcare

In-depth symptom

and QOL assessment

(n = 148)

Fig. 1. Study schema.

measures were chosen because of their reliabil-ity, validity, and frequent use in the cancerliterature. Patients completed all of the ques-tionnaires, regardless of the specific symptom(s)they rated at or above the threshold.

In addition to the symptom inventories,patients completed a questionnaire to assesstheir satisfaction with the clinical servicesthey had received related to cancer symptommanagement. Patients also participated ina semi-structured interview to learn about theirinteractions with their health care team. (Satis-faction and interview data not reported here.)

Data AnalysisWe plotted ROC curves to determine opti-

mal cut scores for the symptom screeningquestions. ROC curves allow for visual inspec-tion of the trade-off between the sensitivityand specificity of a test to identify ‘‘cases.’’ Inthe present study, ‘‘cases’’ were defined byscores on the comprehensive symptom indicesexceeding thresholds from the literature. Spe-cifically, clinically significant cases of fatiguewere defined by a Functional Assessment of

Chronic Illness TherapydFatigue subscale(FACIT-Fatigue subscale) score of 42 or lower;9

pain was identified by a BPI Interference cutscore of 45 or higher.39 In accordance with rec-ommendations for clinical use of the HADS,a score of 8 or higher on the HADS Anxietyand Depression scales identified clinically sig-nificant distress.38,40 Finally, clinically signifi-cant appetite loss was identified with FAACT-Anorexia/Cachexia Scale (ACS) scores of 21or lower.41 In an ROC curve, sensitivity (i.e.,the true positive rate) is plotted against1� specificity (i.e., the false positive rate) foreach potential cut score on the instrumentbeing evaluated. An optimal cut score is ulti-mately determined by the nature and purposeof screening; in the present case, we selectedcut scores that optimized both sensitivity andspecificity.

Another common way to quantify the useful-ness of a test is given by the area under thecurve (AUC), which theoretically ranges from0 to 1 (perfect discrimination). In most cases,the effective lower bound for the AUC is0.50, which is the AUC value for chance

24 Vol. 35 No. 1 January 2008Butt et al.

discrimination. Chance discrimination is oftendepicted by a diagonal line that runs from theorigin to perfect true positive and false positiverates. An item/test with a curve near or belowthis comparison curve does no better thanchance levels in correctly identifying respon-dents. To the extent that the screening instru-ment’s curve approaches the upper left cornerof the graph, the more accurately the item candifferentiate ‘‘cases’’ from ‘‘non-cases.’’

ResultsSample Characteristics

A total of 597 patients completed the screen-ing items, and of this sample, a total of 404(67.7%) patients endorsed the fatigue, pain,distress, or anorexia screening item with a scoreof 4 or higher on a 0e10 scale. As seen in Table 1,the subset of patients that scored 4 or higher onat least one screening question was comparableto the total sample of screened patients. Of the404 eligible patients, 148 (36.6%) provided in-formed consent to participate in the more com-prehensive assessment. A demographic andclinical summary of the in-depth sample ispresented in Table 2.

Screening QuestionsTable 3 summarizes the findings from the

two screening questions for each symptom ofinterest (i.e., pain, fatigue, emotional distress,

Table 1Summary of Screening Sample

All Patients(n¼ 597)

Scored $4on At Least

One Symptom(n¼ 404)

Age in years, M (SD) 57.8 (13.4) 57.9 (13.1)Range 18e93 21e89n¼ 5 missing

Gender, %female 65.3 68.1n¼ 1 missing

Cancer diagnosis, %Breast 30.8 30.0Colorectal 12.7 13.4Non-Hodgkin’s lymphoma 10.4 9.7Lung 9.4 11.1Other 36.7 35.8

‘‘Have you receivedany treatments for yourcancer in the pastmonth?,’’ %yes

86.8 91.4

n¼ 13 missing

and appetite loss) in the complete screeningsample (n¼ 597). Along with ratings of eachsymptom over the past three days, the propor-tion of patients who reported that symptom re-lief would significantly improve their life is alsoshown. For those who reported any fatigue,82.6% reported that fatigue relief would signif-icantly improve their life (69.2% of total sam-ple). Similarly, 73.3% with any pain (35.0%of total), 72.3% of those with sadness or dis-tress (47.6% of total), and 60.8% of those

Table 2Summary of In-Depth Sample (n¼ 148)

n %

Age in yearsM (SD) 56.2 (12.7)Median (range) 57 (22e83)

Female 106 71.6

Race/ethnicityWhite 122 82.4African American 16 10.8Asian 2 1.3Other 8 5.4

Marital statusMarried 89 60.1Divorced 21 14.2Never married 21 14.2Other 17 11.5

Education<12th grade 5 3.4High school diploma

or equivalent23 15.5

Some college 37 25.0Bachelor’s degree 46 31.1Postgraduate 37 25.0

Cancer diagnosis, %Breast 48 32.4Colorectal 19 12.8Non-Hodgkin’s lymphoma 13 8.8Lung 18 12.2Other 50 33.8

ECOG Performance Status RatingPatient-rated

0 20 13.51 77 52.02 44 29.73 7 4.7

Physician-rated (n¼ 34 missing)0 48 42.11 50 43.92 14 12.33 2 1.8

Current extent of disease (n¼ 6 missing)Local 38 26.8Regional 34 23.9Metastasis 60 42.3N/A (i.e., hematologic

malignancy pts.)10 7.0

Vol. 35 No. 1 January 2008 25Single-Item Screening

Table 3Results of Symptom Screening Questions (n¼ 597)

Symptom at Its WorstOver The PastThree Days

Fatigue Pain Sadness/Distress Appetite Loss

n %Yes n %Yes n %Yes n %Yes

0 (none) 83 1.2 294 1.7 186 2.7 258 1.91 39 30.8 67 11.9 68 19.1 54 11.12 60 50.0 42 57.1 66 63.6 54 35.23 89 70.8 67 77.6 93 73.1 64 60.94 70 82.9 37 89.1 49 79.6 43 79.15 85 97.6 30 96.7 38 78.9 32 65.66 51 96.1 20 95.0 28 89.3 26 96.27 53 94.3 12 100.0 24 91.7 28 75.08 38 100.0 19 100.0 19 84.2 15 80.09 15 100.0 3 100.0 10 90.0 15 86.710 (worst) 14 100.0 6 83.3 15 100.0 7 100.0

Summary 597 69.2 597 35.0 596 47.6 596 33.8

Patients were asked to rate their symptoms, if present, on a 0e10 scale over the past three days. If present, patients were asked a follow-up ques-tion: ‘‘Would relief of your (symptom) significantly improve your life?’’ The %Yes columns represent the percentage of patients at each symptomlevel (and overall) who endorsed that symptom relief would improve their quality of life.

with appetite loss (33.8% of total) reported thatsymptom relief would significantly improvetheir quality of life. Participants’ responses tothese questions highlight the impact thesesymptoms have on the lives of cancer patients.

Comprehensive Assessmentof Patient-Reported Outcomes

Notably, 67.7% (n¼ 404) of the screeningsample met the a priori threshold of 4 or high-er on the 0e10 screening question for at leastone of the four symptoms. The subset of 148individuals who agreed to complete the morecomprehensive battery of outcome measureswas comparable with respect to age, gender,and diagnosis to the full sample of 404 (seeTables 1 and 2) and to eligible patientswho did not complete the battery (n¼ 256;all P-values> 0.05). Furthermore, patientswho completed the in-depth assessment scoredsimilarly to the 256 patients that declinedparticipation on the fatigue, pain, distress, andanorexia screening items (all P-values> 0.50).

Of those patients who completed the com-prehensive battery, the proportion who metcase definition based on their scores on thecriterion measures was as follows: fatigue(85.5%); pain (9.9%); anxiety (25.5%); de-pression (24.6%); and appetite loss (9.0%).Responses to the more comprehensive batterywere used to assess the utility of the screeninginstruments to detect clinically meaningfulsymptoms. The correlations between theone-item screening questions and the morecomprehensive outcomes measures were allstatistically significant, ranging from 0.38 to0.67. Descriptive statistics for the comprehen-sive assessment scales along with the correla-tion of each instrument with the screeningitem are shown in Table 4.

Establishing Cut ScoresFig. 2 and Table 5 present results of the ROC

analyses that were used to determine optimalcut scores on the screening items. As shownin Table 5, depending on the symptom,

Table 4Descriptive Statistics for Comprehensive Battery Questionnaires (n¼ 148)

n Mean (SD) RangeCorrelation with

Respective Screena

FACT-G (range 0e108) 144 76.34 (13.15) 44e104.6 dFACIT-Fatigue subscale (range 0e52) 145 30.40 (11.29) 4e52 0.56BPI Interference (range 0e70) 142 13.80 (16.96) 0e61 0.67HADS Anxiety (range 0e21) 141 5.48 (3.71) 0e16 0.56HADS Depression (range 0e21) 142 5.08 (3.70) 0e17 0.38FAACT-ACS (range 0e48) 144 33.12 (8.39) 8e48 0.55

a1) Because high scores are interpreted differently for the scales, absolute values of the correlation coefficients are presented; 2) all correlationcoefficients are statistically significant, P< 0.0001; 3) the distress screening question was used in correlation calculations for both HADSsubscales.

26 Vol. 35 No. 1 January 2008Butt et al.

optimal cut scores on the screening itemsranged from 4 to 6. Examination of the sensi-tivities and specificities of the screening itemsat the cut scores shows that screening wasmost successful (in decreasing order) for iden-tifying cases of anxiety, appetite loss, pain,fatigue, and depression. Although there arefew empirically based guidelines for the inter-pretation of AUC,42 the values obtained inthe present investigation should be consideredadequate to quite good.

Evaluating Cut Scores in Full SampleWe investigated how the cut scores per-

formed with regard to patients’ evaluation oftheir own symptoms. Specifically, we used thecut scores for each symptom to determinehow they were rated in the full (n¼ 597)sample with regard to quality of life. A largesample of patients scored above the ROC-established cut score for fatigue (n¼ 256),with 97.3% indicating that relief of this symp-tom would improve their quality of life. A totalof 127 patients rated their pain at the cut scoreof 4 or above, with 94.5% of those individualsstating that relief of pain would significantlyimprove their quality of life. The cut scoresfor anxiety and depression were both

determined to be 5; for the distress item(that combines both symptoms), this cut scoreidentified 134 patients, 87.3% of them statingthat relief of emotional distress would signifi-cantly improve their quality of life. Finally,the cut score for weight loss identified 90 pa-tients in the original sample; 86.7% of these in-dividuals reported that relief of anorexiawould result in improved quality of life.

DiscussionIn this sample of ambulatory outpatients

with solid tumors, we were able to demonstratethe utility of single-item, self-report screeninginstruments to detect clinically significantfatigue, pain, emotional distress (anxiety,depression), and anorexia. Screening itemsensitivity and specificity were calculatedagainst more comprehensive, well-establishedsymptom assessments using ROC curves ina subsample of 148 patients. For the 0e10 rat-ings on the screeners, these values were maxi-mized at cut scores of 4 (pain), 5 (fatigue,anxiety, depression), and 6 (anorexia). Thesecut-off scores were evaluated in the full sample(n¼ 597) and were found to identify patients

0.75 1.00

Tru

e P

os

itive

Tru

e P

os

itive

Pain Anxiety

Appetite Loss

0.00 0.25 0.50 0.75 1.00False Positive

0.00 0.25 0.50 0.75 1.00False Positive

0.75 1.00 0.00 0.25 0.50 0.75 1.00False Positive

AUC=0.76 AUC= 0.88

AUC = 0.86

4

5

0.00 0.25 0.50False Positive

0.00

0.25

0.50

0.75

1.00

0.00

0.25

0.50

0.75

1.00

Fatigue

Depression

0.00 0.25 0.50False Positive

AUC = 0.71

AUC = 0.68

5

5

6

Fig. 2. ROC curves.

Vol. 35 No. 1 January 2008 27Single-Item Screening

who reported that symptom relief would signif-icantly improve their life. The results of thepresent study suggest that use of a single-itemscreening instrument may assist routine clini-cal assessment of fatigue, pain, distress, andanorexia in ambulatory oncology practice byidentifying patients who may significantlybenefit from targeted symptom management.

These symptoms are common among indi-viduals with cancer, and despite their preva-lence in this population, fatigue, pain,distress, and anorexia remain under-recog-nized and subsequently undertreated. Further-more, the presence of these symptoms hasbeen associated with impairments in health-related quality of life. For example, amonga group of women with breast cancer, fatiguewas identified as one of their most distressingcancer-related symptoms.43 Cancer-related fa-tigue has also been associated with functionallimitations and impaired quality of life.1,3,8 Incontrast, patients who denied difficulties withfatigue had high average scores on the samequality of life instrument. A brief and accurateapproach to screening for clinically significantsymptoms holds the potential to improve qual-ity of care and could ultimately lead to a de-crease in the burden associated with cancerand its treatment.

The use of screening instruments is not new.For example, several of the NCCN PracticeGuidelines for Symptom Management (e.g.,pain, fatigue, and distress) recommend use ofsimilar single-item screening as part of theirtreatment algorithms. NCCN algorithms havesuggested a cut-off of 4 or higher based on con-sensus; however, this report is one of the first toprovide empirically derived cut-off scores. Itshould be stressed that if an individual ratestheir symptom severity in excess of an

Table 5Optimal Cut Scores and Associated Test

Sensitivity and Specificity for Each Symptom

Symptom

OptimalCut Score

(0e10) Sensitivity Specificity AUC

Fatigue 5 0.694 0.714 0.71Pain 4 0.714 0.719 0.76Anxietya 5 0.861 0.771 0.88Depressiona 5 0.629 0.682 0.68Appetite loss 6 0.923 0.824 0.86

aAnxiety and depression discrimination was calculated using thesame screening question, indexed against respective subscales ofthe HADS.

established threshold/cut-off score, the patientmay be experiencing significant symptomatol-ogy. However, a positive screening should befollowed up with a more comprehensive assess-ment to fully explore symptom etiology and pos-sible treatment routes. For example, manypatients under-report pain to physicians. Theuse of screening tools may help to overcomethis barrier. In turn, it may be easier to addressmild to moderate symptoms, identified earlier,than more severe symptoms that have goneundetected. It is interesting to note that evenbelow the cut-off scores that we establisheda sizeable percentage of patients report thatsymptom relief would significantly improvetheir quality of life. Even at mild levels, patientsreport that their symptoms adversely impacttheir quality of life. In environments withlimited resources, effective screening toolshold the promise for identifying those patientswho would most benefit from more thoroughevaluation (and in the process, identifyingthose patients who may be less likely to beexperiencing significant symptoms).

Interestingly, Jacobsen et al.30 recently re-ported the utility of the Distress Thermometerfor the detection of clinically significant,HADS-identified distress. The screening ques-tion we used is similar to the Distress Thermom-eter; however, our question asked patientsabout sadness/distress over the past threedays, not the previous week. Nonetheless, ourscreening item performed similarly to the Dis-tress Thermometer in identifying significantcases of psychological distress. While Jacobsenet al. did not distinguish between types of emo-tional distress, in our sample, we evaluated theemotional distress-screening item in terms ofits ability to identify significant anxiety anddepression, separately.

Although the NCCN recommends a cut-offscore of 4, we found that a cut score of 5 andhigher on our instrument identified patientswith significant distress. However, we did notexpect that the distress screen would discrimi-nate anxiety so much better than depression(AUC¼ 0.88 versus 0.68, respectively). Thisdifference was also reflected in the correlationcoefficients of the screen to the HADSsubscales, which was higher for anxiety thandepression. ‘‘Sadness’’ may not adequatelyreflect the phenomenology of the depressedpatient with cancer. Alternatively, patients

28 Vol. 35 No. 1 January 2008Butt et al.

may interpret the term ‘‘distress’’ as beingmore consistent with symptoms of anxietythan with depression. Interestingly, Trasket al.44 have also found that the Distress Ther-mometer is more sensitive to the presence ofclinically significant anxiety than to depres-sion. Given that depression may be underde-tected in cancer,45 it may be useful for futureresearch to examine the utility of separate oradditional items to screen for this importantsymptom.

Although the present sample was relativelylarge and heterogeneous with respect to diag-nosis and most sociodemographic variables,these results should be interpreted with somecaution. Specifically, care should be exercisedin using the screening items and cut scoresin samples significantly different from thepresent one. For example, the present sampleis representative of the population with regardto race, but it remains relatively homogeneous.As suggested by recent reviews,46,47 there maybe differential rates of diagnosis, referral tospecialty care, and lower survival for minoritypatients. Care should be taken in the transla-tion of these findings to populations with ra-cial distributions significantly different fromthe present sample. Given that the presentsample of patients was within 12 weeks of start-ing cancer therapy, care should also be takenin extrapolating the present findings topatients in palliative or terminal care.

As with all ROC curves, sensitivity and speci-ficity calculations are related to the base rateof the condition being studied. Because onlythose patients scoring 4 or higher on a screen-ing instrument were invited to complete the cri-terion measures, symptom base rate values mayhave been impacted accordingly. The preva-lence rates of fatigue and emotional distressin our sample were similar to those reportedin the literature; however, less than 10% ofpatients met our case definitions for pain andappetite loss. This may reflect suboptimalmeasurement of pain and anorexia or possibleself-selection bias among patients with thesesymptoms. That said, the screening thresholdspresented here may need to be adjusted insamples with varying symptom rates. In otherwords, depending on the symptom prevalence(expected or known) and the purposes ofscreening, corrections may need to be madeto the empirical cut scores presented here.

The present study is unique in that it is oneof few reports that have evaluated several symp-tom screening instruments in the same popu-lation. Many patients in our sample reportedsignificant difficulties with multiple symptoms.Although co-occurring symptoms are fre-quently reported by patients in clinical prac-tice, research on symptom clusters has beena relatively recent phenomenon.48e50 Single-item instruments may assist in assessing thesynergistic impact of multiple symptoms51

and may be used to screen for multiple symp-toms relatively efficiently. Future research isneeded to determine the impact of routinescreening on symptom detection, treatment,and outcome over time. Routine screeningmay be useful for symptom control, objectivefunctional status, and quality of life.

AcknowledgmentsThe authors wish to thank Drs. Jin-Shei Lai

and Stuart Quirk for helpful comments onan earlier version of this manuscript.

References1. Cella D. Factors influencing quality of life in

cancer patients: Anemia and fatigue. Semin oncol1998;25(3Suppl7):43e46.

2. Cella DF, Tulsky DS, Gray G, et al. The Func-tional Assessment of Cancer Therapy scale: develop-ment and validation of the general measure. J ClinOncol 1993;11(3):570e579.

3. Wagner LI, Cella D. Fatigue and cancer: causes,prevalence and treatment approaches. Br J Cancer2004;91(5):822e828.

4. Gordon DB, Dahl JL, Miaskowski C, et al. Amer-ican Pain Society recommendations for improvingthe quality of acute and cancer pain management:American Pain Society Quality of Care Task Force.Arch Intern Med 2005;165(14):1574e1580.

5. Cella DF, Bonomi AE, Leslie WT, Von Roenn J,Tchekmedyian NS. Quality of life and nutritionalwell-being: measurement and relationship. Oncol-ogy Supplement 1993;7(11):105e111.

6. Holland JC, Reznik I. Pathways for psychosocialcare of cancer survivors. Cancer 2005;104(11Suppl):2624e2637.

7. Van Cutsem E, Arends J. The causes and conse-quences of cancer-associated malnutrition. Eur JOncol Nurs 2005;9(Suppl 2):S51eS63.

8. Irvine D, Vincent L, Graydon JE, Bubela N,Thompson L. The prevalence and correlates of

Vol. 35 No. 1 January 2008 29Single-Item Screening

fatigue in patients receiving treatment with chemo-therapy and radiotherapy. A comparison with the fa-tigue experienced by healthy individuals. Can Nurse1994;17(5):367e378.

9. Yellen SB, Cella DF, Webster K, Blendowski C,Kaplan E. Measuring fatigue and other anemia-re-lated symptoms with the Functional Assessment ofCancer Therapy (FACT) measurement system.J Pain Symptom Manage 1997;13(2):63e74.

10. Vogelzang NJ, Breitbart W, Cella D, et al.Patient, caregiver, and oncologist perceptions ofcancer-related fatigue: results of a tripart assessmentsurvey. The Fatigue Coalition. Semin Hematol 1997;34(3 Suppl 2):4e12.

11. Von Roenn JH, Paice JA. Control of common,non-pain cancer symptoms. Semin Oncol 2005;32(2):200e210.

12. Winningham ML, Nail LM, Burke MB, et al. Fa-tigue and the cancer experience: the state of theknowledge. Oncol Nurs Forum 1994;21(1):23e36.

13. Prue G, Rankin J, Allen J, Gracey J, Cramp F.Cancer-related fatigue: a critical appraisal. Eur JCancer 2006;42(7):846e863.

14. Foley KM. The treatment of cancer pain. N EnglJ Med 1985;313(2):84e95.

15. Goudas LC, Bloch R, Gialeli-Goudas M, Lau J,Carr DB. The epidemiology of cancer pain. CancerInvest 2005;23(2):182e190.

16. Foley KM. Pain syndromes in patients withcancer. Med Clin North Am 1987;71:169e184.

17. Cleeland CS, Gonin R, Hatfield AK, et al. Painand its treatment in outpatients with metastatic can-cer. N Engl J Med 1994;330(9):592e596.

18. Massie MJ, Holland JC. Overview of normal re-action and prevalence of psychiatric disorders. In:Holland JC, Rowland JH, eds. Handbook of psy-cho-oncology. New York: Oxford University Press,1990: 273e282.

19. Massie MJ, Popkin MK. Depressive disorders. In:Holland J, ed. Psycho-oncology. New York: OxfordUniversity Press, 1998: 518e540.

20. Hardman A, Maguire P, Crowther D. The recog-nition of psychiatric morbidity on a medical oncol-ogy ward. J Psychosom Res 1989;33:235e239.

21. Meyer TJ, Mark MM. Effects of psychosocialinterventions with adult cancer patients: a meta-anal-ysis of randomized experiments. Health Psychol1995;14:101e108.

22. Uitterhoeve RJ, Vernooy M, Litjens M, et al. Psy-chosocial interventions for patients with advancedcancerda systematic review of the literature. Br JCancer 2004;91(6):1050e1062.

23. Stepp L, Pakiz TS. Anorexia and cachexia in ad-vanced cancer. Nurs Clin North Am 2001;36(4):735e744. vii.

24. Nixon DW, Heymsfield SB, Cohen AE, et al.Protein-calorie undernutrition in hospitalizedcancer patients. Am J Med 1980;68:683e690.

25. Wilcock A. Anorexia: a taste of things to come?Palliat Med 2006;20(1):43e45.

26. Gagnon B, Bruera E. A review of the drug treat-ment of cachexia associated with cancer. Drugs1998;55:675e688.

27. Illman J, Corringham R, Robinson D Jr, et al.Are inflammatory cytokines the common link be-tween cancer-associated cachexia and depression? JSupport Oncol 2005;3(1):37e50.

28. Kirsh KL, Passik S, Holtsclaw E, Donaghy K,Theobald D. I get tired for no reason: a singleitem screening for cancer-related fatigue. J PainSymptom Manage 2001;22(5):931e937.

29. Jensen MP. The validity and reliability of painmeasures in adults with cancer. J Pain 2003;4(1):2e21.

30. Jacobsen PB, Donovan KA, Trask PC, et al.Screening for psychologic distress in ambulatorycancer patients. Cancer 2005;103(7):1494e1502.

31. Chang VT, Xia Q, Kasimis B. The Functional As-sessment of Anorexia/Cachexia Therapy (FAACT)Appetite Scale in veteran cancer patients. J SupportOncol 2005;3(5):377e382.

32. Oken MM, Creech RH, Tormey DC, et al. Toxic-ity and response criteria of the Eastern CooperativeOncology Group. Am J Clin Oncol 1982;5(6):649e655.

33. National Comprehensive Cancer Network. NCCNClinical Practice Guidelines in Oncology (v.1.2007):cancer-related fatigue. Available at http://www.nccn.org/professionals/physician_gls/PDF/fatigue.pdf.Accessed February 6, 2007.

34. National Comprehensive Cancer Network.NCCN Clinical Practice Guidelines in Oncology (v.1.2006): adult cancer pain. Available at http://www.nccn.org/professionals/physician_gls/PDF/pain.pdf.Accessed February 6, 2007.

35. National Comprehensive Cancer Network.NCCN Clinical Practice Guidelines in Oncology(v.1.2007): distress management. Available athttp://www.nccn.org/professionals/physician_gls/PDF/distress.pdf. Accessed February 6, 2007.

36. Cella D, Eton DT, Lai JS, Peterman AH,Merkel DE. Combining anchor and distribution--based methods to derive minimal clinically impor-tant differences on the Functional Assessment ofCancer Therapy (FACT) anemia and fatigue scales.J Pain Symptom Manage 2002;24(6):547e561.

37. Daut RL, Cleeland CS, Flanery RC. Develop-ment of the Wisconsin Brief Pain Questionnaire toassess pain in cancer and other diseases. Pain1983;17(2):197e210.

30 Vol. 35 No. 1 January 2008Butt et al.

38. Zigmond AS, Snaith RP. The Hospital Anxietyand Depression Scale. Acta Psychiatr Scand 1983;67(6):361e370.

39. Caraceni A, Martini C, Zecca E, et al. Break-through pain characteristics and syndromes in pa-tients with cancer pain. An international survey.Palliat Med 2004;18(3):177e183.

40. Herrmann C. International experiences withthe Hospital Anxiety and Depression Scaleda re-view of validation data and clinical results. J Psycho-som Res 1997;42(1):17e41.

41. Ribaudo JM, Cella D, Hahn EA, et al. Re-valida-tion and shortening of the Functional Assessment ofAnorexia/Cachexia Therapy (FAACT) question-naire. Qual Life Res 2000;9(10):1137e1146.

42. Kraemer HC, Morgan GA, Leech NL, et al.Measures of clinical significance. J Am Acad ChildAdolesc Psychiatry 2003;42:1524e1529.

43. Longman AJ, Braden CJ, Mischel MH. Side ef-fect burden in women with breast cancer. CancerPract 1996;4:272e280.

44. Trask PC, Paterson A, Riba M, et al. Assessmentof psychological distress in prospective bone mar-row transplant patients. Bone Marrow Transplant2002;29(11):917e925.

45. Van Fleet S. Assessment and pharmacotherapyof depression. Clin J Oncol Nurs 2006;10(2):158e161.

46. Betancourt JR. Eliminating racial and ethnicdisparities in health care: what is the role of aca-demic medicine? Acad Med 2006;81(9):788e792.

47. Institute of Medicine. Unequal treatment: Con-fronting racial and ethnic disparities in health care.Washington, DC: National Academy Press, 2002.

48. Miaskowski C, Dodd M, Lee K. Symptom clus-ters: the new frontier in symptom management re-search. J Natl Cancer Inst Monographs 2004;32:17e21.

49. Paice JA. Assessment of symptom clusters inpeople with cancer. J Natl Cancer Inst Monogr2004;32:98e102.

50. Walsh D, Rybicki L. Symptom clustering in ad-vanced cancer. Support Care Cancer 2006;14(8):831e836.

51. Theobald DE. Cancer pain, fatigue, distress,and insomnia in cancer patients. Clin Cornerstone2004;6(Suppl 1D):S15eS21.