Dairy consumption and risk of testicular cancer: a systematic review

Abstract

Testicular cancer (TC) is the most common cancer to occur among young men – and yet the factors that cause this cancer remain poorly understood. One of the most investigated exposures is dairy consumption, but it remains unclear whether this is an important exposure in the development of TC. In this systematic review, we aimed to answer whether a) men who consume high amounts of dairy products are at increased risk of TC; b) the risk of TC depends on the type of dairy product consumed; c) the risk of TC increases with increasing levels of dairy product consumption; and d) if dairy consumption during certain periods during the lifecourse are more strongly associated with TC development than others. There was no strong evidence that high levels of dairy consumption are associated with risk of TC, and conflicting evidence of a dose-response relationship. We found inconsistent evidence on whether certain types of dairy are more strongly associated with TC than others, as well as conflicting evidence that exposure during certain life-course periods affects TC risk more than other periods. There is no consistent evidence to support the premise that dairy product consumption is associated with the risk of TC development.

Introduction

Testicular cancer (TC) is the most common cancer among young men aged 15–40 years.[1] Cryptorchidism, prior unilateral TC, family history and increased adult height are known risk factors.[2] However while it is thought that a combination of genetic and environmental factors may increase the likelihood of TC,[3–6] these risk factors remain poorly understood despite being highly-researched.[2]

Among the most highly-researched exposures investigated in this context is dairy consumption. In case-control studies, several authors have found associations between certain dairy products[7–9] or components of dairy products and TC;[10] for example, Garner et al.[7] observed that a high consumption of cheese was associated with an elevated risk of TC (adjusted odds ratio [OR] = 1.87; 95% CI, 1.22–2.86). Comparably, Davies et al[11] observed that a higher consumption of milk was associated with an elevated TC risk with an adjusted odds ratio of 1.39 (95% CI, 1.19–1.63) for each additional quarter pint of milk drunk at age 17. However, different studies have measured dairy consumption in different ways and at different life-course stages, making comparisons difficult. Additionally, there is conflicting evidence regarding an association between dairy consumption and subsequent development of TC, with several authors finding an association[7, 8, 11] and others finding no[12, 13] or weak/limited evidence of an association.[14]

Why has dairy consumption been so frequently implicated as a potential risk factor for the development of testicular cancer? The reason for this may be partly due to the numerous attractive ecological observations that have implicated dairy as a potentially-important driver of increasing global rates of testicular cancer. Consumption of dairy is highest in developed countries,[1, 2] and these countries also have higher risks of TC.[2, 8] Furthermore, the incidence of TC has risen markedly in a number of countries since the 1930’s, at the same time as consumption of dairy products has risen largely due to the advent of refrigeration and pasteurisation.[8, 15] Finally, populations with the highest incidence rates of TC – white Northern Europeans[16] – are also the populations least likely to suffer from lactose intolerance, and so most likely to eat dairy throughout their lives.^[2]

Each of these ecological observations are consistent, in that, if in fact consumption of dairy products does increase the risk of TC, they help to explain why rates of TC are increasing over time. However, they each have the same problems common to all ecological research, including an inability to account for confounding and ecological fallacy – where an assumption at a population level is misinterpreted as an association at an individual level.[17] Therefore, this evidence cannot be considered more than hypothesis-generating.

The following questions remain unanswered: are men who consume high amounts of dairy products at increased risk of TC compared to those who consume low amounts or none at all? Does the risk of TC depend on the type of dairy product consumed? Does the risk of TC increase with increasing levels of dairy product consumption (i.e. is there a dose response relationship)? Finally, is there a certain period during one’s lifetime (e.g. childhood, adolescence, adulthood) at which consumption of dairy products affects the risk of TC? Given the widespread consumption of dairy products within many countries, the answer to these questions is an area of public health importance.

In order to address these questions, we have conducted a systematic review of studies that have investigated the association between the consumption of dairy products and development of TC.

Material and Methods

Search strategy

This study was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.[18] In addition, we used previous relevant systematic reviews and meta-analyses as exemplars.[19–21]

Protocol and registration.

We registered this review in the International Prospective Register of Systematic Reviews (PROSPERO, registration No. CRD42016052039), describing in advance the aims and methods of our investigation.[22]

Eligibility criteria.

The PICOS (Patient/Participant, Intervention, Comparator, Outcome, Study design) criteria used to construct this review are presented in the appendices (A1). Abstracts included in the final analysis included studies which reported associations between consumption of dairy products and TC. Studies were only included if data was provided from which summary associations (odds ratio or relative risks) and their 95% confidence limits were able to be calculated, or if these summary associations were provided by the authors themselves.

Information sources.

A systematic review was conducted on 15 November 2016 for all articles published up until that time. No limits were set in terms of language used or study design during the initial abstract search. The search was conducted using Ovid Medline, Embase, Scopus and Web of Science. References were collected and logged in EndNote vX7.1 (Thomson Reuters, New York, U.S.A.).

In those cases where an eligible study contained inadequate information, the corresponding author for the given paper was emailed with a request for further data. If key data were not provided, the study was deemed ineligible and removed from further analysis.

Search.

Using a Boolean approach, we searched the electronic databases for each possible combination of the following keywords (Table 1). The reference list of those studies which were considered eligible for inclusion were scanned for additional relevant studies.[23]

Table 1:

Search terms used during systematic review of the literature

Exposure-related keywords	Outcome-related keywords
Dairy Product	Cancer of the test*
Milk, Milk fat	Seminoma*
Cream	Non-seminoma*
Yoghurt, Yogurt	Germ cell*
Cheese	Teratoma
Butter	Carcinoma
Casein, whey, curd	Choriocarcinoma
Galactose, lactose	Neoplasm, Tumour

Study selection

Screening of abstracts.

Figure 1 shows the flow chart for study identification and inclusion. Duplicate records were removed prior to abstract screening. Abstracts were screened by one reviewer (VS) to remove irrelevant studies, with a 10% random sample of these verified by a second reviewer (SH). Any disagreements about inclusion were resolved by referral to a third reviewer (JG). The full text of all remaining papers were obtained and assessed by two reviewers (VS and SH) to identify those which met our inclusion criteria, with the same secondary review process completed as per abstract screening.

Figure 1:

Flow chart of literature search

Table of citations included and excluded, with justification for why they were excluded.

All papers that were considered relevant during the abstract screening process but ineligible for inclusion in our final analysis are listed in the appendices, along with justification(s) for why they were ultimately excluded (A2).

Data collection process and data items

For each included study, two reviewers (VS and SH) independently extracted study meta-data. In those cases where the two reviewers disagreed with respect to a data item, the same secondary review process was completed as per abstract screening.

Assessment of risk of bias (individual and across studies).

There is no gold standard measure of study quality for observational research despite the assessment of study quality and potential for bias being an essential feature of any systematic review. However, it has been recommended that any tools used to measure study quality should be as specific as possible to the given topic, and involve a simple checklist as opposed to a scale or score.[24] Given these factors, we assessed study quality and potential for bias using the criteria outlined in the Newcastle-Ottawa Quality Assessment Scale,[25, 26] but did not determine a quality score.[27] Two reviewers (VS and SH) independently assessed study quality against these criteria, with any disagreements resolved by referral to a third reviewer (JG).

Results

Our systematic review identified eight studies that reported associations between dairy consumption and risk of TC (Figure 1). Meta-data for each of the eight included studies are presented in Table 2. All eight studies were case-control studies, four of which were population-based[7, 9, 11, 14] and four hospital-based.[10, 13, 28, 29] The sample sizes of these eight studies ranged markedly, with the smallest number of cases being 50 and the highest 754 (range for controls: 48 to 928). Four of these studies were conducted in North America (USA[10, 13, 14] and Canada[7]), three in Continental Europe (Italy[28, 29] and Germany[9]) and one in the United Kingdom.[11] All studies measured exposure status by self-report utilising either self-completed, computer-assisted or interviewer-administered questionnaires.

Table 2:

Papers included in association of dairy and testicular cancer, with meta-data

Author	Year of publication	Study design	Study period	Location of study	Sample size	Source of case/outcome data	Source of control/total cohort	Exclusion criteria	Method of outcome measurement	Method of exposure measurement
Bonner et al.	2002	CCS	1982 – 1998	Buffalo, New York, USA	117 cases /334 age strata matched controls	Roswell Park Cancer Institute	Roswell Park Cancer Institute – treated with condition other than cancer	Non-germ cell tumours, <18 and <70 years old, non-Caucasian, diagnosis of condition that may affect dietary intake (controls)	Clinical diagnosis	Retrospective self-report questionnaire
Davies et al.	1996	CCS	1981 – 1991	East Anglia, United Kingdom	200 cases/ 400 age matched population controls /400 year of diagnosis matched cancer controls	East Anglian Cancer Registry	Population Controls – 82 GP registries Cancer Controls - East Anglian Cancer Registry	Non-germ cell tumours, not alive at start of study cases and cancer controls, not well enough to answer questionnaire (according to GP)	Clinical diagnosis	Retrospective self-report questionnaire Mothers self-report responses used for validation
Garner et al.	2003	CCS	1994 – 1997	Canada	601 cases/744 controls age and locality matched	Cancer registries of 8 Canadian provinces	Cancer registries of 8 Canadian provinces	Non-germ cell tumours, >55 years old, non-White, missing exposure information, cases with incorrect ICD topography code, questionnaire completed by proxy or incomplete, people with daily energy intake >5,000kcal or <700kcal	Clinical diagnosis	Retrospective self-report questionnaire
Giannandrea et al.	2011	CCS	2006 – 2008	Italy	50 cases /48 age, BMI and locality matched controls	Department of Seminology and sperm bank, University of Rome. Pts with orchiectomy for sperm banking	Healthy men in same Department for andrological exam	Non-germ cell tumours, <18 and >45 years, no other cancer	Clinical diagnosis	Face to face interview with structured questionnaire
McGlynn et al.	2007	CCS	2002 – 2005	USA	767 cases/928 controls 720 age, ethnicity, serum sample date matched pairs	Department of Defence Serum Repositary (DoDSR), Participants with at least one serum sample stored	DoDSR Participants with at least one serum sample stored	Non-germ cell tumours, >18 and <45 years, non-servicemen, non-active duty, deployed to combat zone	Clinical diagnosis	Computer assisted telephone interview with structured questionnaire
Paoli et al.	2015	CCS	Information not supplied	Italy	125 cases/103 BMI and locality matched controls	Single hospital study - Pts with orchiectomy for sperm banking, University of Rome	Single hospital study - Men in same Dept for andrological exam to ascertain fertility	Non-caucasian	Clinical diagnosis	Face to face interview with structured questionnaire
Sigurdson et al.	1999	CCS	1990 – 1996	Texas, USA	160 cases/136 age, ethnicity and locality matched controls	University of Texas, MD Anderson Cancer Centre tumour registry and Genitourinary Oncology Centre	Healthy friends of cases with same State of residence and of same ethnicity, age +/− 5 years	<18 and >50 years, non-resident of Texas, Louisiana, Arkansas, Oklahoma, not alive at data collection	Clinical diagnosis	Retrospective self-report questionnaire
Stang et al.	2006	CCS	1995 – 1997	Germany	269 cases/797 age and locality matched controls	Active reporting of clinical and pathology departments, and Hamburg Cancer Registry	Mandatory residence lists in same 5 cities as cases	Non-germ cell tumours, <15 and >69 years, unable to complete interview	Clinical diagnosis	Face to face interview with structured questionnaire Mothers self-report responses used for validation

Study quality

The assessment of study quality via the Newcastle-Ottawa criteria is presented in the appendices (see A3). The assessment revealed some variation in response rate between studies. In order to maximise the comparability of cases and controls, all eight studies either matched or controlled for age, and controlled for other factors (including five studies that controlled for cryptorchidism). Two studies reported a similar response rate between cases and controls, while for three studies there was some difference in response rate (ranging from 16% to 43% difference between cases and controls). Three further studies did not provide this information. Each of the eight studies relied on retrospective recall to ascertain the level of exposure to dairy products. The potential for recall (and misclassification) bias is discussed later in this paper. Small sample size led to poor precision for some studies, particularly when analyses were stratified by histological subtype. As such, we were limited in our ability to make meaningful inferences regarding differential associations for seminoma and non-seminoma/mixed tumours.

Measurement of dairy consumption

The way in which consumption of dairy products was examined differed markedly across studies, in terms of types of dairy consumed, the unit of measurement for consumption and timing of exposure. This is shown in appendices (A4). Because of the disparate means by which dairy product consumption was measured, meta-analyses of the data could not be performed as originally planned. However we were able to describe the current state of the evidence with respect to the association between consumption of dairy products and subsequent development of TC. Adjustment for confounding had little impact in papers where both crude and adjusted measures were presented, so in this manuscript adjusted risk estimates are used for comparison between studies.

High vs. Low Dairy Consumption

In order to compare dairy consumption across studies we have categorised measures into highest vs. lowest consumption of dairy, regardless of the unit of measurement used. In total, seven studies presented data on an association between high relative consumption of dairy products (compared to low consumption) and risk of TC.[7, 9, 10, 13, 14, 28, 29] Six of these studies investigated dairy consumption during adulthood, one during adolescence alone, and one across during adolescence and adulthood. Pertaining to adulthood, four of these studies investigated total dairy, [7, 14, 28, 29] three milk, [7, 10, 13] and one cheese and butter.[7] Pertaining to adolescence, one study investigated specific components of dairy consumed at age 17 years and their association with TC,[9] while the other, a large study by McGlynn et al. investigated total dairy consumption during both adulthood and adolescence.[14] We present the findings in a forest plot (Figure 2), and below according to type of dairy product and by life-course timing.

Figure 2:

Forest plot, showing measures of association between high and low levels of dairy consumption, by timing during life-course.

Total dairy in adulthood.

In an Italian study of 125 cases and 103 controls, Paoli et al.[29] reported an adjusted OR of 2.37 (95% CI 1.01–5.52) for high consumption of dairy and TC. Similarly, Giannandrea et al.[28] reported a possible association between high consumption of dairy products when compared with low consumption and risk of TC among 50 cases and 48 controls in Italy, although their estimates are imprecise (adjusted OR 2.55, 95% CI 0.76–8.50). In a larger Canadian study of 601 cases and 744 controls, Garner et al.[7] divided total dairy intake into quintiles and found evidence of higher odds of TC among those with intakes in quintiles 3 and 4 compared with those in the lowest quintile (ORs quintile 3 vs. 1: 1.79, 95% CI 1.23–2.62; quintile 4 vs. 1: 2.56, 95% CI 1.73–3.78). The association for those with the highest consumption in quintile 5 (compared with 1) was less clear (1.40, 95% CI 0.93–2.12). In contrast, in a large US study of 767 cases and 928 controls, McGlynn et al. [14] found no evidence of association between high consumption of total dairy products at any age and risk of TC (OR 0.64, 95% CI 0.29–1.45).

Milk in adulthood.

Similarly to their findings relating to total dairy intake, Garner et al.[7] found evidence of an association with TC for moderately high consumption of milk (quintile 4 vs. quintile 1 OR 1.52, 95% CI 1.08–2.16), but less clear associations for those in the highest consumption group (OR quintile 5 vs. quintile 1, 1.23, 95% CI, 0.79–1.92). However, in their US study of 117 cases and 334 controls, Bonner et al. [13] found no association between high consumption of milk and total TC, with an adjusted OR of 0.93 (95% CI 0.40–2.18) for men who consumed two or more glasses a milk per day vs. those who drank none. In a US study of 136 cases and 160 controls, Sigurdson et al.[10] also found no association between high consumption of milk and TC. These authors only investigated TC stratified by histological subtype (i.e. no total TC is reported); seminoma (adjusted OR 0.6, 95% CI 0.2–1.7), non-seminoma (OR 0.5, 95% CI 0.2–1.6) or mixed cell (OR 0.4, 95% CI 0.1–1.7).

Cheese and butter in adulthood.

In terms of other types of dairy during adulthood, Garner et al. [7] found that high levels of cheese consumption were associated with the development of TC. This was apparent for quintile 5 vs. quintile 1 (OR 1.87, 95% CI 1.22–2.86) as well as quintiles 4 and 3 vs. quintile 1 (ORs 1.92, 95% CI 1.26–2.91 and 1.60, 95% CI 1.13–2.25 respectively). In this study, high consumption of butter was not associated with TC risk, although insufficient data prevented comparisons between quintile 5 and quintile 1.

Childhood or Adolescence.

Two studies investigated high vs. low dairy consumption during childhood or adolescence.[9, 14] In a German study of 269 cases and 797 controls, using mother-son pairs Stang et al.[9] asked mothers directly about their sons’ consumption at age 17, while the sons’ consumption at age 17 was estimated from their responses regarding current consumption. In comparison, McGlynn et al.[14] directly asked participating men about their consumption of dairy products at specific time intervals during their childhood and adolescence.

Stang et al. [9] observed that high consumption of milk fat at aged 17 was associated with TC development based on the sons responses (quartile 4 vs. quartile 1, adjusted OR 1.80, 95% CI 1.12–2.89), but not on mothers responses (quartile 4 vs. quartile 1, OR 1.25 95% CI 0.60–2.62). High intake of galactose was not associated with TC for either sons or mothers (quartile 4 vs. quartile 1, OR 1.46 95% CI 0.89–2.39 and 1.36 95% CI 0.66–2.79, respectively). However, McGlynn et al.[14] did not find an association between high consumption of dairy products and TC at any time during the life-course. The authors report adjusted OR of 0.64 (95% CI 0.29–1.45) for dairy consumed at any age, 0.92 (95% CI 0.49–1.76) for birth to kindergarten, 0.73 (95% CI 0.35–1.50) for grades 1–5, 0.53 (95% CI 0.26–1.09) for grades 6–8 and 0.80 (95% CI 0.48–1.34) for dairy consumed during grades 9–12.

Dose Response

Five studies investigated the relationship between varying levels of dairy consumption and development of TC.[7, 9–11, 13] Three of these studies found a weak or no association,[7, 10, 13] and two[9, 11] found that increasing consumption of dairy was associated with increasing TC risk (i.e. a positive association).

Bonner et al.[13] investigated milk consumption in adulthood and found no association between increasing levels of milk consumption and subsequent risk of TC (Total TC adjusted ORs: <1 glasses/day vs 0, 0.76, 95% CI 0.31–1.89; 1–2 glasses/day vs 0, 0.56, 95% CI 0.25–1.23; >2 glasses/day vs 0, 0.93, 95% CI 0.40–2.18). Similarly, no dose response patterns were observed for seminoma, non-seminoma or mixed cell TC.

Garner et al.[7] investigated the level of consumption of total dairy and milk in the usual diet two years before the men completed the questionnaire, and observed some evidence of increasing odds of TC as consumption increased, although this dropped-off at the last quintile and was not statistically significant (quintile 2 vs. quintile 1, adjusted OR 1.39, 95% CI 0.96–2.02; quintile 3, OR 1.79, 95% CI 1.23–2.62; quintile 4, OR 2.56, 95% CI 1.73–3.78; quintile 5 OR 1.40, 95% CI 0.93–2.12; p for trend 0.12). This pattern was similar for seminoma and non-seminoma tumour sub-types. Milk consumption showed a similar pattern to that of total dairy, in which increasing milk intake was not consistently associated with an increased risk of TC. In each sample group the association was strongest for quintile 4 compared with quintile 1 – total TC (adjusted OR 1.52, 95% CI 1.08–2.16), seminoma (adjusted OR 1.34, 95% CI 0.89–2.02) and non-seminoma (adjusted OR 1.75, 95% CI 1.00–3.08) – with a weaker association for quintile 5 (total TC adjusted OR 1.23, 95% CI 0.79–1.92, seminoma 1.35, 95% CI 0.80–2.27, non-seminoma 1.09, 95% CI 0.52–2.26). Similarly, Sigurdson et al.[10] investigated adult milk consumption in the year before diagnosis for cases and the year before questionnaire completion for controls, and found no consistent trend in terms of dose-response. For men diagnosed with non-seminoma adjusted ORs showed a decreasing trend from 1.5, 95% CI 0.6–3.6 (quartile 2 vs quartile 1) to 1.3, 95% CI 0.5–3.3 (quartile 3 vs quartile 1) and to 0.5, 95% CI 0.2–1.6 (quartile 4 vs quartile 1), but again this trend was not statistically significant (p for trend 0.32). A similar pattern was observed for men with mixed germ cell TC with adjusted OR of 1.4 (95% CI 0.4–4.5 quartile 2 vs quartile 1), 1.3 (95% CI 0.4–4.2, quartile 3 vs quartile 1) and 0.4 (95% CI 0.1–1.7, quartile 4 vs quartile 1) (p for trend 0.26). For men diagnosed with seminoma, adjusted OR’s were variable 0.7, 95% CI 0.3–2 (quartile 2 vs quartile 1), 0.9, 95% CI 0.3–2.4 (quartile 3 vs quartile 1) and 0.6, 95% CI 0.2–1.7 (quartile 4 vs quartile 1) (p for trend 0.41).

Conversely, in a UK study of 129 cases and 184 population controls, Davies et al.[11] investigated milk consumption in adolescence and reported a positive association with an adjusted OR increase of 1.39 (95% CI 1.19–1.63) for each quarter pint of milk consumed at age 17 compared to population controls. A weaker association was observed when TC cases were compared with cancer controls (OR 1.14 95% CI 1.00–4.04). A positive dose-response association was also reported by Stang et al.[9] with an adjusted OR increase of 1.37 (95% CI 1.12–1.68) for each additional 20 serves of milk (200 mL/serve) per month during adolescence, as reported by the sons themselves. However, when mothers were asked the same question, this association dropped to 1.21 (95% CI 0.86–1.70). A similar result was seen for low-fat milk, with an apparent dose response based on son’s responses but not for mothers. However no dose response relationship was seen for yoghurt in either mother’s or son’s responses. In contrast to Garner et al.,[7] Stang et al.[9] observed similarly-conflicting results in the context of cheese consumption, reporting an adjusted RR of 1.27 (95% CI 0.99–1.63) for sons and 0.83 (95% CI 0.56–1.22) for mothers for each additional 20 serves of cheese per month.

Ever-use of Dairy Products

One study assessed no-use versus ever-use of dairy products and risk of TC.[13] In a small hospital-based study Bonner et al.[13] investigated milk consumption in adulthood, and found no association between the consumption of milk and subsequent development of TC (adjusted overall OR for no consumption vs. any consumption of milk: 0.80, 95% CI 0.39–1.63).

Timing of Dairy Consumption

While Davies et al.[11] found that for each quarter pint of milk consumed at age 17 there was a 39% increased risk of TC (adjusted OR 1.39, 95% CI 1.19–1.63) overall the results of this review do not provide strong evidence that life-course timing of dairy consumption is important to TC risk. Stang el al.[9] observed conflicting results when asking sons and their mothers about the sons’ adolescent dairy consumption. However the largest study of the association between dairy consumption and TC at any age between birth, grade 12 and adulthood was conducted by McGlynn et al.[14] The authors found no evidence of an association at any stage of the life-course.

Tumour histology

Five studies investigated the association between dairy consumption and risk of TC by tumour histology, comparing seminoma and non-seminoma[7, 9, 14] and in two studies also mixed cell TC.[10, 13] One paper only investigated TC by tumour histology and did not report any results for total TC.[10] Results were mixed with overall results inconclusive of an association between dairy and the histological subtypes of TC.

For example, Stang et al.[9] reported some evidence of an association between milk consumption and seminoma based on son’s (adjusted RR 1.66, 95% CI 1.30–2.12) and mother’s responses (RR 1.45, 95% CI 0.96–2.21) but none for non-seminoma, with similar findings for low-fat milk, cream and cheese. McGlynn et al.[14] reported that consumption of 2% milk seems to be associated with total TC (adjusted OR 1.53, 95% CI 1.11–2.12) as well as its histological sub-types: seminoma (OR 1.56, 95% CI 1.00–2.43) and, non-seminoma (OR 1.50, 95% CI 1.04–2.17). However, the authors found no such association for any other dairy product, nor dairy consumption overall.

Discussion

The eight studies included in this review measured and analysed the exposures in a highly heterogeneous way, and even in those instances where we attempted to determine if patterns existed by pooling results from studies with similar measures of exposure, the evidence remained inconclusive. For example, while one study observed an association for a moderate consumption of milk[7] this association was not observed across other studies that investigated high vs. low milk consumption.[10, 13] Results were similar for total dairy products. As shown in the forest plot (Figure 2), the results comparing high vs. low consumption of dairy varied from an apparent positive effect to an apparent protective effect. Given the inconsistency of results, it appears that there is not good evidence to support the premise that high consumption of dairy products is associated with the risk of TC.

Regarding types of dairy products consumed there is some evidence that a high intake of cheese is associated with the subsequent development of TC. However, this was observed in one study[7] but not another.[9] Likewise, there is some evidence that consumption of certain types of milk may be associated with TC risk, with 2% milk associated in one study[14] and low fat milk in another[9] although overall the evidence towards milk was contradictory and uncertain. Regarding dose-response relationships, two[9, 11] of five studies found a positive association in which increasing consumption of dairy was associated with increasing TC risk, but the remaining three studies found no consistent trend in terms of dose-response. Similarly, the evidence for dairy consumption during adolescence and development of TC was conflicting. One study that did find a positive association[11] estimated adolescent exposure based on reported current consumption, while another which used a similar method of estimated adolescent exposure to dairy had mixed findings – with exposure data sourced from mothers not matching data sourced from their sons.[9] However, perhaps the best evidence comes from the large study by McGlynn et al.[14] who found no evidence of an association between intake of dairy at any life-course stage and TC or its histological sub-types.

So why may there be a lack of clear evidence linking dairy products and the subsequent development of TC? First, we must consider the possible impact of the highly-heterogeneous manner in which dairy consumption was measured, categorised and reported by the included studies. It is unlikely that associations would be similar in magnitude given these underlying methodological differences; however, despite them, we might still expect associations to be in the same direction if underlying associations indeed exist. In fact, in no instance are there unambiguous associations between dairy as a whole, or within sub-categories of dairy, or within particular periods of the life course.

Second, dairy consumption is a near-ubiquitous exposure, particularly in Western contexts where testicular cancer risk is greatest. If the majority of a population is exposed to a risk factor, it becomes very difficult to disentangle the true relationship between the risk factor and the outcome. The one study that included non-dairy consumers as an exposure group found no strong evidence of differential testicular cancer risk compared to dairy consumers – however, this was based on a small number of participants.[13] If an association between dairy consumption and TC risk was strong, it might be possible to observe a dose-response relationship, even in populations where dairy consumption is ubiquitous; however, we observed mixed (or at best weak) evidence that such a relationship exists.

Third, it is likely that the included studies were subject to (potentially substantial) misclassification of the exposure. Most studies asked about dairy consumption that occurred several decades prior (thus opening the possibility of poor recall), and other studies based exposure during childhood on dairy consumption during adulthood. Provided that exposure misclassification did not occur differentially between those who developed testicular cancer and those who did not, then the impact of misclassification bias on the resulting estimates would be likely to move estimates of any associations toward the null. However, since all included studies were case-control studies (and thus case status was known prior to the collection of data on the exposure), recall bias cannot be ruled out. In this context, recall bias may exaggerate any association between dairy consumption and testicular cancer risk – since men (and/or their mothers) that have been diagnosed with this disease may be systematically more likely to report (correctly or incorrectly) high levels of exposure to dairy.

Lastly, we must note that all of the included studies measured dairy consumption in adulthood, with some inferring childhood exposure based on adult exposure. By measuring consumption during adulthood, it is likely that the included studies missed the most important window of exposure in terms of the potential impact of dairy on TC development – i.e. early-life.

Biological plausibility of dairy as an exposure

There are multiple avenues by which dairy consumption has been purported to cause testicular cancer, one of which is consumption of organochlorine compounds through dairy consumption. Dairy products are an important source of human exposure to OC.[30] Historically, organochlorine compounds were widely used in pesticides until they were banned in many countries in the 1970’s and 1980’s, yet they can persist in the environment. Organochlorine compounds act as endocrine disrupting chemicals and interfere with normal hormonal functioning; although the exact process of this remains unknown. Interestingly, the studies by Giaandrea et al.[28] and Paoli et al.[29] both observed increased serum levels of organochlorine compounds among cases when compared to controls; however, this does not mean necessarily that these compounds were derived from dairy consumption.

Second, dairy products contain significant amounts of female sex hormones – estrogen and progesterone – which may increase the risk of TC development.[31] Further research is needed to quantify both the role of these hormones in TC development, and the extent to which dairy consumption is an important source of these hormones for men.

In summary, while there have been conflicting results with respect to the relationship between dairy consumption and risk of testicular cancer, the lack of clarity may be largely derived from heterogeneity in the way in which this relationship has been investigated. The study with the largest sample and a stud design which included investigating dairy product intake across childhood, adolescence and adulthood did not find any evidence to link dairy and TC risk [14] – which is the perhaps the strongest clue that there is not a strong or important causal relationship between dairy consumption and testicular cancer development. In their review of the adolescent and adult risk factors for testicular cancer, McGlynn and Trabert [2] suggest that diet – and specifically dairy product consumption – is indeed unlikely to increase the risk of developing TC.

Strengths and limitations of review

The strengths of this review include a comprehensive literature search, the use of a PICOS statement and adherence to PRISMA guidelines, and a thorough assessment of studies for risk of bias against Newcastle-Ottawa Quality Assessment Scale.

A limitation of this review is that, due to heterogeneity in exposure measurement and group definition in previous studies, we were unable to group studies in a way that would have enabled meta-analyses to be performed. We recognise that meta-analyses would be a useful addition to the literature, but instead chose to provide forest plots without pooled estimates (Figure 2).

Future recommendations

Examining the hypothesis that dairy consumption causes testicular cancer is a difficult task, which is perhaps reflected in the heterogeneity of the studies that have been undertaken in this area. In order to robustly examine this relationship, we would require a longitudinal cohort study – probably a birth cohort study – which collected dietary exposure data over a long period of time, and then linked this cohort to TC outcomes 30–40 years later. It does not seem feasible to conduct such a study to examine this one exposure, or even to examine only TC; rather, to be worthwhile such a study would need to be part of a larger study examining the role of diet (and specifically dairy consumption) on all cancer outcomes (and possibly other health outcomes). The European Prospective Investigation in Cancer and Nutrition (EPIC) study may provide some insight in this regard in the future.[32] Alternatively, identification of long-term biomarkers that robustly indicated early-life dairy exposure would allow us to examine this relationship free of recall bias, and would not require the decades of data collection of a cohort study; however, such long-term biomarkers are not presently available (and may never become available).

Conclusions

We observed no strong evidence that high levels of dairy consumption are associated with risk of TC. We observed conflicting evidence of a dose-response relationship between dairy consumption and risk of TC. In addition we found conflicting evidence that certain types of dairy are more strongly associated with TC than others, as well as conflicting evidence that exposure during certain life-course periods affects TC risk more than other periods. At the current time, there is no consistent evidence to support the premise that the consumption of dairy products is associated with the risk of TC development.

Appendices

Appendix 1:

PICOS (Patient/Participant, Intervention, Comparator, Outcome, Study design) criteria for inclusion of studies.

PICOS Element	Inclusion Criteria	Exclusion Criteria
P articipant	Males	Animal studies
I ntervention (i)	Consumption - ever-use
(ii)	Frequency of consumption – high, dose response
(iii)	Timing of consumption - childhood
C omparator (i)	No consumption
(ii)	Frequency of consumption – low, dose response
(iii)	Timing of consumption - adulthood
O utcome (i)	Any testicular cancer
(ii)	Seminoma	Non-seminoma
(iii)	Non-seminoma	Seminoma
S tudy design	No limit placed on initial search criteria	Studies where no association between exposure and outcome were reported

Appendix 2:

List of excluded papers with reason for exclusion

Author	Year	Title	Reason for exclusion
Dieckmann, K. P et al.	2008	Tallness is associated with risk of testicular cancer: evidence for the nutrition hypothesis	No assessment of dairy consumption exposure
Ganmma et al.	2002	Incidence and mortality of testicular and prostatic cancers in relation to world dietary practices	No association between dairy consumption and outcome investigated or reported
Ganmma et al.	2003	The experience of Japan as a clue to the etiology of testicular and prostatic cancers	No association between dairy consumption and outcome investigated or reported
Hu, J. et al.	2011	Dietary transfatty acids and cancer risk	Does not report dairy consumption exposure
Krumwiede et al	1999	Nutritional factors in the aetiology of testicular tumours	Review manuscript (no primary data)
Lerro, McGlynn & Cook	2010	A systematic review and meta-analysis of the relationship between body size and testicular cancer	Review manuscript (no primary data), no assessment of dairy consumption exposure
Li, X. M. et al	2002	Relationship between the incidence rates of testicular and prostatic cancers and food consumptions	No full-text, same study as Ganmaa et al 2002
Li, X. M. et al	2003	The effects of estrogen-like products in milk on prostate and testes	No full-text, same study as Ganmaa et al 2002
McGlynn & Trabert	2012	Adolescent and adult risk factors for testicular cancer	Review manuscript (no primary data)
Roveda, A. M. et al	2006	Exposure to polychlorinated biphenyls (PCBs) in food and cancer risk: recent advances	Review manuscript (no primary data)
Walcott et al	2002	A case-control study of dietary phytoestrogens and testicular cancer risk	No assessment of dairy consumption exposure

Appendix 3:

Assessment of study quality against Newcastle-Ottawa criteria for case-control studies.

Author	Year	Adequacy of case definition	Representativeness of cases	Selection of controls	Definition of controls	Comparability of cases and controls	Ascertainment of exposure	Same ascertainment for cases and controls	Non-response rate
Bonner1	2002	Yes, with independent validation	Consecutive or obviously representative series of cases	Hospital controls	No history of disease	Cases and controls comparable (study controls for age and other factors)	Written self-report	Yes	No designation
Davies2	1996	Yes, with record linkage	Consecutive or obviously representative series of cases	Community controls	No history of disease	Cases and controls comparable (study controls for age and other factors)	Written self-report	Yes	Rate different (Response rate: cases 73% / cancer controls 65%/ population controls 57%)
Garner3	2003	Yes, with record linkage	Consecutive or obviously representative series of cases	Community controls	No history of disease	Cases and controls comparable (study controls for age and other factors)	Written self-report	Yes	Same rate for both groups (Response rate: cases 63.0% / controls 63.9%)
Giannandrea 4	2011	Yes, with independent validation	Consecutive or obviously representative series of cases	Hospital controls	No history of disease	Cases and controls comparable (study controls for age and other factors)	Interview not blinded to case/control status	Yes	No designation
McGlynn 5	2007	Yes, with independent validation	Consecutive or obviously representative series of cases	Community controls	No history of disease	Cases and controls comparable (study controls for age and other factors)	Structured interview where blind to case/control status	Yes	Same rate for both groups (Response rate: cases 91% / controls 81%)
Paoli 6	2015	Yes, with independent validation	Consecutive or obviously representative series of cases	Hospital controls	No history of disease	Cases and controls comparable (study controls for age and other factors)	Interview not blinded to case/control status	Yes	No designation
Sigurdsen 7	1999	Yes, with independent validation	Consecutive or obviously representative series of cases	Community controls	No history of disease	Cases and controls comparable (study controls for age and other factors)	Written self-report	Yes	Rate different (Response rate: cases 30.1% and 55.5% (cases contacted through genitourinary department) / controls 73.3%)
Stang 8	2006	Yes, with independent validation	Consecutive or obviously representative series of cases	Community controls	No history of disease	Cases and controls comparable (study controls for age and other factors)	Structured interview where blind to case/control status Written self-report by Mothers of cases/controls	Yes	Rate different (Response rate: cases 76% /controls 46%) and (Response rate: Mothers of cases 62.5% / Mothers of controls 42.9%)

Appendix 4:

Extracted data relating to dairy consumption and testicular cancer

Author	Study design	Exposure	Level of exposure / comparator	Measure of relative risk (OR/RR/HR), or absolute exposure (mean)	Reported mean (SD)	Reported crude OR/RR/HR (95% CI)	Reported adjusted OR/RR/HR (95% CI)	Number of exposed vs. non-exposed cases	Number of exposed vs. non-exposed controls/cohort	Adjustment for confounding a
Bonner1	CCS	Milk	Consumed in usual diet several years before diagnosis Mean consumption: Milk glasses/day	Mean	Seminoma 2.3 (SD) 2.3	n/a	n/a	n/a	n/a	Controls matched for age strata. Additional adjustment for energy intake, cryptorchidism, age, date of admission
					Non-seminoma 2.6 (SD) 2.4
					Mixed 2.3 (SD) 2.0
					Controls 2.0 (SD) 2.0
			Milk glasses/day 0, <1, 1–2, <2	OR	n/a	n/a	Total TC <1 vs 0: 0.76, 95% CI 0.31 – 1.89 1–2 vs 0: 0.56, 95% CI 0.25 – 1.23 <2 vs 0: 0.93, 95% CI 0.40 – 2.18	Total TC 12 non-exposed vs. <1: 17 exposed 1–2: 45 exposed >2: 43 exposed	28 non-exposed vs. <1, 51 exposed 1–2, 161 exposed >2, 94 exposed
							Seminoma <1 vs 0: 0.54, 95% CI 0.18 – 1.68 1–2 vs 0: 0.24, 95% CI 0.01 – 0.66 <2 vs 0: 0.58, 95% CI 0.20 – 1.69	Seminoma 8 non-exposed vs. <1: 8 exposed 1–2: 14 exposed >2: 17 exposed
							Non-seminoma <1 vs 0: 0.79, 95% CI 0.16 – 3.96 1–2 vs 0: 0.79, 95% CI 0.21 – 3.08 <2 vs 0: 1.29, 95% CI 0.31 – 5.35	Non-seminoma 3 non-exposed vs. <1: 4 exposed 1–2: 14 exposed >2: 14 exposed
							Mixed Cell <1 vs 0: 2.37, 95% CI 0.25 – 22.16 1–2 vs 0: 2.41, 95% CI 0.30 – 19.45 <2 vs 0: 2.74, 95% CI 0.32 – 23.68	Mixed Cell 1 non-exposed vs. <1: 5 exposed 1–2: 17 exposed >2: 12 exposed
Davies2	CCS	Milk, Cheese, Cream, Butter	Present consumption and whether at age 17 this was more, about the same or less which was used to estimate consumption in adolescence Mean consumption in adolescence: Milk pints/day. Cheese, Cream and Yoghurt times/week.	Mean	Milk pints/day. Cases 1.00 (SD)0.51 Cancer Controls 0.94 (SD) 0.56 Pop Controls 0.80 (SD) 0.44	n/a	n/a	n/a	n/a	Population controls matched for age / Cancer controls matched for year of diagnosis. Additional adjustment for UDT, age, social class, height, weight.
					Cheese times/week. Cases 3.67 (SD) 3.01 Cancer Controls 3.31 (SD) 2.58 Pop Controls 3.57 (SD) 2.80
					Cream times/week. Cases 0.47 (SD) 0.85 Cancer Controls 0.36 (SD) 0.54 Pop Controls 0.39 (SD) 0.58
					Yogurt times/week. Cases 1.32 (SD) 2.34 Cancer Controls 1.13 (SD)1.98 Pop Controls 1.24 (SD) 1.94
			Milk consumption at 17yrs, for each extra ¼ pint	OR	n/a	n/a	Cases vs all controls 1.23, 95% CI 1.09 – 1.38	Information not provided	Information not provided
							Cases vs Pop controls 1.39, 95% CI 1.19 –1.63
							Cases vs cancer controls 1.14, 95% CI 1.00 – 4.04
Garner3	CCS	Total dairy, milk, cheese, butter	Total dairy, milk, cheese, butter. Frequency of consumption in usual diet 2 years before questionnaire, converted to servings per week then divided into quintiles of intake	OR	n/a	n/a	Total Diary Total TC Q 2 vs Q 1: 1.39, 95% CI 0.96 – 2.02 Q 3 vs Q 1: 1.79, 95% CI 1.23 – 2.62 Q 4 vs Q 1: 2.56, 95% CI 1.73 – 3.78 Q 5 vs Q 1: 1.40, 95% CI 0.93 – 2.12	Information not provided	Information not provided	Controls matched for age and locality. Additionally adjusted for BMI, smoking and total energy intake
							Seminoma Q.2 vs Q 1: 1.22, 95% CI 0.77 – 1.91 Q 3 vs Q 1: 2.04, 95% CI 1.31 – 3.17 Q 4 vs Q 1 1.93, 95% CI 1.23 – 3.03 Q 5 vs Q 1: 1.38, 95% CI 0.84 – 2.25
							Non-seminoma Q 2 vs Q 1 1.52, 95% CI 0.78 – 2.95 Q 3 vs Q 1: 1.21, 95% CI 0.61 – 2.40 Q 4 vs Q 1: 3.17, 95% CI 1.65 – 6.09 Q 5 vs Q 1: 1.36, 95% CI 0.67 – 2.79
							Milk Total TC Q 2 vs Q 1: 1.39, 95% CI 0.97 – 1.99 Q 3 vs Q 1: 1.18, 95% CI 0.82 – 1.68 Q 4 vs Q 1: 1.52, 95% CI 1.08 – 2.16 Q 5 vs Q 1: 1.23, 95% CI 0.79 – 1.92
							Seminoma Q 2 vs Q 1: 1.36, 95% CI 0.89 – 2.06 Q 3 vs Q 1: 1.26, 95% CI 0.83 – 1.89 Q 4 vs Q 1: 1.34, 95% CI 0.89 – 2.02 Q 5 vs Q 1: 1.35, 95% CI 0.80 – 2.27
							Non-seminoma Q 2. vs Q 1: 1.30, 95% CI 0.72 – 2.35 Q 3 vs Q 1: 0.69, 95% CI 0.36 – 1.34 Q 4 vs Q 1: 1.75, 95% CI 1.00 – 3.08 Q 5 vs Q 1: 1.09, 95% CI 0.52 – 2.26
							Cheese Total TC Q 2 vs Q 1: 1.30, 95% CI 0.87 – 1.95 Q 3 vs Q 1: 1.60, 95% CI 1.13 – 2.25 Q 4 vs Q 1: 1.92, 95% CI 1.26 – 2.91 Q 5 vs Q 1: 1.87, 95% CI 1.22 – 2.86
							Seminoma Q 2 vs Q 1: 1.19, 95% CI 0.74 – 1.90 Q 3 vs Q 1: 1.47, 95% CI 0.98 – 2.19 Q 4 vs Q 1: 1.68, 95% CI 1.03 – 2.75 Q 5 vs Q 1: 1.43, 95% CI 0.87 – 2.37
							Non-seminoma Q 2 vs Q 1: 1.39, 95% CI 0.68 – 2.81 Q 3 vs Q 1: 1.56, 95% CI 0.85 – 2.86 Q 4 vs Q 1: 2.37, 95% CI 1.21 – 4.67 Q 5 vs Q 1: 1.97, 95% CI 0.95 – 4.05
							Butter Total TC Q 2 vs Q 1: 1.13, 95% CI 0.78 – 1.63 Q 3 vs Q 1: 0.94, 95% CI 0.70 – 1.27 Q 4 vs Q 1: 1.11, 95% CI 0.83 – 1.48 Q 5. - exposure not varied enough to divide into quintiles
							Seminoma Q 2 vs Q 1: 1.21, 95% CI 0.79 – 1.85 Q 3 vs Q 1: 0.93, 95% CI 0.66 – 1.32 Q 4 vs Q 1: 1.02, 95% CI 0.73 – 1.44 Q 5. exposure not varied enough to divide into quintiles
							Non-seminoma Q 2 vs Q 1: 0.79, 95% CI 0.42 – 1.49 Q 3 vs Q 1: 0.83, 95% CI 0.51 – 1.35 Q 4 vs Q 1: 1.24, 95% CI 0.78 – 1.98 Q 5. exposure not varied enough to divide into quintiles
Giannandrea 4	CCS	Dairy Products	Frequent use (at least once/day) vs. rare use (once/week or less)	OR	n/a	2.15, 95% CI 0.67 – 6.83	2.55, 95% CI 0.76 – 8.50	10 exposed (frequent) vs. 40 non-exposed (rare)	5 exposed (frequent) vs. 43 non-exposed (rare)	Controls matched for age, BMI and locality. Additionally adjusted for mothers age at birth, education and parity
McGlynn 5	CCS	All dairy products excluding eggs and soy Milk: whole, 2%, 1%, skim	Consumption of milk or dairy products at least 3 times a week in 1 year before TC diagnosis (or comparable time period for controls) and at least 3 times a week at any point in their lives between birth and grade 12 vs. less than three times/week	OR	n/a	Total TC1 any age 0.61	0.64, 95% CI 0.29 – 1.45	755 exposed vs. 9 non-exposed	910 exposed vs. 18 non-exposed	Controls matched for age, ethnicity, serum sample date. Additionally adjusted for UDT and family history of TC
						Birth to kindergarten 0.90	0.92, 95% CI 0.49 – 1.76	735 exposed vs. 17 non-exposed	896 exposed vs. 23 non-exposed
						Grades 1–5 0.69	0.73, 95% CI 0.35 – 1.50	751 exposed vs. 12 non-exposed	905 exposed vs. 21 non-exposed
						Grades 6–8 0.51	0.53, 95% CI 0.26 – 1.09	751 exposed vs. 11 non-exposed	899 exposed vs. 26 non-exposed
						Grades 9–12 0.77	0.80, 95% CI 0.48 – 1.34	738 exposed vs. 25 non-exposed	889 exposed vs. 39 non-exposed
						Seminoma any age 0.51	0.54, 95% CI 0.18 – 1.63	320 exposed vs. 4 non-exposed	910 exposed vs. 18 non-exposed
						Birth to kindergarten 0.73	0.77, 95% CI 0.32 – 1.84	311 exposed vs. 7 non-exposed	896 exposed vs. 23 non-exposed
						Grades 1–5 0.74	0.79, 95% CI 0.32 – 1.91	316 exposed vs. 7 non-exposed	905 exposed vs. 21 non-exposed
						Grades 6–8 0.49	0.51, 95% CI 0.20 – 1.29	317 exposed vs. 6 non-exposed	899 exposed vs. 26 non-exposed
						Grades 9–12 0.67	0.70, 95% CI 0.35 – 1.41	313 exposed vs. 11 non-exposed	889 exposed vs. 39 non-exposed
						Non--seminoma any age 0.62	0.67, 95% CI 0.24 – 1.85	434 exposed vs. 5 non-exposed	910 exposed vs. 18 non-exposed
						Birth to kindergarten 0.93	0.95, 95% CI 0.44 – 2.06	423 exposed vs. 10 non-exposed	896 exposed vs. 23 non-exposed
						Grades 1–5 0.52	0.56, 95% CI 0.21 – 1.51	434 exposed vs. 5 non-exposed	905 exposed vs. 21 non-exposed
						Grades 6–8 0.44	0.47, 95% CI 0.18 – 1.24	433 exposed vs. 5 non-exposed	899 exposed vs. 26 non-exposed
						Grades 9–12 0.75	0.80, 95% CI 0.42 – 1.54	425 exposed vs. 13 non-exposed	889 exposed vs. 39 non-exposed
			Type of milk consumed most: whole, 2%, 1%, skim	OR	n/a	Total TC Whole milk 1.00 (ref)	1.00 (ref)	394 exposed vs. n/a	457 exposed vs. n/a	n/a
						2% milk 1.53	1.53, 95% CI 1.11 – 2.12	113 exposed vs. n/a	84 exposed vs. n/a
						1% milk 0.20	0.21, 95% CI 0.03 – 1.76	1 exposed vs. n/a	6 exposed vs. n/a
						Skim 1.55	1.51, 95% CI 0.65 – 3.51	13 exposed vs. n/a	10 exposed vs. n/a
						Seminoma Whole milk 1.00 (ref)	1.00 (ref)	164 exposed vs. n/a	457 exposed vs. n/a
						2% milk 1.54	1.56, 95% CI 1.00 – 2.43	38 exposed vs. n/a	84 exposed vs. n/a
						1% milk n/a	n/a	0 exposed vs. n/a	6 exposed vs. n/a
						Skim 2.40	2.17, 95% CI 0.79 – 5.99	7 exposed vs. n/a	10 exposed vs. n/a
						Non-seminoma Whole milk 1.00 (ref)	1.00 (ref)	230 exposed vs. n/a	457 exposed vs. n/a
						2% milk 1.52	1.50, 95% CI 1.04 – 2.17	75 exposed vs. n/a	84 exposed vs. n/a
						1% milk 0.28	0.30, 95% CI 0.04 – 2.53	1 exposed vs. n/a	6 exposed vs. n/a
						Skim 1.11	1.18, 95% CI 0.42 – 3.33	6 exposed vs. n/a	10 exposed vs. n/a
Paoli 6	CCS	Milk and dairy	Frequent use (at least once/day) vs. rare use (once/week or less)	OR	n/a	2.33, 95% CI 1.02 –5.29	2.37, 95% CI 1.01 – 5.52	23 exposed vs. 102 non-exposed	9 exposed vs. 94 non-exposed	Controls matched for BMI and locality. Additionally adjusted for age and education level
Sigurdsen 7	CCS	Milk	Daily consumption gm/1,000kcal reported as Mean	Mean	Total TC 115.7 (SD) 8.5	n/a	n/a	n/a	n/a	Controls matched for age, ethnicity and locality. Matching dissolved in analysis. Seminoma and mixed cell additionally adjusted for age, education, income, UDT and total daily calorie intake. Non-seminoma additionally adjusted for age, education, income, ethnicity, UDT and total daily calorie intake
					Seminoma 112.4 (SD) 14.7
					Non-seminoma 122.2 (SD) 13.9
					Mixed cell 103.8 (SD) 10.9
					All controls 130.5 (SD) 9.2
			Daily consumption gm/1,000kcal reported in Quartiles: 1.none – 43.7 2.43.8 – 112.8 3.112.9 –183.9 4. 184 – 427.2	OR	n/a	n/a	Seminoma Q 2 vs Q 1: 0.7, 95% CI 0.3 – 2 Q 3 vs Q 1: 0.9, 95% CI 0.3 – 2.4 Q 4 vs Q 1: 0.6, 95% CI 0.2 – 1.7	Exposed (Q 2–4) Non-exposed (Q 1) Seminoma 33 exposed vs. 13 non-exposed	Information not provided
							Non- seminoma Q 2 vs Q 1: 1.5, 95% CI 0.6 – 3.6 Q 3 vs Q 1: 1.3, 95% CI 0.5 – 3.3 Q 4 vs Q 1: 0.5, 95% CI 0.2 – 1.6	Non-seminoma 63 exposed vs. 19 non-exposed
							Mixed germ cell Q 2 vs Q 1: 1.4, 95% CI 0.4 – 4.5 Q 3 vs Q 1: 1.3, 95% CI 0.4 – 4.2 Q 4 vs Q 1: 0.4, 95% CI 0.1 – 1.7	Mixed cell 25 exposed vs. 7 non-exposed
Stang 8	CCS	Milk, low fat milk, yoghurt, cheese, cream. Milk fat, and galactose	Frequency of consumption in usual diet during adolescence, using categorical shift (as below) • Asked sons about the frequency of consumption in usual diet during the 1 year before questionnaire(in 7 categories) • converted into usual diet at 17 years by asking if consumption at age 17 was ‘more’, ‘about the same’ or ‘less’ than present consumption • then used a categorical shift method to estimate consumption at age 17 (i.e. + 1 category if answer was ‘more’ and -1 category if answer was ‘less’ • converted to average servings per month at 17 years old • Also assessed by asking participants Mothers of their sons consumption aged 17 years	RR for each additional 20 food item servings per month during adolescence for both sons (categorical shift) and Mothers	n/a	n/a	Milk Total TC 1.37, 95% CI 1.12 – 1.68 Seminoma 1.66, 95% CI 1.30 – 2.12 Non-seminoma 0.96, 95% CI 0.70– 1.31 Mothers Total TC 1.21, 95% CI 0.86 – 1.70 Seminoma 1.45, 95% CI 0.96 – 2.21 Non-seminoma 1.00, 95% CI 0.62 – 1.61	Information not provided	Information not provided	Controls matched for age (5yr groups) and residence (5 strata). Additionally adjusted for highest degree and height 1 year before interview
							Low-fat milk Total TC 1.25, 95% CI 0.97 – 1.62 Seminoma 1.38, 95% CI 1.03 – 1.86 Non-seminoma 0.99, 95% CI 0.64 – 1.53 Mothers Total TC 1.13, 95% CI 0.71 – 1.80 Seminoma 1.09, 95% CI 0.62 – 1.92 Non-seminoma 1.29, 95% CI 0.66 – 2.49
							Yoghurt Total TC 0.82, 95% CI 0.62 – 1.08 Seminoma 0.77, 95% CI 0.55 – 1.07 Non-seminoma 0.82, 95% CI 0.54 – 1.24 Mothers Total TC 0.89, 95% CI 0.59 – 1.32 Seminoma 1.03, 95% CI 0.64 – 1.67 Non-seminoma 0.73, 95% CI 0.40 – 1.31
							Cream Total TC 1.31, 95% CI 0.90 – 1.92 Seminoma 1.55, 95% CI 1.01 – 2.38 Non_seminoma 0.92, 95% CI 0.47 – 1.79 Mothers Not collected
							Cheese (servings) Total TC 1.42, 95% CI 0.94 – 2.16 Seminoma 1.35, 95% CI 0.81 – 2.25 Non_seminoma 1.60, 95% CI 0.86 – 2.97 Mothers (curd) Total TC 0.89, 95% CI 0.59 – 1.36 Seminoma 0.77, 95% CI 0.45 – 1.30 Non-seminoma 1.12, 95% CI 0.64 – 1.96
							Cheese (spread) Total TC 1.27, 95% CI 0.99 – 1.63 Seminoma 1.35, 95% CI 1 – 1.81 Non-seminoma 1.11, 95% CI 0.75 – 1.64 Mothers Total TC 0.83, 95% CI 0.56 – 1.22 Seminoma 0.85, 95% CI 0.53 – 1.37 Non-seminoma 0.79, 95% CI 0.46 – 1.38
			Milk fat consumption at 17 based on categorical shift estimates reported in gm/month quartiles for sons 1. </= 411 2. 411 – 603 3. 603 – 833 4. >/= 833 Based on Mothers estimates reported in gm/month in quartiles 1. </= 281 2. 281 – 469 3. 469 – 615 4. >/= 615			n/a	Milk Fat Total TC Q 2 vs Q1: 0.84, 95% CI 0.49 – 1.42 Q 3 vs Q1: 1.12, 95% CI 0.67 – 1.85 Q 4 vs Q1: 1.80, 95% CI 1.12 – 2.89 Mothers Q 2 vs Q1: 1.37, 95% CI 0.68 – 2.74 Q 3 vs Q1: 0.94, 95% CI 0.44 – 2.00 Q 4 vs Q1: 1.25, 95% CI 0.60 – 2.62			Additionally adjusted for highest degree, height 1 year before interview, family history of TC and history of UDT
							Seminoma Q 2 vs Q1: 0.81, 95% CI 0.42 – 1.54 Q 3 vs Q1: 1.07, 95% CI 0.58 – 1.99 Q 4 vs Q1: 2.48, 95% CI 1.41 – 4.37 Mothers Q 2 vs Q1: 1.36, 95% CI 0.56 – 3.29 Q 3 vs Q1: 1.11, 95% CI 0.43 – 2.88 Q 4 vs Q1: 2.11, 95% CI 0.83 – 5.35
							Non-seminoma Q 2 vs Q1: 0.86, 95% CI 0.38 – 1.93 Q 3 vs Q1: 1.14, 95% CI 0.53 – 2.46 Q 4 vs Q1: 1.15, 95% CI 0.54 – 2.46 Mothers Q 2 vs Q1: 1.50, 95% CI 0.56 – 4.02 Q 3 vs Q1: 0.69, 95% CI 0.23 – 2.05 Q 4 vs Q1: 0.73, 95% CI 0.25 – 2.17
			Galactose consumption at 17 based on categorical shift estimates reported in gm/month quartiles 1. </= 981 2. 81 – 168 3. 168 – 254 4. >/= 254 Based on Mothers estimates reported in gm/month in quartiles 1. </= 138 2. 138 – 245 3. 245 – 318 4. >/= 318				Galactose Total TC Q 2 vs Q1: 1.02, 95% CI 0.60 – 1.72 Q 3 vs Q1: 1.03, 95% CI 0.61 – 1.73 Q 4 vs Q1: 1.46, 95% CI 0.89 – 2.39 Mothers Q 2 vs Q1: 1.28, 95% CI 0.64 – 2.57 Q 3 vs Q1: 0.73, 95% CI 0.34 – 1.55 Q 4 vs Q1: 1.36, 95% CI 0.66 – 2.79
							Seminoma Q 2 vs Q1: 1.26, 95% CI 0.66 – 2.42 Q 3 vs Q1: 1.59, 95% CI 0.82 – 3.06 Q 4 vs Q1: 2.68, 95% CI 1.44 – 4.99 Mothers Q 2 vs Q1: 1.40, 95% CI 0.57 – 3.40 Q 3 vs Q1: 0.56, 95% CI 0.20 – 1.59 Q 4 vs Q1: 2.36, 95% CI 0.94 – 5.97
							Non-seminoma Q 2 vs Q1: 0.68, 95% CI 0.31 – 1.49 Q 3 vs Q1: 0.46, 95% CI 0.20 – 1.04 Q 4 vs Q1: 0.55, 95% CI 0.25 – 1.19 Mothers Q 2 vs Q1: 1.43, 95% CI 0.53 – 3.86 Q 3 vs Q1: 0.92, 95% CI 0.33 – 2.54 Q 4 vs Q1: 0.79, 95% CI 0.28 – 2.23
			Milk fat, and galactose consumptions at 17 years reported as a RR per 200gm/month from both sons and Mothers			Milk Fat Total TC 1.19, 95% CI 1.07 – 1.32 Seminoma 1.30, 95% CI 1.15 – 1.48 Non-seminoma 1.05, 95% CI 0.89 –1.24 Mothers Total TC 1.07, 95% CI 0.85 – 1.35 Seminoma 1.24, 95% CI 0.93 – 1.66 Non-seminoma 0.91, 95% CI 0.66 – 1.27	Milk Fat Total TC 1.17, 95% CI 1.05 – 1.31 Seminoma 1.29, 95% CI 1.13 – 1.47 Non-seminoma 1.03, 95% CI 0.87 –1.22 Mothers Total TC 1.07, 95% CI 0.84 – 1.36 Seminoma 1.24, 95% CI 0.91 – 1.69 Non-seminoma 0.92, 95% CI 0.66 – 1.30
						Galactose Total TC 1.33, 95% CI 0.99 – 1.78 Seminoma 2.01, 95% CI 1.41 – 2.86 Non-seminoma 0.62, 95% CI 0.38 –1.02 Mothers Total TC 1.19, 95% CI 0.80 – 1.78 Seminoma 1.54, 95% CI 0.93 – 2.56 Non-seminoma 0.92, 95% CI 0.51 – 1.66	Galactose Total TC 1.32, 95% CI 0.97 – 1.78 Seminoma 2.02, 95% CI 1.41 – 2.91 Non-seminoma 0.58, 95% CI 0.35 –0.97 Mothers Total TC 1.20, 95% CI 0.79 – 1.83 Seminoma 1.58, 95% CI 0.92 – 2.72 Non-seminoma 0.93, 95% CI 0.51 – 1.68

¹Matched for age, ethnicity and serum sample date