Table 2.
Study design and planning strategies to minimise the problem of missing PRO data
|
Category |
|||||
|---|---|---|---|---|---|
| Design | Topic | Specific recommendation | N recommendations* | Potential drawbacks | Source/s: first author (year). Full citations are provided as Online Supplementary Appendix C |
| Assessment schedule | Specify PRO assessment time points | Specify the required PRO assessment time points | 2 | None | Bernhard, Gusset (1998), Beitz (1996) |
| Specify the minimum PRO data requirements (eg, ‘baseline, on and off treatment, and and/or end of study’ (ref. 5, p. 524) | 3 | May create impression that additional PRO assessments are not important | Bernhard, Cella (1998) | ||
| PRO assessment schedule if treatment schedule is disrupted (ie, will the PRO assessment schedule be altered if the treatment schedule is altered?) | 1 | None | Fairclough (2010) | ||
| Time point selection (guidance on how to select PRO assessment time points) | Align PRO assessments to clinic visits so that data may be captured while the patient visits the clinic | 16 | Clinic visits may not be most informative to capture particular treatment effects (eg, chemotherapy toxicity) | Bernhard, Cella (1998), Moinpour (1998), Movsas (2003), Aaronson (1990), Land (2007), Walker (2003), Calvert (2004), Sprague (2003), Revicki (2005), Fairclough (2010), Kyte (2013), Blazeby (2003), Simes (1998) | |
| May be burdensome to participants to attend clinic for regular assessments | |||||
| Align assessment schedule to a fixed reference point (for ease of calculating when PRO assessments are due) | 1 | May be burdensome to participants to attend regular assessments | Bernhard, Cella (1998) | ||
| Allow sufficient breaks between PRO assessments | 1 | May not be feasible if investigators wish to capture acute disease/treatment effects or their frequency via PROs | Sherman (2005) | ||
| Assess PROs of palliative care patients weekly | 4 | Does not consider when PRO assessment would be most meaningful | Tang (2002) | ||
| Balance the number of required PRO assessments (not too few, not too many) | 4 | None | Revicki (2005), Fairclough (2010) | ||
| Consider patient treatment and expected survival when planning assessment schedule (added note: avoid PRO assessments beyond the point of expected median survival) | 3 | None | Kaasa (2002), Hahn (1998), Atherton (2006) | ||
| Select clinically meaningful time points (ie, ensure that PRO assessments will be taken at clinically informative times, ie, to capture the trajectory of treatment and recovery) | 4 | Clinically meaningful PRO assessment time points may not align with clinic visits, which may require alternative modes of administration | Ganz (2007), Jordhoy (1999), Tang (2002) | ||
| Event-driven PRO assessment for a subsample (ie, rather than subjecting entire sample to detailed PRO assessments if they experience certain clinical events, it may minimise staff effort and resources to restrict these additional assessments to a subsample only) | 2 | Event-driven PRO assessment can be logistically challenging to implement | Bernhard, Cella (1998), Simes (1998) | ||
| Focus on short-term outcomes in patients with advanced disease (focusing on long-term outcomes in such samples will lead to high rates of missing PRO data, and uninformative data) | 1 | May not be clinically meaningful to assess short-term outcomes in all studies | Ganz (2007) | ||
| Justify chosen PRO assessment time points | 1 | None | Ganz (2007) | ||
| Minimise PRO assessment time points (select fewer time points to minimise burden and resource usage) | 3 | May sacrifice important information by omitting time points, for example, differences between treatment arms5 | Bernhard, Cella (1998), Macefield (2013), Cella (1995) | ||
| Shorter follow-up duration (avoid following up patients for a longer period of time as participants are more likely to drop out over time) | 1 | May sacrifice important information by ceasing PRO assessment too early in some studies. Some studies may be interested in long-term follow-up/survival outcomes. | Little, Cohen (2012) | ||
| Treatment failure/cessation | Continue PRO assessments after treatment failure | 6 | May be difficult to engage or contact participants beyond point of treatment failure | Hao (2010), Little, D'Agnostino (2012), Sprangers (2002), Chassany (2002), Cella (1995), Cella (1994) | |
| Specify procedures for contacting participants after treatment cessation | 3 | None | Cella (1994), Revicki, Hao (2010) | ||
| Specify the PRO assessment stopping rule (ie, under what circumstances should PRO assessments discontinue) | 3 | None | Bell (2014), Kaasa (1992), Young, de Haes (1999) | ||
| Time windows | Define PRO assessment time windows (ie, baseline assessment time window should always end before the intervention/treatment commences. Follow-up assessment time windows should border the period in which treatment effects of interest are anticipated, for example, if the time point is 1 week postsurgery, a valid assessment may occur anytime between 4 and 12 days postsurgery). | 12 | None | Bernhard, Cella (1998), Cella (1994), Wisniewski (2006), Blazeby (2003), Hopwood (1996), Bernhard, Peterson (1998), Fayers (1997), Hopwood (1998), Revicki (2005), Fairclough (2010), Cella (1995) | |
| Flexible/large time windows (very narrow time windows may be logistically infeasible to implement and so risk of missing PRO data may be reduced by setting larger time windows) | 3 | Not all time windows can be flexible, particularly when assessing acute effects of treatment | Bernhard, Cella (1998), Little, Cohen (2012), McMillan (2003) | ||
| Collect additional/supporting data (which can be used during PRO data analysis and interpretation) | Auxiliary data (to assist interpretation if there are some missing PRO data). Suggestions of types of auxiliary data in the next column | Additional information about non-responders (type of additional information unspecified) | 1 | Requires prespecification, and additional time and resources to collect | Kim (2004) |
| Clinical data | 1 | Requires additional time and resources to collect | Newgard (2010) | ||
| Health status (clinician-rated quality of life index, Karnofsky or ECOG performance status) | 6 | Requires additional clinician time | Coates (1998), Bell (2014), Bernhard, Cella (1998), Simes (1998), Revicki (2005), Fairclough (2010) | ||
| Comorbidity data | 1 | Requires additional time and resources to collect | Bernhard, Cella (1998) | ||
| Concomitant medications | 1 | Requires additional time and resources to collect | Beitz (1996) | ||
| Observation data | 1 | Requires additional time and resources to collect | Kaasa (2002) | ||
| Participant clinical data | 1 | None | Newgard (2010) | ||
| Participant demographics | 2 | None | Altman (2007), Newgard (2010) | ||
| Proxy† reports when participant is no longer able to self-complete | 21 | Proxy reports are not always concordant with participant self-reports. Care must be taken when interpreting proxy data. This is a specialist subject and additional reading is recommended for investigators considering to use proxy assessment.64 | Bernhard, Cella (1998), Chassany (2002), Fayers (1997), Jordhoy (2010), Kleinpell-Nowell (2000), Kyte (2013), Machin (1998), Moynihan (1998), Peruselli (1997), Revicki (2005), Rock (2007), Simes (1998), Sprangers (2002), Stewart (1992), Taphoorn (2010), Walker (2003) | ||
| Toxicity data | 2 | Requires additional time and resources to collect, if not already being collected as part of the study | Fairclough (2010), Revicki (2005) | ||
| Unspecified (use an alternative to PRO in final weeks of life) | 1 | Requires additional time and resources to collect. Additional drawbacks may be apparent depending on specific alternative measure/s used. | Jordhoy (2010) | ||
| Collect reasons for missing PRO data | – | – | See ‘cover sheet’ section in administration procedures in table 3 | ||
| Eligibility criteria for PRO study (suggestions of specific eligibility or inclusion criteria) | Consider the participants' ability to complete PROs | Include—‘participant must be able to complete PROs’ as an inclusion criterion | 2 | Ability to complete PRO assessments may change over the course of treatment. Results may not be generalisable to all patients. | Bernhard, Cella (1998), Huntington (2005) |
| Exclude patients with language/cognitive barriers from the PRO study only (ie, these participants are able to take part in other aspects of the trial, but will not be included in the PRO study) | 2 | May reduce the sample size/power of PRO study. Results may not be generalisable to all patients. | Hopwood (1998), Sprague (2003) | ||
| Baseline PRO completion (some sources recommended include baseline PRO completion as an eligibility criterion) | 29 | None | Bernhard, Cella (1998), Bernhard, Peterson (1998), Calvert (2004), Cella (1994), Cella (1995), Chassany (2002), Conroy (2003), Fayers (1997), Hayden (1993), Hopwood (1998), Hurny (1992), Kaasa (1998), Movsas (2003), Osoba (1992), Osoba (2007), Sadura (1992), Simes (1998), Sprangers (2002), Walker (2003), Young, Maher (1999), Young de Haes (1999) | ||
| Include patients with minimal level of impairment (as per baseline PRO) to ensure inclusion of patients with severe disease | 1 | May lead to selection bias | Chassany (2002) | ||
| May impact generalisability of results | |||||
| Surviving long enough to complete PROs (palliative care) | 3 | Difficult to estimate in some cases, so prognostic cues predictive of death may be more practical; may introduce selection bias. | Bakitas (2009), Jordhoy (1999), Chassany (2002) | ||
| Participants’ willingness to complete PROs | 3 | May result in selection bias; patients more willing to take part in PRO study may differ systematically from non-participants. | Fayers (1997), Sprague (2003) | ||
| Feasibility issues of PRO studies | Pilot study | Determine feasibility of PRO study (potential issues, resources required and/or sample size), and acceptability by conducting a pilot study | 9 | Requires time and resources | Cella (1994), Cella (1995), Groenvold (1999), Hurny (1992), Moinpour (1989), Kleinpell-Nowell (2000), Young, de Haes (1999), Sherman (2005), Wisniewski (2006) |
| Determine compliance targets by conducting a pilot study | 1 | Requires a long pilot study to determine; significant time and resources | Hahn (1998) | ||
| Conduct a pilot study to determine average time to complete PRO measures | 1 | Requires time and resources | Kleinpell-Nowell (2000) | ||
| Use the PRO pilot study as a training opportunity for less experienced staff | 1 | Requires time and resources | Cella (1995) | ||
| PRO resources | Ensure there is sufficient funding for the PRO study and that the PRO study is included in study budget | 5 | Funding can be difficult to obtain; however, it is possible to minimise costs of PRO studies at no cost to high-quality PRO research | Bernhard, Cella (1998), Cella (1995), Coates (1998), Gotay (2005), Moynihan (1998) | |
| Resource allocation—ensure recruiting sites are sufficiently resourced for the PRO study | 15 | Funding can be difficult to obtain across all sites especially if recruiting internationally or trans-nationally. | Bernhard, Cella (1998), Bernhard, Peterson (1998), Hayden (1993), Hopwood (1998), Hopwood (1996), Kaasa (1992), Moinpour (1998), Moynihan (1998), Revicki (2005), Scott (2004), Sprague (2003), Walker (2003), Wisniewski (2006), Young, de Haes (1999) | ||
| Ensure adequate staff at potential sites | 2 | Funding to employ new staff can be difficult to obtain | Revicki (2005), Scott (2004) | ||
| Minimise resources required for the PRO study | 4 | Care must be taken not to sacrifice quality of data or performance | Bernhard, Cella (1998), McMillan (2003) | ||
| Selection of recruiting sites | Select sites with good compliance record | 1 | May limit the number of participants recruited; may overly burden particular sites; potential for selection bias65 | Bernhard, Cella (1998) | |
| Select sites with adequate resources | 2 | May limit the number of participants recruited | Hurny (1992) | ||
| Sites with adequate resources may not necessarily be sites with best compliance record. | |||||
| Provide PRO-specific guidance for the research team | PRO administration guidance (for site staff) | General administration guidance aiming to standardise administration of PROs | 27 | None | Bernhard, Peterson (1998), Calvert (2004), Cella (1994), Cella (1995), Fayers (1997), Friedman (1998), Ganz (1988), Hahn (1998), Hayden (1993), Hopwood (1998), Kaasa (1998), Kaasa (1992), Land (2007), Newgard (2010), Osoba (1996), Osoba (1992), Sprangers (2002), Taphoorn (2010), Vantongelen (1989), Walker (2003), Wisniewski (2006) |
| Flexible processes across sites (There may be local variations in who is responsible for PRO data collection at different sites; therefore, procedures should be flexible to accommodate such differences.) | 1 | May introduce bias if procedures differ too much between recruiting sites | Bernhard, Peterson (1998) | ||
| Importance of complete data must be stressed in PRO administration guidance | 1 | None | Fayers (1997) | ||
| Instructions to give to participants must be specified in PRO administration guidance | 1 | None | Wisniewski (2006) | ||
| Procedures for missed assessments must be specified in PRO administration guidance | 1 | None | Calvert (2004) | ||
| Staff roles must be specified in PRO administration guidance | 3 | None | Poy (1993), Young de Haes (1999) | ||
| Procedures for handling special situations must be specified in PRO administration guidance | 5 | Not all difficult situations can be predicted in advance | Hahn (1998), Hopwood (1998), Hopwood (1996), Revicki (2005) | ||
| Protocol guidance60 61 63 | Follow PRO protocol guidance (investigators) | 2 | None | Bernhard, Cella (1998), Osoba (2007) | |
| Develop protocol guidance for investigators (trials groups) | 1 | None | Osoba (1996) | ||
| MOA (ie, is the questionnaire administered in hardcopy (pen and paper), electronically, over the phone, etc) | Choice of MOA | Consider costs involved with each MOA | 1 | None | Macefield (2013) |
| Consider impact of MOA on participants’ willingness to disclose information | 1 | The most acceptable MOA for participants may not be the most cost-effective or feasible | Hallum-Montes (2014) | ||
| Consider potential impact MOA on response rate | 3 | None | Hallum-Montes (2014), Cantrell (2007) | ||
| Consider inclusion of remote participants (web-based modes may be more accommodating to remote patients than face-to-face administration) | 1 | None | Cantrell (2007) | ||
| Mode preferred by sample | 1 | Requires additional pilot work to gauge participant preferences. Requires additional staff time and costs. Need to ensure equivalence of modes66 | Basch (2012) | ||
| Electronic modes of administration ‘e-PROs’, for example, using a computer, tablet, smart phone, etc | Allow participants to complete on their preferred electronic device | 1 | Requires resources to ensure compatibility of database across many types of electronic devices | Jansen (2013) | |
| Allows real-time compliance monitoring | 1 | None | Basch (2012) | ||
| Avoid fancy layouts | 1 | None | Cantrell (2007) | ||
| Avoid mandating completion of all items | 2 | May lead to missing item-level data if questions are of sensitive nature67 | Cantrell (2007), Hanscom (2002) | ||
| Present items one at a time | 1 | May be burdensome for participants considering cumulative time required to click between screens | Hanscom (2002) | ||
| Avoid question presentation one at a time (to reduce response burden) | 2 | None | Cantrell (2007), Hallum-Montes (2014) | ||
| Dialogue boxes for missed items | 1 | May be costly to develop | Wisniewski (2006) | ||
| Electronic dictation of questions | 1 | May be costly to develop | Hallum-Montes (2014) | ||
| Email PRO assessment reminders to participants | 1 | Requires time/resources to implement | Cantrell (2007) | ||
| e-PROs encouraged | 1 | e-PRO assessment may not be acceptable to some patient populations. | Basch (2012) | ||
| May be subject to technical fault/data protection/connectivity issues | |||||
| Keep assessment simple to reduce risk of technical fault | 1 | None | Hjermstad (2012) | ||
| Make all items mandatory | 1 | May lead to incomplete questionnaires if questions are of a sensitive nature | Cantrell (2007) | ||
| Flexible MOA | Follow-up missed assessments with alternate mode (eg, if participant misses a face-to-face visit in which hardcopy PRO assessment was scheduled, consider calling the participant to complete PRO over the phone, or posting the questionnaire to their home address with reply-paid envelope to return completed questionnaire) | 4 | Requires additional staff time and resources | Bernhard, Cella (1998), Blazeby (2003) | |
| Interview-administered questionnaires for very sick participants | 4 | Requires additional staff time | Kaasa (1998), Stewart (1992), Moynihan (1998), Chassany (2002) | ||
| Offer more than one MOA | 2 | May complicate data entry procedures or procedures for returning PRO data | Bernhard, Cella (1998), Gotay (2005) | ||
| Negotiate with the site as to their preferred MOA | 1 | May be infeasible to implement different modes between sites—some sites may have to compromise | Simes (1998) | ||
| Interview-administered MOA | Interview-administered MOA may improve response rates. | 1 | Requires additional staff time and resources | Fowler (1996) | |
| Postal MOA | Complete the baseline assessment in clinic and subsequent assessments by post | 1 | None | Kaasa (1998) | |
| Include postage-paid, self-addressed envelope for easy return of completed questionnaires (when using postal MOA) | 3 | Requires additional staff time and postage costs. May be burdensome for participants to send questionnaires back to researchers. | Kleinpell-Nowell (2000), Poulter (1997) | ||
| Patient burden—minimise | Minimise patient burden (general statement) | 8 | None | Aaronson (1990), Hahn (1998), Little, D’Agostino (2012), Macefield (2013), McMillan (2003), Revicki (2005), Walker (2003) | |
| Offer assistance to participants to complete PROs (to reduce burden PRO completion) | Additional assistance—childcare (offer to provide child care for participants’ children so that participants can attend clinic visits in which PRO assessments are scheduled) | 1 | Requires additional resources | Bell (2014) | |
| Additional assistance—travel (offer to arrange or fund travel of participants to the clinic for scheduled PRO assessments) | 1 | Requires additional resources | Bell (2014) | ||
| Avoid the need for a clinic visit where possible | 1 | May be difficult to engage participants away from the clinic | Little, Cohen (2012) | ||
| Offer assistance to complete questionnaire if needed | 1 | Requires additional staff time and resources | Sprague (2003) | ||
| Content | Clear/simple content and instructions of questionnaires | 1 | None | Young, de Haes (1999) | |
| Reduce overlap in questionnaire items | 3 | None | Fallowfield (1998), Walker (2003), Young, de Haes (1999) | ||
| Collect relevant PRO data only | 2 | None | Bernhard, Cella (1998), Little, Cohen (2012) | ||
| Format | Avoid using multiple questionnaires | 1 | None | Chassany (2002) | |
| Avoid written (free text) answers | 1 | None | Friedman (1998) | ||
| Clear/simple format | 6 | None | Conroy (2003), Little Cohen (2012), Kleinpell-Nowell (2000), Bernhard, Cella (1998), Revicki (2005), Sloan (2007) | ||
| Large/clear font | 1 | May increase printing costs if larger font adds pages to the questionnaire booklet | Fairclough (2010) | ||
| Professional format (eg, use study letterhead on printed questionnaires, use consistent formatting, etc) | 3 | None | Kleinpell-Nowell (2000), Revicki (2005), Sloan (2007) | ||
| Single-sided printing (some reports suggest that participants are more likely to overlook the underside of questionnaires printed double-sided) | 2 | Environmental burden. May increase printing costs due to additional pages in the questionnaire booklet | Fairclough (2010), Revicki (2005) | ||
| Uniform presentation format (a consistent formatting approach appears more professional and may be easier for participants to follow, potentially reducing risk of participants skipping items inadvertently or due to lack of understanding) | 2 | May not be possible if using more than one questionnaire | Bernhard, Peterson (1998), Hurny (1992) | ||
| Length of assessments | Consider participant health—sicker participants will not be able to complete long PRO assessments | 3 | None | Moinpour (1989), Stewart (1992), Young, de Haes (1999) | |
| Fewer assessment time points (ie, PRO assessments that occur regularly may be overly burdensome) | 10 | May sacrifice important information by assessing PRO less often | Bernhard, Cella (1998), Little, Cohen (2012), Chassany (2002), Ganz (1988), Jansen (2013), Revicki (2005), Fallowfield (1998), Hurny (1992), Hao (2010), Steinhauser (2006) | ||
| Fewer pages in e-PROs (eg, minimising the number of clicks between pages may reduce burden) | 1 | None | Cantrell (2007) | ||
| Shorter questionnaire | 18 | Limits the amount of information that can be assessed using PROs | Basch (2012), Basch (2014), Bell (2014), Bernhard, Cella (1998), Bernhard, Peterson (1998), Chassany (2002), Fairclough (2010), Hjermstad (2012), Hurny (1992), Moinpour (1989), Revicki (2005), Rock (2007), Sadura (1992), Siddiqui (2014), Young, de Haes (1999) | ||
| Use CAT/screening questions (allows for targeted question content and fewer items, to minimise burden) | 1 | Requires additional set-up costs. Can be difficult to introduce a second, non-electronic MOA if using CAT as questions administered will differ between participants | Hjermstad (2012) | ||
| Use validated questionnaires | Questionnaire items or formatting that participants find burdensome may be addressed in response to feedback obtained during questionnaire validation process | 1 | None | Kaasa (1992) | |
| Participant education and engagement (also see table 3) | Continued participant engagement—use strategies to keep participants engaged throughout the life of the study/trial | Adapt procedures to participant cultural group—conduct background research about the cultural groups involved | 2 | Requires time and resources | Wilcox (2001) |
| Participant incentives for participating/completing PRO questionnaires | Offer participants access to care via/after trial/study | 3 | Requires time and resources | Blazeby (2003), Little, Cohen (2012), Little D'Agnostino (2012) | |
| Offer participants financial incentives | 13 | Requires time and resources. Conflicting evidence about the effectiveness (in general population samples)68 and ethical issues in patient populations | Dykema (2012), Gates (2009), Jansen (2013), Kleinpell-Nowell (2000), Little, Cohen (2012), Meyers (2003), Sherman (2005) | ||
| Offer participants non-financial incentives | 8 | Requires time and resources. Conflicting evidence about the effectiveness (in general population samples)68 and ethical issues in patient populations | Dykema (2012), Little, Cohen (2012), Sherman (2005), Hellard (2001) | ||
| Reimburse participants for their time/costs involved in participating (factor into study budget) | 3 | Requires time and resources | Hellard (2001), Little, Cohen (2012), Senturia (1998) | ||
| Selecting a PRO measure | Acceptable measures for participants | 5 | None | Chassany (2002), Jordhoy (2010), Kaasa (1992), Revicki (2005) | |
| Clinically relevant measures (select PRO measures that are clinically appropriate, that is, include questions about relevant issues to specific disease/treatment) | 7 | None | Bernhard, Cella (1998), Friedman (1998), Ganz (2007), Gheorghe (2014), Hahn (1998), Revicki (2005) | ||
| Features to avoid in prospective PRO measures | Avoid overlapping content/highly correlated items | 2 | None | Beitz (1996), Taphoorn (2010) | |
| Avoid sensitive item content (ie, participants are more likely to skip items addressing sensitive issues such as sexuality or finances; so by avoiding such items you may minimise risk of missing PRO data) | 4 | Participants may have different views about what constitutes sensitive data. Some key issues for particular studies are considered sensitive, for example, sexual function | Fallowfield (1998), Jansen (2013), Pijls-Johannesma (2005), Simes (1998) | ||
| Translated (validated) questionnaires | 2 | Complicates trial set up and implementation, particularly when using e-PROs | Kaasa (1998), Kleinpell-Nowell (2000) | ||
| Validated measures (these are likely to be more clinically relevant and acceptable to patients) | 6 | None | Bernhard, Cella (1998), Blazeby (2003), Fallowfield (1998), Kaasa (1992), Siddiqui (2014) | ||
| Other | Ordering questionnaire items chronologically may speed up completion time and be easier for patients to complete | 1 | We strongly recommend that researchers do not change the item order of validated questionnaires. Questionnaires should be administered in the exact format as validated. | Dunn (2003) | |
| Strategies for measuring sensitive issues (please see Chassany 2002 for a description of various strategies) | 1 | None | Chassany (2002) | ||
| PROs part of trial/larger study | Research team should commit to the PRO substudy (eg, when part of larger trial) | 11 | Requires time and resources | Bernhard, Cella (1998), Bernhard, Peterson (1998), Cella (1994), Cella (1995), Chassany (2002), Hayden (1993), Kiebert (1998), Moynihan (1998) | |
| Incorporate PROs in trial/main study design | PROs should be a mandatory/integral part of the trial/ larger study (ie, PRO data are not an optional extra) | 10 | None | Aaronson (1990), Bernhard, Cella (1998), Hayden (1993), Hurny (1992), Kaasa (1992), Movsas (2004), Osoba (2007), Sadura (1992), Siddiqui (2014), Young, de Haes (1999) | |
| Consider logistic factors when designing PRO study | 4 | None | Chassany (2002), Little, D’Agostino (2012), Wisniewski (2006), Young, de Haes (1999) | ||
| PRO content in the study protocol60 61 63 | Define end points/hypotheses (ensure PRO end point is scientifically compelling) | 5 | None | Cella (1994), Fallowfield (2005), Little, Cohen (2012), Taphoorn (2010), Walker (2003) | |
| Specify how missing data will be handled | 1 | May not be possible to fully plan how missing data will be handled prospectively | Calvert (2004) | ||
| Specify the importance of PRO assessment compliance | 1 | None | Fayers (1997) | ||
| Include/plan PRO aspects of the study carefully | 13 | None | Bell (2014), Fayers (1997), Ganz (2007), Hahn (1998), Hao (2010), Land (2007), Moinpour (1998), Movsas (2003), Poy (1993), Revicki (2005), Sloan (2007),Walker (2003) | ||
| Specify plans for minimising missing data (such as those listed in this review) in the protocol | 11 | None | Beitz (1996), BIQSFP (2012), Calvert (2004), Fairclough (2010), Kaasa (1998), Moinpour (1998), Revicki (2005), Simes (1998), Young, de Haes (1999) | ||
| Specify PRO assessment schedule | 2 | None | Hopwood (1996), Moinpour (1998) | ||
| Specify the rationale for PRO assessment (understanding why PROs are being measured and the value the information will bring to the trial is useful for all members of the trial team, and reinforces the importance of high-quality PRO data collection) | 11 | None | Aaronson (1990), Bell (2014), Cella (1994), Cella (1995), Conroy (2003), Fayers (1997), Hopwood (1998), Sadura (1992) | ||
| Include PROs in the SAP† | Specify potential problems with PRO analysis in SAP | 2 | May not be possible to predict and prepare for all potential problems with PRO analysis when developing the SAP | Taphoorn (2010), Walker (2003) | |
| Plans for addressing missing data in SAP | 2 | May not be possible to fully plan how missing data will be handled prospectively | Bell (2014), Bernhard, Peterson (1998) | ||
| PROs in other trial/study documents | Include PRO study in relevant sections of procedural documents | 1 | None | Land (2007) | |
| QA | QA—planning ahead | Consider logistic factors when designing PRO study | 1 | None | Fallowfield (2005) |
| Create study databases with QA in mind (ie, consider how PRO data completion rates will be monitored using the database) | 5 | Requires time and resources | Bernhard, Cella (1998), Land (2007), Moinpour (1998), Wisniewski (2006) | ||
| Manage PROs with other trial/study end point data (ie, in a single database) | 2 | Data managers will require additional training for PROs—which requires additional time and resources | Bernhard, Peterson (1998), Hurny (1992) | ||
| Describe QA procedures in protocol | 3 | None | Cella (1995), Gheorghe (2014), Revicki (2005) | ||
| Specify QA procedures in a manual | 2 | None | Cella (1994),Cella (1995) | ||
| Establish target PRO compliance rates (ie, quotas that must be achieved, eg, a target of 95% indicates that no more than 5% of missing PRO questionnaires will be tolerated) | 6 | None | Hahn (1998), Little, Cohen (2012), Little, D’Agostino (2012), McMillan (2003), Sloan (2007) | ||
| Sample (for PRO data collection) | PRO subsample (if study power permits and if the study budget or logistics limit capacity to collect PROs from all participants, consider collecting PROs from a subsample only) | PRO data from representative subsample of the trial population | 2 | May be difficult administratively, particularly for site staff to implement | Simes (1998) |
| Do not collect PROs from patients with advanced disease | 1 | QOL issues are often of very important in patients with advanced disease. | Bernhard, Cella (1998) | ||
| Allow patients/sites to opt in to the PRO study | 1 | May lead to selection bias if sites or participants opt-in to PRO study | Simes (1998) | ||
| May also lead to impression that PRO study is of lesser importance than other study outcomes | |||||
| Recruit motivated patients only | 2 | May lead to selection bias if only motivated participants take part in PRO study | Bernhard, Cella (1998), Simes (1998) | ||
| Separate (additional) consent for PRO study | 1 | Requires additional time and resources | Simes (1998) | ||
| Sample size | Increase sample size to allow for attrition | 7 | The rate of missing data is important, regardless of whether the available data meet sample size requirements. Although increasing sample size will improve study power in the case of low PRO completion rates, the outcomes of participants with missing PRO data may differ to those with complete PRO data—which may lead to bias. | Altman (2007), Kaasa (2002), Little, D’Agostino (2012), Sherman (2005), Stewart (1992), Tang (2002), Jordhoy (2010) | |
| Team—design/protocol development | Involve committees (to review PRO study) | Ethics review | 1 | None | Movsas (2003) |
| PRO committee (ie, some trials groups have a dedicated PRO committee, comprised of PRO research specialists who review and provide feedback on PRO aspects of trials) | 6 | Requires access to a trials group with resources for a PRO committee | Hahn (1998), Osoba (1992), Osoba (2007), Revicki (2005) | ||
| Multidisciplinary team involved in design/protocol development (each party brings unique and complementary expertise and experiences to improve the design of the PRO study) | Involve a multidisciplinary team in PRO study design | 6 | None | Bernhard, Cella (1998), Cella (1994), Cella (1995), Kiebert (1998), Moinpour (1998) | |
| Involve experienced investigators in PRO study design (to offer strategies for maximising compliance, selection of informative measures and time points, and other key aspects of study design) | 2 | None | Little, Cohen (2012), Little, D’Agostino (2012) | ||
| Involve nurses in PRO study design (to offer expertise about patient experiences and relevant QOL issues, clinic environment, data collection, etc) | 1 | None | Hayden (1993) | ||
| Involve patients in PRO study design (to comment on the acceptability and relevance of PRO questionnaires, suitability of assessment time points in capturing desired outcomes, patient burden, strategies to educate and engage participants, and many other important aspect of study design) | 3 | None | Bernhard, Peterson (1998), Hurny (1992), Moynihan (1998) | ||
| Involve PRO experts in PRO study design (to offer strategies for maximising compliance, selection of informative measures and time points, analysis and interpretation strategies and other key aspects of study design) | 3 | None | Fallowfield (1998), Kiebert (1998), Basch (2014) | ||
| Involve site coordinators in PRO study design (to offer expertise about logistics of PRO assessment, patient experiences and relevant QOL issues, data collection strategies, etc) | 4 | None | Bernhard, Cella (1998), Hayden (1993), Larkin (2012), Moinpour (1998), Cella (1995) | ||
| Support the site staff | Minimise institution/staff burden (an overly burdensome PRO assessment schedule or procedure for site staff is likely to lead to high rates of missing data) | 6 | None | Aaronson (1990), Young, de Haes (1999) |
*Some sources may have provided a recommendation more than once.
†This review only covers proxy reporting as a strategy to facilitate interpretation of missing PRO data. If considering using proxies, please consult the literature for a review of additional challenges and implementation strategies.
CAT, computer-adaptive testing; ECOG, Eastern Cooperative Oncology Group; ePRO, PROs administered electronically; MOA, mode of administration; PRO, patient-reported outcome; QA, quality assurance; QOL, quality of life; SAP, statistical analysis plan.