Introduction

During the last half century, life expectancy has been rising all over the world, which has resulted in increases in the frequency of surgery involving elderly patients [1, 2]. However, aging causes various physiological alterations, such as tissue fragility, metabolic dysfunction and deterioration of the immune response [3]. Although surgical outcomes have also improved during the last half century [4–6], the risks associated with surgery in elderly patients have not been fully elucidated.

Liver resection is often used to treat malignant tumors such as hepatocellular carcinoma, cholangiocellular carcinoma and metastatic liver cancer (e.g., from colorectal cancer) [7, 8]. The mortality rate of liver resection was reported to be less than a few percent in recent reports [9–12], although the morbidity rate was reported to range from 20 to 30 % [13–16]. The surgical outcomes of liver resection are largely dependent on the complexity of the procedure and the host liver function [8, 17]. Hepatocellular carcinoma usually develops in damaged livers, such as those of chronic hepatitis and liver cirrhosis patients, whose liver function has already deteriorated [17]. The histological background of the liver and liver function might also differ between young and old patients, which could affect the surgical outcomes. In addition, although metastatic liver tumors can develop in normal liver tissue, elderly patients often possess physiological defects that might make them more susceptible to adverse outcomes.

In this systematic review and meta-analysis, we searched the literature published from January 2000 to March 2013 to minimize the historical bias. Although a selection bias is inevitable during the selection of elderly patients, we considered that performing a cumulative meta-analysis might be the only strategy that would provide clear data on the outcomes of liver resection that would allow us to assess the effects of aging on such outcomes. Furthermore, as we considered that it was unlikely that there would be many randomized controlled trials (RCT) comparing the outcomes of liver resection in young and old patients, we selected the studies using the Methodological Index for Non-Randomized Studies (MINORS) scoring system [18]. The aim of this study was to examine the effects of aging on the outcome of liver resection by reviewing relevant papers published since 2000.

Patients and methods

Study selection (Fig. 1)

The methodology used for this study adhered to the guidelines outlined in the Preferred Reporting Items for Systematic reviews and Meta-analysis (PRISMA) statement (Fig. 1); [19, 20]. A search of all the published comparative studies, including evidence-based medicine reviews, examining the outcomes of young and old patients who underwent liver resection for malignant tumors was carried out using the PubMed and Ovid Medline databases. Studies that were published from January 2000 to March 2013 were reviewed, and the following “MeSH” search terms were used: “liver resection” and “hepatectomy.” In addition, the following text filters were applied: “elderly”, “aged” and “young”. Duplicate and non-English or non-human studies were excluded.

Fig. 1
figure 1

A flow chart showing how we conducted the literature search and the quality appraisal prior to the meta-analysis. A total of 16 studies were extracted based on our inclusion and exclusion criteria from among 24 full-text studies that met our Methodological Index for Non-Randomized Studies (MINORS) score criterion

Full size image

As a result, 24 full papers were extracted and had their eligibility assessed. All of the studies were non-randomized studies, and therefore, were evaluated using the modified MINORS scoring system [18]. Of the 24 papers, 16 met our selection criteria, and were included in the final analysis. Among these 16 studies, five examined patients with hepatocellular carcinoma (HCC) (Table 1); [9–12, 21], six investigated patients with colorectal metastatic (CRM) cancer (Table 2) [13–16, 22, 23] and five focused on patients with miscellaneous tumors (Table 3); [24–28].

Table 1 Characteristics and quality assessment scores of studies comparing young and old patients who underwent liver resection for hepatocellular carcinoma
Full size table
Table 2 Characteristics and quality assessment scores of studies comparing young and old patients who underwent liver resection for colorectal metastatic liver cancer
Full size table
Table 3 Characteristics and quality assessment scores of studies comparing young and old patients who underwent liver resection for miscellaneous mixed liver tumors, including primary liver cancer and secondary liver cancer
Full size table

Data extraction

Two reviewers (T.M. and M.K.) independently extracted the following parameters from each study: first author, year of publication, study population characteristics, study design, inclusion and exclusion criteria and matching criteria. There was 100 % agreement between the two reviewers.

Inclusion criteria

To be included in the analysis, each study had to have: (1) compared the outcomes of young and old patients who underwent liver resection for malignant tumors, (2) involved human subjects, (3) reported morbidity and mortality data, (4) been written in English and (5) been published in 2000 or later.

Exclusion criteria

Studies were excluded from the analysis if: (1) the outcomes of interest were not clearly reported, (2) it was impossible to extract or calculate the appropriate data from the published results, (3) they displayed considerable overlap with another study with regard to the authors, centers or patient cohorts evaluated.

Outcomes of interest and definitions

The following outcomes were compared between the young and old patients: the type of liver tumor, morbidity, mortality and the frequencies of single tumors and major resections.

Statistical analysis

A meta-analysis was carried out using the MedCalc software package (Ver 8.0.1.0, Mariakerke, Belgium) and a comprehensive meta-analysis software package (Biostat, Englewood, NJ). Statistical analyses of dichotomous variables were carried out using odds ratios (OR) as a summary statistic, and the data were reported together with 95 % confidence intervals (CI). The odds ratios reported in this paper represent the odds of an adverse event occurring in the old group compared with the young group. The Mantel–Haenszel method was used to combine the OR for the outcomes of interest using the “random effect” meta-analytical technique. Heterogeneity was assessed by graphic exploration, with funnel plots used to evaluate the publication bias.

Results

We reviewed comparative studies published since 2000 involving young and old patients who underwent liver resection, as described in the Methods section (Fig. 1). All studies that achieved less than 15 MINORS points were excluded; hence, a total of 16 studies were analyzed in this study [29–36]. The target tumors in each study were HCC in five studies (Table 1), CRM in six studies (Table 2) and miscellaneous tumors in five studies (Table 3).

Meta-analysis of the HCC studies

Although none of the five HCC studies were RCT [9–12, 21], one of the excluded studies included a propensity score matching-based analysis, which did not detect any differences in the clinical outcomes between the young and old patients [30]. The morbidity in the five included studies did not differ significantly between the old and young patients (Fig. 2a, P = 0.471), although heterogeneity was detected among the studies (Fig. 2b, P = 0.007). No differences in the mortality rates were detected between the groups (Fig. 3, P = 0.888), and the test for heterogeneity was negative (Fig. 3b, P = 0.219). The frequencies of single tumors (Fig. 4a, P = 0.774) and major hepatectomy (Fig. 5a, P = 0.630) did not differ significantly among the groups, and no heterogeneity was detected among the studies (Figs. 4b, P = 0.251; 5b, P = 0.079).

Fig. 2
figure 2

An annotated forest plot obtained via a meta-analysis of the morbidity in old vs. young hepatocellular carcinoma (HCC) patients (a). A Mantel–Haenszel fixed-effects model and random-effects model were used for the meta-analysis. The odds ratios (Odds) are shown together with 95 % confidence intervals (CI). An annotated funnel plot of the SE according to the log odds ratio for the meta-analysis of young vs. old HCC patients (b). Open circles show the original data, and the diamond below the figure indicates the overall mean, as well as the 95 % CI, of the standardized mean difference. The P values are for the heterogeneity test

Full size image
Fig. 3
figure 3

An annotated forest plot obtained via a meta-analysis of mortality in old vs. young hepatocellular carcinoma (HCC) patients (a). A Mantel–Haenszel fixed-effects model and random-effects model were used for the meta-analysis. The odds ratios (Odds) are shown together with the 95 % confidence intervals (CI). An annotated funnel plot of the SE according to the log odds ratio for the meta-analysis of young vs. old HCC patients (b). The open circles are the original data, and the diamond below the figure indicates the overall mean, as well as the 95 % CI, of the standardized mean difference. The P values are for the heterogeneity test

Full size image
Fig. 4
figure 4

An annotated forest plot obtained via a meta-analysis of the frequency of single tumors in old vs. young hepatocellular carcinoma (HCC) patients (a). A Mantel–Haenszel fixed-effects model and random-effects model were used for the meta-analysis. The odds ratios (Odds) are shown together with the 95 % confidence intervals (CI). An annotated funnel plot of the SE according to the log odds ratio for the meta-analysis of young vs. old HCC patients (b). The open circles are the original data, and the diamond below the figure indicates the overall mean, as well as the 95 % CI, of the standardized mean difference. The P values are for the heterogeneity test

Full size image
Fig. 5
figure 5

An annotated forest plot obtained via a meta-analysis of the frequency of major hepatectomy in old vs. young hepatocellular carcinoma (HCC) patients (a). A Mantel–Haenszel fixed-effects model and random-effects model were used for the meta-analysis. The odds ratios (Odds) are shown together with the 95 % confidence intervals (CI). An annotated funnel plot of the SE according to the log odds ratio for the meta-analysis of young vs. old HCC patients (b). The open circles are the original data, and the diamond below the figure indicates the overall mean, as well as the 95 % CI, of the standardized mean difference. The P values are for the heterogeneity test

Full size image

Meta-analysis of the colorectal metastatic liver cancer (CRM) studies

A total of six studies of CRM were eligible for the final meta-analysis (Table 2); [13–16, 22, 23]. The morbidity rate of the CRM patients did not differ significantly among the groups (Fig. 6a, P = 0.881), although heterogeneity was detected among the studies (Fig. 6b, P = 0.019). However, the mortality rate of the old patients was significantly higher than that of the young patients (Fig. 7a, OR: 2.662, P = 0.002), and the test for heterogeneity was negative (Fig. 7b, P = 0.143). Single tumors were significantly more common among the old patients than the young patients (Fig. 8a, OR: 1.310, P = 0.049), but heterogeneity was detected among the studies (Fig. 8b, P = 0.033). Conversely, the old patients underwent major hepatectomy significantly less frequently than did the young patients (Fig. 9a, OR: 0.812, P = 0.000), and no heterogeneity was detected among the studies (Fig. 9b, P = 0.938).

Fig. 6
figure 6

An annotated forest plot obtained via a meta-analysis of the morbidity in old vs. young colorectal metastatic liver cancer (CRM) patients (a). A Mantel–Haenszel fixed-effects model and random-effects model were used for the meta-analysis. The odds ratios (Odds) are shown together with the 95 % confidence intervals (CI). An annotated funnel plot of the SE according to log odds ratio for the meta-analysis of young vs. old CRM patients (b). The open circles are the original data, and the diamond below the figure indicates the overall mean, as well as the 95 % CI, of the standardized mean difference. The P values are for the heterogeneity test

Full size image
Fig. 7
figure 7

An annotated forest plot obtained via a meta-analysis of mortality in old vs. young colorectal metastatic liver cancer (CRM) patients (a). A Mantel–Haenszel fixed-effects model and random-effects model were used for the meta-analysis. The odds ratios (Odds) are shown together with the 95 % confidence intervals (CI). An annotated funnel plot of the SE according to the log odds ratio for the meta-analysis of young vs. old CRM patients (b). The open circles are the original data, and the diamond below the figure indicates the overall mean, as well as the 95 % CI, of the standardized mean difference. The P values are for the heterogeneity test

Full size image
Fig. 8
figure 8

An annotated forest plot obtained via a meta-analysis of the frequency of single tumors in old vs. young colorectal metastatic liver cancer (CRM) patients (a). A Mantel–Haenszel fixed-effects model and random-effects model were used for the meta-analysis. The odds ratios (Odds) are shown together with the 95 % confidence intervals (CI). An annotated funnel plot of the SE according to the log odds ratio for the meta-analysis of young vs. old CRM patients (b). The open circles are the original data, and the diamond below the figure indicates the overall mean, as well as the 95 % CI, of the standardized mean difference. The P values are for the heterogeneity test

Full size image
Fig. 9
figure 9

An annotated forest plot obtained via a meta-analysis of the frequency of major hepatectomy in old vs. young colorectal metastatic liver cancer (CRM) patients (a). A Mantel–Haenszel fixed-effects model and random-effects model were used for the meta-analysis. Odds ratios (Odds) are shown together with the 95 % confidence intervals (CI). An annotated funnel plot of the SE according to the log odds ratio for the meta-analysis of young vs. old CRM patients (b). The open circles are the original data, and the diamond below the figure indicates the overall mean, as well as the 95 % CI, of the standardized mean difference. The P values are for the heterogeneity test

Full size image

Meta-analysis of miscellaneous mixed tumors, including primary liver cancer and secondary liver cancer

A total of five studies of miscellaneous mixed tumors were eligible for the final meta-analysis (Table 3); [24–28]. In this group, all but one of the studies compared the clinical outcomes between old and young patients following major hepatectomy. The morbidity (Fig. 10a, P = 0.307) and mortality (Fig. 11a, P = 0.103) rates did not differ significantly among the old and young patients, and the test for heterogeneity was negative (Figs. 10b: P = 0.691 and 11b: P = 0.965, respectively).

Fig. 10
figure 10

An annotated forest plot obtained via a meta-analysis of the morbidity in old vs. young miscellaneous tumor patients (a). A Mantel–Haenszel fixed-effects model and random-effects model were used for the meta-analysis. Odds ratios (Odds) are shown together with the 95 % confidence intervals (CI). An annotated funnel plot of the SE according to the log odds ratio for the meta-analysis of young vs. old miscellaneous cancer patients (b). The open circles are the original data, and the diamond below the figure indicates the overall mean, as well as the 95 % CI, of the standardized mean difference. The P values are for the heterogeneity test

Full size image
Fig. 11
figure 11

An annotated forest plot obtained via a meta-analysis of the mortality in old vs. young miscellaneous tumor patients (a). A Mantel–Haenszel fixed-effects model and random-effects model were used for the meta-analysis. The odds ratios (Odds) are shown together with 95 % confidence intervals (CI). An annotated funnel plot of the SE according to the log odds ratio for the meta-analysis of young vs. old miscellaneous cancer patients (b). The open circles are original data, and the diamond below the figure indicates the overall mean, as well as the 95 % CI, of the standardized mean difference. The P values are for the heterogeneity test

Full size image

Discussion

We reviewed comparative studies published since 2000 involving old and young patients who underwent hepatectomy for various tumors and evaluated their findings in a meta-analysis. In addition, we also conducted the funnel plot analyses to determine the degree of heterogeneity among the studies analyzed. Although neither the morbidity nor mortality rate of the HCC patients and miscellaneous mixed tumor patients differed significantly among the old and young patients, those of the CRM patients differed among the age groups. This meta-analysis suggests that older age is associated with increased risks after liver resection in patients with CRM.

Definition of elderly patients

The definition of old patients varied among the reports. However, the cut-off age used in all of the HCC studies [9–12, 21], five of the six CRM studies [13–15, 22, 23] and three of the five miscellaneous tumor studies was 70 years old [24, 25, 28]. Although Melloul et al. [28] also found that there was no clear cut-off age for defining old patients, most studies published after 2000 defined old patients as those aged 70 years and over. The effect of aging on liver function is largely unknown, but the size of the liver and hepatic blood flow were previously to have negative correlations with age [13, 37, 38]. Although aging has inevitable physiological effects, a cut-off age of 70 might not be old enough to detect marked clinical effects of aging. The ideal cut-off age for examining the effect of aging might be ≥75 [16], but studies that use such cut-off values might have insufficient statistical power to detect clinical effects due to the limited numbers of patients that can be recruited, which likely explains why a cut-off age of 70 has been used in many studies of liver resection. If we could recruit a sufficient number of patients aged ≥75 years as elderly patients, we might obtain different results and see more marked effects of aging on the surgical outcomes.

Differences in the morbidity and mortality rates based on the tumor type

One of the interesting findings of this study was the differences among the various types of tumors. Although the morbidity and mortality rates of HCC and the miscellaneous tumors did not differ significantly between the young and old patients, those of CRM did. In the CRM analysis, it was found that the old patients exhibited a higher number of tumors, but underwent major hepatectomy less often than did the young patients. Therefore, in the cases of CRM, the higher mortality rate of the old patients might be related to tumor progression, rather than the extent of hepatectomy. On the other hand, in the cases of HCC, neither the number of tumors nor the percentage of patients selected for major hepatectomy differed between the groups. Therefore, the similarities in the oncological characteristics and surgical approaches of these cases might have resulted in the similar clinical outcomes seen in the two age groups. Another possible reason is that the mortality rate varies among the different types of tumors. The mean mortality rate of HCC was around 5–10 %, whereas that of CRM was only a few percent for the young patients and around 5 % for the old patients. The higher mortality rate of HCC might be associated with damage to the liver itself, e.g., due to chronic hepatitis or liver cirrhosis [5]. Although minor resections were preferred for HCC management, the risks associated with poor liver function might outweigh the age-related risk factors. Our meta-analysis clearly demonstrated that the clinical outcomes differed among the examined tumor types; however, the reason for these differences remains unclear.

Indications for hepatectomy in elderly patients

Age itself has never been considered a valid reason to change the basic surgical indications for any tumor [16, 22], although the mortality and morbidity rates of old patients might be higher than those of young patients. The evaluation and management of co-existing co-morbidities, which tend to become more common with age, plays a major role in achieving positive clinical outcomes [22]. Even in 75-year-old patients, survival for 2 years would be unlikely without surgical resection. Therefore, the surgical indications for old patients largely depend on their systemic physiological condition. The eligibility of elderly patients for liver resection should therefore be clarified in a future study by evaluating the effects of co-existing co-morbidities.

Besides the physiological condition, the psychological condition should be considered in the elderly patients. The risk factors for delirium have been reported to include an older age, poor liver function and advanced cancer stage [39]. Providing that an old patient is eligible for liver resection in terms of their physiological and psychological findings, a reasonable clinical outcome can be expected.

Implications for elderly patients who undergo hepatectomy

When interpreting the results of this meta-analysis, it is important to pay attention to the clinical profiles of each type of tumor. In HCC, the mean age of the old patients ranged from 72 to 75 years, whereas that of the young patients ranged from 35 to 60. In addition, the frequency of single tumors was 75 %, and that of major resection was about 20 %. On the other hand, in cases with CRM, the mean age of the old patients was between 73 and 77, whereas that of the young patients ranged from 57 to 62. The frequencies of single tumors and major hepatectomy were both 50 %. Therefore, HCC patients who are aged ≤75 years are more likely to have single tumors, and are eligible for minor resection that can provide similar clinical outcomes after hepatectomy to those obtained in younger patients. On the other hand, the older CRM patients, who were around 75 years old, are at a 2.7fold higher risk of mortality than patients aged around 60 years old, regardless of the number of tumors and the type of liver resection.

In conclusion, we reviewed studies published since 2000 that compared the outcomes of young and old patients who underwent hepatectomy. The rapid growth of the aged population requires clinicians to examine the surgical risk profiles of old patients. The most important finding of our study is that the morbidity and mortality rates after hepatectomy differ according to tumor type. Therefore, the indications for hepatectomy for old patients should be based on the type of tumor. Future studies should focus on elucidating the effects of aging on the patients’ systemic physiological profiles and the associations between these profiles and the oncological characteristics.