`Prostate Cancer Treated With Prednisone Alone
`or Mitoxantrone and Prednisone
`By David Osoba, Ian F. Tannock, D. Scott Ernst, and Alan J. Neville
`
`Purpose: A combination of mitoxantrone plus pred-
`nisoneispreferabletoprednisonealoneforreductionof
`pain in men with metastatic, hormone-resistant, prostate
`cancer. The purpose of this study was to assess the effects
`ofthesetreatmentsonhealth-relatedqualityoflife(HQL).
`PatientsandMethods: Menwithmetastaticprostate
`cancer (n ⴝ 161) were randomized to receive either
`dailyprednisonealoneormitoxantrone(every3weeks)
`plus prednisone. Those who received prednisone alone
`could have mitoxantrone added after 6 weeks if there
`was no improvement in pain. HQL was assessed before
`treatment initiation and then every 3 weeks using the
`European Organization for Research and Treatment of
`Cancer Quality-of-Life Questionnaire C30 (EORTC QLQ-
`C30) and the Quality of Life Module–Prostate 14 (QOLM-
`P14), a trial-specific module developed for this study.
`An intent-to-treat analysis was used to determine the
`mean duration of HQL improvement and differences in
`improvement duration between groups of patients.
`Results: At 6 weeks, both groups showed improve-
`ment in several HQL domains, and only physical
`
`functioning and pain were better in the mitoxantrone-
`plus-prednisone group than in the prednisone-alone
`group. After 6 weeks, patients taking prednisone
`showed no improvement in HQL scores, whereas those
`taking mitoxantrone plus prednisone showed signifi-
`cant improvements in global quality of life (P ⴝ .009),
`four functioning domains, and nine symptoms (.001 F
`PF .01), and the improvement (G 10 units on a scale of
`0 to100) lasted longer than in the prednisone-alone
`group (.004 FPF .05). The addition of mitoxantrone to
`prednisone after failure of prednisone alone was asso-
`ciated with improvements in pain, pain impact, pain
`relief, insomnia, and global quality of life (.001 FP F
`.003).
`Conclusion: Treatment with mitoxantrone plus pred-
`nisone was associated with greater and longer-lasting
`improvement in several HQL domains and symptoms
`than treatment with prednisone alone.
`J Clin Oncol 17:1654-1663. 娀 1999 by American
`SocietyofClinicalOncology.
`
`IN A RANDOMIZED STUDY conducted in multiple
`
`Canadian institutions, therapy with a combination of
`mitoxantrone and prednisone was shown to be preferable to
`treatment with prednisone alone for the palliation of pain in
`men with hormone-resistant, metastatic prostate cancer.1 A
`palliative response was defined as a decrease in pain on a
`self-assessment pain scale without an increase in analgesic
`medication, which was maintained for two consecutive
`evaluations at least 3 weeks apart. A secondary end point
`was a decrease in analgesic use without an increase in pain,
`and other health-related quality-of-life (HQL) parameters
`were evaluated by patient self-assessment using the Prostate
`
`From the Department of Medicine, University of British Columbia,
`Vancouver, British Columbia; Princess Margaret Hospital and Univer-
`sity of Toronto, Ontario; Tom Baker Cancer Centre and University of
`Calgary, Calgary; and Hamilton Regional Cancer Centre and McMas-
`ter University, Hamilton, Canada.
`Submitted August 10, 1998; accepted January 21, 1999.
`Supported by Lederle Laboratories, Immunex Corporation, Seattle, WA.
`Address reprint requests to David Osoba, MD, Quality of Life
`Consulting, 4939 Edendale Court, West Vancouver, British Columbia,
`Canada V7W 3H7; email david_osoba@bc.sympatico.ca.
`娀 1999 by American Society of Clinical Oncology.
`0732-183X/99/1706-1654
`
`Cancer–Specific Quality-of-Life Instrument (PROSQOLI),1,2
`the European Organization for Research and Treatment of
`Cancer (EORTC) Quality-of-Life Questionnaire C30 (QLQ-
`C30),3,4 and a trial-specific module of questions (Quality of
`Life Module–Prostate 14 [QOLM-P14]) designed for this
`study.
`In the previous report, the emphasis was placed on the
`primary and secondary end points, and the quality-of-life
`results were described only briefly.1 As shown by changes in
`the PROSQOLI Linear Analog Self-Assessment scores and
`the QLQ-C30 scores,
`treatment with mitoxantrone plus
`prednisone was favored over prednisone alone for pain
`relief, physical activity or function, constipation, and mood.
`Patients in both treatment arms who met the criteria for a
`palliative response had improvement in most HQL domains,
`but details of the nature of the improvements and their
`duration were not described.
`Here we present a detailed analysis of HQL as measured
`by the QLQ-C30 and the trial-specific module. Analysis of
`quality-of-life data was performed to obtain information
`about the following questions: (1) Was there a difference in
`HQL between the patients treated with mitoxantrone plus
`prednisone compared with prednisone alone at 6 weeks of
`therapy, ie, before patients treated with prednisone were
`
`1654
`
`
`
`JournalofClinicalOncology,Vol 17, No 6 (June), 1999: pp 1654-1663
`
`Downloaded from ascopubs.org by 173.72.4.37 on January 16, 2017 from 173.072.004.037
`Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
`
`MYLAN PHARMS. INC. EXHIBIT 1101 PAGE 1
`
`
`
`HQL AND METASTATIC PROSTATE CANCER
`
`eligible for the addition of mitoxantrone? (2) Was HQL
`improved in patients who continued on treatment with
`mitoxantrone and prednisone, and if so, which domains
`continued to show improvement and for how long? (3) Was
`HQL maintained in patients who continued treatment with
`prednisone alone? (4) Was there an improvement in HQL in
`patients in whom mitoxantrone was added after the failure of
`prednisone alone?
`
`PATIENTS AND METHODS
`
`Patient Characteristics and Treatment
`
`The characteristics of the patient population were described previ-
`ously.1 Briefly, 161 men with metastatic adenocarcinoma of the
`prostate, who had symptoms that included pain and disease progression
`despite standard hormonal therapy, were randomized to receive mitox-
`antrone (12 mg/m2 given intravenously every 3 weeks) and prednisone
`(5 mg orally twice daily) (80 patients) or prednisone alone (81 patients).
`In addition to other eligibility criteria, the patients had to be willing and
`able to complete pain and HQL questionnaires.
`The treatment of the patients in the study was described in detail
`previously.1 Briefly, patients continued their primary androgen ablation
`therapy, and most patients had discontinued additional antiandrogen
`treatment. Midway through the study, flutamide withdrawal response
`was recognized; thereafter, patients were monitored for at least 4 weeks
`after stopping flutamide before being entered onto the study. Two
`responding patients who entered earlier in the study discontinued
`flutamide less than 4 weeks before entry.
`Patients were examined in the clinic every 3 weeks. Analgesic
`medication was adjusted to give optimum pain control before entry and
`throughout the study. All analgesic medications were recorded by the
`patients in a diary, which was used to compute an analgesic score
`representing mean daily analgesic use during the previous week. Use of
`dexamethasone and other corticosteroids was not allowed for antiemetic
`medication. Dosages of mitoxantrone were adjusted according to
`granulocyte and platelet nadirs. Patients with no improvement in pain or
`those with progressive symptoms after treatment with prednisone alone
`at 6 weeks were eligible to have mitoxantrone added to the prednisone.
`The study was not blinded because it was deemed inappropriate to give
`control patients sham injections of colored fluid.
`
`Primary End Point
`
`The primary indicator of palliation was pain relief as measured by a
`two-point reduction in the six-point Present Pain Intensity Scale of the
`McGill Pain Questionnaire3 or complete relief of pain if 1⫹ initially,
`without an increase in analgesic score maintained on two consecutive
`visits at least 3 weeks apart. A decrease of 50% in the analgesic score for
`at least two consecutive evaluations 3 weeks apart was used as a
`secondary criterion in a response. The results of the primary and
`secondary criteria for the primary end point were reported elsewhere.1
`This report deals entirely with the results of the HQL measurements
`using the QLQ-C30 and QOLM-P14.
`
`Quality-of-Life Questionnaires
`
`Patients completed two questionnaires, the EORTC QLQ-C30 and a
`module of 14 items designed for men with hormone-resistant, meta-
`static prostate cancer (QOLM-P14), without assistance in the clinic at
`each 3-week visit. The QLQ-C30 is a 30-item, previously validated
`
`1655
`
`questionnaire with five domains pertaining to functioning, ie, physical,
`role, emotional, cognitive, and social, as well as a separate global
`quality-of-life domain.4,5 In addition, there are three symptom domains
`consisting of fatigue, pain, and nausea/vomiting and six single items
`about appetite, insomnia, diarrhea, constipation, dyspnea, and financial
`impact.
`The QLQ-C30 was supplemented by the QOLM-P14, which con-
`sisted of 14 items related to symptoms from metastatic prostate cancer
`and side effects of analgesics and chemotherapy that are not covered by
`the QLQ-C30 (Appendix). These items were derived by interviewing 10
`men with advanced prostate cancer who were attending a support group
`to obtain information about their concerns. A semistructured format was
`used to ask what was of importance to their HQL and to list their issues.
`Twenty issues, many related to each other but not covered by the
`QLQ-C30, were identified. From these issues, 12 items were written
`and reviewed by oncologists who care for patients with prostate cancer.
`An item on difficulty passing urine and one on the extent to which
`having to pass urine at night disturbed sleep were added to the
`questionnaire halfway through the study, after it was discovered that
`they had been inadvertently left out at the beginning. As a first step,
`before an analysis of the HQL data was performed, an evaluation of the
`psychometric properties of the original 12 QOLM-P14 items was
`conducted on the data set of responses obtained in the trial. Question-
`naire structure and scaling errors were examined by a principal
`components factor analysis, including orthogonal varimax rotation,6
`and by a multitreat scaling analysis.7 Three scales (impact of pain on
`mobility, pain relief, and drowsiness) and two single items (hair loss and
`change in taste) were identified. Internal consistency estimates as
`assessed by Cronbach’s alpha coefficient were 0.88 for impact on pain
`mobility, 0.72 for pain relief, and 0.68 for drowsiness. The two items
`added during the study (nocturia and effect of nocturia on sleep) were
`not included in the factor analysis and were analyzed as single items.
`The QLQ-C30 was scored according to previously described method-
`ology.4,8 The raw scores were transformed to provide a range between 0
`and 100; higher scores for the functioning and global quality-of-life
`scales indicated better functioning, whereas higher scores for the
`symptom scales and items indicated more of each symptom. After factor
`and scaling analysis of the QOLM-P14, the scoring of the resulting
`scales and single items was performed in a similar fashion to that of the
`QLQ-C30.
`Patients also completed a nine-item linear analog self-assessment
`instrument, the PROSQOLI.1,2 A comparison of this assessment method
`with the QLQ-C30 was reported elsewhere.2
`
`HQL Analyses
`
`An SAS program (SAS Institute, Cary, NC) was used to perform the
`statistical analyses. Baseline QLQ-C30 and QOLM-P14 scores were
`determined for each patient before therapy was started. New baseline
`scores were calculated for patients who began treatment with predni-
`sone but had mitoxantrone added at 6 weeks or later by using the last
`score before the addition of mitoxantrone. Subsequently, a description
`of the scores for patients on each treatment arm was obtained at the
`completion of two cycles (6 weeks) of chemotherapy or 6 weeks of
`prednisone. Comparisons were then made between the scores at a given
`cycle and the baseline scores for the same patients within each treatment
`group. Both parametric and nonparametric statistics were used for these
`comparisons, but because these two methods did not give rise to any
`significant discrepancies, only the results of the parametric methods
`(analysis of variance) are reported. In the aforementioned comparisons,
`
`Downloaded from ascopubs.org by 173.72.4.37 on January 16, 2017 from 173.072.004.037
`Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
`
`MYLAN PHARMS. INC. EXHIBIT 1101 PAGE 2
`
`
`
`OSOBA ET AL
`
`Table 1. Patient Characteristics
`
`Prednisone
`(n ⫽ 81)
`
`Mitoxantrone and
`Prednisone
`(n ⫽ 80)
`
`No.
`
`%
`
`No.
`
`67
`64-74
`
`69
`63-75
`
`%
`
`98
`22
`4
`9
`
`6
`57
`26
`10
`1
`1
`38
`38
`19
`5
`
`77
`15
`3
`8
`
`3
`47
`22
`8
`1
`1
`23
`37
`15
`5
`
`2.9
`1.5-4.6
`
`3.0
`1.6-5.1
`
`95
`19
`4
`10
`
`4
`59
`28
`10
`1
`1
`28
`46
`19
`6
`
`78
`18
`3
`7
`
`5
`45
`21
`8
`1
`1
`30
`30
`15
`4
`
`Characteristic
`Age, years
`Median
`Interquartile range
`Site of metastases
`Bone
`Lymph nodes
`Visceral
`Other
`Time from diagnoses, years
`Median
`Interquartile range
`ECOG performance status
`0
`1
`2
`3
`Unknown
`Present pain intensity
`0
`1
`2
`3
`4
`Analgesic score*
`18
`14
`Median
`10-30
`6-24
`Interquartile range
`NOTE. The protocol required patients to be symptomatic with pain at entry
`onto the study, but case report forms indicated 2 patients without pain, and 8
`patients were judged to havean Eastern Cooperative OncologyGroup (ECOG)
`performance status of 0.
`*Calculatedfromapaindiaryortheaveragedailyintakeoverthepriorweek
`using a score of 2 for standard doses of narcotics (eg, morphine 10 mg) and a
`score of 1 for standard doses of nonnarcotics (eg, acetaminophen 325 mg).
`
`1656
`
`formal corrections were not applied for multiple significance testing.
`Instead, we used a P value of .01 to identify probable differences
`between groups of data, although P values less than .05 are reported.
`The proportions of patients in each treatment group who had a change
`of at least 10 units in either domain or item scores (possible range, 0 to
`100) on at least two successive measurements were calculated.9 This
`degree of change is probably meaningful to patients because it is easily
`perceptible to them and may therefore be clinically important. Another
`study using different methodology10 also supports the contention that a
`change of greater than 10 units in QLQ-C30 domain scores is clinically
`important, and studies in other illnesses indicate that changes of this
`magnitude are clinically meaningful.11-13
`The mean duration of change (⫾ 2 SE) for each group was
`determined by an intent-to-treat analysis. To calculate the duration of
`change, each patient who experienced a change of greater than 10 units
`lasting for at least two treatment cycles (6 weeks) was counted only
`once, and data from patients who experienced a change at any cycle
`followed by no further data (off treatment or death) were excluded. The
`mean duration of change was calculated by summing the duration of
`change for all patients and dividing by the number of patients
`randomized to the group. Statistical comparisons for the mean duration
`of change between the intent-to-treat groups were not attempted
`because the data were not normally distributed. We did not use
`nonparametric methods for these analyses because it seemed unreason-
`able to assign median values of zero to a group in which the majority of
`patients did not have an improvement of greater than 10 in a given HQL
`score.
`In addition to assessing the mean duration of change in HQL scores
`for the treatment groups on an intent-to-treat basis, we also assessed the
`duration of a change of greater than 10 only in patients who experienced
`it. This was accomplished by plotting the proportions of patients who
`were still experiencing this degree of change as a function of time,
`rounded to 3-week intervals (the timing of the assessments), and
`comparing the resulting curves by the log-rank test
`in a manner
`analogous to that used for comparison of survival curves.
`Participating investigators were unaware of individual HQL scores
`while the study was being conducted, and the analysis of HQL data was
`performed only after the study was closed to accrual.
`
`Characteristics at Baseline
`
`RESULTS
`
`The characteristics of the 161 men who participated in the
`trial that are of importance for the quality-of-life assessment
`are listed in Table 1. Most characteristics were well bal-
`anced, except that there was a trend for patients who had
`been randomized to the mitoxantrone-plus-prednisone arm
`to have a higher analgesic score than those treated with
`prednisone alone. Details of the baseline QLQ-C30 and
`QOLM-P14 scores indicated no apparent differences be-
`tween the groups in any of the functioning scales, global
`quality-of-life scale, and symptoms (Table 2).
`
`Intergroup Comparison of HQL After 6 Weeks of Therapy
`
`Patients who were randomized to receive only prednisone
`were required to take it for at least 6 weeks before becoming
`eligible for the addition of mitoxantrone because of symptom-
`
`atic progression or lack of improvement in pain. However,
`three patients in the prednisone group had mitoxantrone
`added before the completion of 6 weeks of treatment with
`prednisone. These three patients were included in the
`prednisone group for the 6-week and intent-to-treat analy-
`ses.
`Of the 81 patients randomized to the prednisone arm, 62
`remained in this arm after 6 weeks, compared with 71 of the
`80 patients randomized to the combined treatment arm. The
`difference in attrition rates (23% v 11%) creates bias when
`attempting a between-groups analysis in the conventional
`manner, because the patients who were doing well were
`more likely to remain on study. Therefore, a statistical
`analysis of these differences was not attempted, but a
`description of the HQL results in patients who completed 6
`weeks of prednisone or of mitoxantrone plus prednisone
`(two cycles) is provided, using two different methods.
`
`Downloaded from ascopubs.org by 173.72.4.37 on January 16, 2017 from 173.072.004.037
`Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
`
`MYLAN PHARMS. INC. EXHIBIT 1101 PAGE 3
`
`
`
`HQL AND METASTATIC PROSTATE CANCER
`Table 2. Baseline (Pretreatment) QLQ-C30 and QOLM-P14 Scores
`Mitoxantrone and
`Prednisone
`Mean ⫾ SD*
`
`Prednisone
`Mean ⫾ SD*
`
`No.
`
`No.
`
`1657
`
`Compared with the prednisone-only group, the mitoxantrone-
`plus-prednisone group apparently showed improvement in
`physical functioning (mean change from baseline ⫾ 2 SE,
`⫹8.4 ⫾ 2.3 v ⫺1.1 ⫾ 3.2) but greater hair loss (⫹6.8 ⫾ 2.5
`v –0.6 ⫾ 1.4) and greater nocturia leading to sleep distur-
`bance (⫹4.6 ⫾ 4.6 v ⫺9.0 ⫾ 3.5).
`
`Changes in HQL as Compared With Baseline
`
`Further analyses were conducted in patients who contin-
`ued prednisone treatment after 6 weeks, those who contin-
`ued on mitoxantrone plus prednisone, and those who had
`crossed over from prednisone to mitoxantrone plus predni-
`sone at any time during the study. These analyses required
`calculation of a new baseline value for each patient who had
`crossed over,
`ie,
`the score obtained on the last HQL
`assessment just before the cross-over regardless of when it
`occurred. A total of 48 patients crossed over to the mitoxan-
`trone-plus-prednisone group, but one of these patients had
`missing baseline HQL scores and another had baseline
`scores but did not provide any further HQL scores after
`baseline. The new baseline scores for the cross-over group
`were not significantly different from the original baseline
`scores for either the original prednisone or the mitoxantrone-
`plus-prednisone groups, except for a higher pain score. This
`would be expected in the cross-over group because failure of
`improvement in pain was the primary reason for adding
`mitoxantrone.
`In the following analysis, the differences between the
`baseline scores and scores after each even-numbered (sec-
`ond, fourth, and sixth) cycle of treatment were calculated for
`each of the three treatment groups, ie, patients who started
`and continued on prednisone alone or on mitoxantrone plus
`prednisone and those who had mitoxantrone added (the
`cross-over group). The comparison between baseline and
`each treatment cycle assesses the degree and duration of
`improvements in HQL at specific time points in patients who
`continued on treatment. Thus, it is not a comparison between
`groups by intent to treat but, rather, a comparison of scores
`within the groups of patients remaining on study and their
`own pretreatment scores.
`Of the patients who remained on treatment after 6 weeks
`of therapy (two cycles), the group treated with prednisone
`alone (n ⫽ 62) showed improvements in social functioning,
`global quality of life, nausea and vomiting, anorexia (.003 ⬍
`P ⬍ .007), and the impact that pain had on their mobility
`(P ⫽ .01) compared with baseline scores (Fig 1). The group
`treated with mitoxantrone plus prednisone (n ⫽ 71) showed
`an improvement in physical functioning, social functioning,
`global quality of life, pain, anorexia, constipation,
`the
`impact of pain on mobility, the degree of pain relief, and
`
`Scales and
`Items
`QLQ-C30
`Functioning†
`81
`Physical
`78
`Role
`81
`Emotional
`80
`Cognitive
`78
`Social
`81
`Global QL
`Symptoms‡
`81
`Fatigue
`80
`N/V
`80
`Pain
`80
`Dyspnea
`81
`Insomnia
`79
`Anorexia
`81
`Constipation
`81
`Diarrhea
`81
`Financial Impact
`QOLM-P14‡
`80
`Pain impact
`75
`Pain relief
`79
`Drowsiness
`74
`Hair loss
`79
`Change in taste
`39
`Dysuria
`38
`Nocturia/sleep
`*On a scale of 0 to 100.
`†For the functioning and global quality-of-life scales, higher scores indicate
`better functioning.
`‡For the symptom scales and items, higher scores indicate more of the
`symptom.
`§P values for comparisons between groups for each scale and item
`were ⬎ .05.
`
`53.3 ⫾ 26.8
`44.2 ⫾ 33.2
`70.7 ⫾ 19.7
`78.1 ⫾ 20.3
`60.0 ⫾ 28.5
`45.4 ⫾ 21.3
`50.7 ⫾ 23.6
`17.5 ⫾ 24.4
`51.0 ⫾ 24.8
`28.3 ⫾ 27.1
`30.5 ⫾ 30.4
`34.6 ⫾ 32.7
`39.1 ⫾ 34.1
`7.8 ⫾ 16.9
`18.1 ⫾ 27.4
`40.1 ⫾ 24.0
`26.2 ⫾ 21.3
`38.4 ⫾ 20.1
`0.9 ⫾ 7.7
`16.0 ⫾ 26.1
`14.5 ⫾ 23.9
`48.2 ⫾ 34.4
`
`80
`78
`80
`79
`80
`79
`80
`80
`79
`79
`80
`79
`80
`79
`80
`80
`77
`80
`79
`79
`45
`45
`
`50.0 ⫾ 25.8§
`45.5 ⫾ 38.8
`68.1 ⫾ 19.9
`73.4 ⫾ 23.0
`58.1 ⫾ 28.7
`44.2 ⫾ 21.2
`50.1 ⫾ 20.1
`18.3 ⫾ 22.0
`50.4 ⫾ 27.3
`26.6 ⫾ 25.3
`26.3 ⫾ 29.4
`38.4 ⫾ 36.6
`46.3 ⫾ 32.0
`9.3 ⫾ 22.0
`13.8 ⫾ 25.8
`43.0 ⫾ 23.7
`29.0 ⫾ 22.4
`40.3 ⫾ 19.3
`3.0 ⫾ 12.2
`23.2 ⫾ 33.1
`11.1 ⫾ 20.1
`37.8 ⫾ 26.2
`
`In the first method, the mean scores were calculated for
`each treatment group. The mean values for each score in
`both groups were similar, although patients who received
`mitoxantrone plus prednisone apparently experienced more
`distress from hair loss (mean scores ⫾ SD, 9.0 ⫾ 19.6 v
`1.1 ⫾ 6.0) and a change in taste (mean scores ⫾ SD, 24.6 ⫾
`28.4 v 12.5 ⫾ 22.6), whereas those who received only
`prednisone had more difficulty with insomnia (mean scores ⫾
`SD, 28.0 ⫾ 28.4 v 19.2 ⫾ 25.6). In the second method, we
`examined the differences in the change in scores from
`baseline to the 6-week time point in both groups. The
`baseline scores for each patient were subtracted from the
`scores at 6 weeks and the means of differences were
`calculated for each group. Thus, each patient served as his
`own control. The advantage of this method over the previous
`one is that it accounts for variations in individual baseline
`values between patients (whether or not the group baseline
`values are statistically significantly different between groups).
`
`Downloaded from ascopubs.org by 173.72.4.37 on January 16, 2017 from 173.072.004.037
`Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
`
`MYLAN PHARMS. INC. EXHIBIT 1101 PAGE 4
`
`
`
`drowsiness (.0001 ⬍ P ⬍ .009). This group experienced an
`increase in hair loss as the only statistically significant
`deleterious effect (P ⫽ .009; Fig 1). Six weeks after the
`addition of mitoxantrone, the cross-over group (n ⫽ 35)
`experienced an improvement in pain, insomnia, and the
`impact of pain on mobility (.0001 ⬍ P ⬍ .01; Fig 1). Only
`comparisons with differences reaching a significance level
`of P ⫽ .01 are emphasized, although several other compari-
`sons of differences, particularly in the mitoxantrone-plus-
`prednisone and cross-over groups, showed P values of .01 ⬍
`P ⬍ .05.
`After 12 weeks of therapy (four completed cycles), there
`was no statistically significant improvement, as compared
`with baseline, in any of the HQL scores in the group still
`being treated with prednisone (n ⫽ 42), although there was
`an insignificant decrease in pain (P ⫽ .05; Fig 1). However,
`the group that continued to be treated with mitoxantrone
`plus prednisone from randomization (n ⫽ 54) showed
`continuing improvement over baseline in four functioning
`scores (.0001 ⬍ P ⬍ .004), global quality of life (P ⫽ .009),
`and nine symptoms (.0001 ⬍ P ⬍ .01). The only deteriora-
`
`tion was in hair loss (P ⫽ .001). The cross-over group (n ⫽
`25) had an improvement in global quality of life (P ⫽ .003)
`as well as in pain relief (P ⫽ .0001).
`After 18 weeks of therapy (six completed cycles), the
`small number of patients still receiving prednisone (n ⫽ 19)
`had improvement only in the impact of pain on mobility
`(P ⫽ .004) as compared with baseline (Fig 1). Those
`receiving mitoxantrone plus prednisone (n ⫽ 43) continued
`to have an improvement in 11 of the 14 function and
`symptom scales that had improved after cycle 4. The small
`number of cross-over patients (n ⫽ 17) still remaining on
`study after 18 weeks of mitoxantrone therapy continued to
`have improvement in pain, impact of pain on mobility, and
`pain relief (.001 ⬍ P ⬍ .003) but had significantly greater
`hair loss (P ⫽ .01).
`
`Duration of Change in HQL Scores
`
`In addition to indicating which groups of patients experi-
`enced the greatest magnitude of improvement in HQL, the
`data in Fig 1 show that the change in some domains lasted
`longer in the mitoxantrone-plus-prednisone group than in
`
`
`
`HQL AND METASTATIC PROSTATE CANCER
`
`the prednisone-alone group. To obtain a more detailed
`analysis, the mean duration of a change of greater than 10
`units as compared with baseline was determined by a
`time-to-event analysis in the intent-to-treat populations for
`each domain and symptom score.11
`In the mitoxantrone-plus-prednisone group, duration of
`improvement
`in several domains and symptoms ranged
`from 11 to 19 weeks, whereas duration of improvement in
`the same domains and symptoms in the group treated
`with prednisone alone ranged from 3 to 7 weeks (Table 3).
`In patients who had mitoxantrone added to prednisone
`after worsening of symptoms on prednisone alone, duration
`of improvement ranged from 4 to 26 weeks after the
`addition of mitoxantrone. Hair loss lasted longer in the
`mitoxantrone-plus-prednisone group than in the prednisone-
`alone group. Although some patients experienced deteriora-
`tion in their taste, it was noted that some patients (23.5%)
`who received mitoxantrone plus prednisone had an improve-
`ment in taste.
`When duration of improvement (⬎ 10 points) was
`assessed by the log-rank test, a longer duration of improve-
`ment was found in social functioning, pain, impact of pain
`on mobility, pain relief, insomnia, and drowsiness (.004 ⬍
`P ⬍ .048) in patients treated with mitoxantrone plus
`prednisone compared with prednisone alone (Fig 2). In the
`cross-over patients, a longer duration of improvement was
`found in pain, pain relief, and drowsiness compared with the
`prednisone-alone group.
`
`1659
`
`DISCUSSION
`
`The Canadian study of mitoxantrone plus prednisone
`versus prednisone alone for men with symptomatic hormone-
`resistant prostate cancer has provided a valuable database
`for analyzing HQL during treatment.1 The purpose of the
`present report is to provide a more detailed analysis of the
`HQL data obtained from administration of the QLQ-C30 and
`QOLM-P14. Our approach was to address questions about
`the extent, if any, to which HQL improved in the treatment
`arms, which domains showed improvement, how long
`improvement lasted, and whether patients in whom mitoxan-
`trone was added after failure of prednisone alone derived
`benefit in HQL after the addition of mitoxantrone.
`A general limitation of the analyses of HQL that com-
`pared groups of patients with unequal attrition rates was that
`the resulting missing data led to difficulties in the interpreta-
`tion of any differences between the groups. Unequal rates of
`attrition in each group may lead to bias, usually in favor of
`the group with the largest proportion of patients remaining
`on study. Thus, the comparison of mean values during the
`treatment and follow-up periods is particularly subject to
`bias. An analysis based on the last recorded observation
`suffers from the same criticism.11
`The use of mean change scores between baseline and a
`later time to compare groups is advantageous because it is
`not influenced directly by attrition because each patient acts
`as his own control; however,
`this approach is limited
`
`Table 3. Mean Weeks (ⴞ2 SE) of Duration of Change in HQL
`Prednisone and Mitoxantrone
`Prednisone
`(n ⫽ 80)*
`(n ⫽ 81)*
`%
`%
`
`Mean ⫾ 2 SE
`
`No.
`
`Mean ⫾ 2 SE
`
`HQL
`
`No.
`
`Cross-Over Group
`(n ⫽ 48)†
`%
`
`Mean ⫾ 2 SE
`
`41.7
`22.9
`45.8
`45.8
`31.3
`56.3
`45.8
`45.8
`35.4
`35.4
`37.6
`12.5
`18.8
`22.9
`33.3
`
`4.6 ⫾ 2.6
`4.1 ⫾ 2.5
`6.7 ⫾ 4.7
`7.2 ⫾ 3.0
`4.6 ⫾ 2.5
`25.7 ⫾ 9.9
`14.0 ⫾ 7.8
`14.5 ⫾ 8.0
`14.0 ⫾ 7.8
`6.2 ⫾ 5.0
`14.2 ⫾ 8.2
`1.1 ⫾ 0.9
`1.9 ⫾ 1.5
`2.2 ⫾ 1.5
`3.2 ⫾ 2.0
`
`No.
`
`20
`11
`22
`22
`15
`27
`22
`22
`17
`17
`18
`6
`9
`11
`16
`
`Better
`3.8 ⫾ 1.8
`32.1
`26
`15.2 ⫾ 5.7
`50.0
`40
`Physical functioning
`3.7 ⫾ 2.2
`23.5
`19
`10.9 ⫾ 4.9
`36.3
`29
`Emotional functioning
`5.6 ⫾ 2.4
`42.0
`34
`13.7 ⫾ 5.3
`45.0
`36
`Social functioning
`6.1 ⫾ 3.0
`42.0
`34
`15.0 ⫾ 5.7
`62.5
`50
`Global QL
`5.5 ⫾ 2.6
`38.3
`31
`12.4 ⫾ 5.2
`58.8
`47
`Fatigue
`4.9 ⫾ 2.2
`39.5
`32
`17.4 ⫾ 6.1
`56.3
`45
`Pain
`6.9 ⫾ 3.1
`38.3
`31
`19.1 ⫾ 6.9
`58.8
`47
`Pain impact
`4.7 ⫾ 2.4
`42.0
`34
`13.1 ⫾ 5.4
`47.5
`38
`Pain relief
`3.0 ⫾ 2.0
`23.5
`19
`13.5 ⫾ 5.5
`36.3
`29
`Insomnia
`2.8 ⫾ 1.4
`27.2
`22
`13.9 ⫾ 5.6
`43.8
`35
`Constipation
`4.9 ⫾ 1.1
`35.8
`29
`13.5 ⫾ 4.8
`47.5
`38
`Drowsiness
`2.5 ⫾ 1.5
`13.6
`11
`9.8 ⫾ 4.9
`23.8
`19
`Unpleasant taste‡
`1.3 ⫾ 0.8
`12.3
`10
`5.2 ⫾ 3.6
`17.5
`14
`Financial impact
`Worse
`1.1 ⫾ 0.5
`7.4
`6
`5.1 ⫾ 2.2
`27.5
`22
`Hair loss
`1.2 ⫾ 0.6
`23.5
`19
`8.4 ⫾ 3.3
`38.8
`31
`Unpleasant taste‡
`*The only domains and symptoms presented are those in which the upper and lower limits of 2 SE did not overlap the mean.
`†For the group crossed over from prednisone to mitoxantrone the numbers are the duration of change after the cross-over.
`‡While some patients experienced a worsening in unpleasant taste, some experienced an improvement.
`
`Downloaded from ascopubs.org by 173.72.4.37 on January 16, 2017 from 173.072.004.037
`Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
`
`MYLAN PHARMS. INC. EXHIBIT 1101 PAGE 6
`
`
`
`
`
`HQL AND METASTATIC PROSTATE CANCER
`
`because it can be applied only to patients who are still
`undergoing treatment or follow-up. Thus, there may still be
`some apparent selection bias because patients who experi-
`ence some benefit from treatment are more likely to remain
`on it longer.
`Another potential bias is that better HQL at pretreatment
`also leads to patients remaining on treatment longer. How-
`ever, this is less important when each patient acts as his own
`control because the patients with the highest pretreatment
`HQL scores are also the ones with the least room for
`showing improvement with time.
`To avoid these biases, some investigators have used
`various imputation methods12 and growth curve analysis/
`mixed effects analysis of variance models13 in attempts to
`make optimal use of serial HQL data with missing observa-
`tions. However, these latter sophisticated methods are based
`on making assumptions about the values to be assigned to
`the missing data, are complex to perform, and are not
`intuitively appealing.
`In attempting to determine differences in duration of
`changes from baseline between groups with unequal attri-
`tion rates, some of the aforementioned difficulties may be
`overcome by an intent-to-treat analysis in which the mean
`duration of a predetermined amount of change (eg, ⬎ 10 on
`a scale of 0 to 100) is calculated by summing the length of all
`individual changes for each group and then dividing by the
`number of patients randomized to that group. This approach
`is commonly used to compare duration of tumor response
`and survival between groups in oncology clinical trials. It
`addresses the important question of whether the patients
`assigned to a particular treatment protocol benefit from it
`regardless of subsequent censoring. Thus, a longer period of
`improvement in HQL scores in one group would constitute
`strong evidence for better HQL in that group. The limitation
`of this method in our study is that most of the patients in the
`group that began on prednisone alone had mitoxantrone
`added after prednisone failure. This cross-over group neces-
`sitated a separate analysis. Thus, each patient who began
`treatment with prednisone only and then had mitoxantrone
`added could be considered as part