Background

In Asia, lung cancer is the most common cancer in men (age-standardized rate [ASR; per 100,000] = 35.2) and the third most common cancer in women (ASR = 12.7). The number of patients with lung cancer has increased rapidly by the year [1, 2]. The discovery and development of therapeutics targeting epidermal growth factor receptor (EGFR), namely tyrosine kinase inhibitors (TKIs), in the past decade was an important clinical advance in non–small cell lung cancer (NSCLC) treatment [3, 4]. Recommended by clinical guidelines, both gefitinib (Iressa) and erlotinib (Tarceva) are now widely accepted as standard-of-care therapy for patients with NSCLC whose tumors harbor activating EGFR mutations, especially patients with certain clinical characteristics (Asian descent, female gender, never-smoker, adenocarcinoma) [5,6,7,8]. The EGFR TKIs gefitinib and erlotinib both achieve a higher response rate for treating NSCLC in East Asian countries than in the Western countries [9]. However, which EGFR TKI can achieve better efficacy is controversial. In a phase III randomized controlled trial (RCT), Urata reported a higher incidence of grade 3–4 skin rash but less alanine aminotransferase/aspartate aminotransferase elevation in the erlotinib arm. Progression-free survival (PFS), overall survival (OS), and objective response rate (ORR) were similar between the two groups [10]. In another phase III RCT, Yang reported that gefitinib and erlotinib had similar efficacy (PFS, OS, ORR) in NSCLC, with similar toxicities [11]. Some studies have shown that gefitinib has better anti-tumor efficacy or less toxicity for NSCLC [12, 13]. However, other studies have reported opposite results and have suggested that erlotinib is more effective [14, 15].

To resolve this controversy, we conducted a meta-analysis of related studies to compare the anti-tumor efficacy and adverse effects (AEs) of gefitinib and erlotinib for treating East Asian populations with NSCLC.

Methods

We conducted this meta-analysis according to PRISMA (Preferred Reporting Items for Systematic Review and Meta-Analysis) guidelines.

Search strategy

The relevant literature was retrieved using the following electronic databases: (1) PubMed; (2) ScienceDirect; (3) The Cochrane Library; (4) Scopus; (5) Web of Science; (6) Embase; (7) Ovid MEDLINE; and (8) Google Scholar. The last search was on February 14, 2018. The following terms were used: “gefitinib”, “erlotinib”, and “Lung cancer”. The complete search we used for PubMed was: (gefitinib [MeSH Terms] OR gefitinib [Text Word] OR IRESSA [Text Word] OR ZDl839 [Text Word]) AND (erlotinib [MeSH Terms] OR erlotinib [Text Word] OR Tarceva [Text Word] OR OSI-774 [Text Word]) AND (lung cancer [MeSH Terms] OR lung cancer [Text Word] OR lung carcinoma [Text Word] OR lung neoplasm [Text Word] OR NSCLC [Text Word]). The references of retrieved articles were also searched for further eligible articles. No language restriction was imposed.

Selection criteria

Articles that met the following criteria were included: (1) East Asian population with histologically or cytologically confirmed NSCLC based on the Eastern Cooperative Oncology Group; (2) compared gefitinib versus erlotinib; (3) outcomes were PFS, OS, ORR, disease control rate (DCR), and AEs. We excluded reviews without original data, meta-analyses, animal experiments, abstracts only, and studies with duplicated data.

Data extraction

Two investigators extracted the following data independently: first author, publication year, country, number of participants, participant characteristics (age, sex, stage of cancer, pathological type, line of treatment), anti-tumor efficacy indices (PFS, OS, ORR, DCR), and number of AEs (total AEs, grade 3–5 AEs). A third investigator resolved disagreements on all terms.

Quality assessment

The quality of RCTs was assessed using the 5-point Jadad scale, which contains questions on three main items: randomization, masking, and accountability of all patients. High-quality studies score ≥ 3 points [16].

The quality of cohort studies was assessed using the Newcastle-Ottawa Scale (NOS, 9 points), which also contains questions on three main items: selection, comparability, and exposure. High-quality studies score 8–9 points; medium-quality studies score 6–7 points [17].

Statistical analysis

The meta-analysis was conducted using Review Manager (version 5.3, The Nordic Cochrane Centre) and STATA (version 12.0, Stata Corp). Hazard ratios (HR) with 95% confidence intervals (CI) were used to analyze the PFS and OS (HR > 1 favors the erlotinib group; HR < 1 favors the gefitinib group). The HR data were extracted directly from some studies or from Kaplan–Meier curves according to Tierney et al. [18] from other studies. Pooled risk ratios (RR) with 95% CIs were used to analyze the ORR, DCR, and AEs (RR > 1 favors the gefitinib group; RR < 1 favors the erlotinib group). Subgroup analysis of PFS, OS, and ORR was conducted to determine whether the results would change according to EGFR mutation status, ethnicity, line of treatment, histology, tumor stage, and study design. Heterogeneity was evaluated using the χ2 test and I2 statistic. If I2 > 50% or p <  0.1 for the χ2 test, reflecting significant heterogeneity, the random-effects model was used; otherwise, the fixed-effects model was used. Publication bias was explored using Begg’s rank correlation and Egger’s linear regression tests. P <  0.05 indicated statistical significance.

Results

Search results and study quality assessment

We initially identified 5829 potentially eligible studies. After screening, 31 studies involving 8054 patients (gefitinib group, 4907 patients; erlotinib group, 3147 patients) were included for the final analysis (Fig. 1) [10,11,12,13,14,15, 19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43]. Of the 31 studies, three were RCTs and 28 were retrospective studies. Twenty-two studies were of high quality (three RCTs scored 4–5 points, five retrospective studies scored 9 points, 14 retrospective studies scored 8 points) and nine studies were of medium quality (seven retrospective studies scored 7 points, two retrospective studies scored 6 points) (Table 1). Table 2 summarizes the baseline characteristics and main evaluation indices of the included studies.

Fig. 1
figure 1

Flow chart of study selection

Full size image
Table 1 Quality assessment of all included studies
Full size table
Table 2 Characteristics of included studies
Full size table

Anti-tumor efficacy

We assessed anti-tumor efficacy between the gefitinib and erlotinib groups based on PFS, OS, ORR, and DCR.

Twenty-four studies compared PFS (heterogeneity: p = 0.03, I2 = 38%). No significant difference was found between the two groups (95% CI: 0.97–1.10, p = 0.26; Fig. 2).

Fig. 2
figure 2

Forest plot of HR of PFS associated with gefitinib versus erlotinib

Full size image

Twenty-one studies compared OS (heterogeneity: p = 0.0004, I2 = 58%). No significant difference was found between the two groups (95% CI: 0.89–1.21, p = 0.61; Fig. 3).

Fig. 3
figure 3

Forest plot of HR of OS associated with gefitinib versus erlotinib

Full size image

Thirteen studies compared ORR (heterogeneity: p = 0.24, I2 = 20%). No significant difference was found between the two groups (95% CI: 1.00–1.18, p = 0.06; Fig. 4a).

Fig. 4
figure 4

Forest plots of RR of ORR (a) and DCR (b) associated with gefitinib versus erlotinib

Full size image

Eleven studies compared DCR (heterogeneity: p = 0.17, I2 = 29%). No significant difference was found between the two groups (95% CI: 0.93–1.05, p = 0.68; Fig. 4b).

Toxicity

We compared toxicity between the gefitinib and erlotinib groups based on total AEs, grade 3–5 AEs, and subgroup analysis of the 10 most reported AEs.

Five studies compared total AEs (heterogeneity: p = 0.0007, I2 = 79%). No significant difference was found between the two groups (95% CI: 0.87–1.13, p = 0.94; Fig. 5a).

Fig. 5
figure 5

Forest plots of RR of all-grade AEs (a) and grade 3–5 AEs (b) associated with gefitinib versus erlotinib

Full size image

Seven studies compared grade 3–5 AEs (heterogeneity: p = 0.001, I2 = 73%). The gefitinib group had a significantly lower incidence rate of grade 3–5 AEs than the erlotinib group (95% CI: 0.27–0.71, p = 0.0008; Fig. 5b). Some patients had drug discontinuations/reductions due to the occurrence of serious AEs. Two studies compared drug discontinuations; there was no significant difference between the two groups (95% CI: 0.40–1.80, p = 0.68; Fig. 6a). Four studies compared drug reductions; the erlotinib group had more drug reductions (95% CI: 0.13–0.65, p = 0.002; Fig. 6b).

Fig. 6
figure 6

Forest plots of RR of drug discontinuations (a) and drug reductions (b) associated with gefitinib versus erlotinib

Full size image

In subgroup analysis of the 10 most reported AEs (skin rash, diarrhea, nausea/vomiting, fatigue, anorexia, interstitial lung disease, stomatitis, elevated liver enzymes, infection, neutropenia), the results for all-grade AEs showed no significant differences in anorexia, interstitial lung disease, elevated liver enzymes, infection, neutropenia and nausea/vomiting between the two groups. For all-grade AEs, erlotinib induced significantly higher rates of skin rash (95% CI: 0.74–0.94, p = 0.003), diarrhea (95% CI: 0.73–0.95, p = 0.005), fatigue (95% CI: 0.23–0.95, p = 0.04), and stomatitis (95% CI: 0.15–0.54, p = 0.0001) (Table 3). The results for grade 3–5 AEs showed no significant differences in anorexia, interstitial lung disease, elevated liver enzymes, infection, and neutropenia between the two groups. For grade 3–5 AEs, erlotinib induced significantly higher rates of skin rash (95% CI: 0.12–0.41, p < 0.00001), diarrhea (95% CI: 0.29–0.74, p = 0.001), nausea/vomiting (95% CI: 0.11–0.49, p = 0.0001), fatigue (95% CI: 0.09–0.87, p = 0.03), and stomatitis (95% CI: 0.08–0.99, p = 0.05) (Table 4).

Table 3 Top 10 adverse effects (all grade) associated with gefitinib versus erlotinib
Full size table
Table 4 Top 10 adverse effects (grade 3–5) associated with gefitinib versus erlotinib
Full size table

Subgroup analysis

To determine whether the anti-tumor efficacy of gefitinib versus erlotinib was consistent across subgroups, the pooled efficacy for PFS, OS, and ORR was estimated within each category of the following classification variables: country, tumor stage, histology, line of treatment, EGFR mutation status, and study design. All subgroup differences were not statistically significant in terms of PFS, OS, and ORR between the gefitinib and erlotinib groups (Table 5).

Table 5 Subgroup analysis for progression-free survival, overall survival and objective response rate
Full size table

Sensitivity analysis

Significant heterogeneity was found in the analysis of OS, total AEs and grade 3–5 AEs. The influence of each study on the pooled results was evaluated to evaluate stability and sensitivity. The results suggested that the outcomes of OS, total AEs and grade 3–5 AEs were reliable and stable (Fig. 7).

Fig. 7
figure 7

Meta-based influence analysis for comparisons of OS (a), total AEs (b) and grade 3–5 AEs (c)

Full size image

Cumulative meta-analysis

Analyses of PFS (Additional file 1: Figure S1), OS (Additional file 2: Figure S2), ORR (Additional file 3: Figure S3), DCR (Additional file 4: Figure S4) and total AEs (Additional file 5: Figure S5) demonstrated that the RRs of the final results became robust within a narrow range and remained not significant as publication years increased and as recent high-quality studies were included. After inclusion of Shin et al.’s study [12], the RR and 95% CI for grade 3–5 AEs decreased to < 1 and became stable (Additional file 6: Figure S6). Although there was no significantly reduced risk in ORR, clear evidence showed that the confidence interval was becoming narrow, and trended toward significance (favors gefitinib).

Publication bias

There was no evidence of publication bias for PFS (Begg’s test p = 0.585; Egger’s test p = 0.477, Fig. 8a) and OS (Begg’s test p = 0.880; Egger’s test p = 0.798, Fig. 8b).

Fig. 8
figure 8

Begg’s and Egger’s tests for comparisons of PFS (a) and OS (b)

Full size image

Discussion

Gefitinib and erlotinib are two similar small molecules with different binding capabilities and pharmacokinetic and pharmacodynamic properties related to their differing molecular structures [44,45,46]. Whether the differences between these first-generation EGFR TKIs can cause different anti-tumor efficacy is controversial [10, 11, 47]. By analyzing 31 high-quality studies, we directly compared the anti-tumor efficacy and safety of gefitinib and erlotinib for treating NSCLC [10,11,12,13,14,15, 19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43]. Our meta-analysis provides the most current medical evidence and shows that anti-tumor efficacy (PFS, OS, ORR, DCR) is comparable between gefitinib and erlotinib for treating East Asian patients with NSCLC. Subgroup analysis according to country, tumor stage, histology, line of treatment, EGFR mutation, and study design did not change the results. However, erlotinib toxicity was significantly greater than that of gefitinib, especially in all-grade/grade 3–4 skin rash, nausea/vomiting, fatigue, and stomatitis.

The greater drug toxicity is an critical problem regarding erlotinib. In our analysis, we found high incidences of drug reduction, skin rash, diarrhea, nausea/vomiting, fatigue, and stomatitis in the erlotinib arm. Although it might not decrease survival time, it greatly reduces patients’ quality of life [48, 49]. We believe there are two reasons for these results: (1) the oral dose of erlotinib (150 mg/day) was closer to the maximum tolerated dose (150 mg/day) as compared with gefitinib (oral dose, 250 mg/day; maximum tolerated dose, 600 mg/day) [50, 51]; (2) The two EGFR TKIs have different pharmacokinetics. After absorption, more gefitinib accumulates in tumor tissue than in plasma; the opposite is true for erlotinib [52]. In the published literature, more severe AEs have been reported in East Asian patients as compared with patients from Europe and America [9, 53]. Interstitial lung disease is one of the most important AEs, and can cause worse prognosis and increased risk of death [54]. However, our analysis and other published studies show that most cases of interstitial lung disease are reported in East Asian populations and that it is rare in Western populations. This might be attributed to the smaller physiques of Asians in general. In a retrospective study, Yeo reduced the erlotinib dose to 25 mg/day and achieved similar or even better prognosis as compared with the standard dose [55]. Other retrospective studies have reported similar results [13, 56,57,58]. Accordingly, we suggest that individualized drug dose based on weight or body surface area might be more appropriate than a fixed oral dose for treating advanced NSCLC. More large-sample, well-designed RCTs are needed to confirm the best dose of gefitinib and erlotinib for East Asian patients with advanced NSCLC.

Almost all of the included studies did not show any differences in all anti-tumor efficacy indices, which formed the basis of our results. Only one study reported an unfavorable result for erlotinib, with both lower PFS and OS, which might relate to the erlotinib group having more patients with non-adenocarcinoma NSCLC as based on government regulations [14]. Our results also showed a trend for prolonged median PFS (gefitinib group, 7.1 months vs. 4.9 months; erlotinib group, 7.7 months vs. 3.4 months) and OS (gefitinib group, 19.1 months vs. 14.0 months; erlotinib group, 15.5 months vs. 12.7 months) in patients with adenocarcinoma as compared with squamous-included NSCLC. However, no difference was found between gefitinib and erlotinib in this subgroup.

In the EGFR mutation status subgroup, we also found no difference between the anti-tumor efficacy of gefitinib and erlotinib. However, our results indirectly prove that both gefitinib and erlotinib are more suitable for treating EGFR mutation–positive NSCLC. Both median PFS (gefitinib group, 11.4 months vs. 4.9 months; erlotinib group, 9.6 months vs. 3.1 months) and OS (gefitinib group, 22.6 months vs. 16.0 months; erlotinib group, 20.9 months vs. 12.0 months) were longer in the EGFR mutation–positive subgroup than in the partial EGFR mutation–positive subgroup. Accordingly, we observed that the proportion of EGFR mutations increased by the year in EGFR TKI treatment (Table 1). Multiple EGFR mutation isoforms (exon 19, exon 21, others) were found, although the isoform most susceptible to gefitinib or erlotinib remains unclear. A phase III RCT compared gefitinib and erlotinib treatment in EGFR mutation–positive NSCLC and found significantly higher RR and longer median OS for patients with EGFR exon 19 mutations than for patients with EGFR exon 21 mutations following erlotinib or gefitinib treatment. However, no difference was found between gefitinib and erlotinib for both mutations [11]. Another RCT involving more EGFR mutation isoforms (exon 19, exon 21, T790 M) reported similar results [10]. However, Kuan suggested that erlotinib is associated with significantly longer PFS and lower risk of progression than gefitinib in patients with EGFR exon 19 deletions [15]. Limited by the quantity of published studies and included patients, further large-sample, well-designed RCTs focusing on single EGFR mutations are warranted to identify the best EGFR TKIs.

The line of treatment in which EGFR TKIs should be used in NSCLC remains controversial. Mainstream thinking considers EGFR TKIs second-line or later treatment after chemotherapy failure or first-line treatment for patients unable to tolerate chemotherapy. However, Table 1 shows that an increasing number of studies have used gefitinib and erlotinib as first-line treatment for advanced NSCLC [15, 33, 42]. However, no differences were found for PFS, OS, and ORR between gefitinib and erlotinib in each line of treatment subgroup. Wu et al. conducted a phase III RCT and suggested that first-line erlotinib can significantly improve PFS as compared to gemcitabine+cisplatin in patients with EGFR mutation–positive NSCLC [59]. Another phase III RCT suggested that PFS is significantly longer with gefitinib treatment in patients with mutation-positive NSCLC as compared with carboplatin+paclitaxel [60]. Several other high-quality RCTs have reported similar results [61,62,63]. Based on these positive results, the US Food and Drug Administration approved gefitinib as first-line treatment for EGFR mutation–positive NSCLC [64]. In the 2017 National Comprehensive Cancer Network (NCCN) guideline on NSCLC, both gefitinib and erlotinib are suggested as first-line treatment for EGFR mutation–positive NSCLC [65].

Several limitations should considered when interpreting our results. First, only high-quality studies published in English were included, which might result in language bias. Second, only three RCTs were included, which would weaken the quality of the results. Third, there was significant heterogeneity for some comparisons (OS and total/grade 3–5 AEs), which would weaken the reliability of these results. Fourth, the type and rate of EGFR mutations differed between the included studies, which might increase heterogeneity and weaken the quality of the results. Fifth, we obtained data from only three East Asian countries (China [Mainland and Taiwan], Japan, Korea), which might reduce the representativeness of the study. Sixth, quality of life and survival time are two equally important evaluating indicators for a treatment. Quality of life cannot simply be replaced by the number of AEs. However, the included studies did not compare quality of life between treatment with the two EGFR TKIs. Accordingly, we suggest that quality of life be considered an essential indicator in future drug evaluation studies.

Conclusion

Our results show that both gefitinib and erlotinib are effective for treating advanced NSCLC in East Asian patients, with comparable PFS, OS, ORR, and DCR. Erlotinib induces a significantly higher rate and severity of skin rash, nausea/vomiting, fatigue, and stomatitis, which might cause a higher rate of dose reduction. Therefore, we suggest that individualized drug dose based on weight or body surface area might be more appropriate than a fixed oral dose for both agents in treating East Asian patients with advanced NSCLC. However, due to the inherent limitations of our meta-analysis, more large-scale, high-quality RCTs are warranted to confirm this conclusion.