The authors are grateful to the Scottish Government for funding the Scottish Hepatitis C Clinical Database.
Our objective was to address two shortfalls in the hepatitis C virus (HCV) literature: (1) Few data exist comparing post-treatment liver-related mortality/morbidity in HCV-sustained virologic response (SVR) patients to non-SVR patients and (2) no data exist examining liver-related morbidity among treatment response subgroups, particularly among noncirrhotic SVR patients, a group who in the main are discharged from care without further follow-up. A retrospective cohort of 1,215 previously naïve HCV interferon patients (treated 1996-2007) was derived using HCV clinical databases from nine Scottish clinics. Patients were followed up post-treatment for a mean of 5.3 years. (1) By Cox-regression, liver-related hospital episodes (adjusted hazard ratio [AHR]: 0.22; 95% confidence interval [CI]: 0.15-0.34) and liver-related mortality (AHR: 0.22; 95% CI: 0.09-0.58) were significantly lower in SVR patients, compared to non-SVR patients. (2) Rates of liver-related hospitalization were elevated among all treatment subgroups, compared to the general population: Among noncirrhotic SVR patients, adjusted standardized morbidity ratio (SMBR) up to 5.9 (95% CI: 4.5-8.0); among all SVR patients, SMBR up to 10.5 (95% CI 8.7-12.9); and among non-SVR patients, SMBR up to 53.2 (95% CI: 49.4-57.2). Considerable elevation was also noted among patients who have spontaneously resolved their HCV infection (a control group used to gauge the extent to which lifestyle factors, and not chronic HCV, can contribute to liver-related morbidity), with SMBR up to 26.8 (95% CI: 25.3-28.3). Conclusions: (1) Patients achieving an SVR were more than four times less likely to be hospitalized, or die for a liver-related reason, than non-SVR patients and (2) although discharged, noncirrhotic SVR patients harbor a disproportionate burden of liver-related morbidity (i.e., up to six times that of the general population). Furthermore, alarming levels of liver-related morbidity in spontaneous resolvers is an important finding warranting further study. (HEPATOLOGY 2011;)
With an estimated worldwide prevalence of 2.35%1 (equating to 160 million persons chronically infected), the hepatitis C virus (HCV) presents a global public health challenge. In cohorts of HCV patients, up to 24% develop liver cirrhosis within 20 years of infection,2 and thereafter, the annual incidence of hepatocellular carcinoma, and decompensated cirrhosis, is 3.5% and 6.5%, respectively.3 In Scotland, McDonald et al. found liver-related mortality (LRM) to be 25 times higher4 and liver-related morbidity to be 41 times higher5 among persons diagnosed anti-HCV positive, compared to the general population.
Chronic HCV infection can, however, be treated via interferon (IFN)-based therapy; the optimal outcome is a sustained virologic response (SVR) (defined by the absence of viral RNA for at least 6 months after termination of treatment). However, the current treatment regimen (pegylated IFN and ribavirin for 24-48 weeks) is far from ideal, given that (1) it is expensive (UK drug costs are between $7,900 to $11,000 for 24 weeks, and $15,800 to $22,000 for 48 weeks of therapy6), (2) 84% of initiates experience at least one treatment-related adverse effect,7 severe enough to cause 11%-14% of patients to discontinue therapy prematurely,8-11 and (3) there are limited success rates, such as in clinical trials, with more than 50% of genotype (GT) 1 patients (who account for an estimated 45% of infections in the United Kingdom12 and >70% in the United States13) failing to achieve an SVR.8-10 When justifying the resourcing and widespread use of treatment, these drawbacks should be considered in the context of the improvement in prognosis that an SVR brings. However, there is a surprising lack of robust data in the scientific literature quantifying the benefit of this improvement in prognosis, particularly in terms of all-cause and non-liver-related outcomes. Further, although SVR patients without cirrhosis tend to be discharged from care without further follow-up (FU), liver-related morbidity in this population, relative to the general population, until now, remains entirely unexplored.
Thus, through linkage of health databases in Scotland, we examined the post-treatment risk of LRM and liver-related hospital episodes among HCV patients treated with an IFN-based regimen at nine clinics in Scotland, between 1996 and 2007. To explore the onward care needs of noncirrhotic SVR patients, we additionally utilized liver-related hospital episode data in both (1) the underlying Scottish population and (2) Scotland's HCV-diagnosed, spontaneously resolved population (a group more commensurable to HCV treatment patients than the general population, in terms of key liver-disease–related lifestyle factors). Our principal aims, relating to the first 5 years post-treatment, were 2-fold: (1) to quantify the benefit of attaining a SVR (relative to non-SVR) and (2) explore excess liver-related morbidity, particularly among noncirrhotic SVR patients.
The Scottish HCV clinical database, collated and held at Health Protection Scotland (Glasgow, Scotland UK), contains clinical FU data on HCV-infected patients attending antiviral treatment clinics in Scotland. This includes a history of antiviral treatment episodes. In addition to the clinical database, Health Protection Scotland holds a national HCV diagnosis database, containing data on all new anti-HCV positive diagnoses (from 1991 onward) and subsequent polymerase chain reaction (PCR) tests conducted on these individuals (from 1994 onward). These PCR tests can be used to distinguish persons who have spontaneously resolved HCV from those who have developed chronic infection. The diagnosis database (up to March 2009) was linked to Scottish death and hospital discharge records (both up to December 2009), in line with methods described by McDonald et al.4, 5 To obtain mortality and morbidity data on treated patients, a dataset of these individuals derived from the clinical database was linked probabilistically to the diagnosis database (itself linked to Scottish death and hospital admissions records, as previously described), using the following set of common identifiers: forename initial, surname soundex, date of birth, gender, district of residence, and hospital reference number.
A retrospective cohort of treatment patients (N = 1,400) were included in these analyses, if treatment naïve, and initiated their first course of IFN-based therapy (initiated defined as receiving at least 1 dose of therapy) at one of nine Scottish treatment clinics (Gartnavel General Hospital, Glasgow Royal Infirmary, Southern General Hospital, Ninewells Hospital, Monklands Hospital, Royal Infirmary Edinburgh, Crosshouse Hospital, Stirling Royal Infirmary, and Victoria Hospital) since January 1, 1996 (relating to the introduction of comprehensive PCR testing to gauge SVR status), and had terminated their first course of therapy before July 1, 2007 (to allow a minimum of 1 year of potential FU after the determination of SVR status at 6 months post-treatment termination). In total, 185 of 1,400 (13%) patients were later excluded (Fig. 1), including 45 SVR patients, who, although indicated on our treatment database to be SVRs, were discovered to have had at least one PCR-positive test post-treatment recorded in the national HCV diagnosis database (N.B. in a sensitivity analysis, whereby these 45 patients were retained in the cohort; the interpretation of our results did not change; see Discussion). Thus, the number of patients considered in our final analyses was 1,215.
Furthermore, to treatment patients, persons diagnosed with HCV antibodies in Scotland between January 1, 1996 and December 31, 2008, who have subsequently been tested at least once for viral RNA (but have never tested positive) and have no record of an IFN-based treatment episode in Scotland (as determined from the HCV clinical database) were, in these analyses, considered to be spontaneous resolvers of HCV (N = 3,690).
The two outcomes of primary interest were LRM and liver-related hospital episodes. Hospital episodes were used as a measure of morbidity; thus, we use “morbidity” and “hospital episodes” interchangeably. A hospital episode is defined as an unbroken period spent as an inpatient, regardless of change in consultant, significant facility, speciality, and/or hospital. As previously described by McDonald et al.,4, 5 a liver-related death or hospital episode was defined on the basis of International Classification of Disease (ICD)-9 or -10 codes (Table 1. Hospital episodes were considered to be liver-related under two scenarios, on the basis of either (1) the main discharge code(s) only (i.e., if a liver-related discharge code was present in the main position of any of the admissions underlying the episode) or (2) all discharge codes (i.e., if a liver-related discharge code was present in either the main or supplementary position of any of the admissions underlying the episode).
Table 1. International Classification of Disease (ICD) 9/10 Codes Used in Analyses to Define (1) Liver-Related Hospital Episode and (2) Alcohol-Related Hospital Episode
No corresponding ICD-9 code.
Liver-related hospital episode (principal outcome variable of interest)
Disease of the liver (relating to alcohol)
Disease of the liver (not relating to alcohol)
570, 571.4-571.9, 572-573
Sequelae of viral hepatitis
B94.2, R17, R18, I85.0, I98.2
Alcohol-related hospital episode (used to derive time-dependent exposure variable)
The primary exposure variable of interest for treatment patients was a SVR (SVR is the optimum virological outcome of treatment). SVR (and non-SVR) was defined as PCR negative (versus PCR positive) for viral RNA at least 6 months after termination of treatment. Other exposure variables considered in these analyses were the following: gender, age at study entry, ethnicity, ever injected drugs, genotype, diagnosed cirrhotic at study entry, alcohol-related hospitalization, and mean post-treatment alanine aminotransferase (ALT). A diagnosis of cirrhosis was made on the basis of one or more of the following: (1) liver biopsy, (2) radiology, (3) endoscopy, (4) laboratory tests, and (5) clinical examination. Patients' mean post-treatment ALT was calculated from values obtained 0-6 months after terminating therapy.
Alcohol-related hospital episodes were used as a proxy indicator of excessive alcohol consumption. The definition of an alcohol-related hospital episode was based on the presence of one or more of the ICD-9 and -10 discharge codes (in either the main or supplementary discharge diagnosis positions) outlined in Table 1. Using past (i.e., occurring before the start of FU) and current (i.e., occurring during FU) alcohol-related hospital episodes, a time-dependent variable with three states was created, relating to risk periods: (1) before a current alcohol hospital episode (if any) in patients without a past alcohol-related hospital episode; (2) before a current alcohol-related hospital episode (if any) in patients with a past alcohol-related hospital episode; and (3) after a current alcohol-related hospital episode. Under this time-dependent variable, a patient can only move from state 1 to state 3 or from state 2 to state 3.
Calculating FU Time: Mortality Data.
Each treatment patient's FU time began 6 months from being administered their final treatment dose (ensuring consistent comparability between SVR and non-SVR treatment groups) and ended at the date of liver-related death (i.e., primary end-point) or censoring date. FU time was censored for the following: (1) death for non-liver-related causes (3%; 33 of 1,215) or (2) reaching the right censor date (January 1, 2009) (93%; 1,127 of 1,215).
Calculating FU Time: Morbidity Data.
In recognition of multiple episodes per subject, observations were divided into distinct risk periods. The first risk period began 6 months after terminating therapy. Subsequent risk periods (if any) began with the discharge date from a previous liver-related hospital episode occurring during FU. All risk periods ended at the admission date of a liver-related hospital episode (if any) or the censoring date. FU time was censored for the following: (1) all cause death (7%; 88 of 1,215) or (2) reaching the right censor date (January 1, 2009) (93%; 1,127 of 1,215).
For spontaneous resolvers, observations were analogously divided into distinct risk periods. The first risk period began 30 days after the date of HCV diagnosis (to reduce bias resulting from an increased risk of hospitalization around the time of diagnosis4, 5). Subsequent risk periods (if any) began with the discharge date from a previous liver-related hospital episode occurring during FU. All risk periods ended at the admission date of a liver-related hospital episode (if any) or the censoring date. FU time was censored for the following: (1) all-cause death (8%; 279 of 3,690) of patients or (2) reaching the right censor date (January 1, 2009) (92%; 3,411 of 3,690).
To meet our prestated objectives, two statistical approaches were adopted: (1) Cox regression and (2) calculation of standardized morbidity ratios (SMBRs).
Cox regression was used to determine the association between a SVR and the risk of liver-related mortality and hospitalization after adjustment for gender, age at study entry, ethnicity, ever injector, genotype, diagnosed cirrhotic, alcohol-related hospital episodes (according to the time-dependent variable previously described), and mean post-treatment ALT, as appropriate. Variables crudely associated with time to outcome (at the <10% threshold level) were selected for inclusion and retention in multivariate models. Hospital episodes (unlike mortality) are recurrent events (i.e., can be experienced multiple times during FU). Thus, to account for within-patient clustering in the likelihood of a hospital episode, an Andersen-Gill model for recurrent events, with robust variance estimation, was used. The proportional hazards assumption, underwriting Cox regression, was checked graphically (via Nelson-Aalen plots) and quantitatively (by calculating Schoenfeld residuals).
Rates of (1) liver-related hospital episodes, (2) non-liver-related hospital episodes, (3) alcohol-related hospital episodes, and (4) all-cause hospital episodes, for Scotland's general population, were obtained from the Information Services Division, National Services Scotland. These rates, stratified by age (<30, 30-39, 40-49, 50-59, and 60+ years), gender, and calendar year (1996-1998, 1999-2001, 2002-2004, 2005-2007, and 2008-2009), were compared to corresponding stratified rates in subgroups of our HCV treatment cohort (subgroups were noncirrhotic SVR patients, all SVR patients, and all non-SVR patients) and spontaneous resolver cohorts via the calculation of SMBRs and their associated 95% confidence intervals (CIs). In this way, morbidity in our treatment cohort was compared to the underlying Scottish population after adjustment for age, gender, and calendar year. Standardized mortality ratios were not determined as the observed number of liver-related deaths in noncirrhotic SVR patients (the subgroup of principle interest) would have been too small to draw meaningful comparisons with the general population. All SMBRs were calculated according to all hospital episode discharge codes (i.e. all main discharge code(s) and all supplementary codes), but also through restricting to the main discharge code(s) only.
Sample Characteristics and FU Duration.
In the final cohort, 46% (560 of 1,215) attained an SVR (Table 2). Sixty-nine percent (843 of 1,215) were male, and the mean age of the cohort at study entry was 41.8 years (standard deviation [SD], 9.7). Almost half of this cohort had ever injected drugs (49%; 596 of 1,215), and 14% (173 of 1,215) had been diagnosed with cirrhosis.The majority of patients were treated with pegylated and ribavirin combination therapy (with ribavirin: 735 of 1,215, 61%; without ribavirin: 11 of 1,215, 1%). The remainder was treated with standard IFN therapy (with ribavirin: 250 of 1,215, 21%; without ribavirin: 219 of 1,215, 18%). The contribution of each clinic to the final cohort was not uniform: For example, patients of the Royal Edinburgh Infirmary, Gartnavel General, Glasgow Royal Infirmary, and Ninewells Hospital made up 79% (954 of 1,215) of our treatment cohort. Treatment patients were followed up for a mean duration of 5.32 years (range: 27 days to 12.4 years).
Table 2. Description of Final Analysis Cohort* in Terms of Key Variables by Treatment SVR Status
Furthermore, a total of 3,690 spontaneous resolvers were identified. Their mean age at study entry (30 days after diagnosis of HCV antibodies) was 32.9 years (SD, 10.0), 57.5% (2,123 of 3,690) were male, and 62.8% (2,317 of 3,690) had ever injected drugs. The mean follow-up duration of this cohort was 5.7 years (range: 3 days to 12.9 years).
Mortality and Hospital Episode Rates of Treatment Patients.
A total of 2,962 hospital episodes were observed during FU of our treatment cohort. Of these, 1,005 (34%) were liver-related (based on main and supplementary discharge codes) hospital episodes (103 episodes from 47 SVR patients and 902 episodes from 266 non-SVR patients). Eighty-eight patients died during FU, of which 55 deaths (63%) were liver related (based on main and supplementary discharge codes).
Rates of liver-related, non-liver-related, and all-cause outcomes were all lower among SVR patients, compared to non-SVR patients (Table 3). This was most apparent, however, for liver-related outcomes (hospital episode crude hazard ratio [CHR]: 0.20; 95% CI: 0.13-030; mortality CHR: 0.19; 95% CI: 0.08-0.48).
Table 3. Mortality Rates and Hospital Episode Rates (Per 100 Person Years) by SVR Status Observed Among 1,215 Post-Treatment HCV Patients in Scotland, 1996-2007
Hospital Episodes (Including Multiple Episodes Per Patient)
In univariate analyses of treatment patients (Table 4), variables were significantly associated (P < 0.10) with time to a liver-related hospital episode (defined on the basis of main and supplementary discharge codes), and thus included in multivariate analyses, were as follows: SVR, age group at study entry, Asian ethnicity, ever injector, diagnosed cirrhotic, alcohol-related hospitalization, and mean ALT post-treatment >50 IU/L. Variables not significantly associated with time to a liver-related hospital episode in univariate analyses included gender and genotype. In multivariate regression (Table 4), individuals with a significantly reduced risk of a liver-related hospital episode included those with an SVR, compared to a non-SVR (adjusted hazard ratio [AHR]: 0.22; 95% CI: 0.15-0.34), and those who had ever injected, compared to never injected (0.70; 95% CI: 0.50-0.98). Although those with a significantly increased risk of a liver-related hospital episode included those older in age at study entry (30-39 years, compared to <30; 1.68; 95% CI: 0.89-3.19; 40-49 years, compared to <30: 2.39; 95% CI: 1.33-4.32; 50-59 years, compared to <30: 2.81; 95% CI: 1.46-5.40; and >=60 years, compared to <30: 2.84; 95% CI: 1.36-5.94), of Asian ethnicity (2.13; 95% CI: 1.27-3.58), diagnosed cirrhotic (3.38; 95% CI: 2.42-4.71), and with an alcohol-related hospitalization during FU (4.27; 95% CI: 2.69-6.77). Our results did not significantly differ when a liver-related hospital episode was defined on the basis of main discharge code(s) only.
Table 4. Results of Andersen-Gill Cox Regression of the Patient Factors Associated With Time to Liver-Related Hospital Episodes Among 1,215 Post-Treatment HCV Patients in Scotland, 1996-2009
In univariate analyses of treatment patients (Table 5), variables significantly were associated (P < 0.10) with time to a liver-related death (defined on the basis of codes for main and supplementary causes), and thus included in multivariate analyses, were as follows: SVR, age at study entry, diagnosed cirrhotic, and alcohol-related hospitalization. Given that there were only 55 liver-related deaths in our cohort, age at study entry was modeled as a linear variable to prevent overfitting (a likelihood ratio test comparing inclusion of age as a categorical versus continuous variable was not significant; P = 0.35).
Table 5. Results of Cox Regression of the Patient Factors Associated With Time to Liver-Related Mortality Among 1,215 Post-Treatment HCV Patients in Scotland, 1996-2009
Outcome of Liver-Related Mortality
Hazard Ratio (HR)
95% CI (Lower)
95% CI (Upper)
95% CI (Lower)
95% CI (Upper)
“\” means parameter estimate could not be derived because of there being no outcome events among persons of other ethnicity.
In multivariate regression (Table 5), individuals with a significantly reduced risk of a liver-related death included those with an SVR, compared to a non-SVR (AHR: 0.22; 95% CI: 0.09-0.58), whereas those with a significantly increased risk of a liver-related hospital episode included those older in age at study entry (linear increase over <30, 30-39, 40-49, 50-59, and >=60 years age group categories: 1.70; 95% CI: 1.27-2.29), diagnosed cirrhotic (3.63; 95% CI: 1.99-6.60), and with an alcohol-related hospitalization during FU (6.82; 95% CI: 3.79-12.26). Our results did not significantly differ when a liver-related death was defined on the basis of the main cause of death only.
Adjusted liver-related SMBRs (Tables 6 and 7) were higher when the main and supplementary discharge codes were collectively considered, compared to when only the main discharge code was considered. Adjusted liver-related SMBRs were highest among individuals with a non-SVR—up to 53 (based on main and supplementary codes: 53.17; 95% CI: 49.43-57.23) times greater than that of the general Scottish population. They were lowest among noncirrhotic SVR patients, but still between two times (based on main discharge code[s] only: 2.19; 95% CI: 1.12-4.92) and six times (based on main and supplementary codes: 5.92; 95% CI: 4.49-7.95) greater than the general Scottish population. Furthermore, there was no evidence that the risk of an alcohol-related hospital episode in noncirrhotic SVR patients differed from that of the general population (based on main and supplementary codes: 1.26; 95% CI: 0.89-1.84).
Table 6. Age, Sex, and Calendar Period Adjusted Standardized Morbidity Ratios (SMBRs) Comparing (1) Liver-Related, (2) Non-Liver-Related, (3) Alcohol-Related Hospital, and (4) All-Cause Hospital Episodes in Hepatitis C Treatment Patients, Compared to the Underlying Scottish Population
Standard error for 95% CI was derived using the jackknifing technique.
Abbreviations: ICD, International Classification of Disease; CI, confidence interval; SVR, sustained virologic response.
Noncirrhotic SVR patients (N = 503)
Main and supplementary
All SVR (N = 560)
Main and supplementary
All non SVR (N = 638)
Main and supplementary
Table 7. Age, Sex, and Calendar Period Adjusted Standardized Morbidity Ratios (SMBRs) Comparing (1) Liver-Related, (2) Non-Liver-Related, (3) Alcohol-Related, and (4) All-Cause Hospital Episodes in Persons Known to Have Spontaneously Resolved HCV Infection in Scotland (i.e., diagnosed HCV Ab+ve, PCR -ve) (N = 3,690) to That of the Underlying Scottish Population
Standard error for 95% CI was derived using the jackknifing technique.
Abbreviations: HCV, hepatitis C virus; PCR, polymerase chain reaction; ICD, International Classification of Disease; CI, confidence interval; IDU, injection drug use.
All resolved patients (N = 3,690)
Main and supplementary
Resolved and known IDU (N = 2,317)
Main and supplementary
Resolved and not known IDU (N = 1,373)
Main and supplementary
The risk of a liver-related hospital episode in patients who had spontaneously resolved their HCV infection was between 18 (based on main discharge code[s] only: 18.25; 95% CI: 16.52-20.20) and 27 (based on main and supplementary discharge codes: 26.75; 95% CI: 25.29-28.31) times greater than that of the general Scottish population. Furthermore, their risk of an alcohol-related hospitalization was up to 10 times higher than the general population (based on main and supplementary codes: 9.50; 95% CI: 8.64-10.48).
In terms of non-liver-related morbidity, SMBRs for non-liver-related hospital episodes in all SVR patients were between 29% (based on main and supplementary discharge codes) and 41% (based on main code[s] only) lower than that of non-SVR patients.
In a post-HCV treatment cohort with a mean patient FU of 5.3 years, our analyses show that treatment-naïve patients attaining a SVR were five times less likely both to die a liver-related death (AHR: 0.22; 95% CI: 0.09-0.58) and experience a liver-related hospital episode (0.22; 95% CI: 0.15-0.34), compared to patients not attaining an SVR. The size of this SVR effect was considerable and is consistent with other studies.14-16 Furthermore, our observations of reduced all-cause and non-liver-related SMBRs among SVR patients (compared to non-SVR patients) extends the findings of Backus et al., who reported the SVR to be associated with reduced all-cause mortality.17 Given that the durability of an SVR has been shown not to vary according to treatment type,18 the impending introduction of novel treatment regimens should not outdate these findings. Future work will, however, be required to explore whether the magnitude of this SVR effects changes over longer periods of FU time (i.e., beyond 5 years post-treatment).
Our finding that noncirrhotic SVR patients (a group who, in the main, are discharged from clinical care without further FU) have liver-related morbidity two to six times higher than the general population is important. This excess morbidity, in the main, may relate to the following: (1) liver damage (i.e., mild to moderate fibrosis) incurred before SVR, that has not fully ameliorated, and/or (2) post-SVR progression of liver disease through exposure to liver-disease–related lifestyle factors, which will not be accounted for by merely adjusting SMBRs for age, gender, and calendar period.
Compared to the general population, persons ever infected with HCV are a chaotic group. For example, in Scotland, it has been previously shown that (1) 57% of all HCV-diagnosed persons have ever injected drugs, representing 89% of those with a known risk factor12 (this is in stark contrast to the Scottish general population, where an estimated 0.76% ever injected drugs19), and (2) 29% of injectors drink alcohol to excess (personal communication; Maureen O'Leary, 2011). We, therefore, surmised a priori that such lifestyle disparities between HCV patients and the general population would likely not be resolved in SMBRs adjusted merely for age, sex, and calendar period. Thus, given that (1) spontaneous resolvers of HCV typically harbor viral RNA for less than 1 year20 and (2) median duration of HCV infection for progression to cirrhosis is 30 years,21 HCV-induced liver damage in this population should be negligible, and thus any liver damage apparent should be largely attributable to lifestyle factors (and not past HCV infection), we chose to explore excess morbidity among spontaneous resolvers to gauge the extent to which lifestyle factors in themselves can cause liver damage.
On this basis, although the rate of liver-related hospital episodes (compared to the general population), in noncirrhotic SVR patients, were two to six times higher, this rate was far greater (i.e., 18-27 times) among Scotland's spontaneous resolvers. However, as our data indicate considerably higher alcohol consumption among spontaneous resolvers, compared to noncirrhotic SVR patients, ultimately, it is difficult to tease out the extent to which excess morbidity observed in noncirrhotic SVR patients (our principal treatment subgroup of interest) could be attributed to previous chronic HCV infection versus lifestyle factors instead.
Nevertheless, our finding that noncirrhotic SVR patients still harbor a disproportionate burden of liver disease, and still (despite attaining the optimal virological outcome of treatment and despite no evidence of excessive alcohol consumption) incur a healthcare cost (irrespective of whether attributable to previous chronic HCV infection or not), is of key importance to healthcare policy makers. Perhaps of even more significance, these analyses surreptitiously highlight alarming levels of liver-related morbidity in spontaneously resolved patients, likely fueled by alcohol—a novel observation. More work is required to elucidate the independent contribution of chronic HCV to liver damage to ensure that what appears to be a sizeable baseline risk (particularly in injecting drug users) is accounted for particularly in studies of the cost effectiveness of HCV treatment.
There are several noteworthy limitations to highlight. In relation to the classification of patients as “spontaneous resolvers,” spontaneous resolvers could, in reality, be chronically infected with HCV, if, after initially resolving, they were reinfected with HCV and thence developed chronic infection, but were never retested for viral RNA. However, it is unlikely that any such patients would be overly contributing to liver-related hospital episodes, as clinicians would surely retest for viral RNA if such a patient was admitted to hospital for a liver-related cause. Furthermore, chronic infection following reinfection may be less likely for past spontaneous resolvers.22
In this analysis, patients were considered cirrhotic if their clinician had diagnosed them as such by the time FU was commenced. The presence (or absence) of such a diagnosis was available for all persons in this treatment cohort. Current guidelines suggest that a liver biopsy may be unnecessary in selected patients.23, 24 In our cohort, 38% (394 of 1,042) of noncirrhotic patients and 49% (85 of 173) of cirrhotic patients had a record of a liver biopsy held on the Scottish clinical database, highlighting that in the routine clinical setting, liver biopsies are administered to patients judiciously and not routinely (as is common in clinical trials). Nevertheless, because (1) the liver biopsy is considered the gold-standard means of determining liver cirrhosis and (2) up to 62% of noncirrhotic patients did not receive at least one liver biopsy within 2 years of study entry, the accuracy of the clinicians' diagnosis is called into question. Selective use of the liver biopsy could (to a greater or lesser extent) explain the noted excess liver-related morbidity of our noncirrhotic SVR subgroup. However, if it is the case that cirrhosis in SVR patients is frequently missed, and such persons are accordingly discharged, and go on to disproportionately contribute to the excess morbidity of discharged SVR patients, then this highlights an important point: Better methods to diagnose cirrhosis (particularly compensated) in the routine clinical setting are required.
A proportion of patients initially attaining a non-SVR were retreated during FU (182 of 655; 33%). Thus, some non-SVR patients (for a proportion of their FU time) were, in fact, negative for viral RNA, either temporarily (through a transient response attained during retreatment) or permanently (through having attained a SVR upon retreatment). However, the proportion of FU time under which a SVR through retreatment had been attained in our non-SVR cohort was minimal (∼6%).
Finally, results of PCR tests performed in Scotland (for viral HCV RNA) are held in the national HCV diagnosis database. We examined the test history of SVR patients in the period after termination of treatment. On this basis, we identified and subsequently excluded 45 SVR patients who, although were indicated to have attained an SVR (from the clinical database), had at least one positive test record for viral RNA after terminating treatment (from the national HCV diagnosis database). In 14 of these SVR patients (with a positive result in the first 6 months after terminating treatment), this must be attributable to incorrect classification of SVR on the HCV clinical database. For the remaining 31 patients, reinfection, or late viral relapse, are other possible explanations.25 We performed a sensitivity analysis, whereby the 14 cases of possible incorrect SVR classification were retained and treated as non-SVR patients, and the 31 cases of possible reinfection/late viral relapse were retained and considered as SVR patients. In this analysis, adjusted log hazard ratios (for SVR versus non-SVR) and adjusted SMBRs (for SVR subgroups) differed by less than 8% from the results presented. Thus, our decision to omit these 45 patients does not undermine our principal conclusions. Finally, it is important to note that cross-checking SVR status against national PCR data is a diligent check not performed in similar studies, to date.14-17
In conclusion, compared to patients with chronic HCV, an SVR is associated with a considerable clinical benefit in the first 5 years post-treatment. However, healthcare planners and patients alike should be aware that although discharged from clinical care, noncirrhotic SVR patients still harbor a disproportionate burden of liver morbidity, relative to the general population.
Participating members of the Hepatitis C Clinical Database Monitoring Committee during 2010-2011 were as follows: Bill Carmen, John Dillon, Ray Fox, Andrew Fraser, David Goldberg, Peter Hayes, Sharon Hutchinson, Hamish Innes, Nick Kennedy, Peter Mills, Adrian Stanley, and David Wilkes. The Hepatitis C Clinical Database Monitoring Committee would like to extend their thanks to Elaine Cadzow, Fiona Elliot, Susan Gilfillan, Jane Holmes, Shirley McLeary, Wendy Mitchell, Grace Thomson, and Toni Williams for their roles in the maintenance of the data included in these analyses. The authors thank also Toby Delahooke for his role in the design of the Scottish hepatitis C Clinical Database. Finally, the authors are grateful to the Scottish government for funding the Scottish Hepatitis C Clinical Database.