2.1 Data sources

This study cohort was identified from the Surveillance, Epidemiology, and End Results (SEER)—Medicare Health Outcomes Survey (MHOS) data linked to population-based data providing detailed information about older Americans with cancer. The MHOS, supported by the Centers for Medicare & Medicaid Services, is a yearly survey administered to a random sample of 1000 to 1200 Medicare beneficiaries from each managed care plan. The selected participants completed a survey at baseline and a 2-year follow-up survey if they still remained in the same plan. The SEER program, supported by the National Cancer Institute, collected information about patients with newly diagnosed cancer within specific geographic regions. A description of the SEER-MHOS data resource has been published previously.^{12, 13}

We included 15 MHOS cohorts with baseline assessments from the period 1998 to 2012 and follow-up assessments from the period 2000 to 2014. Response rates on the MHOS baseline survey, linked to SEER data, ranged from 66.1% in 1998 to 51.6% in 2012.¹⁴ Follow-up response rates ranged from 87.0% in 2000 to 70.3% in 2015, resulting in a sample size of 81,642 MHOS respondents linked to SEER across the 15 cohorts. The extent of potential nonresponse bias was studied previously, suggesting that response bias would be minimal in our study.¹⁵

To prospectively assess the impact of HCC diagnosis on HRQL, Medicare beneficiaries (aged ≥65 years) who developed HCC between the baseline and the follow-up MHOS are included. We identified 62 patients with HCC who completed both baseline and the follow-up MHOS. Controls were selected using propensity score-matched analysis (5:1 control to case ratio) from the same data resource and the same period of time from subjects without cancer but completed both baseline and follow-up MHOS. Matching was based on patient demographics, survey characteristics, and preexisting chronic medical conditions other than cancer. Our final study cohort included 365 control cases matched to 62 patients with HCC.

2.2 Data collection

The MHOS provides self-reported data on sociodemographic, survey characteristics, chronic medical conditions, and HRQL. HRQL was assessed using the Medical Outcomes Study Short Form-36 (SF-36) from 1998 to 2005 and the Veterans RAND 12-Item (VR-12) from 2006 to 2014. The MHOS used the algorithm to recode the 8 VR-12 scales to match the SF-36.¹⁶ SF-36/VR-12 is a generic HRQL instrument designed to assess well-being, functional status, and patient's perception of their health. It has eight subscales (Physical Functioning [PF], Role-Physical [RP], Bodily Pain [BP], General Health [GH], Vitality [VT], Social Functioning [SF], Mental Health [MH], Role-Emotional [RE]), two summary scores: Physical Component Summary (PCS) and Mental Component Summary (MCS). Each of the eight scale measures as well as two summary scales was transformed to a T-score metric, normed to the general the U.S. population, with a mean of 50, standard deviation of 10, and range of 0–100. Higher scores indicate better health. The minimal clinically important difference of 2 for summary measures and 4 for subscales was utilized.¹⁷

2.3 Statistical methods

We compared socio-demographic characteristics, preexisting comorbid conditions. and HRQL measures at baseline between HCC cases and controls using a non-parametric Kruskal–Wallis test for continuous variables and chi-square test for categorical variables.

To investigate whether there is a difference in HRQL after diagnosis of HCC, the longitudinal mixed models based on generalized estimating equations (GEE) were performed, allowing the adjustments for socio-demographic characteristics as well as time-variant covariates between surveys. The models included time-invariant covariates (sex, race, and education) as well as time-variant covariates: age, assessment mode, Medicaid, active smoker, and self-reported comorbid conditions (hypertension, heart disease, stroke, Crohn's disease, arthritis, sciatica, diabetes, and depression). Changes in HRQL measures between surveys were tested by the use of orthogonal contrasts using the fully adjusted model. The adjusted estimates of HRQL measures are reported in this paper since similar findings were observed in unadjusted estimates.

To understand the effect of time since HCC diagnosis on HRQL measures, time since diagnosis was categorized into two groups: time since diagnosis of ≤12 months and >12 months. Univariable logistic regression is applied for evaluating predictors of a meaningful decrease in HRQL. All analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC).

3.1 Baseline characteristics

Demographics characteristics and comorbidities for the HCC cases and the matched non-cancer controls were compared (Table 1). The mean (SD) age of the entire study population is 72.57 (9.19) years, 57% were Caucasians, and 40% were females. In the study group, 61% represented the western region of the United States, and 12% each from the northeast, Midwest, and eastern regions. 73% of the entire cohort had a history of hypertension, 60% had arthritis, and 26% had heart disease. 40% had a history of diabetes and 40% with depression diagnosis. 85% of surveys were assessed using the mailed method and remaining by phone. There were no significant differences between the two groups on demographics or comorbidities because of the propensity score matching (p > 0.05). The median (Interquartile range) time between surveys was 25.1 (23.8 to 26.5) months for HCC cases and 25.3 (24.4 to 27.1) months for matched controls.

3.2 Baseline SF-36 survey results

The adjusted mean of PCS and MCS for HCC cases and controls are presented in Table 2. There were no differences in the adjusted mean PCS between HCC cases and controls (44.4 [95% CI, 33.9 to 54.8] vs. 44.8 [34.8 to 54.7]) and MCS (53.7 [45.7 to 61.7] vs. 51.6 [44.0 to 59.2]), all p-values >0.50 at baseline. The same result was observed for the eight subscale measurements. These baseline means are consistent with the 1988 U.S. population norms for adults aged 65 years or over (PCS, males = 42.0 and females = 41.0; and MCS, males = 52.5 and females = 51.4).¹⁸ Similar patterns were observed in the unadjusted SF-36 scores at baseline (Table S1).

3.3 Change in HRQL over time (cases vs. controls)

Compared to their baseline HRQL scores, after diagnosis of HCC, subjects were more likely to report declines in PCS and MCS as compared to non-cancer controls (mean decline = −2.63 [−6.60 to 1.34] vs. 0.11 [−1.17 to 1.39], p = 0.198 and −6.36 [−9.94 to −2.77] and 0.02 [−1.25 to 1.29], p = 0.001, respectively) (Figure 1). These trends differed substantially across the eight subscale measures. Compared to the controls, the greatest declines after diagnosis of HCC were observed in social function/SF (mean decline = −7.05 [−11.41 to −2.68] vs. 0.33 [−1.24 to 1.89] p = 0.002), followed by general health/GH (−6.62 [−10.66 to −2.58] vs. −0.11[−1.45 to 1.24], p = 0.003), mental health/MH (−5.24 [−8.69 to −1.79] vs. −0.82 [−2.05 to 0.42], p = 0.018), role emotional/RE (−5.13 [−9.80 to −0.46] vs. 0.27 [−1.41 to 1.94], p = 0.033), and vitality/VT (−4.60 [−7.87 to −1.32] vs. −0.05 [−1.47 to 1.37], p = 0.013) scores (Table 2 and Figure 1). In contrast, no significant changes were noted in physical function/PF, role physical/RP, and bodily pain/BP (−2.87 [−7.53 to 1.80] vs. −0.99 [−2.52 to 0.54]; −2.16 [−6.62 to 2.30] vs. 1.28 [−0.20 to 2.75]; and −1.64 [−5.28 to 2.00] vs. −0.09 [−2.52 to 0.54], all p > 0.100). Notably, 52% HCC patients had more than 4-point decrease in PCS score and 61% on MCS (Table S2). This reflects a clinically meaningful decrease in score from baseline to follow-up. In the unadjusted model, HCC patients with diabetes and depression were more likely to have clinically meaningful declines in MCS (Odd ratio = 1.94 [0.60–6.31] and 4.02 [1.07–15.07]) (Data not shown). Given the small sample size, multivariable analysis was not considered.

3.4 Change in HRQL over time by time since diagnosis and by stages Among HCC cases

The median time from HCC diagnosis to follow-up completion of the survey was 8.5 months (Interquartile range, IQR [5 to 16 months]) (Tables 3 and 4).

Of the 62 HCC cases, although there is no statistical difference on the mean change in PCS as well as BP, PF, RP, and GH score by time since diagnosis (p > 0.400), numerically higher decline for physical health scales in HCC patients within 1 year of diagnoses versus post 1 year of diagnosis for all domains except bodily pain was found (PCS, −4.39 vs. −0.33; PF, −7.52 vs. −5.64; RP −2.09 vs. −1.44; and GH, −7.32 vs. −6.21). In contrast, the adjusted mean declines in MCS and MH score among HCC patients within 1 year of diagnosis and post 1 year of diagnosis were both clinically and statistically significant (MCS −0.24 [−4.72 to 4.25] vs. −10.16 [−14.66 to −5.66], p < 0.001; MH (0.53 [−3.77 to 4.83] and −8.03 [−12.16 to −3.91], p < 0.005). This clearly shows that the negative effects of HCC for all the physical health scales, except bodily pain, stabilized over time, whereas those on all the mental health scales continued to progressively decline over time. Staging was defined using Stage derived from the American Joint Committee on Cancer (AJCC) staging system. Of the 53 HCC cases (with AJCC stage available), compared to HCC patients with stage 1 or 2, HCC patients with stage 3 or 4 had no statistical difference but numerically higher adjusted mean declines in PCS (−0.49 vs. −2.47); RF (−0.26 vs. −4.59); GH (−4.95 vs. −7.5); MCS (−3.44 vs. −6.23); MH (−0.88 vs. −6.54); and SF (−2.78 vs. −6.17). In contrast, the adjusted mean declines in VT scores between stage 1 or 2 and stage 3 or 4 were both clinically and statistically significant (0.33 [−4.16 to 4.81] vs. −6.65 [−10.71 to −2.60], p = 0.017) (Table 4). The unadjusted mean HRQL scores by the AJCC stage are reported in Table S3.