Cancer Clinical Trials

today's call isbeing recorded. if you have anyobjections, you may disconnect at this time. now, i would like to turnthe call over to diane st. germain. >> diane st.germain: thank you. hello. this is diane st. germain with the national cancer institute.

on behalf of theinternational society of quality of life researchand the national cancer institute, i would liketo welcome you to today's webinar series. the series is titled "bestpractices for integrating patient-reported outcomesin oncology clinical trials." the webinar series iscomprised of six webinars, each approximately 45minutes in length. the webinars are designedto be viewed independently

to meet the individuallearning needs of the participants, or theseries can be viewed sequentially inits entirety. today's webinar is thethird in the series, and is titled "how to designa high-quality study with pro endpoints,"presented by dr. madeleine king, dr. michelle naughton,and dr. lari wenzel. i will now turn itover to dr. king. >>madeleine king: hello,and welcome

to this webinar. we're going to provide yousome guidance on how to design a high-qualitystudy with pro endpoints. my name's madeleineking, and i'm from the university ofsydney, australia. it's my very greatpleasure and privilege to be co-presenting professormichelle naughton from wake forest university,north carolina, and professor lari wenzel from

the university of the california, irvine. next. here are the pointswe will cover in this webinar. i'll start with pro aims,and i'll touch on why it's so important to beclear in specifying and justifying them. i'll then provide somepointers on things to think about when planningyour pro assessments

in terms of when, where,how, and who to assess. i'll finish by consideringthe use of auxiliary data, that is, additional to prodata to supplement missing pro data. i'll then hand over tolari wenzel, who will explain confidentiality issuesin pro data, and she'll present the firstof our examples to illustrate how our generalpoints apply to randomized control trials.

professor naughton willthen present our second example, illustratinghow our general points apply to a survivorshipcohort study. but before we get started,we wanted to make this general point: thetenets of sound research methodology are universal. they apply to studies withpro endpoints, just as they do to any othertype of outcome. so you shouldn't besurprised by anything

that we say in this webinar,and we hope that you'll see how clarity, care,and common sense are all that's required todesign a high-quality study with pro endpoints. the first step indesigning the pro component of your study,whether or not your pros are primary orsecondary outputs, is to clearly specify your pro aims and theassociated hypotheses. this step isabsolutely critical.

your pro hypotheses andaims need to be really clearly defined. it might help you to thinkabout the pico framework here: your patientpopulation, your intervention, yourcomparator, or control if it's a randomizedsetting, and your outcomes. you'll also need tojustify your aims and hypotheses in terms of theclinical relevance, that is, how will they improvethe position

of health care, and also interms of importance to patients or survivors, thatis, how they improve the process and/or outcomes ofcare for those patients. and you'll need to makereference to previous literature. what's already known aboutpros in this or similar clinical contexts,and what's the gap in knowledge that yourstudy will fill? specific pro demands ofinterest also need

to be measurable, that is,validated measures of the constructs of interestneed to be available. and finally, the proassessment methods that you're going to use needto be feasible, either in the clinic orin the field. having defined andspecified your aims and hypotheses, you nowneed to plan your pro assessment schedule. that is, you've got toselect

your pro measures and the time points. now, this will involvethinking about a range of characteristics andconstraints of your patient populationin clinical context. you need to think aboutthe type of study that's going to be conducted,so considerations for a cohort study will bedifferent from those for a clinical trial, andconsiderations for a medical intervention willbe different

from those for a behavioralintervention. these sorts of issues willobviously relate quite directly with yourspecific aims, once you've defined those interms of pico: patient, intervention,comparator, and outcome. you need to think aboutthe nature of the cancer, and the treatmentor intervention. so, for example, themeasures and time points for a clinical trial ofradiotherapy

for prostate cancer will bequite different from those for assessing patients atrisk of poor psychosocial outcomes in a survivorshipstudy of ovarian cancer. the thinking about yourpatients and participant characteristics will help you in deciding which measures and time points. the issue of which promeasures is a really big one, and indeed, there'san entire webinar on this point by ethan dosch[spelled phonetically]

and bryce reed [spelled phonetically], and we encourage you tolisten to that webinar. the remaining points here,who, where and when, and how, are topics that weconsider in detail now. let's think about"who" first. now, this basically isyour pro inclusion and exclusion criteria. ideally, it's best forthose involved in the pro assessments to bejust the same subjects

as those evaluated for allthe study endpoints. this is a practicalapproach, and it's easy to implement. and in terms of thescience, it increases your credibilityand the interpretability of your results. however, in practice,there may be some limitations. some folks may have somephysical limitations,

such as poor vision, or aninability to hold a pen or to use a keyboard. others may have somecognitive problems. others may have adifferent language or literacy for areas that make self-assessment infeasible. but please try to keepyour exclusions down to a minimum. so, for example, theeortc,

fesa [spelled phonetically], andpromis questionnaires are all available in numerouslanguage translations, and these can be utilized toreduce exclusions due to language. the next thing to thinkabout is where will your pros be assessed. now, if you are doing aclinic-based study, then the clinic is a veryconvenient place to assess your pros, becausepatients are attending

for treatment and follow-up. however, when they come toclinic might not be the most informative time. so, for example, if you'vegot cyclic acute toxic effects like nausea, youmight miss them if you assess the pros justbefore the dose of chemotherapy, which iswhen the patient comes to clinic. online pro assessment ismaking completion

at more informative times muchmore feasible, and it's also making pro assessment in non-clinic-based populations,such as survivorship cohorts, more feasible. so, that new technology iscertainly very promising for giving us moreflexibility about where pros are assessed. you also need to thinkabout how pros will be assessed, and bythis,

i mean modes of administration. traditionally we've hadpencil and paper, and telephone, and a commonlycalled computer-assisted telephone interview, orcati, and increasingly these days we're using electronic self-complete, so touchscreen, online, and web-basedapproaches. i'm going to go throughthe pros and cons of each of these modesof administration

in the next few slides. let's think abouthard-copy questionnaires first. the pros of these arethey're very convenient, the patients can completethem while at clinic, they can have very highcompletion rates if you have staff at those siteswho are handing them out and ensuring that patientsgive them back, and they have that humanface-to-face contact that really helps withhigh completion rates.

and also, thequestionnaires can be completed and collectedin one place, which reduces the risk of losing forms. however, on the con side,the data entry is by hand, and this is time-consumingand prone to human error. if you want to usescanning, then this has set-up costs, andit adds work to already full workloads that our datamanagers and nurses and other cros have.

also, when completed athome, returning by post is costly, and you have therisk of non-return or loss in the mail. let's think about onlinequestionnaires now. the pros of these are thatlinks to questionnaires and reminders can be sentby email, data entry and checking isautomated, which reduces cost, and the results are a quickerreturn and efficient data management, and here i'llrefer you to a couple

of papers: the greene-moore[phonetic sp.] and the frica [phonetic sp.],which you can find and read if you're interestedin more details. also, onlinequestionnaires can optimize the useof skip logic. that's a situationlike, "if 'yes' to this question, then we want toexplore in more detail on the next question, but if'no', skip that question." now, skip logic inhard-copy questionnaires

is very cumbersome, soit's quite a benefit of online and othercomputer-based questionnaires. online questionnairesalso allow a thing called computer-adapted testing,or cat, methods, which you may have heard about, and are increasingly being implemented. on the con side, patientscan easily ignore emailed reminders, and i'm surewe're all guilty of ignoring some ofour emails, right?

the cons also are thatonline questionnaires require computer literacyand online access, and we can still have technicalerrors, so when the server is down or there are other connectivity issues. let's think abouttouchscreens and tablets now. on the pro side,touchscreens in the clinic have proved to bea great success. numerous studies haveshown they're feasible

in oncology clinics, andthey've been used to great success in screeningfor distress in cancer. you see the same benefitshere as for online and in-clinic data collection. indeed, a meta-analysis of65 studies has shown that paper-completed,pencil and paper, and computer-based versionshave demonstrated equivalence of data of thetwo modes, which is really great news, andi refer you here

to the meta-analysis by gwaltney. on the con side, though,touchscreens and tablets don't eliminate the riskof technical fault, and they do include expenses,such as tablets and ipads and software, and theadditional time that staff needs to help patients use these data collection methods. the final point now inmode of administration is computer-assistedtelephone interview,

and these have been used inthe past to assess at times that are not linkedwith clinic visits. they can help incompletion rates, again because of thathuman-context element. and, importantly, theyallow inclusion of respondents who have lowliteracy with written words, or who arenot computer literate, or who have other physicaldifficulties with self-completion, eitheron hard copy or computer.

they're also goodfor slightly more context-cognitive tasks, like preference-based measures. also data entry can be done at the time of data collection, which makesthem very efficient, and they're great if theresearch is being contracted out to acontract research organization. on the con side,however, they're costly. there's postage forsending

out the questionnaires topatients so they have them in front of them during thetelephone interview. there's the phone costsand the software, and also the additionalstaff costs. and they'retime-consuming. there's also some evidencethat there's potential bias, and you'll notethere that there are three studies which show thatthere is the tendency for higher scores for pros bytelephone administration

due to a sort of socialdesirability effect. but on the other hand, twostudies have shown that telephone prosare equivalent to self-complete. and again, i've given youthe names and years of publication, so you cancheck out those papers if you're interestedin the details. okay, so that's everythingwe need to say about pro administration.

just to review, we talkedabout who, and ideally, you will include allpatients in your study and minimize exclusions. we've talked about where,and how the clinic is a convenient place for aresearch staff, but you also need to thinkabout if it's the most informative time, and homemay be a more suitable place, particularly forsurvivorship studies. we've also talked abouthow, and we've talked

about paper-based versus computer-based versus telephone-assisted. we've thought aboutlogistics and cost, set-up, data collection,data entry, follow-up and reminder ofnon-respondents, and you need to think about allof these in your own context, and in the context of thefunding that you have and the otherresourcing you have. and we've also talkedabout equivalence.

and you need to look inthe literature to see whether equivalenceis proven. and once it is proven, andthere is certainly some promising evidence comingout these days, then this is going to open theway for mixed-mode designs to enhance participation andresponse rates for our pros. okay, so now we're movingon to the next big topic in this webinar, which istiming of pro assessments.

now this isabsolutely critical. getting the timingright really matters. the sorts of things youneed to think about are: are there acute effects,what are they, and are there late effects,what are they. what are the trajectoriesat the time of acute and late effects? this will help youdecide on the timing and frequency of proassessments,

both during the active treatment phase and the survivorship phase, andalso how long you want to be following patients up. the final point isacceptable time windows. you also need to thinkabout whether or not to assess patients afterthey've discontinued therapy, and this is just in the clinical trials context. the first issue in the"when" question

is baseline, and theanswer is "always." always have a baseline. the reason is that there'salways a lot of relatively stable between-personvariation in pro measures. that is, a fifthof the time. so adjusting the baselineis a great way to improve the power of yourtreatment comparisons. so what is asensible baseline? well, you really needto think about what's

clinically sensible and what's logistically possible. so, for example, ifyou have a surgical intervention, then yourbaseline really should be pre-surgery. but sometimes that canbe a little bit tricky, particularly if theinitial surgery is actually part of thediagnostic workup. for studies involving interventions,

before the intervention startsis always the best baselining possible. and if you're runningan rct, or a randomized controlled trial, thenconsider making baseline pro assessment a criterionfor randomization. that's a really greatway to get really good compliancerates and baseline. your survivorship studies,it's a very different sort of situation, and baselinetypically

is at the recruitment point. okay, so let's think aboutwhen to assess pros after baseline. now we need to think aboutwhat are the critical time points that need to beassessed to answer your specific researchquestions and to address your aims. we need to think aboutboth the positive effects of your treatment, and theadverse effects

of the illness and/or of thetreatment, and the time course of events. when are theylikely to occur? is the effect or thesymptom constant or intermittent? is the effect or thesymptom related to a specific treatment, orto the illness itself? does the symptom or theeffect of treatment get worse, better, or staythe same over time?

when i'm consulting with aresearcher or clinician on designing their studies, ialways ask them to think about the trajectoriesof the time, and to draw me a little graph. so i'm going -- i've given you this hypothetical to think about. here you see a graphwith two traces. this is like two differenttreatment arms in an rct. and on the vertical axis,we have

a toxicity where higher is worse. on the horizontal axis,we have time, going from baseline, and then intwo-month increments. and you can see that inarm a, we have a peak, an early peak intoxicity which then dissipates. in arm b, the peak intoxicity is a little later. now let's just think aboutthose red-dotted vertical lines. those are at three months,six months, and 12 months.

these are typical proassessment time points. however, in this case, ifwe assessed at those time points, we wouldnot get the full picture. we would miss the earlypeak of toxicity in arm a, we would miss the latepeak of toxicity in arm b, and we would really notget an accurate picture of what's going on here. so what i want you to getout of this, is that you have to think aboutwhen are going

to be the highs and the lows,and make sure that you assess pros at those time points. the next thing to thinkabout is how frequently you are going to assess, and this will depend on the pro. so, for example, weeklyassessment might be required if you are tryingto catch a changing or an unstable symptom, whereasmonthly or even longer assessment periods mightbe required

for more stable pros. and every six or 12 monthsmight be suitable for survivorship studies. now let's thinkabout the duration of the pro follow-up. how long should yourfollow-up continue? if your primary outcome issurvival and you're going out to five years, then itmight be tempting to do annual pro assessmentsup to five years.

however, you've got aproblem, which is dropout, and this will be worsewith advanced disease. so it's often helpful tothink about what your expectedattrition rate is going to be in yourpatient population, and at what point is the pro data going to be so diminished that the sample is no longer really of interest. a suggestion is to think about expected median survival.

this might provide a goodguideline for ending your planned pro assessment,because at that stage you're likely only to have50 percent of the target sample that youstarted with. okay, so let's thinkabout a thing called time windows. this is a time periodwithin which the effect of interest will beobserved and not diluted. you need to think aboutthis and define

it in your protocol. in this table, i've justgot an example from the [unintelligible]radiotherapy trial that i was involved in,and i've used it in this webinar because it's such a nice,clear example of how to define the beginning and end of acceptable time windows. and you'll notice that westart at pre-registration, then we go topre-treatment,

during treatment, end oftreatment, two weeks after treatment, thensix, three months, and six months after treatment. and i won't go into thedetails, but i'll just notice that the timewindows become wider and wider and wider as wemove away from treatment, because we're moving awayfrom the unstable peaks and troughs and gettingout into the stable end of the response trajectory.

okay, so now let's think about event-triggered assessment. this is in relation toepisodic pros, that is, acute events such aspain or post-chemo nausea, and that's where you getpeaks and troughs. and it might be betterto assess these with an event-triggered datacollection schedule, or more intensivediary-type assessments. now, event-triggered datacollection requires

close monitoring and activepatient participation to capture therelevant events. diaries can be useful inthese situations; they certainly are responsiveto symptoms and side effects, but you canhave problems with compliance, and missing dataare an issue. but there are somepromising new developments with electronic diaries,mobile phones, that will make these muchmore feasible.

so do keep this in mind if you are dealing with episodic pros. survivorship studies areof increasing interest as we get better attreating cancer. increasingly, cancer isbeing seen as a chronic disease, andthere's interest in the experience of patientsafter they move beyond the active treatment period,which is typically the context for your randomized controlledtrials,

and moving out into thesurvivorship phase. researchers designing asurvivorship study have particular challenges,including when, where, and how to best assess pros, and they may need to think about much longer time periods. typically pros areassessed at regular intervals, such as every three months after recruitment,but recurrence

is a really major event, and youmight want to trigger an additional pro assessmentat that time, because pros are going to bechanging then, not only the physical ones of symptoms, but also the psychologicalimpact, the disappointment that cancerhas come back. treatment may alsobe triggered by the recurrence, so thenthat's going to also affect your pros.

so you might then need toalso plan to assess pros at the end of thetreatment to capture the treatment effects. but all of this requires a lot of careful study coordination. when, where, and howyou're best going to assess your pros, willthere be regular clinic assessments, or will youneed to figure out how to assess at home?

you've got to think aboutlonger time periods. you're going to need tothink about, is there going to be achange of care plan? so, for example, thismay include moving on to palliative care, which --where obviously there's a big potential confounderand need to be considered in your analysis plan. and you need to thinkabout reasons for loss to follow-up.

all of these issueswill be illustrated by professornaughton in example two. just moving back to therct context for a moment, because we need to thinkabout what to do with pro assessments after patientsdiscontinue their therapy. now, diane fairclough, inher book, emphasizes the need for a clear policyfor these kind of patients who can no longerfollow the treatment protocol, for example, becausetoxicity is too great.

off-treatment assessmentsmay be really difficult to obtain, because thepatients may no longer be coming back toclinic, or they may have moved on to a different place of care. but practical issuesreally shouldn't determine this decision. you need to think aboutwhether or not these patients need to befollowed for scientific robustness.

now, one of the issues isthat, if discontinuation of treatmentactually limits future therapy to even more toxic orintensive treatment, or to no further treatment dueto disease progression, then if you fail tocontinue pro assessments, this might lead toselection bias. for example, a treatmentarm with a higher rate of dropout may seem to beartificially beneficial. now, why would that be?

because the sickerpatients are the ones who are dropping out, right? okay. now we should be on theslide that says "our therapy goals." i hope they're stillall synced together. the off-therapy worldapproach that diane fairclough suggests is aconservative one, which is to continue proassessments,

because then you'll have the data. off-therapy assessmentscan always be excluded later, but they cannotbe retrieved later. and the reason why shebelieves in this is because it's in linewith intention to treat principles, which areadhered to even when a patient withdraws from theprotocol therapy due to progression or toxicity. okay, that's it forpro timing in rcts.

let's just review whatwe've considered. baseline pro assessmentshould always be done, and it should be donepre-randomization. and it should be aneligibility criterion for your pro study. subsequent pro timepoints should be based on treatment cycles,expected toxicities, and benefits, for example, inpalliative therapies, and the trajectories of theexpected changes

at other times, either recoveryfrom toxicities, or loss of quality of life dueto palliative -- oh, sorry, recovery fromtoxicities, or recovery due to palliative benefits, orindeed loss of quality of life oncepalliative benefit is lost. i've made a bit ofa botch-up of that. diane, are you hearing me? anyway, the samegoes for all arms. so describe your policyfor pro assessment

after discontinuation of yourtrial therapy as well. just to review now protiming for all study types. remember to match yourassessment times to your research questions, yourobjectives, and your hypotheses. consider all the relevantdomains, consider the timing andfrequency of pro assessments, during your therapy for youracute effects,

during follow-up for your lateeffects, and out into your survivorship time. remembering that the[unintelligible] and consequence treatment willalso need to be taken into account. the key time points willbe the ones that show the important changes onyour expected trajectory. and finally, defineacceptable time windows for each one of those.

the final point that i'dlike to make relates to missing data, which arecovered in detail in two other of these webinars,and see on the next slide for details ofwhich those are. but the point i'd like tomake now, is that missing pro data are inevitable. so you need to plan forthem when you're designing your study. to do this, you can planto collect auxiliary data,

that is, additional tonon-pro data that is selected tosupplement pro data. and it's really usefulto statisticians if it's correlated to the proand if it's predictive of missing ones. so, some common examplesin cancer, patient performance status,which is rated by a clinician or anurse. so you've probably heardof the karnofsky

or the who or the ecogperformance status measures, and theseclearly correlate with physically based pros. you can also considerhaving a proxy for the target pro given by theclinician or a caregiver. the other thing that you can do in planning for missingpro data is to record the reasons for the missing pro data. the statisticians can usethis to assess

the missing data mechanisms andthe plausibility of the model assumptions. and the reasons andrates can be reported in resulting papers, alongwith any sensitivity analyses to aid theinterpretation of the pro results. here are the details ofthe other webinars that we recommend you watch toget a better picture of missing data,

which are really one of thebiggest threats to pro study quality and interpretability. i will now hand overto professor wenzel. >> lari wenzel: thankyou, madeleine. i am going to be talkingabout two subjects. one, confidentiality ofpro data, and second, an example of an ovariancancer clinical trial. so, first onconfidentiality. given the nature ofpro assessment

and the multiple methods inwhich we can capture these data, as previously reviewedby professor king, it's important to addresslevels of confidentiality, which may or may not beavailable within these data sources. so i would ask you toimagine, for example, utilizing a measure toscreen for emotional distress within a busy oncology outpatient clinic.

if the patient scores ator above a particular threshold, this wouldtrigger a follow-up conversation during thisclinic visit between a member of the treatmentteam with the patient. this is enhancing goodclinical practice and clinical care, andsomething most people would appreciate. however, if these data arealso part of the clinical trial or asurvivorship study,

has the patient been made awareof, or consented to, use of the pro scores forclinical care? and does the patient knowthat, in addition to trial data, this may beincluded as part of the medical record? it has been customary forpatients to be informed that their data areconfidential and will not be shared withothers outside of the trial team or treatment team, andwill be securely stored.

however, when you'relooking at additional considerations for useof pros, you would also need to haveadditional conversations and assurances that may benecessary, such as the example just shared. so the level of dataprotection may indeed depend on the modeof administration. so again, it's importantto ask, who has access to the information, how arethe data stored,

how are they transferred, andhow secure is this method? so, for example, ourpaper-pencil pros in the clinic. is it de-identifiedand should it be? who has access, and howoften do these data change hands, from thepatient, to a clerk, to a nurse, to a doctor, to dataentry, others? and could theseforms be lost? so that is one concernrelated

to confidentiality with paper-pencil pros. with computer-assistedinterviews, there may be greater protection,particularly if it's entered directlyduring the interview. and then, finally, yourcomputer-based self-entry might allow the greatest level of protection of data. and, again, the level ofdata may depend on the mode and who has accessto confidentiality.

so, a couple scenarios inwhich confidentiality may be of concern, althoughrare, may be a situation in which thepatient does not want family members, for example, to seetheir responses to pros. this may occur if theydon't want someone to know their actual levelof pain, or actual level of despair. so how these data areprotected is important as they're being collected.

similarly, if a proxyis actually providing information on behalf ofthe patient, is this truly confidential? and, similarly, is there arisk of misinterpretation of these data due to proxy level provision of response? so to summarize theconfidentiality discussion, considerwithin clinical trials and survivorship studies, thatwe tell the patient

that their information andresponses to questions is confidential, but thatlevel of confidentiality can be challenged ormaximized, depending on the mode of datacapture and the measures taken to secure and store the dataduring collection, and then also the importance of considering data transfer and analyses. now moving on to adifferent topic, is our example of the phase 3randomized control trial.

i'm now going to describehow use of pros has influenced development ofthe phase 3 randomized trial in thesetting of advanced ovarian cancer, and this is a studyconducted by the gynecologiconcology group. so with this disease,intravenous chemotherapy has been the standard ortraditional treatment for first line advancedovarian cancer. however, the use ofintraperitoneal,

or ip, therapy, in which aport is placed into the abdomen, has conferreda survival benefit, compared to iv-only treatmentin some trials. thus, these are twovery different modes of treatment delivery, makingthe question of quality of life difference verygermane in this study. so in this study, we'vecompared pros on quality of life, neurotoxicity,fatigue, and nausea and vomiting in orderto determine

if one or both of our chemotherapy ip regimens improve progression-free survivalcompared to iv therapy. our rationale for prosinclusion was really constructed on a priorstudy, in which we demonstrated that iptreatment provided the longest median survivalfor this ovarian cancer patient population,however, with significantly worsequality of life during the active treatment period,and worse neurotoxicity

developing during activetreatment, and also one year post-treatment. so we asked them, whatwould the survival benefit of ip advantage would beif we were able to reduce toxicity and some ofthe logistic issues concerned with its use? so in fact, the survivalbenefit of ip would have made this approachstandard of care, except that it was too toxic andthere were concerns

of logistic issues. so subsequently, thetreatment was modified and we justifiedthe modifications to assess benefits and untowardeffects of both therapies, and we did this in partthrough retaining many of our quality of lifemeasures, and for this we used the factmeasurement tool. we also added fatiguebecause we recognized that this was aparticular problem

for this population. so the null hypothesis wasthat the mean scores at the specified time pointswould be independent of the randomized treatment. and then our question,or we would hypothesize, perhaps, that the pros aresignificantly different between treatment armsduring active treatment, with ip arms having poorerquality of life during active treatment,

but without significant difference subsequently. however, the informedimportant clinical question is, since thetreatment arms were modified in this trial tobe less toxic, perhaps pss and os, or overallsurvival, remains superior in the ip arm, but the pros are no longer significantly different. so what we're examiningis how in the treatment alterations in this trialreduce

the ip versus iv differences onthe pros, and whether or not these differencespersist over time. so in order to examineboth differences associated with acutetreatment as well as longer treatment,long-term differences, we wanted to captureassessments that were between regimen,during active treatment, persistent long-termdifferences, and potential emerging late effects.

and so you can seethe timing of these assessments. to answer questionsrelated to both active treatment and persistentdifferences, we obtained assessments at fivespecific time points. it's important to note,the later time points also allowed us to capture any benefits associated with maintenance therapy. as dr. king has mentioned,our interpretation

of results is only as good asthe data we've captured. you'll see that with thispopulation, valid and comprehensive data isrelatively easy to obtain at baseline, but unfortunately decreases over time. this population rarelymisses answers, so missed items are very few. but missing data, whichmeans missing the entire assessment, oftenbecause the patient

is too sick, or she's beentaken off study. and indeed, institutionsoften make the mistake that off-study means they no longer collect pro data. this is an institutionalerror, since these data are still neededand still analyzed. so we need to continue towork with sites to capture assessments in atimely and valid manner. so we don't haveresults of that trial.

results will be availablein december of 2014, but i wanted to share with youthis last slide, which is an example of anevidence base for symptoms or domains important in ovarian cancer clinical trials. so, since it helps toinform measurement selection per domainof interest in ovarian cancer, what you'llsee is that certain symptoms or domains may beparticularly important

for this population and foryour consideration as you construct a trial. so these include, forexample, aspects of disease-relatedsymptoms such as abdominal discomfort, difficultybreathing, fatigue, and pain, as well as theprimary treatment side effects for thispopulation, which is unfortunately neuropathy, or development of neurotoxicity.

so i refer you to the 2014jnci issue which provides a comprehensivediscussion of evidence bases for quality of life and proassessments across many disease sides. so it is now professornaughton, who will speak about asurvivorship cohort study. >> michelle naughton:thank you, dr. wenzel. since much of our focus sofar in this webinar has focused on clinicaltrials, we believed

it important to also providean example using a non-clinical trial design. so our second example is asurvivorship cohort study. the particular aims ofthis study, that was examining neutral cyclemaintenance and quality of life after breast cancertreatment, was to look atchemotherapy-related amenorrhea in menstrual cycling of womendiagnosed with invasive breast cancer between the ages of18 and 45 years of age.

so these are quite youngwomen, and many of you may be familiar withbreast cancer, that the majority -- we havemany studies on breast cancer, but the majority are on thosewho are above the age of 50 and particularly above the ageof 60 years of age. and treatment related toamenorrhea and fertility are major issues forthese young women. and our second aim was totrack the health-related quality of life of thesewomen over time,

looking at long-term physicaland psychological symptoms, and following them atleast through -- from baseline throughfive years. now, the next seriesof slides are going to outline protocol treatmentconsiderations and participantcharacteristics that are relevant to the designof this particular trial. so the protocolconsiderations were, we recruited 836 patients.

they were recruited at seven clinical sites across the u.s. within eight monthsof their diagnosis. the majority of the womenwere diagnosed around eight months from thetime of being diagnosed and the time enrolling in thestudy, so we didn't get them prior totreatment, which is one difference, if you were looking at atreatment trial.

we wanted to see how theywere faring if they were in the midst of chemo,and in other types of treatment moving forward. all of the baseline formswere completed in the clinic, and participantsdid them on pen and paper forms. and then, actually,the staff who were participating with usmailed the forms to this centralizedcoordinating center.

all patients were followedup by mail at six-month intervals, so theydid not come back to the clinic for any of thesetypes of assessments, and it freed us up to some extent, in terms of we did not have to worry about coinciding with particular treatment orparticular side effects. we wanted to look at theirlong-term symptoms and quality of life. so the questionnaireswere mailed

from the coordinating center tothe participants, and then they completed them andmailed them back to the coordinating center. and in terms of treatmentor intervention considerations, this wasan observational study, so we didn't have to makeallowances or design measures that weretrying to look at treatment effects or an interventiontype of effect. we did, though, need tothink about such things

as what were thecharacteristics of this population, what is thesurvival like, and how might they change asthey move from active vasomotor therapy moving on, andthe five-year survival was projected to beabout 85 percent. and the reason that thatis important is that it gives you an idea ofkind of what your losses might be, in terms oflost to follow-up. and also how well, orhow ill,

this patient population might be. so we also -- though thereare some participants and patients who didn'talways go off of treatment or they had maybechemotherapy but they didn't have radiationtherapy, but we tracked all of the followingchemotherapy, if they had it, radiotherapy,hormonal therapy, and any surgeries, as wellas recurrences. in terms of theparticipant

characteristics, a keyfor this study is these are young women. they were all englishspeakers, for the most part. we did enroll some spanishspeakers, but most of them were able to completethe forms in english. we did exclude patientswho had any major mental or cognitive impairment asbaseline, but we did not for people whomight have developed

cognitive difficultiesthroughout the course of the follow-up period or anydepressive symptoms, they were not excludedlater on. the key for us, andactually it was an advantage to theparticular method that we used, is that thesewomen were young enough and capable enough ofcompleting the forms independently, and thenalso returning the forms in the mail.

and a key point is alsothat the majority of participants within abouta year post-diagnosis, or in about an 18-month,a year post-treatment, chemotherapy if theyreceived it, were feeling fairly well and weremoving back into their lives with workor with family. so when we approachedthe task of selecting patient-reported outcomemeasures, we wanted to look for validatedmeasures that had been

used with breast cancerpatients, and the point of that was thatwe wanted to be able to compare it with other cancer studiesover time, or also to compare a younger cohort with,perhaps, an older cohort of women. we also needed to makesure that the directions and the form layout waseasy to follow, because we would not have staffpeople there when they were doing thefollow-up forms at home.

we also knew that we weregoing to be doing repeated measures, and a key was is that we had tohave some balance of measures that would apply equally well tothose who were undergoing active cancer therapy, as wellas following them through five years or more. and some of these womenare now 12 years out. and then also the veryimportant balance of, you want to gather theinformation

that you believe is necessary oressential, but you need to keep yourparticipant burden to a minimum. so this is just an exampleof the quality of life measures thatwere selected. we also looked at thefact, the functional assessment ofcancer therapy. but because, after aperiod of time these women were going to bemore and more removed from active therapy, we alsowanted to put some type

of a generic quality of lifemeasure there as well. and we also, that alsoenabled us to do several different analyses lookingat a cancer-specific measure, which is the factb, but also more of a generic physical andmental health status measure. other, kind of, symptomand lifestyle measures that we included is thatwe did menstrual bleeding calendars because thatwas a major focus of this particular study, isthat we actually,

these women were very faithfulin that they filled out monthly bleeding calendarsfor us up through about five to six years post-diagnosis. and we just basicallywanted to see how their menstrualbleeding co-varied with different types of symptomssuch as hot flushes, vasomotor symptoms and so on, sothat we were looking at kind of a proxy of aphysical measure, and then how the symptoms thatthey might

have might co-vary with themenstrual bleeding calendars. we also tracked if theyhad any additional cancer treatment duringfollow-up. we looked at theirreproductive history, if they became pregnant, ifthey wanted to become pregnant. but the outcomes were --but then we also looked at several other things suchas depressive symptoms, sexual functioning, sleepdisturbance,

which are fairly common relatedthings for people who might be amenorrheic, and then going through menopause. and we also looked atsocial support and lifestyle risk factorssuch as smoking, alcohol, physical activity,and so on. also, just as a side note,that not all of the pro measures might be, youknow, related to symptom or quality oflife assessment.

you might also be testingan intervention at some point, which was not thefocus of this particular study, but you need topay attention to kind of retention activities,and who might drop out differentially duringthe course of a trial. and we were very consciousof trying to build in incentives to retentionfor this particular cohort, and we did somethings, we had incentives for quarterly drawings,for gift certificates

to different things. all of the participantsgot an annual gift. we tried to make thingsas simple as we could. the participants could dothe questionnaires over the phone if they didn'twant to hassle with sending them back andforth, and also if it meant keeping participantsin follow-up, we eliminated some studyforms over time. and the study teamactually

identified up front what werethe pertinent or the essential forms that we'd like to have the participants complete. but it's very important,when you're doing a long-term study, isthat that you have some flexibility and somethought given a priori to what are the essentialcomponents that you would like to have theseparticipants complete. but the main goal is tokeep them involved

and engaged in the study. so, just to recapthe major points. you need to make sure thatyou specify all your pro aims up front, andjustify them with reference to the literature and pastwork in the area. you need to specifythe details of the pro assessments, whenthey would occur, where, how, and who will becompleting the assessments. you might also considercollecting

auxiliary forms and, as dr. kinghad mentioned, this might help further validatesome of the pro instruments, or also importantly, maybe,insist on explaining some of your study findingsthat you might have. also, you need to alwaysconsider confidentiality issues and protecting andsafeguarding the rights of the participants asresearch subjects. but all of these pointsapply equally well to all study designs.

it's more how you applythese types of principles that will differ acrossstudies and also by the patient population andthe clinical context of your study. there are many things --there's a fair amount of literature now onpro assessment. but one reference that wedid want to provide you, which is excellent, isone by diane fairclough, looking at the design andanalysis of quality

of life studies inclinical trials. this is an excellentreference that you can use to guide and also tosupplement some of the material from thisparticular webinar. this ends our webinar,and we hope that the information and guidelineswe've presented will assist you in designinghigh-quality clinical trials with pro endpoints. these slides, inconjunction with

the other isoqol presentationsin this best practices series, provide youwith solid research fundamentals inintegrating pros into oncology trials. we wish you very happy and productive research endeavors.