Time and spatial trends of operated cryptorchidism in France and environmental hypotheses: a nationwide study from 2002 to 2014

Abstract

STUDY QUESTION

Is there an evolution in the risk of operated cryptorchidism in France and does local geographical environment appear as an important trigger for this defect?

SUMMARY ANSWER

We observed an increase of the risk of operated cryptorchidism in boys under the age of 7 years during the period 2002–2014 and a strong spatial heterogeneity, with the detection of spatial clusters suggesting environmental factors.

WHAT IS KNOWN ALREADY

Epidemiologic data on cryptorchidism are scarce and its etiology is poorly understood. As part of the testicular dysgenesis syndrome, cryptorchidism is suspected to be a male genital developmental disorder caused by endocrine disruptor chemical (EDC) exposure during the prenatal period.

STUDY DESIGN, SIZE, DURATION

This was a retrospective and descriptive study using data from the French national hospital discharge database, in the 2002–2014 study period. We built an indicator to reflect incident cases of operated cryptorchidism in boys under the age of 7 years in metropolitan France, with an algorithm using specific codes for diseases (ICD-10 codes) and surgical acts (CCAM codes).

PARTICIPANTS/MATERIALS, SETTING, METHODS

The study population was composed of 89 382 new cases of operated cases of cryptorchidism in boys under the age of 7 years. We estimated the temporal evolution of the incidence rate. We fitted a spatial disease-mapping model to describe the risk of cryptorchidism at the postcode scale. We used Kulldorff’s spatial scan statistic and Tango’s flexibly shaped spatial scan statistic to identify spatial clusters.

MAIN RESULTS AND THE ROLE OF CHANCE

The estimated increase in the incidence of operated cryptorchidism from 2002 to 2014 was equal to 36.4% (30.8%; 42.1%). Cryptorchidism displayed spatial heterogeneity and 24 clusters (P < 0.0001) were detected. The main cluster was localized in a former coal mining and metallurgic area in northern France, currently an industrial area. The cluster analysis suggests the role of shared socio-economic and environmental factors that may be geographically determined and intertwined. The industrial activities identified in the clusters are potentially the source of persistent environmental pollution by metals, dioxins and polychlorinated biphenyls.

LIMITATIONS, REASONS FOR CAUTION

The indicator we used reflects operated cases of cryptorchidism, with an under-evaluation of the health problem. We cannot exclude a possible role of the evolution and local differences in surgical practices in the observed trends. Our inclusion of boys under 7 years of age minimized the biases related to differences in practices according to age. Regarding the environmental hypothesis, this is an exploratory study and should be considered as a hypothesis-generating process for future research studies.

WIDER IMPLICATIONS OF THE FINDINGS

To our knowledge, this is the first descriptive study to address nationwide trends of operated cryptorchidism with detection of spatial clusters, with a very large sample allowing great statistical power. Our results generate plausible environmental hypotheses, which need to be further tested.

STUDY FUNDING/COMPETING INTEREST(S)

This study was entirely funded by Santé publique France, the French National Public Health Agency. All authors declare they have no actual or potential competing financial interest.

TRIAL REGISTRATION NUMBER

N/A.

cryptorchidism, disease spatial clusters, environment, endocrine disruptors, epidemiology, incidence rate, France, reproductive health, birth defects

Introduction

Cryptorchidism is the absence of one or both testes from the scrotum at birth (congenital cryptorchidism) or after birth (acquired cryptorchidism). It is the most frequent male genital defect among infants. Epidemiologic data on cryptorchidism are scarce, because birth defect registers do not systematically register this condition, and standardized registration methods are lacking (Toppari et al., 2001). The frequency of congenital cryptorchidism is estimated to range from 1% to 8% (Virtanen and Adamsson, 2012), and from 2% to 4% in healthy term infants (Barthold and Gonzalez, 2003), as it is much more frequent in premature newborns (Foresta et al., 2008).

An increase in the incidence of congenital cryptorchidism was observed in England between 1950 and 1980, and again from 2001 to 2008 (Acerini et al., 2009), in Denmark between 1960 and 2000 (WHO (World Health Organization)/UNEP (United Nations Environment Programme), 2013; Skakkebæk et al., 2016) and recently in South Korea between 2000 and 2005 (Chul Kim et al., 2011). However, there are conflicting results on the temporal trends in other countries (Gurney et al., 2017).

Testicular descent is a hormone-dependent process (Foresta et al., 2008), but its etiology is poorly understood, and most cases are considered idiopathic. Known risk factors, defining syndromic cryptorchidism, are genetic defects (e.g. Klinefelter syndrome, mutations of the androgen receptor gene), which account for a small percentage of cases, around 3% (Ferlin et al., 2008), and rare polymalformative syndromes (e.g. Kallman-De-Morsier syndrome).

Genetic, socio-economic, perinatal and environmental factors are thought to be intertwined in idiopathic cases. Cryptorchidism has been associated with a low socio-economic status (Moller and Skakkebæk, 1996) and perinatal factors (prematurity, being small for gestational age) (Damgaard et al., 2008). Endocrine disruptor chemical (EDC) exposure is suspected to be a risk factor of cryptorchidism, as part of the testicular dysgenesis syndrome (Skakkebæk et al., 2001, 2016). The weight of evidence is especially documented in experimental studies for chemicals such as phthalates, pesticides, diethylstilbestrol or dioxins but epidemiologic data are more conflicting (Virtanen and Adamsson, 2012).

Santé publique France, the French national public health agency, has developed an epidemiological monitoring program to study reproductive health indicators relating to EDC exposure nationwide using existing databases. Based on the weight of evidence on its possible links to EDC exposure, we selected cryptorchidism as one of the key reproductive health indicators to be monitored (Le Moal et al., 2016).

People are widely exposed to chemicals through all routes via food, water, atmospheric pollution, dust, dwellings, drugs, cosmetics, and so on. In this study, we aimed to focus on potential risk factors that can be geographically determined or enhanced. Therefore, our objectives were to describe the temporal and spatial trends of cryptorchidism nationwide in Metropolitan France at a fine scale during the period 2002–2014, and explore if the local geographical environment may be an important trigger for this defect.

Materials and methods

Health and population data

To study the epidemiology of cryptorchidism in France, we previously built an indicator to reflect the incidence of operated cryptorchidism in boys under the age of 7 years using data from the French national hospital discharge database (Rambourg et al., 2004; Suzan et al., 2011), which is currently available from the National Health Data System. This comprehensive database covers the whole of France since 2002 and includes discharges from both public and private hospitals (Buisson, 2005; Goldberg, 2017). We extracted data according to the authorization of The National Computers and Privacy Commission (CNIL) obtained by Santé publique France (No. 902167).

We selected operated cases of cryptorchidism in boys under 7 years using an algorithm to identify their surgical stay. It was elaborated with clinicians using specific codes for diseases (ICD-10 codes) and surgical acts (CCAM codes).

Incident cases were defined as the first stay of patients in the period 2002–2014 and selected by connecting all stays in the study period using the unique patient identification number. Individual data were available at the date of surgical intervention, and patients were localized at the scale of their postcode of residence. In Metropolitan France, there are 5646 postcodes that define a geographic unit with one or several municipalities.

We analyzed the incident cases globally and according to age. We also analyzed the specific subgroup of bilateral operated cryptorchidism, selected with the corresponding surgical code. Population data per year, age, and municipality were provided by the French National Institute for Statistical and Economic Studies (Insee). The expected number of cases for each postcode and year was computed by applying the national French incidence rates to the population of each postcode, stratified by age.

Temporal and spatial model at the postcode scale

A quasi Poisson log-linear model was defined to estimate the temporal trend of the incidence of operated cryptorchidism among boys aged 6 years and under. A penalized spline function was used to test the linearity of the temporal trend (Wood, 2017).

A Poisson log-linear model was defined to estimate the spatial trend of operated cryptorchidism. It was formulated within a hierarchical Bayesian framework. The Besag–York–Mollié (BYM) model (Besag et al., 1991) was used to model spatially structured and unstructured random effects. For comparative purposes, we also considered the model proposed by Leroux et al. (1999) and the modified BYM proposed by Riebler et al. (2016). Model fit was measured by the Deviance Information Criterion (DIC). Adjacency was used to define neighbors: all postcodes sharing a border with the postcode of interest were defined as neighbors. The expected number of cases was included in the model as an offset term. The integrated nested Laplace approximation (INLA) approach was used to compute the posterior marginals of all the parameters of interest (Rue et al., 2009).

These models were implemented in the R software environment using the mgcv (Wood, 2017), INLA (Rue et al., 2009) and CARBayes (Lee, 2013) packages.

The residuals from the BYM model were tested for spatial autocorrelation using a permutation test based on Moran’s I statistic (based on 10 000 random permutations) (Moran, 1950). We present results as the percentage increase in the risk of cryptorchidism during the study period and its 95% CI as well as postcode-specific relative risks (RR) and their posterior probability to be greater than 1.

Spatial cluster detection at the postcode scale

Kulldorff’s spatial scan statistic was applied to detect spatial clusters (Kulldorff and Nagarwalla, 1995; Kulldorff, 1997). It allowed us to scan the entire territory to search for any particular area that may be associated with a higher incidence of operated cryptorchidism, without any a priori knowledge of the factors that may be involved. A Poisson model was used. The scan statistic is based on overlapping circles centered around each postcode to define a scanning window. In our study, the scanning window is represented by grouping neighboring postcodes within a maximum radius of 15 km. For each window, the likelihood ratio was based on the alternative hypothesis that the incidence rate was higher inside the cluster candidate than outside, while the null hypothesis was that both incidence rates were equal. The window that maximizes the likelihood ratio function is defined as the most likely cluster.

Adjustments were made for the postcode population density. The significance level was given by 9999 Monte Carlo simulations. All spatially non-overlapping clusters were identified. For comparative purposes, we also considered the flexible spatial scan statistic developed by Tango and Takahashi (2005).

This detection study was run in SaTScan version 9.6 (Kulldorff, 2015) and FleXScan version 3.1.2 (Takahashi et al., 2013).

Cluster descriptive analyses

We mapped all the clusters, and listed the municipalities included in each cluster. We documented the observed and expected cases, the RR, the excess of cases and the surface. We described the main cluster in more depth, according to demographic and economic characteristics, and the secondary clusters by focusing on the most populated municipality. Then we searched for common features between the main and secondary clusters to identify potential typologies. Finally, we complemented our spatial analyses using several available indicators in order to further explore the main raised hypotheses and foster the discussion.

Results

Descriptive results

We identified 89 382 cases of operated cryptorchidism in Metropolitan France between 2002 and 2014, including 9703 (11%) bilateral cases. Table I shows the distribution of cases according to age. Most cases were operated during the first and second year. The number of cases increased from 5804 in 2002 to 7950 in 2014 (+37%), while the population of boys under 7 years increased by 5% during this same period from 2 655 088 to 2 786 018 (Table II). The observed incidence rate was 2.2 per 1000 in 2002 and 2.8 per 1000 in 2014.

Table I

Observed cases of all-type operated cryptorchidism according to age in France.

Age (years)	Number of observed cases (% of cases)	Population (2008)
0	4952 (5.5)	388 748
1	20 085 (22.5)	385 415
2	20 833 (23.3)	387 372
3	13 486 (15.1)	389 055
4	10 501 (11.7)	390 590
5	9703 (10.9)	392 869
6	9822 (11.0)	393 451
Total	89 382	2 727 500

Table II

Observed cases of all-type operated cryptorchidism per year in France.

Year	Number of observed cases	Population
2002	5804	2 655 088
2003	5624	2 672 707
2004	6339	2688 954
2005	6179	2 705 305
2006	6264	2 723 916
2007	6553	2 734 038
2008	6981	2 736 937
2009	7006	2 747 420
2010	7138	2 764 297
2011	7613	2 780 956
2012	7885	2 790 273
2013	8046	2 789 972
2014	7950	2 786 018
Total	89 382	35 575 881

The distribution of observed and expected cases per postcode was quite similar. The mean number of observed cases per postcode was 15.8, ranging from 0 to 537.

Temporal trends

In the study period, the estimated increase for the incidence of operated cryptorchidism was equal to 36.4% (95% CI 30.8%; 42.1%). This increase was particularly high for boys aged 2 years and under, being equal to 55.8% (48.8%; 63.1%) (Fig. 1), as well as for the subgroup of bilateral cases (Fig. 2). The estimated increase for the incidence of operated bilateral cryptorchidism was equal to 58.7% (37.0%; 83.9%).

Figure 1.

Estimated linear temporal trend for operated cryptorchidism among boys of all ages and among boys aged 2 years and under. The solid lines show the estimated trend, while the dashed lines are the 95% confidence intervals. The black dots denote the annual observed incidences (per 1000).

Figure 2.

Estimated linear temporal trend for operated bilateral cryptorchidism. The solid line shows the estimated trend, while the dashed lines represent the 95% confidence interval. The black dots denote the annual observed incidence (per 1000).

Spatial trends

Figure 3 shows the map of smoothed estimates for the RR of all-type operated cryptorchidism, that is, the posterior mean for the postcode-specific RR, as well as the map of uncertainty associated with these posterior means, that is, the posterior probability of the RR being greater than 1. We observe a structured spatial heterogeneity, that is, a large part of the variability is explained by the spatial structure. Neighboring spatial units tend to show the same risks. Several areas are at high risk, scattered across different parts of France except in the south west. We obtained similar results using the models of Leroux et al. and Riebler et al. (results not shown). The DIC for these models are presented in Supplementary Table SI.

Figure 3.

Postcode-specific posterior relative risks (top figure) and postcode posterior probability of relative risk >1 (bottom figure) (Besag–York–Mollié model), all-type cryptorchidism.

Cluster detection for all-type cases

We detected 24 clusters (P < 0.0001) using SatScan. The most likely cluster has the smallest probability of occurring by chance. The secondary clusters are ordered according to the likelihood ratio test statistic. Table III shows this ordered list of clusters. A total of 9024 cases (10% of the overall number) is included in the identified clusters. The total estimated excess is 3315 cases. The cluster surfaces correspond to several hundred square kilometers, except for cluster no. 10, which is a single village of only 2 km² with 31 observed cases compared to four expected cases. These clusters are presented in Fig. 4. The detected clusters are located mainly in the north and central east of France. Cluster detection using FlexScan gave very similar results (Supplementary Fig. S1). The main cluster is the same, and 23 clusters are located in the same areas.

Figure 4.

Map of the clusters for all types of operated cryptorchidism in Metropolitan France (SatScan).

Table III

Clusters of all-type operated cryptorchidism.

Ordered cluster number	Number of cases	Estimated case excess	Relative risk	Most populated town in the cluster (inhabitants in 2013)	Overall surface of the cluster (km2)
1	1244	453	1.58	Lens (34 190)	695
2	304	172	2.31	Etampes (23 000)	677
3	397	178	1.81	Boulogne-sur-mer (42 680)	467
4	456	192	1.73	Lorient (57 408)	565
5	503	193	1.62	Mantes la Jolie (42 727)	706
6	201	106	2.11	Montluçon (38 400)	870
7	697	214	1.45	Saint-Etienne (170 049)	399
8	602	190	1.47	Clermont-Ferrand (139 900)	609
9	153	83	2.19	Vesoul (15 800)	615
10	31	27	6.94	Lézennes (3100)	2
11	352	132	1.60	Thionville (40 950)	357
12	451	148	1.49	Le Mans (143 200)	682
13	630	178	1.40	Valenciennes (43 300)	498
14	337	125	1.59	Besançon (115 879)	520
15	143	75	2.09	Saint-Dizier (24 825)	660
16	325	115	1.55	Troyes (60 300)	626
17	369	118	1.47	Montbéliard (25 900)	740
18	162	73	1.82	Pont-l’Abbé (8300)	498
19	318	105	1.50	Annecy (50 400)	502
20	291	100	1.52	Charleville-Mézières (49 800)	597
21	92	50	2.17	Vagney (4000)	386
22	489	125	1.34	Orléans (114 195)	573
23	174	69	1.66	Dreux (30 536)	215
24	303	94	1.45	Sarreguemines (21 500) et Forbach (21 500)	510
Total	9024	3315			13 395

The main cluster (Supplementary Fig. S2) is located in the north around the city of Lens (34 190 inhabitants) and includes 1244 cases, with an estimated excess of 453 cases. The number of observed cases represents 13.8% of the total number of cases included in the clusters, and 1.4% of the cases in Metropolitan France. The cluster includes 101 communes, with two other medium-size cities (31 900 and 26 300 inhabitants), as well as numerous less populated municipalities with less than 1000 inhabitants. It is situated in the heart of the former so-called North-Pas de Calais coal mining area.

It includes the two production sites of a former smelter (society Metaleurop Nord), where most of the local population was previously employed. After more than a century of non-ferrous metal production, it closed in 2003 and induced widespread environmental pollution with metals, especially lead and cadmium (Ilef et al., 2000). This cluster also includes a metallurgic plant, and two industrial areas still in activity.

The demographic and economic characteristics of the main and secondary clusters of all-type cryptorchidism led us to identify three main typologies:

• The majority of clusters (13/24), like the first six clusters, include one or several mid-sized towns (with 10 000s inhabitants), often known for their industrial history, especially in the areas of mining (No. 1, 5, 6, 7, 13, 24), smelting works and/or metallurgy (No. 1, 3, 4, 5, 6, 7, 8, 9, 11, 12, 13, 15, 17, 20, 24), and mechanics and/or car industry (No. 1, 2, 3, 5, 7, 8, 9, 12, 13, 14, 15, 17, 19, 23, 24), and frequently facing a current economic decline. Cluster no. 13, is also located in the former North-Pas de Calais mining area.

• Some clusters include larger municipalities (5/24), with more than 100 000 inhabitants (No. 7, 8, 12, 14, 22). Four of these cities are also known for their former and current industrial activity in the same industries as the former clusters.

• Three clusters (No. 10, 18, 21) are located in areas where the main commune is much less populated with less than 10 000 inhabitants. Their typology seems different and perhaps specific to each area. The area of the cluster No. 21 shows high mining and metallurgic activities.

In total, we observe mining activities in 8/24 clusters, metallurgic activities in 17/24 clusters and mechanic activities in 16/24 clusters.

Cluster detection for bilateral cases

When focusing on the bilateral operated cases, the cluster detection using SatScan identified 21 clusters (P < 0.0001) (Table IV and Fig. 5) with 2292 observed cases (23% of the total number of bilateral cases), with an excess of 1603 cases during the study period. Twelve clusters, including the same main cluster (Lens) as previously identified, are located in the same areas around the same municipality, as with all-type cryptorchidism. However, nine new clusters were identified, including cities situated in the south east of France: Grenoble, Montpellier, Beziers, Nîmes, Alès and Toulon.

Figure 5.

Map of the clusters for bilateral cases of operated cryptorchidism in metropolitan France (SatScan).

Table IV

Clusters of bilateral operated cryptorchidism: the new clusters compared to the clusters for all-type cryptorchidism are shown in red.

Ordered cluster number	Number of cases	Estimated case excess	Relative risk	Most populated town in the cluster (inhabitants in 2013)	Overall surface of the cluster (km2)
1	340	271	5.07	Lens (34 190)	426
2	134	99	3.84	Le Mans (143 240)	602
3	146	100	3.23	St Etienne (170 049)	431
4	104	79	4.13	Besançon (115 879)	708
5	178	107	2.50	Grenoble (150 424)	756
6	175	105	2.54	Montpellier (264 538)	662
7	73	57	4.59	Alès (40 851)	744
8	107	74	3.25	Mantes la Jolie (42 727)	706
9	104	72	3.24	Annecy (51 012)	647
10	94	67	3.42	St Herblain (43 082)	587
11	91	65	3.50	Thionville (40 950)	411
12	66	52	4.58	Dreux (30 536)	392
13	50	42	5.98	St Dizier (24 825)	621
14	109	71	2.91	Nîmes (144 940)	650
15	138	82	2.48	Toulon (163 974)	474
16	48	39	5.46	Nord sur Erdre (7970)	675
17	81	57	3.40	Béziers (71 432)	662
18	74	50	3.06	Boulogne sur Mer (42 680)	501
19	28	24	6.32	Vitry le François (13 106)	442
20	61	41	3.02	Lorient (57 408)	489
21	91	49	2.17	Orléans (114 185)	848
Total	2292	1603			11 728

Owing to the fewer number of cases, RRs are stronger for bilateral cases, with several clusters showing RR > 5, including the main cluster. The cluster surfaces for bilateral cases are close to those of all-type cases.

The typology of the clusters of bilateral cases is very similar to that of the clusters of all-type cases. The same typologies are observed for several new clusters: mining (No. 7, 16) as well as heavy industry and metallurgy (No. 10, 19). However, a new typology is added: agricultural activity, with vineyards and orchards in particular, for the clusters No. 5, 14, 17 and partly No. 15.

Results of complementary analyses are detailed in the Supplementary Data Files S1 and S2, Fig. S3, and Table SII, and mentioned in the discussion.

Discussion

Our study highlights two main results. First, the risk of operated cryptorchidism increased in Metropolitan France during the period 2002–2014, especially for boys under the age of 2 years and for bilateral cases. Second, there was strong spatially structured heterogeneity for this risk during the same period, leading to the identification of several clusters scattered across Metropolitan France.

The strengths and weaknesses of these results are related to the indicator used, which reflects the incidence of operated cases. As a strength, this indicator is highly specific because the algorithm used to select cases is based on hospital codes for diseases and surgical acts, at least one of which must be specific to cryptorchidism (Suzan et al., 2011). Therefore, it is likely to reflect seamlessly and comprehensively the nationwide incidence of operated cases.

Temporal trend

The increasing risk of cryptorchidism during a recent 13-year period is consistent with the scant available literature on the epidemiology of cryptorchidism in other countries (England, Denmark, and South Korea). However, as no study has used the same method as our own to identify the cases, it is difficult to compare the incidence rates. This increasing trend could be consistent with changes in known individual risk factors, especially those of prematurity (Damgaard et al., 2008) and heavy maternal smoking, which is associated with bilateral cryptorchidism (Thorup et al., 2006). In France, as in other countries, prematurity rates have increased since 1995 (Inserm and DREES, 2017), linked to the increase in maternal age, multiple pregnancy rates after ART, and labor induction. The premature birth rate was 5.9% in 1995, 6.8% in 1998, 7.2% in 2003, and 7.4% in 2010, the year of the last available French perinatal study. The prematurity rate seemed to increase more quickly before than during the study period. As for maternal smoking, the rate increased up to 1998 but has since decreased, according to the same source. Therefore, these factors do not explain the increasing trend of cryptorchidism over the study period, even if we cannot exclude their minimal contribution.

Regarding environmental changes, the national temporal trend is consistent with an increase in the exposure to ubiquitous EDC through all routes, involving increasing numbers of chemicals since the 1950s. These chemicals include pesticides, phthalates, and flame-retardants (Balicco et al., 2019; Fillol et al., 2019), which are suspected to be linked to cryptorchidism based on the weight of evidence (Le Moal et al., 2016).

Cluster detection and hypotheses

The identification of spatial trends and spatial clusters could reflect the spatial distribution of individual and environmental risk factors of cryptorchidism, both of which may be intertwined. The data used in this study only allow us to discuss the contribution of factors that may be geographically determined or enhanced.

Genetic factors are involved in a small proportion of cases of congenital cryptorchidism. These may have contributed to spatial clusters, if the distribution of these factors is heterogeneously distributed on the territory, but we do not have data to argue this hypothesis.

According to data available online (Insee, municipalities web sites, wikipedia) several cluster areas, including the main one, faced an economic decline, especially related to the fall of industrial activities since the 1970s, which points to the local contribution of socio-economic factors in our results. The risk of congenital cryptorchidism is correlated with a small weight for gestational age, prematurity (Damgaard et al., 2008) and low socio-economic status (SES) (Moller and Skakkebæk, 1996), the latter being correlated (Lejeune, 2008) through a reduced medical monitoring during pregnancy. These perinatal factors could also reflect early placental dysfunction, possibly related to an unhealthy periconceptional maternal lifestyle, such as smoking, alcohol and under- and over-nutrition (Reijnders et al., 2019), factors also potentially associated to low SES.

As no individual socio-economic data were available, we carried out a complementary analysis adjusting the spatial analyses on a deprivation index defined at the postcode level (Rey et al., 2009). Although a positive association was observed with the risk of cryptorchidism, no change in the spatial distribution was observed and we detected mostly the same clusters with only some changes in their order (Supplementary Fig. S3 and Table SII). We carried out this analysis as a complementary analysis, because social status and especially low income are strongly associated with increased exposure to environmental risks in the private home or related to residential location (Braubach and Fairburn, 2010). So even if we considered the deprivation index as a proxy for individual socio-economic status, at the same time it is an indicator of deprivation at the geographic level and, when adjusting for it, we may be adjusting for environmental exposures.

The clusters of all-type cryptorchidism highlight shared activities in the areas of mining, smelting, and metallurgy. To explore this hypothesis, we included in the spatial model, in addition to the deprivation index, two proxys, one for exposure to mining activities and one for exposure to metallurgic activities: the density of workers (number of workers, defined at the postcode of the workplace, per km²) in these industrial sites available for the year 2006 from Insee. As for the deprivation index, we observed a positive association between the risk of cryptorchidism and both proxy variables (Supplementary Data File S2). Although this needs to be further studied, it allows us to discuss the possible contribution of these factors.

Mining and metallurgic activities lead to environmental pollution with persistent metals such as lead, arsenic, cadmium, mercury, zinc, copper, uranium, and antimony, which are likely to continue contaminating soils and waters. Several metals are regarded as reproductive endocrine disruptors (WHO (World Health Organization)/UNEP (United Nations Environment Programme), 2013) with documented effects on male reproductive functions for cadmium, lead, and mercury (Pizent et al., 2012) in cases of occupational exposure (Bonde, 2010) and also at environmental doses (Wirth and Mijal, 2010). The observed effects in animals and humans mainly relate to fertility and pregnancy outcomes in both males and females (Kumar, 2018). For example, lead exposure has been associated with intrauterine death, prematurity, and low birthweight (Papanikolaou et al., 2005), factors also associated with an elevated risk of cryptorchidism and several birth defects (Bellinger, 2005). The hypothesis linking an increased risk of cryptorchidism to prenatal metal exposure is biologically plausible because metals can cross (Mikelson et al., 2019) and even accumulate (cadmium) (Wirth and Mijal, 2010) in the placenta. In experimental studies, cadmium in particular is gonadotoxic with harmful, long-lasting, and irreversible effects if animals are treated during fetal development (de Angelis et al., 2017). To our knowledge, there is only one study to date suggesting a link between prenatal exposure to metals and the risk of cryptorchidism in humans (Morales-Suarez-Varela et al., 2011).

Metallurgic and mechanic activities are also a source of exposure to dioxins and polychlorinated biphenyls (PCBs) (Citepa, 2019) through occupational or environmental contact (Abballe et al., 2013). These pollutants are highly persistent in the environment and accumulate in the food chain, while they are also known reproductive endocrine disruptors. In animal studies, exposure to dioxins has been associated with the disrupted development of the male reproductive system, including testicular maldescent (Virtanen et al., 2012). In humans, data are scarce. A French study (Brucker-Davis et al., 2008) suggested an association between PCB levels in the umbilical cord blood and colostrum and the risk of cryptorchidism, whereas a Danish study using measures of PCBs and dioxins in the placenta did not (Virtanen et al., 2012).

Regarding metal as well as dioxin/PCB exposure, it cannot be excluded that the current effects in the clusters could be amplified by epigenetic mechanisms, involving the ancestral exposure of one or several generations who lived and worked there (Nilsson and Skinner, 2015; Skakkebæk et al., 2016). Indeed, metal (Martinez-Zamudio and Ha, 2011) and dioxin/PCB (Manikkam et al., 2012; Vaiserman, 2014; Pilsner et al., 2017; Gillette et al., 2018; Mennigen et al., 2018; Patrizi and Siciliani de Cumis, 2018) exposure seem to cause transmissible epigenetic marks. In support of this hypothesis, a Danish study suggested an association between paternal—not maternal—occupational exposure to heavy metals and the risk of cryptorchidism (Morales-Suarez-Varela et al., 2011).

Pesticide exposure can be discussed as a possible geographically determined environmental hypothesis although it emerges more weakly in specific areas. Pesticides are highly used in agricultural areas in France and a recent review suggests that residents in the proximity of these areas are particularly exposed to agricultural pesticides (Dereumeaux et al., 2020). Increasing numbers of studies show that pesticides can be regarded as reproductive endocrine disruptors (WHO (World Health Organization)/UNEP (United Nations Environment Programme), 2013) and are often persistent in the environment, with agricultural soils showing reproductive hormone activity (Zhang et al., 2018). Several studies found an increased risk of cryptorchidism in the sons of mothers (Virtanen and Adamsson, 2012; Garcia et al., 2017) and fathers (Pierik et al., 2004) occupationally exposed to pesticides. However, associations have not consistently been observed between pesticide levels in maternal biological matrices and cryptorchidism frequency (Brucker-Davis et al., 2008; Virtanen and Adamsson, 2012). Lastly, a recent meta-analysis on the link between prenatal or perinatal EDC exposure and male reproductive developmental outcomes, including cryptorchidism, concluded there was an increased risk with exposure to the pesticide p, p′ dichlorodiphenyldichloroethylene, a metabolite of dichlorodiphenyltrichloroethane (DDT) (Bonde et al., 2016). As regards the epigenetic hypothesis for pesticides, several pesticides, especially those previously used in vineyards (DDT, vinclozoline, metoxychlore, atrazine, permethrin), are known to have transgenerational in-vitro reproductive effects (Nilsson and Skinner, 2015; Murase et al., 2018).

Limitations

This study has several limitations. First, the indicator is poorly sensitive: it does not reflect self-resolved cases, which mainly occur before 1 year of age and account for around 50% of congenital cases (Barthold and Gonzalez, 2003). According to some authors, this type of cryptorchidism is assumed to be particularly related to EDC exposure (Brucker-Davis et al., 2008). The sensitiveness is even poorer for bilateral cases, because they could be operated separately and thus coded as unilateral cases.

Changes in the management of cryptorchidism during the study period may be mentioned. In France, the medical treatment of cryptorchidism has been rarely used, and it was abandoned in 2005 following the advice of experts. In 2011, the French Urological Association recommended operating on boys with congenital cryptorchidism before 2 years of age (Association Française d’Urologie, 2011), and the French Society of Pediatric Urology between 1 and 3 years (Société Française de Chirurgie Pédiatrique, 2011). As the temporal trend for all ages and for boys aged 0–2 years is linear during the study period (Fig. 1), it is unlikely that these recommendations have substantially contributed to the increasing trend. Moreover, according to experts, changes in terms of disease detection during the study period are unlikely because the mainly clinical detection methods did not evolve. We therefore argue that the increase in operated cryptorchidism mostly reflects a real increase in cases.

Spatial trends could reflect spatially determined medical and surgical practices. Some surgery teams may have anticipated the international recommendations to operate on boys before 1 year of age when possible. This issue would be mostly resolved with our inclusion of boys under 7 years of age. Nevertheless, we mapped the indicator for boys aged 1–3 years (Supplementary Fig. S4) to reflect the national recommendations, and it is similar to the map for all ages. Taken together, these points suggest the robustness of the national spatial results but we cannot exclude the role of localized specific practices. We questioned whether the surgical practices in the main cluster differed from the national practices (Supplementary Fig. S5) and did not identify differences according to age.

Another limitation of the study is its localization of cases at the date of surgical intervention without taking into account mobility since pregnancy. Localizing the cases during pregnancy or at birth would be better in terms of assessing the role of environmental exposure. According to a recent report on residential mobility in France (Haran and Garnier, 2018), around 10% of children aged 0–5 years move to a new municipality every year, among which only 3–4% move in a new department. Moreover, using the study data, we estimated that the percentage of children moving out of their commune (not necessarily their postcode) increases with age, from 13% at 1 year to 40% at 6 years. Taken together, these data suggest that our results are more robust for the youngest cases given that boys aged 0–3 years account for two-thirds of operated cases.

In conclusion, to our knowledge, this is the first study that addresses the temporal and spatial trends of cryptorchidism in France with the largest sample ever studied. Our results show an increasing trend of operated cryptorchidism between 2002 and 2014 and suggest that local geographic factors may contribute to the risk of operated cryptorchidism. In addition to the known role of a low SES, we identified several environmental hypotheses, providing new insights into the possible contribution of geographic EDC exposure. This part of the study could be regarded as a proof of concept about the interest of the national monitoring of diseases related to EDC exposure, and should be considered as a hypothesis-generating process for future etiological research studies.

Data availability

According to data protection and the French regulation, the authors cannot share the data used in this study. However, any person or structure, public or private, for-profit or non-profit, is able to access the National Health Data System (SNDS, https://www.snds.gouv.fr) upon authorization from the National Computers and Privacy Commission (CNIL), in order to carry out a study, a research or an evaluation of public interest.

Acknowledgments

We thank Jean-Claude Desenclos, Director for Science at Santé publique France, for his support, reviewing and helpful advice about this study, and Pr Patrick Fénichel (Nice Côte d’Azur University) for his helpful expertise in the reproductive and environmental health field and reviewing.

Authors’ roles

JLM and SG contributed to the conception and the design of the study. SG conducted the statistical analysis and JLM the discussion of the results. Both draft the manuscript. AG was in charge of the geographic data management and mapping. AR and JL were in charge of the data management.

Funding

This study was entirely funded by Santé publique France, the French National Public Health Agency.

Conflict of interest

All authors declare they have no actual or potential competing financial interest.

Author notes

J. Le Moal and S. Goria Joint first authorship.