Human papillomavirus infection and gastric cancer risk: A meta-epidemiological review

Abstract

Gastric cancer (GC) is a multifactorial disease, and several modifiable risk factors have been reported. This review summarizes and interprets two previous quantitative systematic reviews evaluating the association between human papillomavirus (HPV) infection and GC risk. The results of two systematic reviews evaluating the same hypothesis showed a statistically significant difference in summary odds ratios and their 95% confidence intervals. Thus, it is necessary to conduct a subgroup analysis of Chinese and non-Chinese studies. Additional meta-analyses that control for heterogeneity are required. Reanalysis showed that all the Chinese studies had statistical significance, whereas the non-national studies did not. The funnel plot asymmetry and Egger's test confirmed publication bias in the Chinese studies. In addition, the proportion of HPV-positive cases in Chinese studies was 1.43 times higher than that in non-Chinese studies and 2.81 times lower in controls. Therefore, the deduced evidence is currently insufficient to conclude that HPV infection is associated with GC risk.

Key Words: Papillomavirus; Stomach neoplasm; Case-control studies; Meta-analysis; Systematic review; Risk factors

Core Tip: Chinese studies showed that human papillomavirus infections increased the risk of gastric cancer; however, non-Chinese studies showed no statistical significance. Therefore, the deduced evidence is currently inadequate to conclude that human papillomavirus infection is associated with gastric cancer risk.

INTRODUCTION

Gastric cancer (GC) is the fifth most common incident cancer according to Global Cancer Statistics 2018[1] and ranks third in absolute years of life lost[2]. GC is a multifactorial disease, and several modifiable risk factors have been reported[3,4].

Infection with Helicobacter pylori or oncogenic viruses has important implications for preventing and managing GC[5]. Helicobacter pylori eradication is one of the reasons behind the steady decline in global GC incidence[6]. Therefore, human papillomavirus (HPV), which is among potential oncoviruses posing GC risk reviewed by Niedźwiedzka-Rystwej et al[7], should be considered to control GC occurrence because HPV vaccines have been used to prevent uterine cervix cancer[8,9].

However, the International Agency for Research on Cancer did not suggest an association between HPV infection and GC risk in a monograph published in 2007[10]. This review summarizes and interprets previous quantitative systematic reviews evaluating the association between HPV infection and GC risk.

PREVIOUS SYSTEMATIC REVIEWS

A PubMed (https://pubmed.ncbi.nlm.nih.gov) search, using "papillomavirus infection" and "stomach neoplasms" as the keywords of the hypothesis, identified two systematic reviews as of December 31, 2020[5,11]. Both selected case-control studies and their results are summarized in Table 1.

Table 1 The summary odds ratio with its 95%CI from two systematic reviews.

Ref.	Search to	Subgroup	Case-control studies	sOR (95%CI)	I2 (%)
Zeng et al[11], 2016	Jun 2016	All	15	7.39 (3.88-14.1)	56.7
Wang et al[5], 2020	Apr 2020	All	14	1.53 (1.00-2.33)	59.8
		Chinese	5	1.98 (1.04-3.75)	73.7
		Non-Chinese	9	1.17 (0.68-2.02)	33.4
		Tissue	11	2.24 (1.13-4.43)	66.5
		Serum	3	1.04 (0.75-1.44)	0.0
		HPV-16	8	2.42 (1.00-5.83)	67.5
		HPV-18	3	1.08 (0.59-1.99)	0.0

HPV: Human papillomavirus; sOR: Summary odds ratio.

Zeng et al[11] reported that in 2016, a total of 15 case-control studies, including 12 studies on Chinese patients, and a meta-analysis showed that HPV infection increased the risk of GC by 7.39 times [95% confidence interval (CI) of summary odds ratio (sOR): 3.88–14.1]. Further, a study by Wang et al[5] published in 2020 selected a total of 14 case-control studies, including five studies on Chinese patients, and the sOR was 1.53 (95%CI: 1.00–2.33).

The results of two systematic reviews evaluating the same hypothesis showed a statistically significant difference in sORs and their 95%CI. These findings can be inferred from the following three reasons. First, there was a difference in selection criteria. Wang et al[5] included three serological studies, in addition to tissue tests. Therefore, it is necessary to limit future research to tissue studies and conduct a meta-analysis again. Second, there was a difference in search databases between the two systematic reviews. Zeng et al[11] and Wang et al[5] selected 12 and five Chinese studies, respectively. Whereas Zeng et al[11] did not report a subgroup analysis, Wang et al[5] showed different subgroup analysis results between Chinese and non-Chinese studies. Therefore, it is necessary to conduct subgroup analyses of Chinese and non-Chinese studies in all selected articles. Finally, potential bias is possible due to heterogeneity. Wang et al[5] found no statistical significance in subgroups with less than 50% of the I-squared value, such as non-Chinese studies, serum studies, and HPV-18 studies (Table 1). Therefore, additional meta-analyses that control for heterogeneity are required.

RE-ANALYSIS OF META-ANALYSIS

Both systematic reviews selected a total of 25 articles. After excluding three serological studies[12-14], three studies had no information on the control group[15-17], and one showed zero HPV positivity in both the case and control groups[18]; hence, 18 articles were selected for reanalysis[19-35].

Table 2 illustrates the information extracted for the reanalysis of each study. Xu et al[25] extracted the results for cardia as well as those for the entire region for use in subgroup analysis by GC site.

Table 2 Extracted information of the 18 selected case-control studies.

Ref.	Year	Nation	Site	Test	Sample	PCa	NCa	PCo	NCo
Sha et al[19]	1998	China	Gastric	PCR	FFPE	27	38	4	61
Dong et al[20]	1999	China	Gastric	PCR	Other	10	27	0	20
Yu et al[21]	1999	China	Gastric	PCR	FFPE	30	102	3	101
Zhou et al[22]	1999	China	Gastric	PCR	FFPE	19	31	0	20
Zhu et al[23]	2000	China	Gastric	PCR	FF	11	31	0	42
Liao et al[24]	2001	China	Gastric	ISH	Other	26	24	2	28
Xu et al[25]	2003	China	Cardia	ISH	FFPE	50	24	10	40
Xu et al[25]	2003	China	Gastric	ISH	FFPE	111	125	10	40
Ma et al[26]	2007	China	Gastric	PCR	FFPE	15	25	2	38
Ma et al[27]	2007	China	Cardia	PCR	FFPE	32	61	0	21
Rong et al[28]	2007	China	Cardia	PCR	FFPE	16	5	2	19
Wang et al[29]	2013	China	Gastric	PCR	FFPE	20	72	4	82
Su et al[15]	2015	China	Gastric	PCR	Other	1	14	0	15
Anwar et al[30]	1995	Japan	Gastric	PCR	FFPE	23	28	2	10
Erol et al[31]	2009	Turkey	Gastric	PCR	FFPE	17	21	33	73
Cândido et al[32]	2013	Brazil	Gastric	PCR	FFPE	4	36	10	30
Türkay et al[33]	2015	Turkey	Cardia	PCR	FFPE	2	17	0	8
Bozdayi et al[34]	2019	Turkey	Gastric	PCR	Other	20	33	5	21
Leon et al[35]	2019	Ethiopia	Cardia	PCR	FF	11	51	0	56

FF: Fresh frozen tissue; FFPE: Formalin-fixed paraffin-embedded tissue; ISH: In situ hybridization; NCa: Negative in cases; NCo: Negative in controls; PCa: Positive in cases; PCo: Positive in controls; PCR: Polymerase chain reaction.

Figure 1 displays a forest plot showing the results of the reanalysis. The sOR for 18 studies was 5.80 (95%CI: 3.27–10.31), showing statistical significance. While the I-squared value was reduced from 60% in all studies to 0% in 12 Chinese studies, their sOR remained statistically significant at 7.86 (95%CI: 5.19–11.89). However, the sOR for six non-Chinese studies was 1.97 (95%CI: 0.79–4.89), which was not statistically significant. In other words, all Chinese studies showed statistical significance; however, the non-national studies did not. This finding was the same in the subgroup analysis by cardiac tissue, formalin-fixed paraffin-embedded tissue, fresh frozen tissue, and polymerase chain reaction (Table 3).

Figure 1 Forest plot for estimating summary odds ratio. CI: Confidence interval.

Table 3 Subgroup analysis by nationality.

		All	Chinese studies	Non-Chinese studies
All		5.80 (3.27-10.31) [60.0] <18>	7.86 (5.19-11.89) [0.0] <12>	1.97 (0.79-4.89) [56.8] <6>
Area
	Gastric	4.83 (2.64-8.83) [62.4] <14>	7.08 (4.60-10.89) [0.0] <10>	1.54 (0.60-3.92) [62.6] <4>
	Cardia	10.88 (5.42-21,8) [0.0] <5>	11.17 (5.34-23.35) [0.0] <3>	8.62 (0.88-84.8) [14.2] <2>
Sample
	FFPE	5.13 (2.55-10.34) [68.4] <12>	8.02 (4.74-13.6) [19.6] <8>	1.38 (0.45-4.16) [58.5] <4>
	FF	27.9 (3.70-211.7) <2>	31.0 (1.76-546.6) <1>	25.2 (1.45-439.1) <1>
Methods
	PCR	5.88 (3.00-11.52) [62.2] <16>	10.93 (6.44-18.5) [0.0] <10>	1.97 (0.79-4.98) [56.8] <6>
	ISH	6.23 (1.56-24.9) [64.0] <2>	6.23 (1.56-24.9) [64.0] <2>	-

Study: Summary odds ratio (95% confidence interval) [I² value (%)] <Number of selected studies>; FF: Fresh frozen tissue; FFPE: Formalin-fixed paraffin-embedded tissue; ISH: In situ hybridization; PCR: Polymerase chain reaction.

Twelve Chinese studies were examined for publication bias. The asymmetry of the funnel plot (Figure 2) and Egger's test (P = 0.013) confirmed publication bias. The trimming sOR from trim-and-fill analysis[36] was 6.78 (95%CI: 4.40–10.45).

Figure 2 Funnel plot in 12 Chinese studies (P value of Egger test = 0.013).

CONCLUSION

To summarize the above reanalysis results, Chinese studies demonstrated that HPV infections increased the risk of GC; nonetheless, non-Chinese studies showed no statistical significance. Therefore, the deduced evidence is currently insufficient to conclude that HPV infection is associated with GC risk.

The following interpretations and suggestions may be made based on the significant associations observed only in Chinese studies. First, there is a possibility that publication bias was involved in the selection of Chinese studies. After checking for publication bias using the funnel plot (Figure 2) and Egger's test, trim-and-fill analysis was performed. However, the trimming sOR in Chinese studies showed that HPV infections persistently increased the risk of GC. This mandated an alternative interpretation. The author attempted to infer that HPV positivity might have been different between Chinese and non-Chinese studies.

Using the information in Table 2, the proportion (%) of HPV positivity (PP) was obtained from both Chinese and non-Chinese studies (Table 4). On combining both the case and control groups, the PPs in Chinese and non-Chinese studies were 27.3% (95%CI: 24.9–29.9) and 24.9% (95%CI: 21.2–28.8), respectively. Their 95%CIs overlapped, showing no statistically significant differences. However, the case-group PP in Chinese studies was 41.9% (95%CI: 38.2–45.6), higher than that in non-Chinese studies (29.3%;95%CI: 23.8–35.2), and their 95%CIs did not overlap, showing a statistically significant difference. In contrast, the control-group PP in Chinese studies was 7.2 % (95%CI: 5.1–9.8), lower than the 20.2 % (95%CI: 15.4–25.7) in non-Chinese studies, and their 95%CIs did not overlap. In other words, the case PP in Chinese studies was 1.43 times (= 41.9/29.3) higher than that in non-Chinese studies and 2.81 times (= 20.2/7.2) lower in controls. This indicates a potentially significant relationship between HPV infection and GC risk in Chinese studies.

Table 4 Proportion of human papillomavirus positivity (%) by nationality.

		Chinese studies	Non-Chinese studies
Total
	Positive/Observe	335/1225	127/511
	PP (95%CI)	27.3 (24.9-29.9)	24.9 (21.2-28.8)
Case
	Positive/Observe	298/711	77/263
	PP (95%CI)	41.9 (38.2-45.6)	29.3 (23.8-35.2)
Control
	Positive/Observe	37/514	50/248
	PP (95%CI)	7.2 (5.1-9.8)	20.2 (15.4-25.7)

PP: Human papillomavirus positivity.

Given that the PP in the control group of the Chinese studies was significantly lower, descriptive epidemiological studies on HPV infection in the Chinese population are warranted. It is also necessary to conduct follow-up studies on whether the GC incidence rate due to HPV infection will change in the future due to the HPV vaccination project currently targeted at the Chinese population.