Marijuana Smoking and Markers of Testicular Function Among men From a Fertility Centre

Feiby L. Nassan; Mariel Arvizu; Lidia Mínguez-Alarcón; Paige L. Williams; Jill Attaman; John Petrozza; Russ Hauser; Jorge Chavarro; for the EARTH Study Team


Hum Reprod. 2019;34(4):715-723. 

In This Article

Materials and Methods

Study population

Men from couples presenting for evaluation at the Massachusetts General Hospital (MGH) Fertility Center between 2000 and 2017 were invited to participate in an ongoing study aimed at identifying environmental determinants of fertility (Meeker et al., 2011; Messerlian et al., 2018). Of the approached men, ~55% agreed to participate. All men signed an informed consent. The studies were approved by institutional review boards at the Harvard T. H. Chan School of Public Health and MGH.

Of the 1011 men recruited, 280 men did not answer questions regarding drug use. Furthermore, 18 men were azospermic and excluded. We also excluded 51 men who did not have complete semen analysis data. The remaining 662 men contributed 1143 semen samples between 2000 and 2017 (Supplementary Figure 1). This included 296 semen samples, from men enroled between 2000 and 2004, that were previously analysed for sperm DNA damage. Of the 662 men, 317 also provided serum samples that were analysed for reproductive hormones. Due to limited resources, not all 1143 semen samples were analysed but selection was unrelated to semen analysis results, type or outcome of any infertility treatments, marijuana smoking status or any other participant's characteristic. Differences in participant characteristics between men included and men excluded from the analysis were minor (Supplementary Table I).

Supplementary Figure S1.

Flowchart of men included in the analysis of marijuana smoking in relation to testicular markers in our study.

Marijuana smoking and covariate assessment

At baseline, men reported marijuana smoking in a self-administered questionnaire. Specifically, they first reported if they had ever smoked marijuana (more than two joints/cigarettes or the equivalent amount of marijuana in your lifetime) and if they were current marijuana smokers. Among ever smokers, we also assessed the average number of joints/cigarettes (or equivalent amount of marijuana) they smoked per week, whether they ever quit and for how many years, age of starting to smoke marijuana, last time they smoked marijuana, and the total duration of marijuana smoking. The questionnaire had similar questions about cocaine use. Men also self-reported demographic information, data on other lifestyle factors and medical history. A research nurse abstracted clinical information from medical records and measured their height and weight to calculate body mass index (BMI) (kg/m2) at the time of enrolment.

Semen analysis

Men provided a semen sample onsite at the MGH andrology laboratory by masturbation into a sterile plastic specimen cup. Men were asked to abstain from ejaculation for 2–5 days before providing the semen sample. Men reported the duration of abstinence before providing the sample. All semen samples were analysed using standardised protocols and quality control was as described previously (Nassan et al., 2016). Before analysis, the sample was liquefied at 37°C for 20 min after collection. Ejaculate semen volume (mL) was measured using a graduated serological pipet. Sperm concentration (million/mL) and % motility were assessed using a computer-aided semen analyser (CASA; 10HTM-IVOS, Hamilton-Thorne Research, Beverly, MA, USA). We calculated the total sperm count (million/ejaculate) as semen volume × sperm concentration. Sperm morphology (% normal) was assessed on two slides per specimen (with a minimum of 200 cells assessed per slide) via a microscope with an oil-immersion ×100 objective (Nikon, Tokyo, Japan). Strict Kruger scoring criteria were used to classify men as having normal or below normal morphology (Kruger et al., 1988). The andrologists participate regularly in internal and external quality control checks.

Sperm DNA integrity

The neutral comet assay was used following the previously described protocol (Meeker, Yang, Ye, Calafat and Hauser, 2011; McAuliffe et al., 2014; Nassan et al., 2018). Briefly, 50 μL of a semen/agarose mixture was embedded between two additional layers of agarose on microgel electrophoresis glass slides. Slides were immersed in a cold lysing solution to dissolve the sperm cell membranes and make sperm chromatin available. After 1 h of cold lysis, slides were transferred to a solution for enzyme treatment with RNAse (Amresco, Solon, OH) and incubated at 37°C for 4 h. Slides were transferred to a second enzyme treatment with proteinase K (Amresco) and incubated at 37°C for 18 h then placed on a horizontal slab in an electrophoretic unit toundergo electrophoresis for 1 h. DNA in the gel was subsequently precipitated, fixed in ethanol and dried. Slides were stained and observed using a fluorescence microscope.

Comet extent (CE), DNA percent in the tail (%tail) and tail distributed moment (TDM) were assessed in 100 sperm cells in each semen sample using the VisComet software (Impulus Computergestutze Bildanalyse GmbH, Gilching, Germany). CE represents the average total comet length in μm from the beginning of the head to the last visible pixel in the tail. %Tail represents the average proportion of DNA that is in the tail of the comet. TDM represents an integrated measure that takes into account the distance and intensity of comet fragments (Nassan et al., 2018). TDM is calculated as Σ(I × X)ΣI, where ΣI is the sum of all intensity measures for the head, body or tail, and X is the x-position of the intensity measure. An additional measure of sperm DNA damage used was the counted number of sperm cells with CE > 300μm, i.e. too long to measure with VisComet.

Reproductive hormones

A non–fasting blood sample was drawn between 9 a.m. and 4 p.m. on the same day of the first semen sample in a subset of the men. Blood was centrifuged, and serum was stored at −80°C until analysis. Serum was thawed and analysed in one batch for follicle-stimulating hormone (FSH), luteinising hormone (LH), estradiol, inhibin-B, total testosterone and sex hormone-binding globulin (SHBG). FSH, LH, and estradiol concentrations were determined by microparticle enzyme immunoassay using an automated Abbot AxSYM system (Abbott Laboratories, Chicago, IL). The assay sensitivities were 1.1 IU/L for FSH and 1.2 IU/L for LH. The intra-assay coefficient of variation (CV) for FSH and LH was <5% and <3%, respectively with inter-assay CVs for both hormones of <9%. The assay sensitivity for estradiol was 20 pg/mL with a within-run CV between 3% and 11%, and the total CV was between 5% and 15%. Total testosterone was directly measured using the Coat-A-Count RIA kit (Diagnostic Products, Los Angeles, CA), which had a sensitivity of 4 ng/dL, inter-assay CV of 12% and intra-assay CV of 10%. Inhibin-B was measured using a double-antibody enzyme-linked immunosorbent assay (Oxford Bioinnovation, Oxford, UK) with inter-assay CV of 20% and intra-assay CV of 8%. SHBG was measured using an automated system (Immulite; DPC Inc, Los Angeles, CA), which used a solid phase two site chemiluminescent enzyme immunometric assay and had an inter-assay CV of <8%.

Statistical analysis

We calculated descriptive statistics for baseline characteristics across categories of marijuana smoking and tested for differences across categories. We natural-log transformed ejaculate volume, sperm concentration, total sperm count, CE, %tail, TDM and serum hormone concentrations. We used linear mixed effect models to evaluate the associations of marijuana smoking with semen parameters and included a random intercept for each man to account for the longitudinal collection of multiple semen samples per man. For sperm DNA fragmentation measures and serum hormones, we used linear regression models. We used Poisson regression to model the number of cells with high DNA damage while accounting for over-dispersion. All results are presented as adjusted marginal means (Searle et al., 1980). The primary analyses consisted of evaluating men's marijuana smoking at enrolment (never/ever and never/past/current) in relation to study outcomes. Among the marijuana smokers, we also analysed the association of joint-years of marijuana smoking (joints/day for the total duration of marijuana smoking in years) with the same outcomes. In addition, we evaluated the association of time since last use of marijuana and sample collection, and age at the start of marijuana smoking. Potential confounders were selected based on prior knowledge and descriptive statistics in the study population. The final model adjusted for age, race, sexual abstinence time, BMI, tobacco smoking, coffee and alcohol intake, cocaine use and calendar year. In the sperm motility models, we further adjusted for duration elapsed between semen sample collection and analysis. We also conducted an additional analysis in which semen parameters were dichotomised as above or below WHO-2010 lower reference limits (WHO, 2010) using the first semen sample per man (closest to marijuana assessment). In this analysis, we used generalised linear models with a binary distribution and logit link adjusting for the same covariates as above.

To assess the robustness of our results, we conducted a series of sensitivity analyses including (1) re-categorising the marijuana smoking status based on last time reported of smoking marijuana (recent if ≤2 years, and past of >2 years), (2) restricting analyses to men who did not receive a male factor infertility diagnosis, (3) restricting analyses to the first semen sample per man which was closest to reporting the marijuana smoking, (4) further adjustment for history of sexually transmitted diseases (STDs), and stress levels as assessed by the standardised perceived stress scale 4 (Cohen et al., 1983; Cohen and Janicki-Deverts, 2012) and (5) further adjusting the testosterone models for time of serum sample collection. In addition, to address the possibility of selection bias, we compared the characteristics at enrolment and the semen parameters between men included in the main analysis versus those who were excluded. Finally, we calculated the E-value (VanderWeele and Ding, 2017) to quantitatively assess the potential impact of unmeasured confounding on the observed associations, conditional on the measured covariates. We conducted all statistical analyses using SAS version 9.4 (SAS Institute Inc., Cary, NC).