Unresolved problems with distress screening

Abstract

Centre for Psycho-Oncology Research and Training, School of Public Health, The University of Hong Kong, 5/f,WMW Mong Building, 21, Sassoon Road, Pokfulam, Hong Kong, ChinaSir,The critique of Coyne, 2013 of the Carlson et al, 2012 paper andresponse by Carlson et al, 2013 raise important issues regardingscreening patients for cancer-related distress that have concernedus for some time. Relevant to the debate are the following:(1) The comment by Coyne, 2013 regarding declines in distressover time following cancer diagnosis and treatment is widelybelieved, but incorrect. Distress trajectories based on group meanvalues are problematic. The problem is, if in a study 50% of thesample score 10 out of 10 on a notional distress scale declining to 0out of 10 over time, whereas the other 50% score 0 out of 10increasing to 10 out of 10 over the same period, the observed groupmean will remain at 5 out of 10 and the conclusion would be thereis no change in distress. This of course is the wrong conclusion.Analyses using Mixture Growth modelling, a method thatdecomposes samples into distinct trajectories, reveal patterns ofdistress distinctly different to the accepted wisdom of a steadydecline over time (Helgeson et al, 2004; Deshields et al, 2006,Henselmans et al, 2010; Lam et al, 2010, 2011, 2012a, 2012b).These studies consistently show that a broad majority of cancerpatients, around 60% are resilient, experiencing persistently lowdistress with only marginal and transient peaks throughout thecancer trajectory. A second subset of patients who have low distressearly in the cancer trajectory report gradually increasing distresslevels that peak around the end of treatment and quickly declinethereafter. A third set follow the classic pattern mentioned byCoyne, 2013, with high levels early in the trajectory declining overtime thereafter. Finally, a fourth group, 5–20%, show stable levelsof high distress persisting over the duration of the cancertrajectory. (Henselmans et al, 2010; Lam et al, 2010, 2011, 2012a,2012b) Patterns of distress in the first year following diagnosispredict distress outcomes up to 6 years following diagnosis(Helgeson et al, 2004; Lam et al, 2011, 2012a). These patternsbroadly mirror the resilience model of response to trauma(Bonanno et al, 2011).(2) Evidence increasingly points to unresolved symptoms asbeing a major predictor of cancer-related distress trajectories (Lamet al 2010; 2012b). This suggests that more effective symptommanagement would be a more cost-effective approach to cancer-related distress. However, there are always going to be patients whoneed distress-support services and such patients need to beidentified. In particular, there is a pressing need to differentiate andidentify the chronic distress patients who would benefit most fromsupportive interventions from those with transient distress.(3) Given the above, the timing of any programme for distressscreening in cancer patients will give quite different resultsdepending on when screening is performed. Currently, evidencesuggests that the greatest proportion of patients (B80%)experience some signs of distress at around 1 month followingprimary treatment. Mostly, this distress is transient, and patientsand their families cope well with it; for some it is part of a longerdecline from an earlier peak, for others it is increasing, only todecline later. Only about 1 in 6 to 1 in 8 of those positivelyscreened as distressed would have chronic distress that is unlikelyto remit on its own. Any one point estimate will not tell you towhich group a screened distressed patient belongs and hencewhether they would, or would not benefit from additionalintervention: this is relevant to Coyne’s point about outcomes.Coyne suggests the effects reported by Carlson et al, 2012are due to recognised declines in distress over time. Of greaterconcern to us is that most patients show very little distress for mostof the cancer trajectory, although they might get picked up ifrepeatedly screened, during a fleeting peak of distress, probably atcritical points in the cancer trajectory (diagnosis, treatmentcessation, recurrence). Hence, including them in any randomisedcontrolled trial is unlikely to confer much benefit and, potentially,both dilute any group-intervention effects and increase dropout, aphenomena clearly evidenced in Carlson et al’s cohort wherearound one out of three patients dropped out. This is misleading,as for the chronically distressed group the benefits of supportiveinterventions may be more substantial than published studiessuggest.(4) It is well recognised that uptake and completion ofinterventions tends to be much greater in randomised controlledtrials than in everyday clinical practice. This is likely to be the casefor therapeutic interventions targeting cancer-related distress. Ineveryday practice, most people cope with cancer. The challenge isto identify and help those who don’t.(5) Frankly put, it seems that most cancer patients are resilientand it is not cost-effective to include them in distress-supportprogrammes, because they will simply not attend or drop out.Identifying the approximately one in eight distressed patients who