E. Personalized Health Care and HTA

Clinical trials and other studies that report only average treatment effects may be misleading. Therapies that yield a statistically significant treatment effect across a study population may not necessarily work for all treated patients; they may be ineffective for some patients and harmful for others. Also, therapies that do not yield a statistically significant treatment effect across a study population―and that may be dismissed as ineffective―may work for certain subsets of the population.

Personalized health care (a broader concept that personalized medicine) refers to the tailoring of health care to the particular traits (or circumstances or other characteristics) of a patient that influence response to a heath care intervention. These may include genetic, sociodemographic, clinical, behavioral, environmental, and other personal traits, as well as personal preferences. Personalized health care does not mean the creation of interventions that are unique to a patient; rather, it recognizes differences in how patient subgroups respond to particular interventions, and uses that information to treat individual patients.

Some examples of technologies used in personalized health care are:

• CYP2C9 and VKORC1 genetic testing for warfarin anticoagulation response for patients with atrial fibrillation, mechanical heart valves, deep vein thrombosis, etc.
• HER-2/neu receptor testing for trastuzumab for breast cancer
• BRCA 1,2 testing for pharmaceutical and surgical prevention options for and surveillance for breast cancer
• KRAS testing for use of EGFR inhibitors (e.g., cetuximab, panitumumab) for colon cancer
• Socioculturally-tailored therapy to treat certain ethnic minority patients with diabetes and depression (Ell 2011)
• Alternative procedure techniques (gastric banding, gastric bypass, etc.) for bariatric (morbid obesity) surgery
• Alternative types of coronary artery revascularization (e.g., coronary artery bypass graft surgery, percutaneous coronary interventions) for symptomatic ischemic coronary artery disease
• Alternative regimens to treat infertility

In order to support personalized health care, information is needed about how alternative technologies affect not just the average patient with a given health problem, but how those technologies affect various subgroups of patients with that health problem. For example, more specific information about how response differs by age, sex, comorbidities (other health problems), or genetic traits can be used by patients with a certain health problem who share those characteristics. Heterogeneity of treatment effects (HTEs) refers to the variation in patient responses observed across levels or types of patient characteristics such as these (Kravitz 2004).

HTA is increasingly oriented to identifying, assembling, and evaluating evidence on HTEs. This applies to all types of technologies, including screening, diagnostic, therapeutic, palliative, and others. Deriving findings about effects on subgroups depends in large part on the availability of data from studies that have been designed to detect such subgroup differences. This depends not only on how well the study population represents various subgroups, but whether the study designs and methods of subgroup analysis are capable of detecting HTEs for the patient subgroups of interest. For example, prospective subgroup analyses (i.e., identification of subgroups prior to rather than after data collection) tend to be more reliable than retrospective ones, and sample sizes for the subgroups under study must be large enough to detect true subgroup differences where they exist (Oxman 1992; (Wang 2007) Meta-analyses and other integrative methods may be used to pool subgroup data from different studies. HTA can also help to strengthen the evidence base for personalized health care by encouraging the development, validation, and use of patient-centered (including patient-reported) outcome measures; involvement of patients in planning clinical trials; and use of alternative data sources, such as health services utilization data (including from insurance claims) and patient registries, to help identify potential subgroup effects.