Introduction

Gastrointestinal cancers (GIC) account for over one-quarter of all cancer cases and one-third of cancer-associated deaths worldwide [1]. Although there has been great advancement in the treatment of GIC, the outcome for the majority of GIC patients remains poor [2]. Thus, exploring a reliable prognostic index for patient survival can enable physicians to adopt better therapeutic and preventative measures.

Numerous studies in recent years have confirmed that systemic inflammation is crucial for the development and growth of GIC [3, 4]. A variety of inflammatory cells and proinflammatory cytokines are activated in the early stages of carcinogenesis, which promote the creation of lymphatic ducts and new blood vessels, causing a pro-cancer microenvironment for growth and differentiation [5]. At later stages, cancer-related inflammation can impair immune cell function, creating a conducive environment for metastasis [6]. Thus, inflammatory indicators are anticipated to be important prognostic biomarkers in cancer. For instance, an elevated neutrophil-to-lymphocyte ratio (NLR) is linked to a weak immunological response and a high inflammatory response [7,8,9]. In cancer patients, the nutritional status of the body is also closely associated with tumor development and clinical outcome. Some common nutritional indicators have been shown to have a high prognostic significance in cancer, such as body mass index (BMI) [10] and serum albumin level [11].

Recently, the advanced lung cancer inflammation index (ALI), a new inflammatory marker that is calculated as BMI (kg/m2) × albumin (g/dL)/NLR, was initially found to be a useful prognostic index in lung cancer [12]. ALI is thought to reflect systemic inflammation better than other biomarkers due to combining the indicators of nutrition and inflammation. To date, some retrospective articles have analyzed the association between ALI and prognosis in GIC patients. However, there has not been a systematic evaluation of whether ALI is a reliable predictive factor for GIC patients. Thus, we conducted the first meta-analysis to identify the predictive significance of pre-treatment ALI in GIC patients, which may help to determine prognosis and formulate an effective treatment strategy that will further minimize mortality.

Methods

Literature search strategies

The current meta-analysis accompanied the PRISMA statement [13]. The protocol for this meta-analysis was available in PROSPERO (CRD42022371374). On December 29, 2022, PubMed, EMBASE, and the Cochrane Library were retrieved using the keyword: “advanced lung cancer inflammation index [All Fields]”. We further searched Google Scholar for grey literature. Additionally, we manually retrieved the reference lists of the publications that qualified.

Inclusion and exclusion criteria

If studies met all the following criteria, they were included: patients diagnosed with GIC; research evaluated the prognostic value of ALI; provided at least one of the outcomes [overall survival (OS), disease-free survival (DFS), progression-free survival (PFS), and cancer-specific survival (CSS)]. The conference abstracts, case reports, or comments were excluded.

Data extraction and quality assessment

Data extraction mainly focused on the author, year, study region, study design, study period, sample size, the number of male and female patients, cancer types, treatment, follow-up duration, cut-off, and outcomes. The Newcastle–Ottawa Scale (NOS) score was utilized to evaluate the quality of the observational studies. High-quality literature was defined as having a score above six. All of the above steps were double-checked by Lilong Zhang and Kailiang Zhao, and any disparities were addressed by Weixing Wang and Wenhong Deng.

Statistical methods

Statistical analysis was conducted by Stata 15.0. The statistical heterogeneity was calculated using the chi-squared test. P < 0.1 and I2 > 50% indicated high heterogeneity, so a random effect model was applied; otherwise, the fixed effect model was used. The tests of Egger’s and Begg’s were employed to evaluate publication bias. If there was significant publication bias, we used the trim-and-fill method to modify the results [14]. Sensitivity analysis was implemented to assess the stability of the results by excluding each study independently.

Results

Characteristics of studies

After the initial search, 67 duplicate studies were removed. Then there were 339 articles deleted after carefully reading the titles and abstracts. Later, the full texts of the remaining 46 articles were further assessed. 18 articles involving 6899 patients were ultimately included [15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32]. The PRISMA flow diagram is provided in Fig. 1.

Fig. 1
figure 1

The flow diagram of identifying eligible studies

Full size image

The main characteristics of the studies included are shown in Table 1. Of the 18 studies, seven were on colorectal cancer (CRC), three on gastric cancer (GC), two on esophageal cancer (EC), two on pancreatic cancer (PC), one on hepatocellular carcinoma (HCC), one on oral cavity cancer (OCC), and one on cholangiocarcinoma (CC). Besides, the study by Ruan et al. included 270 patients with CRC, 245 patients with GC, 145 patients with EC, and 31 patients with hepatobiliary cancer. Eleven studies were carried out in China and four in Japan, plus one each in Austria, Turkey, and Korea. The cutoff point of the ALI was reported as ranging from 13.2–70.4. The NOS scores for 18 articles ranged from 6–8, which represented a low risk of bias (Table 1).

Table 1 Main characteristics of the studies included
Full size table

ALI and overall survival

In total, 16 articles involving 6177 patients explored the association between ALI and OS in cancer patients. The pooled HR was 1.914 (95% CI: 1.514–2.419, P < 0.001), implying that low ALI raised death risk by 91.4% (Fig. 2). Since there was significant heterogeneity, a random effects model was used (I2 = 88.4%, P < 0.001).

Fig. 2
figure 2

Forest plot of the advanced lung cancer inflammation index in relation to overall survival. HR, hazard ratio; CL, confidence interval

Full size image

We then conducted subgroup analyses based on cancer types. The results showed that patients with low ALI had worse OS than those with high ALI in EC (HR = 1.937, 95% CI: 1.204–3.119, P = 0.006, Fig. 3), GC (HR = 1.451, 95% CI: 1.206–1.746, P < 0.001, Fig. 3), and CRC (HR = 2.279, 95% CI: 1.769–2.935, P < 0.001, Fig. 3). We also found no significant heterogeneity between included studies in the GC (I2 = 0.0%, P = 0.824) and EC (I2 = 0.0%, P = 0.816) subgroups; and lower heterogeneity between included studies in the CRC (I2 = 60.6%, P = 0.013) subgroup, so differences in cancer type were a source of heterogeneity (Fig. 3).

Fig. 3
figure 3

Subgroup analysis of overall survival based on cancer types. HR, hazard ratio; CL, confidence interval; EC, esophageal cancer; GC, gastric cancer; CRC, colorectal cancer

Full size image

Subgroup analyses based on study region, sample size, treatment, and ALI cutoff were also performed, and the results for each subgroup were consistent with the above findings (Table 2).

Table 2 Subgroup analysis of the association between ALI and overall and disease-free survival in patients with gastrointestinal neoplasms
Full size table

ALI and disease-free survival

The relationship between ALI and DFS was also examined using prognostic data from 7 studies involving 3,047 participants. Significant heterogeneity was observed in the included studies (I2 = 86.9%, P < 0.001, Fig. 4), so a random effects model was used. We found that patients with low ALI had a shorter DFS than those with high ALI (HR = 1.631, 95% CI: 1.197–2.224, P = 0.002, Fig. 4).

Fig. 4
figure 4

Forest plot of the advanced lung cancer inflammation index in relation to disease-free survival. HR, hazard ratio; CL, confidence interval

Full size image

Subgroup analysis showed that lower ALI was associated with poorer DFS in CRC patients (HR = 1.911, 95% CI: 1.517–2.408, P = 0.002, Fig. 5). No significant heterogeneity was observed in the subgroups (I2 = 0.0%, P = 0.420, Fig. 5), and a fixed effects model was utilized. Therefore, differences in cancer type were the source of heterogeneity.

Fig. 5
figure 5

Subgroup analysis of disease-free survival based on cancer types. HR, hazard ratio; CL, confidence interval; CC, cholangiocarcinoma; GC, gastric cancer; CRC, colorectal cancer; OCC, oral cavity cancer

Full size image

In addition, subgroup analyses based on study region, sample size, and ALI cutoff were also performed, and the results for each subgroup were generally consistent with the above results (Table 2). Notably, ALI was not found to be associated with worse DFS in subgroups with sample sizes ≤ 300 (HR = 1.385, 95% CI: 0.859–2.233, P = 0.182); the opposite was true in subgroups with sample sizes > 300 (HR = 1.783, 95% CI: 1.369–2.322, P < 0.001).

ALI and progression-free survival and cancer-specific survival

A connection between ALI and PFS in cancer patients was observed in a total of 2 studies involving 803 individuals. As shown in Fig. 6A, patients with low ALI had a worse PFS than those with high ALI (HR = 1.679, 95% CI: 1.073–2.628, P = 0.023). Significant heterogeneity was found in studies (I2 = 71.4%, P = 0.061), and a random effects model was applied to this analysis.

Fig. 6
figure 6

Forest plot of the advanced lung cancer inflammation index in relation to progression-free survival (A) and cancer-specific survival (B). HR, hazard ratio; CL, confidence interval

Full size image

The association between ALI and CSS in cancer patients was explored in two articles with 722 individuals (Fig. 6B). Interestingly, we found no significant correlation between ALI and CSS in cancer patients (HR = 1.121, 95% CI: 0.694–1.812, P = 0.640) using a random effects model (I2 = 78.9%, P = 0.030).

Sensitivity analysis

We used the leave-one-out method to do a sensitivity analysis to assess how each study might impact the combined results. We found that the pooled HR for OS was not significantly changed after excluding one study at a time, ranging from 1.853 (95% CI: 1.469–2.337, after omitting Deng et al. 2022) to 1.966 (95% CI: 1.523–2.537, after omitting Zhang et al. 2022, Fig. 7A). Similarly, the pooled HR for DFS was not significantly different in the sensitivity analysis (Fig. 7B). The overall HR ranged from 1.513 (95% CI: 1.123–2.039, after omitting Deng et al. 2022) to 1.708 (95% CI: 1.155–2.526, after omitting Zhang et al. 2022). From the above, we can see that our results are stable and reliable.

Fig. 7
figure 7

Sensitivity analysis of overall survival (A) and disease-free survival (B). CL, confidence interval

Full size image

Publication bias

The publication bias in OS (Egger’s test: P = 0.001, Begg’s test: P = 0.548) and DFS (Egger’s test: P = 0.021, Begg’s test: P = 0.548) was found by Egger's test. Next, the trim and fill method was utilized to calculate the number of missing studies in OS and DFS. By factoring in the missing hypothesis studies, the combined HR of OS and DFS was recalculated but was not substantially different. As a result, the publication bias had little impact, and the outcome was stable.

Discussion

Our goal was to explore the predictive significance of ALI in GIC patients, and the pooled data demonstrated that a lower ALI was remarkably related to shorter OS, DFS, and PFS. Furthermore, these results held steady even after sensitivity analysis and subgroup analysis. This is the first meta-analysis to thoroughly explore the impact of ALI on the prognosis of GIC patients. As an extremely accessible indicator in clinical practice, pre-treatment assessment of patients’ ALI can help physicians more effectively and easily predict clinical outcomes and assist them to adjust treatment in a timely manner, thereby further reducing mortality. However, it is worth noting that our results also found that ALI levels were not associated with CSS in patients with GIC. Considering that this index (including PFS) only integrated the data of two studies, it may lead to instability in the results, which need to be further confirmed by subsequent studies.

Both the systemic inflammatory response and nutritional state are recognized prognostic factors in cancer patients, and mounting research has shown a close relationship between the systemic inflammatory response and nutritional status in various cancers [33]. Furthermore, the latest view is that systemic inflammatory response via host-tumor interaction is now considered to be the 7th hallmark of cancer [34]. Systemic inflammatory response and nutritional status have been assessed using a variety of blood examination-based derivatives up to this point, such as NLR [35], platelet-lymphocyte ratio (PLR) [36, 37], prognostic nutrition index (PNI) [38], BMI [39], and albumin [40], and a number of lines of research have shown that these derivatives have the potential to be employed by patients with malignancies as prognostic markers [35,36,37,38,39,40].

The ALI is a newly defined cancer index, and one of its unique features is as a composite index combining the nutritional state and the inflammatory state [12]. Deng et al. confirmed the predictive ability of the ALI for 5-year OS and 5-year DFS was better than that of the PNI or systemic inflammation index (SII) in CRC patients [20]. Some studies also found that ALI was superior to albumin, NLR, and BMI in predicting complications, 5-year PFS, and 5-year OS in CRC and OCC patients [17, 22]. Interestingly, Wu et al. revealed that ALI outperformed NLR, PLR, monocyte-lymphocyte ratio (MLR), SII, and PNI in predicting OS and DFS in patients with cholangiocarcinoma by using time-dependent ROC analysis [16]. Thus, the ALI may have a higher discriminating value compared to other biomarkers. Taking all the current evidence together, our study found that ALI predicted a poor prognosis in patients with GIC, and the results held true in gastric, oesophageal, and colorectal cancers, according to subgroup analysis.

Surely, this analysis still has some limitations. The absence of ALI dynamics' evaluations, rather than the use of a single time-point value, is a significant limitation. The absence of a correlation between interleukins, chemokines, and ALI prevents us from elucidating the mechanistic relationship between ALI values and clinical outcomes. The use of various salvage maneuvers may, by chance, have altered the results in favor of one group depending on the opportunities at the treatment center. The vast majority of articles were retrospective cohort studies, which possibly limited their statistical power. There is a lack of uniformity in the cut-off values for ALI across studies, and aggregated survival results may deviate from the actual values. Thus, in order to confirm and update our conclusion, more high-quality studies with sizable sample sizes, particularly multicentre RCTs, were urgently required. At the same time, these studies should also include patients of different races and explore the optimal cut-off values to guide the clinic more precisely for the benefit of patients.