Reproductive Factors and Kidney Cancer Risk in 2 US Cohort Studies, 1993–2010

Abstract

Clinical and experimental findings suggest that female hormonal and reproductive factors could influence kidney cancer development. To evaluate this association, we conducted analyses in 2 large prospective cohorts (the National Institutes of Health–AARP Diet and Health Study (NIH-AARP), 1995–2006, and the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO), 1993–2010). Cohort-specific and aggregated hazard ratios and 95% confidence intervals relating reproductive factors and kidney cancer risk were computed by Cox regression. The analysis included 792 incident kidney cancer cases among 283,952 postmenopausal women. Women who had undergone a hysterectomy were at a significantly elevated kidney cancer risk in both NIH-AARP (hazard ratio = 1.28, 95% confidence interval: 1.09, 1.50) and PLCO (hazard ratio = 1.41, 95% confidence interval: 1.06, 1.88). Similar results were observed for both cohorts after analyses were restricted to women who had undergone a hysterectomy with or without an oophorectomy. For the NIH-AARP cohort, an inverse association was observed with increasing age at menarche (P for trend = 0.02) and increasing years of oral contraceptive use (P for trend = 0.02). No clear evidence of an association with parity or other reproductive factors was found. Our results suggest that hysterectomy is associated with increased risk of kidney cancer. The observed associations with age at menarche and oral contraceptive use warrant further investigation.

Kidney cancer accounts for nearly 2% of all new primary cancer cases in the United States (1) and is the seventh and ninth most common malignancy among men and women, respectively (2, 3). Kidney and renal pelvis cancer incidence rates according to the US Surveillance, Epidemiology, and End Results (SEER) program for the period 2005–2009 were 20.7 per 100,000 for men and 10.5 per 100,000 for women (4). Established risk factors for kidney cancer include cigarette smoking, obesity, hypertension, and family history of cancer (1, 4, 5).

The 2-fold higher incidence of kidney cancer among men versus women (1, 2, 4), coupled with experimental findings about the effects of estrogen on renal tumors (5–7), have stimulated interest in the role of female hormonal and reproductive factors in kidney cancer development. Exogenous estrogens have been shown in animals to promote and induce kidney cancer formation (6, 8), and sex hormone receptor expression in both normal and malignant renal tissue suggests that endocrine regulation could directly influence kidney cancer development (9, 10). Anatomical changes to the kidney during pregnancy might make nephrons more vulnerable to inflammation and oxidative stress (11, 12). The anatomical proximity of the ureter to the female reproductive organs also could leave the kidneys vulnerable to injury during surgery to the reproductive system (13). Lastly, hormonal disturbance has been associated with obesity and smoking (14, 15), 2 major risk factors most consistently observed across kidney cancer studies.

Epidemiologic study findings have suggested that parity (16–24) and hysterectomy (16–19, 25–29) could be associated with increased kidney cancer risk, although conflicting results have also been reported (24, 26, 27, 29–32). Other reproductive factors have not been consistently associated with risk. Most studies examining female reproductive factors and kidney cancer have been based on small case numbers, with limited statistical power to detect associations. We conducted analyses in 2 large, prospective cohort studies to more comprehensively investigate the association of female hormonal and reproductive factors with kidney cancer risk.

MATERIALS AND METHODS

Data sources

The cohorts included in this study were the National Institutes of Health–AARP Diet and Health Study (NIH-AARP) and the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO). Both studies were approved by institutional review boards at the National Cancer Institute.

NIH-AARP Study

The NIH-AARP prospective cohort study was initiated in 1995 to investigate the relations among diet, lifestyle, and health (33). A baseline questionnaire that collected information on female reproductive history, use of oral contraceptives, menopausal hormone therapy, and other factors was mailed to approximately 3.5 million members of AARP (formerly the American Association of Retired Persons) who were 50–71 years old and were from 6 US states (California, Florida, Louisiana, New Jersey, North Carolina, and Pennsylvania) and 2 metropolitan areas (Atlanta, Georgia; and Detroit, Michigan); 617,119 participants returned the questionnaire. In 1996, a follow-up risk questionnaire was mailed to approximately 339,000 respondents of the baseline questionnaire to collect additional information on diet and medical history, including history of hypertension.

PLCO Cancer Screening Trial

PLCO was designed to evaluate the effectiveness of prostate, lung, colorectal, and ovarian cancer screening modalities on disease-specific mortality rate (34). Approximately 155,000 participants, 55–74 years of age, were enrolled in the study between 1993 and 2001 from 10 US screening centers (Washington, DC; Detroit, Michigan; Salt Lake City, Utah; Denver, Colorado; Honolulu, Hawaii; Minneapolis, Minnesota; Marshfield, Wisconsin; Pittsburgh, Pennsylvania; St. Louis, Missouri; and Birmingham, Alabama). A baseline questionnaire was used to collect information on a variety of factors, including female reproductive history, use of oral contraceptives, and menopausal hormone therapy.

Subject selection and kidney cancer case ascertainment

Persons were excluded from our study if they were male (NIH-AARP: n = 325,174; PLCO: n = 76,693); were not classified as postmenopausal (NIH-AARP: n = 14,590); had not specified age at or type of menopause (e.g., surgical, natural, radiological/chemotherapy) (PLCO: n = 1,559); had not returned a baseline questionnaire (PLCO: n = 2,094); had questionnaires filled out by proxies (NIH-AARP: n = 15,760; PLCO: n = 790); had reported a previous kidney (PLCO: n = 97) or urinary (NIH-AARP: n = 117) cancer at baseline; had reported end-stage renal disease (NIH-AARP: n = 438); were diagnosed with in situ, squamous, or transitional cell cancers of the kidney (NIH-AARP: n = 16; PLCO: n = 20); or had died of an unknown cause, had an undetermined case status because of loss to follow-up, were missing date of death, or withdrew from the study (NIH-AARP: n = 7; PLCO: n = 5). The analytic population consisted of 283,952 postmenopausal women (NIH-AARP: n = 210,300; PLCO: n = 73,652).

Incident cases of primary kidney cancer (International Classification of Diseases for Oncology, Third Edition, code C649) were ascertained within PLCO by annual follow-up questionnaire with subsequent confirmation through medical records, and within NIH-AARP through record linkage to state cancer registries (35). Overall, 792 incident kidney cancer cases (NIH-AARP: n = 601; PLCO: n = 191) were identified during follow-up among women in the analytic population.

Statistical methods

Information on female reproductive factors was gathered via self-administered questionnaires. In both cohorts, number of live births (nulliparous, 1–2, 3–4, and ≥5), age at first live birth (<20, 20–24, 25–29, ≥30 years), age at menopause (<40, 40–44, 45–49, 50–54, ≥55 years), type of menopause (i.e., natural, surgical, radiation, chemotherapy), oral contraceptive use (ever, never), hormone replacement therapy (HRT) use (ever, never), oophorectomy status (yes, no), hysterectomy status (yes, no), and age at hysterectomy (<40, 40–44, 45–49, ≥50 years) were assessed by use of identical categories. The question on oophorectomy status in the NIH-AARP cohort inquired about removal of both ovaries. In PLCO, both unilateral and bilateral oophorectomy data were obtained. Because risk estimates were virtually identical for unilateral, bilateral, and unilateral/bilateral combined oophorectomy status, we present risk as unilateral/bilateral combined oophorectomy status. Age at hysterectomy in NIH-AARP was derived from information collected on hysterectomy status and age at menopause. Different categories were used by the cohorts to collect data on age at menarche (NIH-AARP: ≤10, 11–12, 13–14, ≥15 years; PLCO: ≤11, 12–13, 14–15, ≥16 years), duration of oral contraceptive use (NIH-AARP: never or <1, 1–4, 5–9, ≥10 years; PLCO: never, 1, 2–3, 4–9, ≥10 years), and duration of HRT use (NIH-AARP: never, 1–4, 5–9, ≥10 years; PLCO: never, 1–5, 6–9, ≥10 years). Categories with similar frequencies for these variables were identified and harmonized across the 2 cohorts for inclusion in the analysis. Information on history of benign ovarian tumors, uterine fibroid tumors, and endometriosis was ascertained in PLCO only.

Statistical tests were determined to be significant at a 2-sided P value <0.05. All analyses were conducted in SAS statistical software, version 9.1.3 (SAS Institute, Inc., Cary, North Carolina), unless otherwise stated. Follow-up started at age at baseline (time when reproductive data were collected) and ended at age at kidney cancer diagnosis or age at censoring. Censoring events were death, loss to follow-up, or end of the study (NIH-AARP: December 31, 2006; PLCO: August 31, 2010). Study-specific hazard ratios and 95% confidence intervals relating female reproductive factors to kidney cancer incidence were calculated with Cox proportional-hazards models, with age (in days) as the time metric. Tests for linear trend of kidney cancer risk with ordinal variables were conducted by treating each category as a continuous term (0, 1, 2 …) in the models and were based on the Wald statistics. Proportional-hazards assumptions were checked by adding interaction terms between age and each of the exposures of interest in Cox models; no evidence of violations against proportionality was found.

Final study-specific models were adjusted for potentially confounding variables. For NIH-AARP, variables included body mass index (BMI; weight (kg)/height (m)²) (<25, 25 to <30, ≥30), highest educational level completed (≤11th grade, 12th grade or completed high school, post–high school training other than college, some college, college or postcollege graduate), race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, and Asian, Native Indian, or Alaskan Pacific Islander), and smoking status (never, former, current). We additionally adjusted for history of diagnosed hypertension (yes, no) within the subset of participants who completed the risk factor questionnaire (n = 126,310, of whom 345 developed kidney cancer); the findings were virtually identical to those from the analyses of all participants and are not reported. For PLCO, the variables (categorized identically to NIH-AARP) adjusted for in the final models included BMI, highest educational level completed, race/ethnicity, history of hypertension, and smoking status. History of diabetes (yes, no) did not appear to influence hazard ratio estimates in either cohort when included as a model covariate and thus was not adjusted for in the final models.

Summary hazard ratios were computed with fixed-effects meta-analytic models by combining the cohort-specific risk estimates; summary results from random-effects models were virtually identical (data not shown). Higgin's I² statistic was used to test heterogeneity in hazard ratio estimates across the 2 cohorts (36). Subsequently, a pooled analysis across the cohorts was conducted by merging all subjects into a single data set, and aggregated hazard ratios were estimated with Cox proportional-hazards regression models adjusted for study, BMI, highest educational level completed, race/ethnicity, and smoking status. Indicator variables were created for missing values. Because risk estimates from the meta-analysis and pooled analysis were essentially the same and virtually no evidence of between-study heterogeneity was observed, we present results for pooled analyses only.

Additional analyses of female reproductive factors were conducted with stratification by BMI, smoking status, history of diabetes, race/ethnicity, and history of hypertension. HRT use and age at menopause also were stratified by hysterectomy status. Heterogeneity across strata was assessed by the likelihood ratio test, comparing models with and without the corresponding interaction term.

Sensitivity analyses were performed to evaluate the possibility of detection bias introduced by more intensive medical surveillance after hysterectomy. Age at hysterectomy was assumed to be the midpoint of the 5-year age at hysterectomy categories. A time-dependent covariate, time since hysterectomy (<10 or ≥10 years), was computed by subtracting age at hysterectomy from baseline age. The risk difference between the parameters for women who had a hysterectomy <10 years versus ≥10 years before baseline age was compared by use of a Wald test.

RESULTS

The NIH-AARP and PLCO analytic sets collectively included 283,952 postmenopausal women (Table 1). Overall, 2,771,440.8 person-years were accrued over the 11.2 years of follow-up in NIH-AARP and 14.2 years of follow-up in PLCO, during which 792 incident kidney cancer cases were identified. Most participants (89.4%) were of white race. The median age at baseline for participants was comparable across the studies (NIH-AARP: 62.3 years; PLCO: 63.1 years), as were the prevalences of obesity, smoking, and history of diabetes, as well as educational level. Similarly, the prevalence of hypertension among NIH-AARP respondents of the follow-up risk factor questionnaire was comparable to that seen in PLCO.

Table 1.

Select Characteristics of Postmenopausal Women From the NIH-AARP and PLCO Cohort Studies, United States, 1993–2010

Characteristic or Risk Factor	NIH-AARP Study				PLCO Study				Pooled Study
	Cases		Noncases		Cases		Noncases		Cases		Noncases
	No.	%a	No.	%a	No.	%a	No.	%a	No.	%a	No.	%a
No. of participants	601		209,699		191		73,461		792		283,160
Person-years of follow-up	3466.6		2,047,170.5		1,046.6		719,787.1		4,513.2		2,766,927.6
Age at enrollment, years
<59	158	26.3	72,185	34.4	47	24.6	24,881	33.9	205	25.9	97,066	34.3
60–64	198	32.9	60,997	29.1	73	38.2	22,292	30.3	271	34.2	83,289	29.4
65–69	216	35.9	68,820	32.8	45	23.6	16,267	22.1	261	33.0	85,087	30.0
≥70	29	4.8	7,697	3.7	26	13.6	10,021	13.6	55	6.9	17,718	6.3
Missing	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0
Body mass indexb
<25	174	29.0	89,380	42.6	49	25.7	29,330	39.9	223	28.2	118,710	41.9
25–30	182	30.3	66,119	31.5	65	34.0	25,176	34.3	247	31.2	91,295	32.2
>30	224	37.3	47,538	22.7	75	39.3	17,994	24.5	299	37.8	65,532	23.1
Missing	21	3.5	6,662	3.2	2	1.0	961	1.3	23	2.9	7,623	2.7
Smoking status
Never smoked	250	41.6	91,384	43.6	104	54.5	40,852	55.6	354	44.7	132,236	46.7
Former smoker	254	42.3	81,739	39.0	61	31.9	25,459	34.7	315	39.8	107,198	37.9
Current smoker	80	13.3	30,357	14.5	26	13.6	7,134	9.7	106	13.4	37,491	13.2
Missing	17	2.8	6,219	3.0	0	0.0	16	0.0	17	2.1	6,235	2.2
Highest educational level completed
≤11 years	57	9.5	13,420	6.4	17	8.9	4,519	6.2	74	9.3	17,939	6.3
12 years or completed high school	171	28.5	53,815	25.7	62	32.5	20,263	27.6	233	29.4	74,078	26.2
Post–high school training other than college	60	10.0	22,480	10.7	25	13.1	9,519	13.0	85	10.7	31,999	11.3
Some college	142	23.6	51,871	24.7	46	24.1	17,094	23.3	188	23.7	68,965	24.4
College graduate or postgraduate	139	23.1	61,289	29.2	41	21.5	21,920	29.8	180	22.7	83,209	29.4
Missing	32	5.3	6,824	3.3	0	0.0	146	0.2	32	4.0	6,970	2.5
Race
Non-Hispanic white	528	87.9	187,916	89.6	174	91.1	65,372	89.0	702	88.6	253,288	89.5
Non-Hispanic black	42	7.0	11,712	5.6	10	5.2	4,017	5.5	52	6.6	15,729	5.6
Hispanic	17	2.8	3,856	1.8	3	1.6	1,144	1.6	20	2.5	5,000	1.8
Asian, Pacific Islander, Native Indian, or Alaskan Native	5	0.8	3,218	1.5	4	2.1	2,902	4.0	9	1.1	6,120	2.2
Missing	9	1.5	2,997	1.4	0	0.0	26	0.0	9	1.1	3,023	1.1
History of diabetes
No	526	87.5	193,451	92.3	165	86.4	64,699	88.1	691	87.2	258,150	91.2
Yes	75	12.5	16,248	7.7	16	8.4	4,793	6.5	91	11.5	21,041	7.4
Missing	0	0.0	0	0.0	10	5.2	3,969	5.4	10	1.3	3,969	1.4
Hypertension statusc
No	141	40.9	69,834	55.4	95	49.7	46,735	63.6
Yes	189	54.8	47,552	37.7	90	47.1	24,848	33.8
Missing	15	4.3	8,580	6.8	6	3.1	1,878	2.6
Estimated time since hysterectomy, years
No hysterectomy	291	48.4	114,722	54.7	103	53.9	46,279	63.0	394	49.8	161,001	56.9
<10	27	4.5	9,685	4.6	11	5.8	3,569	4.9	38	4.8	13,254	4.7
10 to <20	93	15.5	28,103	13.4	26	13.6	8,358	11.4	119	15.0	36,461	12.9
20 to <30	116	19.3	36,672	17.5	40	20.9	12,125	16.5	156	19.7	48,797	17.2
≥30	41	6.8	9,943	4.7	11	5.8	2,917	4.0	52	6.6	12,860	4.5
Missing	33	5.5	10,574	5.0	0	0.0	213	0.3	33	4.2	10,787	3.8

Abbreviations: NIH-AARP, National Institutes of Health–AARP Diet and Health Study; PLCO, Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial.

^a Because of rounding error, some categories might not sum to 100%.

^b Weight (kg)/height (m)².

^c Data for hypertension status in NIH-AARP based on risk factor questionnaire responses.

We observed consistent evidence of an association between hysterectomy and increased kidney cancer risk in both the NIH-AARP (hazard ratio (HR) = 1.28, 95% confidence interval (CI): 1.09, 1.50) and PLCO (HR = 1.41, 95% CI: 1.06, 1.88) cohorts (Table 2); adjustment for additional risk factors like use or duration of HRT produced virtually identical results (data not shown). Analyses stratified by HRT use did not significantly modify results between kidney cancer risk and hysterectomy (data not shown). We saw no significant difference in the association with hysterectomy risk by time from hysterectomy to cohort enrollment. Age at hysterectomy was not associated with kidney cancer in either cohort. Because women who had undergone hysterectomy but had intact ovaries might not be truly “menopausal” if hysterectomy had been performed at a younger age before the permanent cessation of ovarian function, we compared cancer risk among women who had a hysterectomy before age 50 to risk among those ≥50 years of age. No difference in risk was observed (HRs = 1.29 and 1.34, respectively). Associations between oophorectomy and kidney cancer risk appeared elevated (pooled HR = 1.17, 95% CI: 1.01, 1.36); however, after stratification by hysterectomy status, associations were no longer observed for those reporting an oophorectomy only (pooled HR = 0.84, 95% CI: 0.53, 1.31).

Table 2.

Renal Cancer Risk^a and Hysterectomy and Oophorectomy Status Among Postmenopausal Women in the NIH-AARP and PLCO Cohort Studies, United States, 1993–2010

Characteristic or Risk Factor	NIH-AARP Study				PLCO Study				Pooled Study
	No. of Cases	HRb	95% CI	P for Trend	No. of Cases	HRc	95% CI	P for Trend	No. of Cases	HRd	95% CI	P for Trend
History of hysterectomy
No	291	1.00			103	1.00			394	1.00
Yes	309	1.28	1.09, 1.50		88	1.41	1.06, 1.88		397	1.32	1.15, 1.52
Age at hysterectomy, years
<40	114	1.00			34	1.00			148	1.00
40–44	65	0.90	0.66, 1.22		17	0.74	0.41, 1.32		82	0.86	0.66, 1.13
45–49	65	1.12	0.83, 1.52		23	1.21	0.71, 2.07		88	1.15	0.88, 1.49
≥50	33	1.06	0.72, 1.56	0.65	14	0.77	0.41, 1.44	0.77	47	0.96	0.69, 1.34	0.87
Oophorectomy status
No	385	1.00			145	1.00			530	1.00
Yes	207	1.16	0.98, 1.37		46	1.19	0.85, 1.66		253	1.17	1.01, 1.36
Oophorectomy and hysterectomy status
Neither	273	1.00			100	1.00			373	1.00
Oophorectomy only	17	0.86	0.53, 1.40		3	0.73	0.23, 2.31		20	0.84	0.53, 1.31
Hysterectomy only	111	1.23	0.98, 1.53		45	1.47	1.03, 2.10		156	1.30	1.08, 1.57
Yes to both	190	1.28	1.06, 1.54		43	1.40	0.98, 2.01		233	1.32	1.12, 1.56

Abbreviations: CI, confidence interval; HR, hazard ratio; NIH-AARP, National Institutes of Health–AARP Diet and Health Study; PLCO, Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial.

^a Statistical tests were determined to be significant at a 2-sided P value <0.05.

^b Cohort-specific hazard ratios were estimated with Cox proportional-hazards regression models adjusted for body mass index, educational level, race, and smoking status.

^c Cohort-specific hazard ratios were estimated with Cox proportional-hazards regression models adjusted for body mass index, educational level, race, hypertension status, and smoking status.

^d Aggregated hazard ratios were estimated with Cox proportional-hazard regression models adjusted for body mass index, educational level, race, study, and smoking status.

Associations with kidney cancer for other female hormonal and reproductive factors are shown in Table 3. A nominally significant inverse association with increasing age at menarche was observed in NIH-AARP (HRs = 0.72, 0.67, and 0.63 for 11–12 years, 13–14 years, and ≥15 years, respectively; P for trend = 0.02). The association in PLCO was in the same direction but did not reach statistical significance (HRs = 1.27, 0.77, and 0.67 for 12–13 years, 14–15 years, and ≥16 years, respectively; P for trend = 0.17). A history of endometriosis was associated with kidney cancer risk (HR = 1.69, 95% CI: 1.09, 2.62) in PLCO; this association did not materially change after adjustment for hysterectomy in the model (HR = 1.58, 95% CI: 1.02, 2.46) (data not shown). Kidney cancer was not associated with parity, number of live births, age at first live birth, age at menopause, or a history of benign ovarian or uterine fibroid tumors.

Table 3.

Reproductive Factors Among Postmenopausal Women in the NIH-AARP and PLCO Cohort Studies and Risk of Renal Cancer,^a United States, 1993–2010

Characteristic or Risk Factor	NIH-AARP Study				PLCO Study				Pooled Study
	No. of Cases	HRb	95% CI	P for Trend	No. of Cases	HRc	95% CI	P for Trend	No. of Cases	HRd	95% CI	P for Trend
Endogenous  hormones
Age at menarche,  years
≤10e/≤11f	56	1.00			38	1.00			94	1.00
11–12e/12–13f	255	0.72	0.54, 0.96		119	1.27	0.88, 1.84		374	0.87	0.69, 1.09
13–14e/14–15f	238	0.67	0.50, 0.90		28	0.77	0.47, 1.26		266	0.74	0.58, 0.94
≥15e/≥16f	52	0.63	0.43, 0.93	0.02	5	0.67	0.26, 1.71	0.17	57	0.69	0.50, 0.97	0.004
Parity
Nulliparous	71	1.00			13	1.00			84	1.00
Parous	521	1.13	0.88, 1.45		178	1.28	0.73, 2.26		699	1.15	0.92, 1.45
No. of live births
0	71	1.00			13	1.00			84	1.00
1–2	215	1.14	0.87, 1.49		46	1.06	0.57, 1.96		261	1.11	0.87, 1.42
3–4	224	1.10	0.84, 1.44		92	1.45	0.81, 2.61		316	1.17	0.92, 1.49
≥5	82	1.20	0.87, 1.65	0.43	40	1.27	0.67, 2.40	0.20	122	1.21	0.91, 1.61	0.15
Age at first live  birth, years
<20	107	1.00			43	1.00			150	1.00
20–24	275	1.12	0.89, 1.41		89	0.81	0.56, 1.19		364	1.02	0.84, 1.24
25–29	100	1.04	0.78, 1.39		33	0.76	0.47, 1.23		133	0.95	0.74, 1.21
≥30	38	1.21	0.83, 1.77	0.35	11	0.74	0.37, 1.46	0.28	49	1.05	0.76, 1.46	0.88
Age at menopause, yearsg
<40	133	1.00			35	1.00			168	1.00
40–44	97	0.86	0.66, 1.12		29	0.93	0.56, 1.53		126	0.87	0.69, 1.10
45–49	154	0.97	0.75, 1.25		41	0.91	0.56, 1.48		195	0.95	0.76, 1.19
50–54	173	0.90	0.69, 1.17		68	1.09	0.67, 1.78		241	0.94	0.74, 1.19
≥55	40	0.91	0.62, 1.33	0.62	18	1.07	0.57, 2.03	0.63	58	0.95	0.68, 1.32	0.85
Exogenous  hormones
Hormone replacement therapy useg,h
Never	297	1.00			57	1.00			354	1.00
Ever	292	0.83	0.69, 0.99		132	1.30	0.93, 1.80		424	0.90	0.77, 1.05
Duration of  hormone  replacement  therapy use,  yearsg,h
Never	297	1.00			57	1.00			354	1.00
≥4e/≥5f	105	0.90	0.71, 1.13		67	1.46	1.02, 2.10		172	1.01	0.84, 1.22
5–9e/6–9f	53	0.66	0.49, 0.89		24	1.42	0.87, 2.33		77	0.77	0.60, 0.99
≥10	133	0.86	0.68, 1.07	0.05	41	1.00	0.65, 1.55	0.95	174	0.86	0.70, 1.05	0.06
Oral birth control  use
Never or <1  yeare/Neverf	415	1.00			93	1.00			508	1.00
Ever or ≥1  yeare/Everf	182	0.82	0.69, 0.99		98	1.02	0.75, 1.38		280	0.87	0.75, 1.02
Duration of birth control use, years
Never or <1  yeare/Neverf	415	1.00			93	1.00			508	1.00
1–4e/1–3f	86	0.87	0.69, 1.10		55	1.18	0.84, 1.68		141	0.96	0.80, 1.17
5–9e/4–9f	61	0.89	0.67, 1.16		23	0.82	0.51, 1.31		84	0.86	0.68, 1.09
≥10	35	0.66	0.46, 0.93	0.02	20	0.93	0.56, 1.52	0.56	55	0.72	0.55, 0.96	0.02

Abbreviations: CI, confidence interval; HR, hazard ratio; NIH-AARP, National Institutes of Health–AARP Diet and Health Study; PLCO, Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial.

^a Statistical tests were determined to be significant at a 2-sided P value <0.05.

^b Cohort-specific hazard ratios were estimated with Cox proportional-hazards regression models adjusted for body mass index, educational level, race, and smoking status.

^c Cohort-specific hazard ratios were estimated with Cox proportional-hazards regression models adjusted for body mass index, educational level, race, hypertension status, and smoking status.

^d Aggregated hazard ratios were estimated with Cox proportional-hazards regression models adjusted for body mass index, educational level, race, study, and smoking status.

^e Risk factor subgroup associated with the NIH-AARP cohort study.

^f Risk factor subgroup associated with the PLCO cohort study.

^g Also adjusted for hysterectomy status (never, ever).

^h Between-study heterogeneity: 0.02.

Inconsistent evidence of association with kidney cancer was observed for other reproductive factors. Use of oral contraceptives was associated with reduced kidney cancer risk in NIH-AARP (HR = 0.82, 95% CI: 0.69, 0.99), where a dose-response relation with duration of use was observed (HRs = 0.87, 0.89, and 0.66 for 1–4 years, 5–9 years, and ≥10 years, respectively; P for trend = 0.02); adjustment for additional risk factors like parity or number of live births revealed virtually identical findings (data not shown). No such associations were observed in PLCO. HRT use was associated with reduced kidney cancer risk in NIH-AARP (HR = 0.83, 95% CI: 0.69, 0.99), though the trend with duration of use was only nominally significant (P = 0.05). No association with HRT use was seen in PLCO. Additional analyses for hormonal and reproductive factors stratified on selected kidney cancer risk factors (BMI, smoking status, diabetes, race, and hypertension) did not suggest the presence of effect modification (data not shown).

DISCUSSION

In this investigation of reproductive factors and kidney cancer risk in the NIH-AARP and PLCO cohorts, women who had undergone a hysterectomy either with or without an oophorectomy were observed to have an approximately 30%–40% increased risk of kidney cancer. This association did not change materially with time since hysterectomy, which is an argument against detection bias as an explanation for our findings. A history of endometriosis was associated with increased kidney cancer risk in PLCO, the only cohort that collected data on this condition. We also observed statistically significant associations with reduced kidney cancer risk in NIH-AARP, but not PLCO, for older age at menarche and use of oral contraceptives and HRT. The inconsistent evidence of association with kidney cancer for these reproductive factors could reflect power limitations for the (smaller) PLCO cohort. Other reproductive factors, including parity and maternal age at first birth, were not associated with kidney cancer in either cohort.

Hysterectomy has been associated with increased kidney cancer risk in previously published epidemiologic studies (16–19, 25–29), although associations reached statistical significance in only a few (16, 19, 25, 27, 28). The only other study of comparable size to ours was a cohort study linking data from the Swedish Inpatient and Cancer Registers, in which a significantly increased risk of kidney cancer after hysterectomy also was observed (25). The findings from that study have been questioned because of its inability to control for obesity (37). However, our findings, which were adjusted for BMI and other established risk factors, are an argument against such confounding as an explanation for this association. In the Swedish study, stronger associations with kidney cancer were observed for earlier age at hysterectomy, whereas in our study, no such differences with age were observed. Differences between the studies in the method of baseline data collection (linkage to clinical data vs. self-report) or average age at which gynecological procedures were performed might account for the inconsistent findings for age at hysterectomy.

It is unclear what biological mechanisms would mediate an association between hysterectomy and kidney cancer risk. It has been speculated that estrogen and progesterone replacement therapy, commonly used among women who have undergone a hysterectomy, might affect the kidneys adversely. Progesterone has been shown to inhibit the kidneys' ability to filter out toxins (38). Estrogen-mediated cell proliferation has been shown in animals to be an early event in the progression of estrogen carcinogenesis (38, 39). However, previous epidemiologic studies have yielded inconsistent findings with regard to use of menopausal hormones and hysterectomy (21, 27, 29, 32). Another possible explanation is renal damage as a result of unintentional ureteral injury during surgery (25, 40). In an earlier study, a high incidence of postrenal obstruction was reported soon after surgery among women who had undergone hysterectomy (41). In a more recent study, radiologically verified persistent hydronephrosis, without recognized injury to the ureter, was observed in patients for up to 6 months after radical hysterectomy (40). Secondary to pelvic anatomy changes after a hysterectomy, the twisting and constricting of the distal ureter could be associated with renal cell proliferation (25). Increased lipid peroxidation after surgery also has been proposed as a possible mechanism responsible for increased kidney cancer risk (42). The impact of gynecological surgery has been shown to alter the rate of lipid peroxidation levels in women, where increased lipid peroxidation can induce DNA damage and promote mutations in proto-oncogenes and tumor suppressor genes (42–44).

Findings from case-control studies (16, 19, 29, 30), though few, generally have found increasing age at menarche to be nonsignificantly associated with reduced cancer risk (16, 29, 30), whereas results for cohort studies (17, 23, 24, 26) typically have shown trends in the opposite direction (17, 23, 24). Earlier age at menarche has been associated with increased risk of obesity in adulthood (45). Although we did adjust for BMI in our models, residual confounding cannot be discarded as an explanation for this finding. It is plausible that subjects with a younger age at menarche in our study were more likely to have excess BMI, a major risk factor for kidney cancer, than were subjects with a later age at menarche. We caution against any firm conclusions about causation, given the inconsistent findings of the epidemiologic literature.

In recent epidemiologic studies, it has been observed that women with endometriosis have a greater risk of certain cancers, including non-Hodgkin lymphoma and endocrine, ovarian, and breast cancers (46–48). The relation between endometriosis and kidney cancer had been examined previously in a series of epidemiologic studies involving analysis of data from the National Swedish Inpatient Registry (47–49), the latest of which showed a statistically significant 36% increase in risk (49). The carcinogenic potential of endometriosis is poorly understood but is speculated to be multidimensional in cause, involving genetic, hormonal, and immunological factors (48).

The associations with HRT and oral contraceptive use observed in NIH-AARP are not clearly supported by previously published studies (16–19, 21, 23, 24, 26–30, 32). The published epidemiologic evidence involving HRT use and kidney cancer has generally been null (16–18, 23, 24, 26–30, 32). To date, only 1 study has shown HRT use to significantly increase kidney cancer risk (19), though no trend with duration of use was seen. Likewise, a statistically significant inverse association between oral contraceptive use and kidney cancer risk has been reported in only 1 previous study (30); typically, findings for oral contraceptive use and kidney cancer risk have been null (16–19, 21, 23, 24, 26–29). The inverse association observed between kidney cancer risk and use of HRT and oral contraceptives in the NIH-AARP cohort, if real, could be related to estradiol, which has been shown to lower blood pressure levels and to inhibit oxidative stress and lipid peroxidation (42). Still, properties of other sex hormones in these medications cannot be excluded as a plausible mechanism of effect.

To our knowledge, our investigation, involving an analysis of data from 2 large prospective cohorts that collected detailed information on female reproductive factors and potential confounders, represents the most comprehensive investigation of reproductive factors and kidney cancer risk conducted to date. Limitations of our study include the possibility of misclassification due to self-reported hormonal and reproductive factors collected at baseline, which were not validated in either cohort. Such misclassification, however, most likely would have been nondifferential in nature, biasing estimates toward the null. In NIH-AARP, only a portion of prevalent urinary cancers were captured by cancer registries at baseline because cancer registry linkage was conducted only a few years before baseline. Therefore, a small number of prevalent urinary cancers at baseline might have been included in our analyses. However, given that only 0.2% percent of PLCO participants reported a history of kidney cancer at baseline (these subjects were excluded from the analysis), the presence of a similarly small number of prevalent cases in NIH-AARP would not have materially affected our results.

In summary, our analysis of reproductive factors and kidney cancer risk provides consistent evidence of an association between hysterectomy and increased kidney cancer risk. This association, if real, represents a potentially important cause of this cancer, because hysterectomy is the second most common surgical procedure for women of reproductive age in the United States, performed on 1 of every 3 women by age 60 years (50). Our findings that suggest a possible association with kidney cancer for menarche, use of oral contraceptives and HRT, and endometriosis also warrant further investigation.