Coverage of symptoms by the general PROs falls into two categories (Table 22.2). The FACT, PRO-CTCAE, and QLQ-C30 measure most symptoms used in general HRQOL studies, while PROMISE and EQ-5D measure only a small subset of those symptoms but complement them with additional assessments such as physical function and satisfaction with social roles and activities.

There have been difficulties in using PROs and PROMs on both the scientific and administrative fronts. Scientifically, many trialists wanted to measure an average QOL score at the beginning of the trial on all the patients, then compare this to a score 2 months later. At first, this was done by taking the average of all the patients’ scores at time zero and at 2 months, which inevitably showed no difference. This would be like taking the average of the whole group of tumor responses, rather than the changes in each individual’s cancer. Better reporting and better computer entry have fixed this issue.

Another scientific issue has been defining the minimal amount of change in QOL that is clinically significant. While even small benefits in overall and progression-free survival have been used to justify drug use, the amount of “improved quality of life” has less universal acceptance. An early conference [35] agreed on common metrics: first, ask the patients what a clinically important change score would be, for example, 20 % change in pain. Second, ascertain what percentage of patients achieved that level of change. Such easily understandable and reportable metrics have allowed comparison across studies.

An additional issue has been whether to use patient scores, clinician scores, or both; and if both, in what order, and whose scores take precedence. Current data show that patient symptom scores, in addition to clinician scores, more accurately predict survival of patients than either set of scores alone [36]. In a study of 161 lung cancer patients followed for 12 months, patient-reported symptoms and QOL with the Euro-QOL 5 were complementary to clinician-reported toxicity scores using the NCI CTAEC. The clinician-reported scores were more predictive of emergency room visits and death than patient-reported scores for fatigue (p < 0.001), nausea (p = 0.01), constipation (p = 0.038), and Karnofsky Performance Status (p < 0.001), but the patient scores more accurately portrayed day-to-day function [37]. (This study also suggests that when these patient-reported symptoms rise to the level of clinical recognition they are serious and should be addressed.)

The main administrative hurdle to use of PROs and PROMs in clinical trials arises because most trials are done by pharmaceutical companies to satisfy registration (new drug approval) requirements. To date there has been only one drug approved for cancer based on a purely palliative endpoint: mitoxantrone. All others have been approved based on changes in survival, some as small as 2 weeks (median change 0.33 months) for the addition of erlotinib to gemcitabine in advanced pancreas cancer [38]. Most trials report the PRO scores separately from the main study results as a complement; of 77 recent trials in castrate-resistant metastatic prostate cancer only 18 % had PRO or tolerability companion studies [39]. For instance, sipuleucel-T improved survival by about 4 months [40], and had predictable declines in QOL during treatment due to fatigue, but by week 26 and thereafter QOL was identical between treatment and placebo groups [41]. Until the FDA changes its rules, PROs and PROMs will be an “add on” to the trials rather than the main endpoint [42].

A second administrative hurdle is related to the electronic medical record, which should be able to integrate PROs with little or no dif-ficulty, and could then be used for comparative effectiveness research. For instance, does cabazitaxel when given to patients with co-morbidities such as congestive heart failure have the same favorable effects as in clinical trial subjects [43]? The main questions would be what to measure and record for universal access, within the bounds of what is possible and not too expensive. A recent review [44] explored the “architecture” of such decision-making and proposed some rationality before we attempt to add patient portals with multiple instruments, none of which are standardized, to the mix. This will require adoption of one or several instruments capable of cross-talking and interpretation, with heightened privacy requirements and more expense, when the EMR vendors are trying to carve out market share. Such improvements are possible but will likely require major governmental intervention similar to “meaningful use” requirements.