Gastroesophageal reflux disease: A risk factor for laryngeal squamous cell carcinoma and esophageal squamous cell carcinoma in the NIH-AARP Diet and Health Study cohort

Abstract

Background:

Prior studies suggest that gastroesophageal reflux disease (GERD) may be associated with risk of squamous cancers of the larynx and esophagus, however, most of these studies have had methodological limitations or insufficient control for potential confounders.

Methods:

We prospectively examined the association between GERD and esophageal adenocarcinoma (EADC), esophageal squamous cell carcinoma (ESCC), and laryngeal squamous cell carcinoma (LSCC) in 490,605 subjects aged 50–71 years at baseline in the NIH-AARP Diet and Health Study cohort. Exposure to risk factors were obtained from the baseline questionnaire. GERD diagnosis was extracted among eligible subjects via linkage to Medicare diagnoses codes and then multiply-imputed for non-Medicare-eligible subjects. Hazard ratios (HR) and 95% CI of GERD were computed using Cox proportional hazards regression.

Results:

From 1995–2011, we accrued 931 EADC, 876 LSCC and 301 ESCC in this cohort and estimated multivariable-adjusted HRs (95% CIs) for EADC of 2.23 (1.72–2.90), LSCC of 1.91 (1.24–2.94) and ESCC of 1.99 (1.39–2.84). The associations were independent of sex, smoking status, alcohol drinking, and follow-up time periods. We estimated that 22.04% of the US general population aged 50–71 years suffered GERD. Using risk factor distributions for the US from national survey data, 16.92% of LSCC and 17.32% of ESCC cases among those 50–71 years old were estimated to be associated with GERD.

Conclusions:

Future prospective studies are needed to replicate these findings. GERD is a common gastrointestinal disorder. If replicated, these findings may inform clinical surveillance of GERD patients and suggest new avenues for prevention of these malignancies.

Precis for use in the Table of Contents

In a prospective study, we observed that people with a history of GERD had a two-fold higher risk of laryngeal squamous cell carcinoma (LSCC) and esophageal squamous cell carcinoma (ESCC) in a large cohort of 490,000 people followed for 16 years, and estimated 17% of these cancer cases were associated with GERD. If replicated by future prospective studies, it may inform clinical surveillance of GERD patients and suggest new avenues for prevention of these malignancies.

Introduction

Gastroesophageal reflux disease (GERD) was the most prevalent gastrointestinal disorder in the United States in 2009 with nearly 9 million clinic visits ¹, and is a primary risk factor for esophageal adenocarcinoma (EADC) ². The possible carcinogenic mechanisms of reflux in the esophagus include direct effects of acid or bile or indirect effects on the formation of N-nitroso compounds, which may also be relevant to the upper aerodigestive tract (UADT) tissue (e.g. larynx) exposed to refluxate from the stomach ^3–5. Understanding the potential association between GERD and esophageal cancer and the UADT cancers may help identify the at-risk population, improve necessary surveillance, and potentially mitigate the target cancer risk through appropriate medical therapy.

Several previous studies have tested for the association between GERD and laryngeal squamous cell carcinoma (LSCC), but the results have been inconsistent ^6–13. Most previous studies have been limited by size, study design, or insufficient control for important potential confounders (e.g. tobacco smoking and alcohol consumption^14–16), limiting the strength of the inference^17–19. A recent meta-analysis summarized 15 case-control studies and found that there was a 2.37-fold higher risk of LSCC among people with GERD ¹³, while other studies have suggested that this association was largely driven by residual confounding due to unmeasured variables such as smoking and alcohol^{8, 20, 21}. Few studies have investigated the association between GERD and risk of esophageal squamous cell carcinoma (ESCC), which may also be affected by reflux. Thus, large prospective studies are needed to test for the association between GERD and LSCC or ESCC.

Herein, we examine the association between the GERD and risk of LSCC and ESCC in the prospective NIH-AARP Diet and Health Study cohort. With a 16-year follow-up from 1995/1996 to 2011, we have accrued more than 1,100 incident cancer cases and prospectively collected detailed risk factor information in this cohort, which enabled us to expand the investigation by conducting analyses stratified by the potential effect modifiers such as sex, smoking, and follow-up period.

Materials and Methods

Study population

The NIH-AARP Diet and Health Study is a prospective cohort that was established in 1995–1996 by mailing questionnaires to 3.5 million AARP members, aged 50–71 years, living in one of six US states (California, Florida, Louisiana, New Jersey, North Carolina, and Pennsylvania) or one of two metropolitan areas (Atlanta, Georgia, and Detroit, Michigan) ²². The baseline questionnaire queried about demographics, alcohol use, tobacco use, diet, anthropometrics, and medical history. Overall, 17.6% (n= 617,119) of potential participants returned the questionnaire. Of the 566,398 respondents who satisfactorily completed the baseline questionnaire, we further excluded²³: proxy respondents (n= 15,760), subjects with a history of cancer (except nonmelanoma skin cancer) at baseline (n=51,346), those with only information on cancer death (n=4,268), those with energy intake more than two interquartile ranges above or below the sex-specific median, which we have used as a marker for questionnaire completeness and accuracy (n=4,387), and those who died or were diagnosed with cancer on the first day of follow-up (n =32). Our final analytic sample included 490,605 participants of whom 454,038 (92.6%) self-reported their race/ethnicity as non-Hispanic white. The study was approved by the Institutional Review Board of the US National Cancer Institute.

Cohort follow-up and outcomes

Vital status was obtained by linkage to the National Death Index and cancer diagnoses were updated via linkage to the state cancer registries. Incident LSCC, ESCC, and EADC were identified by anatomic site and histologic codes of the International Classification of Diseases for Oncology, Third Edition (ICD-O-3)²⁴. LSCC was classified with site codes C32.0-C32.9, and excluded histologic codes of 8120, 8140, 8500, and 9220. Both ESCC and EADC were defined with site code C15.0-C15.9. ESCC included histologic codes of 8070, 8071, 8072, 8074, and 8083, and EADC included 8140, 8142, 8144, 8260, 8261, 8263, 8310, 8480, 8481, 8490, 8570, 8145, 8211, and 8255.

Risk factor assessment

Known risk factors for UGI cancers including age, sex, race/ethnicity, body mass index (BMI), smoking status, and alcohol drinking status were collected by the baseline questionnaire in 1995–1996. We defined BMI as a categorical variable according to WHO definitions²⁵: underweight (<18.5), normal weight (18.5–24.9), overweight (25.0–29.9), and obese (≥30.0); we defined smoking status as never, former, and current (participants who reported quitting within the past year were considered as current smokers). We categorized total alcohol beverage consumption as: non-drinker, up to or including one drink/day, greater than one to three drinks/day, and greater than three drinks/day²³. One drink corresponded to one serving size according to the US Department of Agriculture’s food guide pyramid: one 12-fluid ounce beer, one 5-fluid ounce glass of wine, or one 1.5-ounce shot of liquor (each approximately 13–14 g of ethanol)^{26, 27}. GERD diagnoses were extracted from the linkage to the Medicare dataset (1994–2008) from 107,258 (21.9%) of 490,605 eligible participants. We considered a subject had GERD if they had two or more Medicare claims which matched the GERD definition using all claims from the Medicare Provider Analysis and Review file (MEDPAR), Outpatient, and the National Claims History file (NCH) during or after their second eligible year using the NCI Comorbidity Index approach ^{28, 29}.

Statistical analysis

Missing values for the risk factors other than GERD and adjustment variables were addressed by replacing the missing value with the most common value. Because GERD was missing for a large percentage of the cohort, we multiply-imputed GERD five times using logistic regression imputation models that included age at baseline, sex, ethnicity, BMI, smoking dose and duration, alcohol intake, educational level, healthy eating index 2010, and the target subtype of cancer. Hazard ratios (HR) and 95% confidence intervals (CI) of GERD were computed using Cox proportional hazards regression with age as the time scale. The testing of proportional hazards assumptions showed no violence. The exit age was defined as the age of the diagnosis of gastric, esophageal, or head and neck cancer (to avoid possible detection bias due to surveillance for second cancers); death; movement out of catchment area; or the end of the follow-up (31 December 2011), whichever occurred first. Estimates of HRs of GERD come from models included sex, ethnicity, BMI, smoking status, and alcohol drinking and used age as the time scale. Likelihood ratio tests were used to assess the interactive effects of smoking status and alcohol intake on risk of ESCC, as well as the effect modification on GERD by sex, tobacco smoking, and alcohol drinking for LSCC and ESCC.

Population attributable risk (PAR) was used to estimate the proportion of each subtype of cancer that was associated with the presence of GERD in the US for individuals 50–71 years old (N=83,170,147). The PAR’s were obtained using the adjust HR’s from the NIH-AARP Cohort with the estimated distribution of risk factors from the 2015 National Health Interview Survey (NHIS). Because GERD was missing for all subjects in the 2015 NHIS, we used the multiple-imputation logistic-regression model for GERD data from the NIH-AARP cohort to multiply-impute missing GERD five times for all the NHIS subjects. The estimation of the PARs and their standard errors take into account the sample weights and other aspects of the complex sample design of the NHIS and the variation across the multiple imputations³⁰ .

In sensitivity analyses, we restricted our analysis to subjects with eligible Medicare data only (n=107,258); further categorized smoking status into 25 different levels according to the cigarette dose (1–10 cigarettes per day, 11–20 cigarettes per day, 21–30 cigarettes per day, 31–40 cigarettes per day, 41–60 cigarettes per day, and > 60 cigarettes per day ) and cessation (never, current smokers, quit 1–4 years previously, quit 5–9 years previously, and quit ≥ 10 years previously); stratified by sex, smoking status, or alcohol drinking; did lag analysis by dividing the follow-up time into four time periods (1995–1999, 2000–2004, 2005–2008, and 2009–2011); none materially changed the results. Analyses were performed with SAS, version 9.4, software (SAS Institute, Cary, North Carolina) and SUDAAN, version 11.0, software (Research Triangle Park, NC). Two-sided p-values of less than 0.05 were considered significant. We did not adjust for multiple comparisons.

Results

During 6.6 million person-years of follow-up (median= 15.5 years, IQR: 13.7–15.6 years), we accrued 876 LSCC cases, 301 ESCC cases, and 931 EADC cases. The multiply-imputed prevalence of GERD was 23.7% in the NIH-AARP cohort, with 116,476 of 490,605 people suffering from GERD. Table 1 shows the distributions of potentially confounding variables in the full NIH-AARP cohort overall and in participants with or without GERD. In general, females and older people (65–71 years) have a higher prevalence of GERD than males and younger people.

Table 1.

Baseline risk factor distributions among participants with or without gastroesophageal reflux disease and the full NIH-AARP Diet and Health Study Cohort (n=490,605)

Characteristics	Gastroesophageal reflux diseasea		Total, (%)
	Yes, (%)	No, (%)
Number	116,476	374,129	490,605
%	23.74	76.26	100.00
Age group (years, %)
50–54	10.00	14.74	13.62
55–59	19.59	23.74	22.75
60–64	29.07	27.86	28.15
65–71	41.34	33.66	35.48
Sex (%)
Male	54.48	61.20	59.60
Female	45.52	38.80	40.40
Race/ethnicity (%)
Non-Hispanic white	93.04	92.39	92.55
Non-Hispanic black	3.89	3.88	3.88
Hispanic	1.80	1.95	1.92
Asian/Pacific Islander/American Indian/Native Alaskan	1.28	1.78	1.66
Education (%)
Less than or high school graduateb	37.72	34.58	35.32
Some college	23.08	23.20	23.17
College graduate	39.19	42.22	41.50
Body mass index (%)
<18.5	0.97	1.03	1.02
18.5–<25	32.80	33.80	33.56
25–<30	44.81	43.67	43.94
≥30	21.42	21.50	21.48
Smoking status (%)
Never smoker	35.68	35.06	35.21
Former smoker	48.81	46.94	47.38
Current smoker	11.78	14.19	13.62
Unknown	3.74	3.81	3.79
Alcohol drinking (%)
Never drinker	26.87	23.68	24.44
>0–1 drink	53.07	52.84	52.90
>1 to 3 drinks	13.47	15.71	15.18
>3 drinks	6.58	7.77	7.49
Healthy eating index 2010 (mean)	66.13	65.91	65.96

Note:

^aWe multiply imputed GERD to the full NIH-AARP cohort using Logistic regression using SAS PROC MI by age at baseline, sex, ethnicity, detailed smoking recency and intensity (25 levels), alcohol use, education, Healthy Eating Index score, and body mass index.

^bThis category includes subjects that reported less than high school, completed high school, or some post-high school training.

As in many prior studies, we found a significant higher risk of EADC among people with GERD overall (HR=2.23, 95%CI: 1.72–2.90) and by stratum of sex, smoking status, or alcohol drinking (Table 2), and the lag analysis showed an overall statistically significant higher risk of EADC across 16-year follow-up. We observed a significant positive association between GERD and the risk of incident LSCC (HR=1.91, 95%CI: 1.24–2.94) and ESCC (HR=1.99, 95%CI: 1.39–2.84) through 16 years of follow-up overall, or by sex, smoking status, or alcohol drinking (Table 3). There was no significant interaction between smoking and alcohol intake on the association between GERD and LSCC (P=0.054) or ESCC (P=0.93). Moreover, we found no statistically significant effect modification on GERD by sex (P=0.9 for ESCC, P= 0.9 for LSCC), tobacco smoking (P=0.8 for ESCC, P= 0.7 for LSCC), or alcohol drinking (P=0.5 for ESCC, P=1.0 for LSCC). We further divided the 16-year follow-up time into four time periods, and the lag analysis showed an overall higher risk (all period point estimates >1) across 16-year follow-up among people with GERD for both LSCC and ESCC.

Table 2.

Hazard ratios and 95% CIs of gastroesophageal reflux disease (GERD) for esophageal adenocarcinoma in the full NIH-AARP Diet and Health Study Cohort (n=490,605, 1995–2011)^a

Group	Esophageal adenocarcinoma
	No. of cases	HR	95% CI		P value for interaction
Totalb	931	2.23	1.72	2.90
Stratified analyses
Malec	862	2.23	1.73	2.87	0.7
Femalec	69	2.32	0.98	5.49
Never smokerd	163	2.60	1.62	4.18	0.6
Former smokerd	566	2.16	1.62	2.89
Current smokerd	161	2.24	1.56	3.22
Never drinkere	180	2.22	1.34	3.69	0.8
>0–1 drinke	481	2.25	1.69	2.99
>1 to 3 drinkse	160	2.41	1.67	3.49
>3 drinkse	110	1.95	1.22	3.13
Period analysis
1995–1999	190	2.02	1.31	3.10
2000–2004	319	2.13	1.51	3.01
2005–2008	252	2.78	1.90	4.07
2009–2011	170	1.86	1.21	2.86

Note: GERD, gastroesophageal reflux disease;

^a, All risk analyses from the Cox proportional hazards regression models accounted for multiple-imputation.

^b, HRs come from Cox models adjusted for sex, ethnic group, BMI, smoking status and alcohol intake with age as the time scale.

^c, HRs come from Cox models adjusted for ethnic group, BMI, smoking status, and alcohol intake with age as the time scale.

^d, HRs come from Cox models adjusted for sex, ethnic group, BMI, and alcohol intake with age as the time scale.

^e, HRs come from Cox models adjusted for sex, ethnic group, BMI, and smoking status with age as the time scale.

Table 3.

Hazard ratios and 95% CIs of gastroesophageal reflux disease (GERD) for laryngeal squamous cell carcinoma and esophageal squamous cell adenocarcinoma in the full NIH-AARP Diet and Health Study Cohort (n=490,605, 1995–2011) ^a

Group	Laryngeal squamous cell carcinoma					Esophageal squamous cell carcinoma
	No. of cases	HR	95% CI		P value for interaction	No. of cases	HR	95% CI		P value for interaction
Totalb	876	1.91	1.24	2.94		301	1.99	1.39	2.84
Stratified analyses
Malec	749	1.91	1.17	3.10	0.9	202	2.02	1.37	2.97	0.9
Femalec	127	1.95	1.07	3.55		99	1.94	1.19	3.17
Never smokerd	85	1.96	1.15	3.33	0.7	38	1.92	0.79	4.66	0.8
Former smokerd	395	1.97	1.18	3.29		141	2.06	1.13	3.75
Current smokerd	351	1.76	1.04	2.99		102	1.77	1.00	3.13
Never drinkere	201	1.91	1.06	3.44	1.0	52	2.14	1.00	4.60	0.5
>0–1 drinke	356	1.96	1.13	3.41		89	1.92	1.17	3.14
>1 to 3 drinkse	160	1.75	1.15	2.68		64	2.74	1.31	5.73
>3 drinkse	159	1.94	1.09	3.44		96	1.55	0.92	2.63
Period analysis
1995–1999	235	1.76	0.86	3.60		72	1.89	1.00	3.57
2000–2004	301	1.89	1.23	2.92		112	1.95	1.21	3.16
2005–2008	201	2.16	1.30	3.60		67	2.29	1.13	4.64
2009–2011	139	1.68	0.97	2.91		50	1.63	0.66	4.06

Note: GERD, gastroesophageal reflux disease;

^a, All risk estimates come from Cox proportional hazards regression models using SAS PROC MIANALYZE accounting for multiple-imputation.

^b, HRs come from Cox models adjusted for sex, ethnic group, BMI, smoking status and alcohol intake with age as the time scale.

^c, HRs come from Cox models adjusted for ethnic group, BMI, smoking status, and alcohol intake with age as the time scale.

^d, HRs come from Cox models adjusted for sex, ethnic group, BMI, and alcohol intake with age as the time scale.

^e, HRs come from Cox models adjusted for sex, ethnic group, BMI, and smoking status with age as the time scale.

Using imputation, the GERD prevalence of the US general population aged 50–71 years was estimated as 22.04% in 2015 (Table 4). Females had a higher GERD prevalence than males (24.47% vs. 19.81%). Former smokers (25.70%) had a higher prevalence than never (21.17%) or current smokers (18.30%), and never drinkers (23.17%) had a higher GERD prevalence than alcohol drinkers. We estimated that 16.92% (95%CI: 5.36%−28.48%) of LSCC and 17.32% (95%CI: 7.34%−27.31%) of ESCC cases in US were associated with GERD. In general, the PARs were similar in males and females for both LSCC and ESCC, but higher in former smokers than never or current smokers for both cancer types. In addition, GERD was associated with a relatively higher percent of LSCC in never drinkers than alcohol drinkers, but contributed more to people who drink greater than one to three drinks/ day than other people for ESCC.

Table 4.

The population attributable risk (PAR) of gastroesophageal reflux disease (GERD) for laryngeal squamous cell carcinoma and esophageal squamous cell adenocarcinoma in US population aged 50–71 years old in 2015 (N=83,170,147)

Groupa	GERD prevalencef, %	Laryngeal squamous cell carcinoma			Esophageal squamous cell carcinoma
		PARg	95% CI		PARg	95% CI
Totalb	22.04	16.92	5.36	28.48	17.32	7.34	27.31
Stratified analyses
Malec	19.81	17.46	4.56	30.36	18.91	7.37	30.45
Femalec	24.47	15.11	−0.36	30.57	15.05	2.01	28.10
Never smokerd	21.17	17.83	2.23	33.44	16.93	−7.14	41.00
Former smokerd	25.70	21.32	7.59	35.05	21.85	3.80	39.91
Current smokerd	18.30	13.13	−0.43	26.69	12.89	−1.93	27.70
Never drinkere	23.17	17.97	1.55	34.39	20.48	−1.17	42.13
>0–1 drinke	21.19	17.66	2.21	33.12	16.50	1.41	31.59
>1 to 3 drinkse	19.86	12.61	−2.14	27.37	24.19	2.73	45.64
>3 drinkse	21.96	15.35	−0.53	31.23	9.86	−3.85	23.58

Note: PAR, Population attributable risks; GERD, gastroesophageal reflux disease;

^a, All risk estimates come from Cox proportional hazards regression models using SAS PROC MIANALYZE accounting for multiple-imputation.

^b, HRs come from Cox models adjusted for sex, ethnic group, BMI, smoking status and alcohol intake with age as the time scale.

^c, HRs come from Cox models adjusted for ethnic group, BMI, smoking status, and alcohol intake with age as the time scale.

^d, HRs come from Cox models adjusted for sex, ethnic group, BMI, and alcohol intake with age as the time scale.

^e, HRs come from Cox models adjusted for sex, ethnic group, BMI, and smoking status with age as the time scale.

^f, GERD prevalence of the general population in the United States, was multiply imputed using Logistic regression model using SAS PROC MI by including age at baseline, sex, ethnicity, smoking status, alcohol use, and body mass index from the NHIS survey in 2015.

^g, Results are based on risk estimates from the NIH-AARP cohort (1995–2011) using Cox proportional hazards regression and the distribution of GERD and other risk factors in the US population from the NHIS survey in 2015 accounting for multiple-imputation.

Moreover, we examined associations between GERD and cancer risk using only the 107,258 subjects with Medicare data and found comparable associations (Table 5) as those seen by multiply-imputation (Table 2 and 3) overall, or by sex, smoking status, or alcohol drinking. An overall significant positive association between GERD and the risk of incident LSCC (HR=1.70, 95%CI: 1.28–2.24), ESCC (HR=1.75, 95%CI: 1.20–2.57), and EADC (HR=2.04, 95%CI: 1.63–2.56) was found in this subset. For sensitivity analyses, we further re-categorized smoking status by doses and cessation into 25 detailed levels to fully adjust for this important risk factor, and found similar results for HRs estimation as those seen by categorizing smoking status into three levels as never, former and current smokers.

Table 5.

Hazard ratios and 95% CIs of gastroesophageal reflux disease (GERD) for laryngeal squamous cell carcinoma, esophageal squamous cell adenocarcinoma, and esophageal adenocarcinoma in the NIH-AARP Diet and Health Study Cohort subset with eligible Medicare data on GERD (n=107,258, 1995–2011)

Group	Laryngeal squamous cell carcinoma				Esophageal squamous cell carcinoma				Esophageal adenocarcinoma
	No. of cases	HR	95% CI		No. of cases	HR	95% CI		No. of cases	HR	95% CI
Total, adjustment a	219	1.70	1.28	2.24	115	1.75	1.20	2.57	313	2.04	1.63	2.56
Total, adjustment b	219	1.70	1.29	2.25	115	1.76	1.20	2.58	313	2.03	1.62	2.55
Stratified analyses
Malec	181	1.75	1.29	2.38	76	2.00	1.25	3.18	294	2.01	1.59	2.54
Femalec	38	1.48	0.75	2.91	39	1.41	0.72	2.76	19	2.51	1.01	6.20
Never smokerd	21	1.61	0.65	4.02	16	1.89	0.68	5.22	66	3.01	1.85	4.89
Former smokerd	89	2.34	1.54	3.56	58	1.97	1.16	3.33	194	1.70	1.27	2.28
Current smokerd	109	1.27	0.83	1.94	41	1.44	0.73	2.83	53	2.44	1.41	4.22
Never drinkere	44	2.19	1.20	3.99	15	2.39	0.83	6.86	69	2.39	1.48	3.87
>0–1 drinke	91	1.75	1.14	2.67	39	1.83	0.95	3.53	152	2.05	1.48	2.85
>1 to 3 drinkse	32	0.81	0.34	1.96	29	2.67	1.27	5.63	59	2.02	1.19	3.41
>3 drinkse	52	1.84	1.04	3.27	32	0.88	0.38	2.01	33	1.53	0.74	3.14

Note: GERD, gastroesophageal reflux disease;

^a, HRs come from Cox models adjusted for sex, ethnic group, BMI, smoking status (never, former, current smokers), and alcohol intake with age as the time scale.

^b, HRs come from Cox models adjusted for sex, ethnic group, BMI, detailed smoking recency and intensity (25 levels), and alcohol intake with age as the time scale.

^c, HRs come from Cox models adjusted for ethnic group, BMI, smoking status, and alcohol intake with age as the time scale.

^d, HRs come from Cox models adjusted for sex, ethnic group, BMI, and alcohol intake with age as the time scale.

^e, HRs come from Cox models adjusted for sex, ethnic group, BMI, and smoking status with age as the time scale.

Discussion

We estimated a 23.7% of GERD prevalence among people in the NIH-AARP cohort and 22.0% in the US general population aged 50–71 years in 2015. People with GERD presented a two-fold higher risk of EADC (HR=2.23, 95%CI: 1.72–2.90) through 16 years. Similarly, we identified an adverse association between GERD and LSCC (HR=1.91, 95%CI: 1.24–2.94) and ESCC (HR=1.99, 95%CI: 1.39–2.84) across 16-year follow-up, and the association was independent with sex, smoking status, alcohol drinking, and follow-up time period. We estimated 16.9% of LSCC and 17.3% of ESCC cases in US were associated with GERD.

Our analysis estimated that 22% of the general US population aged 50–71 years suffered from GERD, and women had a higher prevalence than men. This rate was comparable to a recent systematic review which reported a GERD prevalence of 18.1%−27.8% for the US general population³¹. There were various of methods for the diagnosis of GERD, including Reflux Symptom Index, proton pump inhibitor trial, ambulatory reflux monitoring, upper gastrointestinal endoscopy, patient self-reporting, and study specified questionnaire survey. The current study extracted GERD diagnosis from the linkage to the Medicare dataset by using medical claims, which may result in a lower prevalence than those of symptom-based questionnaire surveys due to self-treatment or use of over-the-counter medicines. Previous data showed that the prevalence of GERD has significantly increased by 50% from 2000 to 2010, and GERD remains the most common gastrointestinal disorder in US^{31, 32}. GERD was also prevalent worldwide with some geographic variation, with a prevalence of 8.8%–25.9% in Europe, 2.5%–7.8% in East Asia, 8.7%–33.1% in the Middle East, 11.6% in Australia and 23.0% in South America³¹. Therefore, understanding the potential association between GERD and cancers of esophagus and larynx is important for the mitigation of these cancer burdens.

We estimated that people with GERD had about a two-fold higher risk of both EADC and ESCC compared to those without GERD, which was independent with sex, smoking status, alcohol drinking, and the follow-up time period. GERD is considered a primary known risk factor for EADC, and this result serves as a positive control finding for our use of medical claims data. Few studies have investigated the association between GERD and ESCC. Our results add robust prospective evidence for an adverse association between GERD and EADC and ESCC. These results suggest that a common pathophysiological mechanism may convey higher risk for these two different histological types of esophageal cancer, which share few risk factors other than tobacco smoking. These mechanisms include the impaired and transient relaxation of the lower esophageal sphincter resting tone, delayed gastric emptying, dysfunctional peristalsis, inadequate esophageal acid clearance, reduced salivation, impaired mucosal resistance, and increased intraabdominal pressure. These factors may lead to exposure of the esophagus to gastric acid and other stomach contents, including pepsin, bile, small intestine fluid, and pancreatic secretions, all potentially injurious to the esophageal mucosa³³. The chronic inflammation of the esophageal mucosa may progress to cellular transformation and metaplasia, which presents as “Barrett esophagus”, a precursor of the esophageal adenocarcinoma. According to the current evidence, the latest European Society of Gastrointestinal Endoscopy position statement recommends endoscopic screening for esophageal adenocarcinoma among people with long-standing gastroesophageal reflux disease symptoms (i. e., > 5 years) and multiple risk factors³⁴. Future prospective studies are required to examine the causality and diagnostic utility of GERD and risk of ESCC in different high-risk areas of ESCC, including Eastern to Central Asia, along the Rift Valley in East Africa, and South Africa. Furthermore, several studies have suggested that laryngeal injury can occur from reflux of acidic and nonacidic gastric contents³⁵, and the laryngeal mucosa may be more sensitive to the effects of reflux compared with the esophagus^{36, 37}.

A recent meta-analysis of previous case-control studies showed a significant adverse association (OR=2.37) between GERD and LSCC¹³. Further analysis showed that the association between reflux disease and LSCC was stronger when diagnoses used objective methods, including 24-hour pH probe monitoring and findings on endoscopy (OR=3.82) than chart review-based methods that used clinical history based on ICD codes (OR=1.81) for the diagnosis of reflux disease¹³. Some other studies found no association between GERD and LSCC in non-smokers or after excluding patients with alcoholism^{8, 20}, which may suggest that the association between GERD and LSCC was largely driven by smoking or alcohol²¹. Although, a meta-analysis showed an adverse association (OR=2.07) between reflux disease and LSCC among only studies that controlled for tobacco smoking and alcohol intake using multivariate logistic regression¹³. Strengths of our study include prospective collection of exposure information, large sample size, control for the primary risk factors for these cancers including tobacco and alcohol consumption and the national representativeness of the estimated PARs. Unlike most prior studies of the association, we were able to comprehensively adjust for established risk factors. We showed that maximally adjusting for tobacco use with a 25-category exposure variable did not materially alter our risk estimates.

Another advantage of our study was its prospective design. Most prior studies on GERD and UADT cancers were retrospective and were significantly limited by the potential reverse causation effect, which means that patients who were diagnosed with GERD may have presented with symptoms caused by the cancer. In this study, GERD diagnoses were extracted from the linkage to the Medicare dataset (1994–2008), and we separately investigated the association between GERD and different cancer subtypes in four time periods from 1995 to 2011. We found a universal higher risk of LSCC, ESCC, and EADC among people with GERD across 16-year follow-up, with all period point estimations > 1 (including 2009–2011). Thus, our data provided strong prospective evidence for a positive association between GERD and LSCC, ESCC and EADC.

Limitations of our study was use of a medical claim’s definition of GERD and the partial and complete missingness of GERD in NIH-AARP and NHIS, respectively, that was addressed by using multiple-imputation.

To our knowledge, this is the largest prospective study to investigate the associations between GERD and LSCC and the first prospective study to report an association between GERD and ESCC. Future prospective studies with validated GERD questionnaires or endoscopy-proven GERD paired with comprehensive control for confounding are needed to replicate these findings. GERD is the most common gastrointestinal complaint in many populations. Thus, if replicated, these findings may inform clinical surveillance of GERD patients and suggest new avenues for prevention of these malignancies.

Abbreviations:

BMI

Body mass index

CI

confidence intervals

EADC

esophageal adenocarcinoma

ESCC

esophageal squamous cell carcinoma

GERD

gastroesophageal reflux disease

HR

Hazard ratios

ICD-O-3

International Classification of Diseases for Oncology, Third Edition

LSCC

laryngeal squamous cell carcinoma

MEDPAR

Medicare Provider Analysis and Review file

NCH

National Claims History file

NHIS

National Health Interview Survey

PAR

Population attributable risk

UADT

upper aerodigestive tract