Objective: The objective of this study was to examine the relationship between a pharmaceutical product's impact on health-related quality of life (HRQOL) and differential product price. Methods: Design: An exhaustive search of the literature was conducted to acquire all HRQOL evaluations of pharmaceutical products that utilized a test-retest experimental approach. Data Collection: Effect sizes were calculated from the data extracted for 31 products. Average wholesale price for each product and similar products in the corresponding therapeutic class were collected, as well as the number of products in the class, availability of a generic and the percent of generics in the class, and whether the product was the lowest cost product in the class. Cost per day of therapy at the recommended starting dose and the ratio of the product cost to the lowest cost product in class were calculated. Analysis: Multivariate linear regression and analysis of variance models were constructed where either average wholesale price or cost per day of therapy was the dependent variable and effect size, therapeutic class, and other cost data were independent variables. Diagnostics were performed to verify model assumptions. Results: Using multivariate linear regression, the number of months on the market, ratio of drug price to lowest price drug in class, number of drugs in the class, and average effect size were significant, with a model R-square = 0.65. In the reduced model, the percent of generics in the therapeutic class was removed and the remaining independent variables were significant, with a model R-square = 0.61. Diagnostics revealed no violations of model assumptions. Conclusions: There is sufficient evidence to suggest that there is a direct positive relationship between a pharmaceutical product's ability to cause improvement in HRQOL and the price of the product. In addition, the number of products within a therapeutic class influenced drug price. This is interpreted as providing evidence for the validity of HRQOL measurement, and for the existence of product competition in the pharmaceutical industry. Further research should be conducted to evaluate the impact of prescription medications on HRQOL, and to identify and characterize the effects of drug and marketplace variables on drug prices.

Key Words

Health-related quality of life;

HRQOL;

Pharmaceutical price;

Linear regression;

New product competition;

Pharmaceutical industry

INTRODUCTION

This report details the results of a study that began as an intellectual exercise. The original question was, "How do you demonstrate the validity of HRQOL evaluation, as a technology?" In other words, how can the process of HRQOL evaluation itself", as opposed to a particular HRQOL instrument, be validated? Discussion about how to address that inquiry led to a decision to limit the exploration to pharmaceuticals.

In order to develop a conceptual framework, an attempt was made to identify and characterize relationships between pharmaceutical product variables, market variables, and price. The model that emerged was one that included relationships between a metric of a pharmaceutical product's impact on HRQOL, several market-related variables, and differential product price (expressed as average wholesale price). The HRQOL metric was the most powerful predictor of variance in the model.

The finding of a statistically significant, and rather substantial, model for this relationship is presented; it will be argued that this provides solid evidence for the validity and value of the use of HRQOL assessment, at least for pharmaceuticals. Finally, the implications of a direct relationship between the HRQOL metric used here and pharmaceutical product price will be discussed as they pertain to the concept of product competition (as opposed to price competition) in the pharmaceutical market.

BACKGROUND

Researchers are familiar with the necessity of evaluating the validity and reliability of measures used for research purposes. In the broad range of psychometrically-derived instruments, well-established procedures exist for determining that the instrument in use measures what it purports to measure (validity) and that the estimates obtained are free from measurement error (reliability). Estimates of convergent and divergent validity assessed via correlation, internal consistency reliability, or test-retest reliability are all examples of familiar techniques for establishing a specific instrument's suitability for use in research.

However, if the question asked is, "How is the validity of an entire field established?" rather than the validity of an individual instrument, no specific tests or procedures are readily apparent. Examination of the literature suggests that, instead of the application of some specific test, consensus on the validity of measurement techniques within a specific area emerges over time, resulting in the development of a "body of evidence" for the usefulness (and, therefore, acceptance) of a particular approach. In the face of what might be considered the slow adoption of HRQOL as a viable technology, the general question that initiated this research was, "Does any means exist for demonstrating the validity of HRQOL measurement as a whole?"

To investigate this question required the adoption of a perspective considerably removed from those used in individual instrument evaluation. What metric of HRQOL could be used across studies? To what would that metric be compared that may be construed as evidence for the validity of the HRQOL approach?

The selection of a metric was easily derived from the considerable literature examining the responsiveness of HRQOL instruments (1,2) A simple effect size, sometimes referred to as g, seemed most suitable (3). This unit-less metric, the change in score divided by the pooled baseline standard deviation of the subjects, would enable comparison of results across studies and instruments. This use of the simple effect size is congruent with its original derivation as a means of combining results from different measures employed in different studies in meta-analysis (4). Other means of calculating an effect size exist, but this form was selected because the information required for its calculation was more likely to exist in published reports than the data required for other forms of effect size appearing in the literature.

What then, is the meaning of an effect size? By definition, an effect size is the ratio of change in score to the inherent variance of the measure in the population under study. If a normal distribution were assumed, then an effect size of one would equate to a change in the mean for the sample such that the percentage of subjects below the baseline mean would change from 50% to 19%. Another way to think of the meaning of an effect size would be as the ratio of signal (change) to noise (inherent variance) as per Guyatt et. al. (5). In either conceptualization, change is expressed as a function of the variance in score, and not as a change in whatever scoring system the measure under study may employ. In general, then, larger effect sizes are indicative of a larger change, or that the observed change is likely to include a smaller proportion of measurement error as the effect size increases. Importantly, because the effect size is unit-less, direct comparison of the magnitude of change across different studies and different instruments is feasible.

If effect size is a measure of the magnitude of change over time, then what does an effect size of HRQOL mean? The general purpose of HRQOL measurement, to obtain some measure of outcomes from the patient perspective, might be conceived of as a systematized method for asking patients how they are doing. If this admittedly crude conceptualization can be accepted for the purposes of general discussion, then an effect size of HRQOL can be considered a measure of how much better the patients see themselves doing from one time period to the next.

Because the effect size is unit-less, this means that how much better (or worse) patients see themselves between time one and time two as measured with the Sickness Impact Profile can be compared to results in a different study utilizing the SF-36, the Asthma Quality of Life Scale, or any other quality of life instrument (6,7,8). Although comparison of results between two HRQOL instruments (for example, one designed for use in asthma and another designed for use in arthritis) would, in most cases, seem nonsensical, it is appropriate for the question under consideration here.