The predictive value of a model can be confirmed by comparing its forecasts with real-world observations. The model was calibrated using HCV prevalence by age and gender in the year 2000, as reported by NHANES.1
The incidence was back-calculated and the model was used to fit reported prevalence in 2000. Total prevalence in other years, prevalence and incidence by sequelae, and mortality were calculated. A 2010 incidence of 16,020 (13,510-19,510) was forecasted versus the reported incidence of 17,000.29
The wide CI for incidence was driven by the large uncertainty in reported prevalence.1
According to the study by Davis et al.,15
HCV incidence peaked in 1989 when it was 11.5 times higher than the incidence in 1950. This corresponded to a peak incidence of 274,000 in a single year. A 2010 prevalence of 2.5 (2.1-3.2) million cases was estimated, matching the most recent NHANES data that showed 2.5 million cases in the 2009-2010.30
In comparison, Davis et al.15
reported an HCV prevalence of about 3.3 million in the same period.
Our analysis predicted that HCV prevalence in the US peaked in 1994 at 3.3 million viremic cases. The overall prevalence is declining, and the 2030 prevalence is expected to be one-third of the peak prevalence. Incidence has dropped significantly since its peak in 1989 due to the implementation of HCV antibody screening of the blood supply in 1992, with full implementation of universal donation screening for viral RNA through nucleic acid testing (NAT) in 1999,31,32
and to a decline in IDU.33
However, disease burden continues to grow. The dichotomy of HCV is that, while the overall number of infections is projected to decline, the number of individuals experiencing advanced liver diseases, liver related deaths, and healthcare costs are expected to increase. This was a key insight provided by this analysis.
A recent study by the CDC2
reported an increased recorded mortality rate in the US HCV-infected population in 1999-2007. Consistent with this study, we forecast that mortality will continue to increase and peak in 2020 (Supporting Appendix E
). After 2020, the decline in the number of HCV infections will outweigh the increase in background mortality, and liver-related deaths and the number of deaths will decrease. Mortality is projected to peak at ~69,440 deaths, with 29,650 deaths attributable to liver disease, including over 9,000 attributed to HCC in 2020.
As shown in , the incidence of more advanced liver diseases will continue to increase, with incidence of decompensated cirrhosis and HCC peaking in 2016-2017. However, not all infected individuals progress to the next stage, and the peak incidence is lower at each consecutive sequelae. The total prevalent population of each sequela is shown in . Over 50% of the HCV prevalent population resides in F0-F3 stage of the disease at any point in time. However, by 2030 compensated cirrhosis cases will account for 37% of all prevalent cases. The HCV compensated cirrhosis population is projected to peak in 2015, while the decompensated cirrhosis population will peak in 2019. A smaller portion of the HCV-infected population will go on to have HCC, but the size of this population does not grow substantially beyond 24,000 due to the very high mortality rate in this population.
A key observation was that peak healthcare costs lag peak prevalence by almost three decades. This is due to the time required for infected cases to progress to more advanced forms of liver disease, which are more expensive to treat.
Sensitivity analysis identified the key drivers of variance in peak healthcare cost. The incidence uncertainty (20,070-28,990), calculated from the uncertainty in NHANES 2000 prevalence, accounted for 52% of the variance in peak cost. Higher incidence led to more prevalent cases and higher cost. Uncertainty in the annual cost of diuretic sensitive ascites ($2,525-$29,860)17,18
accounted for 15% of the total variance. Finally, uncertainty in persistence (32%-80%)34,35
accounted for 13% of the variance. Higher persistence resulted in higher SVR and a greater number of cured patients, which in turn resulted in lower healthcare costs. This highlights the importance of SVR on future costs. In this study, the treatment cost was specifically excluded, and yet the SVR of historically treated cases still turned out to be important. The treated population had to be included in the disease progression portion of the model since it affected the size of prevalent populations. In 2002-2011, we estimated that 322,700 individuals were cured. If persistence in the real world were the same as observed in clinical trials (80%),35
the average SVR would be 46%, resulting in 430,000 cured cases in 2002-2011. This would result in a decrease of $1 billion dollars in peak healthcare costs.
Patients experiencing decompensated cirrhosis accounted for the majority of future costs. In 2011, it accounted for 40% of total costs, and by 2030 it accounted for 47%. This was followed by compensated cirrhosis (22% of 2011 and 19% of 2030 total cost) and HCC (15% of 2011 and 16% of 2030 total cost). The prevalence of decompensated cirrhosis was 20% of compensated cirrhosis, but the annual cost was 12 times higher.17
The average lifetime cost of a patient was estimated at $64,490 as compared to a recent study that reported an average cost of $19,660 per patient in 2002-2010 alone.17
The analysis of cost by age at infection demonstrates a link between life expectancy and healthcare cost. Individuals infected in the 1950s were expected to have lower lifetime costs due to lower life expectancy (and lower medical costs), while newly HCV-infected individuals are expected to cost the healthcare systems more due to the longer life expectancy. This highlights the continued importance of prevention as a means of managing future healthcare expenditure.
The effects of new therapies were excluded from our model. However, if the number of treated patients is doubled and kept constant at 126,000 per year in 2012-2030 and the average SVR is increased to 70%, the 2030 prevalent population is projected to be fewer than 100,000 cases. This illustrates that it is possible to substantially reduce HCV infection in the US through active management.
There were a number of limitations in this study that impact the accuracy of our base projections. There is strong evidence that progression transition rates change with age and gender. A single transition rate was used for all ages and genders. This led to a higher incidence/prevalence in early years and among females, as well as higher liver-related mortality among the younger age groups. However, the CIs in our study did capture uncertainty in the above assumptions.
The model does not explicitly account for alcohol consumption and metabolic syndrome. Frequent heavy intake of alcohol significantly increases fibrosis progression,36,37
and accelerated disease progression has been associated with metabolic syndrome.38,39
The model implicitly takes these factors into account, as the transition probabilities and sequelae cost incorporate some level of alcohol consumption and metabolic syndrome. If an increasing proportion of the prevalent population experiences heavy alcohol intake or metabolic syndrome, progression to advanced liver disease, and the associated costs, will likely increase.
The model does not take into account the persistent risk of fibrosis progression and liver cancer in virologically cured patients. Observational studies have demonstrated that most patients who achieve SVR experience stabilization or regression of fibrosis. After SVR, episodes of cirrhosis decompensation are extremely rare, and instances of HCC are likely to be small in number and not greatly impact overall disease burden or costs.40
A limitation of prevalence measures used in this analysis is that high prevalence populations may be undersampled through the NHANES.41
In particular, undersampling of veterans, prisoners, and the homeless would result in underestimation of the current prevalence, future disease, and cost burden. In addition, while IDU has declined from a peak in the 1970s, there is some evidence of a recent increase in IDU among middle-aged adults, potentially leading to a higher incidence of HCV.33
In all cases, the sequelae prevalence and the healthcare costs will be higher than the estimated base value.
A further limitation is that the model does not consider recent recommendations42
to implement birth cohort screening for HCV. Such screening could reduce the future incidence of advanced liver disease and associated costs, when infected individuals identified through screening receive appropriate treatment and achieve SVR.43
Treatment of HCV prior to 2002 was also ignored. The first pegylated interferon was launched in August of 2001, and the number of patients treated with pegylated interferons was small in that year. Prior to that launch, patients were treated with nonpegylated interferon. The number of individuals cured prior to 2001 was small, and their exclusion did not have a material impact on the outcome of the model.
The rate of SVR used in the model was derived from studies of treatment-naïve patients; however, average SVR is lower in treatment-experienced patients. Because the majority of treated patients are naïve, it is unlikely that the use of a single rate for SVR substantially impacted estimates of treated and cured patients beyond our CIs.
A final limitation is that the future cost of liver transplants is based on the assumption that transplantation will remain at the same rate as today. All other sequelae costs were determined as the result of the disease progression. The number of liver transplants, however, is determined by the clinical guidelines and availability of donors. Thus, the future costs associated with liver transplants could be higher if transplantation rates increase.
In conclusion, our analysis demonstrated that overall HCV prevalence in the US is in decline due to lower incidence. However, the prevalence of advanced liver disease will continue to increase, as will the corresponding healthcare costs. Lifetime healthcare costs for an HCV-infected person are significantly higher than for noninfected persons, and the expected cost is higher among populations with a higher life expectancy. Finally, it is possible to substantially reduce HCV infection in the US through active management.