Differences in health state valuation for small, low-risk thyroid cancer between general population and cancer survivors: a cross-sectional analysis

Small (≤ 1 cm), well-differentiated papillary thyroid cancers are collectively termed papillary thyroid microcarcinomas (PTMC) and are characterized by slow growth and excellent prognosis [1]. Over the last 10 years, management strategies have expanded from surgical resection to include less invasive options like radiofrequency ablation and active surveillance [2, 3]. This shift emphasizes the importance of shared decision-making to choose the right treatment for each patient.

There is a prevailing impression among healthcare practitioners that patients prefer surgical resection for these low-risk cancers, however, whether this preference holds true in reality is unknown [4]. The gold standard measurement of health-related quality of life in health economics is health utility multiplied by time, expressed in quality adjusted life years (QALYs) [5]. Health utility is measured on a scale of 0 to 1, where 0 = death and 1 = perfect health. For thyroid cancer, health state utilities have historically been derived from quality-of-life instruments. These instruments have highly variable results and proven susceptibility to ceiling effect [6, 7].

Conventionally, health utility estimates are acquired from cohorts of general population volunteers, a practice that is founded in the ethical theory that, compared to patients, these individuals are less susceptible to bias and better represent taxpayers who ultimately contribute to payment of medical care [8‐10]. Additionally, “generic” estimates are helpful to compare between diseases. However, the derivation of health utilities from patients has been increasingly explored [11, 12]. Some advocate that the aforementioned “bias” that patients introduce into value assessments is not problematic but useful, as they bring a realistic perspective informed by their lived experience with disease [13]. Indeed, studies in breast cancer have shown that general population health utilities are lower relative to breast cancer survivors [14]. This finding has not been studied in thyroid cancer. Establishing this difference is important because utility estimates ultimately drive comparative- and cost-effectiveness studies that carry implications for societal resource allocation.

The purpose of this study was to compare health state utility estimates for low-risk papillary thyroid cancer as derived from general population volunteers and thyroid cancer survivors. We hypothesized that, similar to data from breast cancer, the general population would report lower health state utilities compared to thyroid cancer survivors.

To our knowledge, this is the first study to compare preference-based health utilities between the general population and thyroid cancer survivors, and the largest to use gold-standard methodology (TTO) to elicit preference-based health utilities within thyroid cancer. The use of TTO in this study is critical, as prior research in thyroid cancer has shown that traditional quality-of-life instruments such as the EQ-5D, SF-6D, and HUI are non-responsive in this condition and highly susceptible to ceiling effect [6]. This may in part be due to the fact that treatments for early-stage thyroid cancer are well-tolerated by most patients. Therefore, traditional components of health utility indexes—pain, fatigue, nausea, etc.—are common in many diseases but less relevant in thyroid cancer. Our data suggest that the general population consistently assigns lower utilities to small, low-risk thyroid cancer health states compared to thyroid cancer survivors, which is consistent with our study hypothesis and echoes similar findings from breast cancer studies. Additionally, as expected, complicated scenarios were assigned lower health utility scores compared to uncomplicated health states.

Differences between health utility assignments by the general population and by patients appear to vary across diseases. In one study using TTO, compared to colorectal and breast cancer patients, the general population assigned higher health utilities to the same SF-6Dv2-defined general health states, and this difference widened with poorer health states [17]. In contrast, a large meta-analysis in 2010 compared 30 studies reporting TTO data for various health states collected from patients and the general population. They found that there was no difference between health state utilities when TTO was used with hypothetical vignettes, but that patients assigned higher health utilities than the general population when the TTO was used with their own health state [18]. Conversely, Kaur et al. reported that general population volunteers undervalue breast cancer health states compared to cancer survivors [14]. Using a standard gamble method for prostate cancer health states, Gries et al. similarly reported a trend that the general public assigned lower health utilities to compared to cancer survivors [19]. Health economists should be aware of these established differences in valuations between patients and the general population when selecting a sample for the derivation of health utilities.

There has long been a debate about whose preferences ought to be utilized for cost-effectiveness analysis. Advocates of the utilization of public preferences emphasize the importance of capturing the societal perspective for resource allocation, the downplaying of impairment severity by patients due to their adaptation to it, and the insurance principle (e.g. individual, insured patients will not bear the cost of care and thus may be more willing to accept expensive, minimal gain therapies than is justified by the associated societal financial burden) [13, 20]. Advocates of the utilization of patient preferences believe the patients’ perspectives of their own lived experience is more informative and accurate than the public’s hypothetical perspective, citing the disability bias phenomenon (e.g. individuals with lived experience with a chronic illness or disability often rate their quality of life higher than those who have not experienced that health state) [13, 20, 21]. Others advocate that both public and patient preferences should be considered, either by creating separate cost effectiveness models or by providing the public with patient-derived preferences during their valuations [20, 22]. The consideration of both valuations may be ideal, but it may not always be practical to obtain preferences from more than one population. This debate has not reached a consensus; our team believes that health utilities should ideally be derived from samples of the population to which they are being applied, and that utilities from the general population should be utilized in health economics models with caution. Importantly, relative advantages of each study population may also vary depending on the disease being studied.

There are several limitations to this study. First, both of our cohorts are subject to sampling bias, which may limit the validity of the derived utilities. On average, our thyroid cancer survivor participants were older, had more advanced disease, and underwent more extensive surgery than is typical of patients with PTMC [23, 24]. Accruing an adequately powered PTMC-exclusive survivor cohort would require a multi-institutional collaborative network. Nevertheless, given the excellent prognosis of non-metastatic thyroid cancer and favorable quality of life among treated patients, we believe that our data should approximate those of a PTMC cohort [25‐27]. Regarding the general population, our sample approximated 2020 census data for age (median: 39 years, though this does not exclude children), income (national median income: $78,719), marital status (51% in our sample married vs. 47% nationally), employment status (69% vs. 61% employed), living in non-rural environments (87% vs. 80%), and White race (81% vs. 75%) [28]. Some features were over-represented such as college degree (71% vs. 35%) and non-Hispanic ethnicity (94% vs. 80%) [28]. This comparison suggests that our sample may not be fully generalizable, though the direction of the potential bias this may have introduced to our utility data is uncertain. Both the general population and cancer survivor groups had substantial non-response rates, which likely contributes to the limited generalizability noted above. Relatedly, there were demographic differences between our study’s participant cohorts. Prior research suggests that higher socioeconomic status (as is the case in our cancer survivor sample) is associated with higher health utility valuations, and that older age (as is the case in our cancer survivor sample) may be associated with lower health utility values [29, 30]. We adjusted for age and income in our multivariable regression model and neither were significantly associated with reported utility. Third, we designed the TTO task based on previously published thyroid cancer utility values, which are predominantly > 0.8 [7]. Therefore, we designed our task with a range of 0.5 to 1.0 to reduce survey fatigue. The potential floor effect of this approach was mitigated by participants’ ability to manually enter any number between 0 and 10 years. Still, this approach may have biased our utilities towards higher values. Fourth, our study vignettes, though well-validated, by nature cannot encompass all aspects of each health state [16, 31]. Examples of missing information include out-of-pocket costs and adverse effects of potential future treatments. Finally, the TTO technique itself has inherent limitations, such as a more pronounced ceiling effect among participants who are parents, the confounding effect of participants whose own life expectancy is less than 10 years, difficulties of cognitive understanding the trade-off task, and interviewer effect [32‐34]. We mitigated the latter two by (a) providing a tutorial and a proctor, and (b) by standardizing training across proctors.

Acknowledging these limitations, this study aligns with prior literature that suggests health utilities may vary by the population from which they are derived. This phenomenon has now been demonstrated for breast cancer and low-risk thyroid cancer health states; future research could apply a similar approach to other disease states such as more advanced thyroid cancers, other malignancies, and benign diseases. A mixed-methods investigation of the drivers of the difference in utility between the general population and cancer survivors could also elucidate which cancer health states may be appropriately approximated by data from a general population sample. Health services researchers should consider the source of utilities integrated into the cost- and comparative-effectiveness models they create and evaluate, as models may be sensitive to the utility differences incurred by utilizing different derivation populations.

For 10 low-risk thyroid cancer health states, general population participants reported significantly lower health utility values than thyroid cancer survivors when adjusting for age, sex, and income. These differences should be considered when integrating health utilities into comparative-effectiveness research.