Human Reproduction Update Advance Access originally published online on August 25, 2005
Human Reproduction Update 2005 11(6):607-611; doi:10.1093/humupd/dmi032
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Using infertile patients in epidemiologic studies on subfecundity and embryonal loss
1 The Danish Epidemiology Science Centre, University of Aarhus, Aarhus C, Denmark, 2 Department of Epidemiology, UCLA School of Public Health, Los Angeles, CA, USA, 3 Department of Occupational Medicine and 4 Department of Gynaecology and Obstetrics, Aarhus University Hospital, Aarhus, Denmark
5 To whom correspondence should be addressed at: The Danish Epidemiology Science Centre, University of Aarhus, Vennelyst Boulevard 6, Building 260, DK-8000, Aarhus C, Denmark. E-mail: jo{at}soci.au.dk
Submitted on October 21, 2004; revised on June 29, 2005; accepted on July 19, 2005.
 |
Abstract |
|---|
 TOP Abstract IntroductionRecruiting infertility patients...External validityDesign options for follow-up...Limitations in the cause-effect...Pros and consConclusionReferences |
|---|
Subfecundity is a frequent and serious problem that may sometimes be preventable, but we need to know more about its determinants. Different epidemiologic designs are available. The best of these use prospectively collected data from the population, but they are time consuming, expensive and often hampered by low-participation rates. Most patients undergoing infertility treatment are closely monitored for clinical reasons, making it feasible to use secondary data to study the period from conception to implantation and pregnancy. In spite that infertility patients are highly selected, there are specific exposure-effect relations that can be studied in cohorts of infertility patients. These patients offer a potentially useful setting for studying exposures that operate late in fertilization, whereas the designs may be inadequate to identify exposures that cause reduced sperm counts, anovulation and total occlusion. The clinical sampling and the treatment set limitations for what can be studied. In certain situations, infertile patients can, however, provide useful epidemiologic evidence for learning about the causes of subfecundity.
Key words: epidemiology / ICSI outcome / infertility / IVF
|
Introduction |
|---|
|
TOPAbstractIntroductionRecruiting infertility patients...External validityDesign options for follow-up...Limitations in the cause-effect...Pros and consConclusionReferences |
|---|
About 15% of all couples trying to become pregnant will not succeed within the first 12 months of trying, and a substantial proportion of these will seek infertility treatment (Rachootin and Olsen, 1982
). Whether couple subfecundity is an increasing problem is not known, but its frequency and the related social, economic and health consequences justify more research into its possible determinants.
Some determinants of subfecundity are known, such as advanced age, some diseases, some environmental exposures (Joffe, 2003) and some lifestyle factors like smoking (Bolumar et al., 1996). For other potential determinants (such as alcohol and exposure to welding), findings have been inconsistent (Baird et al., 1986; Jensen et al., 1998; Hjollund et al., 1998a,b; 2000; Juhl et al., 2003).
Epidemiologic studies on the determinants of subfecundity have mainly addressed two time windows of exposure: (i) downstream causes acting at the period of trying and (ii) upstream causes operating at the period of organogenesis. The so-called estrogen hypothesis (Sharpe and Skakkebaek, 1993) focuses upon estrogen exposure during fetal life and semen quality in the offspring, but other exposures are also of interest (Storgaard et al., 2003). Both male and female fecundity may, at least in part, be shaped during fetal life or in puberty. Studies with this time window of exposure will not be addressed in this article where we focus upon current exposures during fertile age. We discuss the potential for using infertility patients for studying determinants of subfecundity with emphasis on IVF patients.
Fecundity is a function of the timing and frequency of sexual intercourses and the reproductive capacity of the partners. The probability of achieving a pregnancy as a function of a set of determinants is best studied in a prospective setting, especially if the determinants are difficult to record retrospectively in an unbiased way. It is possible, but difficult and expensive, to set up prospective studies (Wilcox et al., 1988; Bonde et al., 1998; Colombo and Masarotto, 2000; Buck et al., 2004).
In a Danish follow-up study (Bonde et al., 1998), it was necessary to contact more than 50 000 childless cohabitating couples in the age group of normal procreation to find about 400 couples who had never tried to become pregnant before, were planning a pregnancy within the following year and were willing to participate in a detailed and time-consuming data collection. All follow-up studies have been highly selected, either by choice or by design (belonging to a certain group, willing to spend time on research or being couples seeking pregnancy planning advice).
To plan a pregnancy is a private decision that is usually not recorded or computerized, and most studies have to use retrospective data because only a clinically recognized pregnancy is easy to ascertain. Like in the follow-up studies, these studies focus upon the time to pregnancy (TTP) or cycles to pregnancy (CTP) as endpoints, but those who did not become pregnant are excluded, making it impossible to study exposures that cause sterility. The additional problems related to these studies are now well described (Weinberg et al., 1994; Barlow, 2003; Olsen and Basso, 2004).
Patients seeking infertility treatment are identified before they conceive, and they often go through many clinical procedures on a routine basis. In some countries, they are well registered and closely followed (Andersen et al., 1999), at least until the end of an achieved pregnancy. During follow-up, sensitive pregnancy tests and ultrasound examinations and other monitoring measures are performed. As part of the fertility workout, we would have detailed data on possible biological causes of their subfecundity, including semen characteristics. Many of these data are difficult to obtain for research purposes only.
The question we address in the following is ‘are infertility patients (especially patients seeking IVF treatment) useful in epidemiologic research focusing upon determinants of subfecundity, especially external determinants?’
|
Recruiting infertility patients to aetiologic research |
|---|
|
TOPAbstractIntroductionRecruiting infertility patients...External validityDesign options for follow-up...Limitations in the cause-effect...Pros and consConclusionReferences |
|---|
Infertility patients seek treatment and usually have no personal benefits from taking part in studies related to prevention. They have an invisible disease that interferes with their role in evolution and with social stereotypes for procreation. Many couples have undergone several treatment attempts and have been ‘medicalized’ in a way that may be stigmatizing and difficult to cope with. It adds to this psychological burden that many consider childlessness from infertility to be a self-imposed and decadent problem in a world with a population growth out of control in many countries.
Our experience with collaboration with both private and public infertility clinics has, however, been good, but obtaining informed consent from patients has proved difficult. Some dislike being reminded of an unsuccessful treatment, they would rather try to forget. Some feel their privacy to be violated even when approached for informed consent. Some feel they do not need to exercise any altruistic deeds because they paid for the treatment themselves (N.H. Hjøllund, personal communication).
|
External validity |
|---|
|
TOPAbstractIntroductionRecruiting infertility patients...External validityDesign options for follow-up...Limitations in the cause-effect...Pros and consConclusionReferences |
|---|
About half the couples with infertility seek medical help (Olsen et al., 1996
), making infertile patients unsuitable to provide representative data on most exposures. In aetiologic research, our interest is, however, not related to representative data but primarily to estimate health effects of certain exposures. To estimate such effects, we make use of comparisons of exposed and unexposed cycles, and the aim is to control for as many factors as possible that could distort this comparison. To achieve this, we need data on confounders and to make sure that infertility patients are not recruited as a function of the exposure pattern we want to study (Olsen, 1994).
Furthermore, we usually want to study cause-effect mechanisms that are not particularistic to infertility patients. It has recently been shown that infertility patients have a 20% higher miscarriage rate than other pregnancy planners (Wang et al., 2004), and causes of spontaneous abortions in infertility patients may thus include causes that are specific for these patients.
Infertility patients are often subject to treatment that may interfere with the determinants under study and by treatments that aim at altering the natural process of fertilization (Schenker and Ezra, 1994; Dew et al., 1998; Silver et al., 1999; Goldfarb et al., 2000; Strom et al., 2000; Steirteghem et al., 2002).
Some of the problems related to using case–control designs may be bypassed by sampling controls within the group of infertile patients. When using treated patients as cases, they determine the study base because only a fraction of infertile couples receive treatment. The underlying cohort under study is now those who would seek treatment if they tried in vain to become pregnant, making it possible to do case–control sampling within infertile couples. Such a study could make use of the difference in the couples’ aetiology of infertility. Women who are biologically ‘normal’ may come in contact with infertility clinics, because their male partner has low sperm counts and they could serve as controls for a study that addresses determinants of female infertility (Rachootin and Olsen, 1983). Caution is, however, needed because different forms of selection may bias the results (Tielemans et al., 2001). The design will, e.g., be overmatched for exposures that cluster in couples, like environmental exposures related to the place of living.
|
Design options for follow-up studies |
|---|
|
TOPAbstractIntroductionRecruiting infertility patients...External validityDesign options for follow-up...Limitations in the cause-effect...Pros and consConclusionReferences |
|---|
Within couple comparisons
An interesting possibility is to make use of the existing exposure variation between different treatment cycles for the same couple. Exposures, such as infections, dietary habits, alcohol intake, medications, traumatic events, psychological distress, and some occupational exposures, may vary even over short-time spans and may thus be subject to scrutiny in this design if the time between exposure and effect is short. Exposed treatment cycles are compared with unexposed cycles for the same couple with a given treatment sequence. This design option was called case crossover by Maclure in 1991. The aim of this design is to make use of the underlying incidence rate of conception/abortion, in exposed cycles and in unexposed cycles. In the case–crossover study, we will estimate how much more frequently the event is seen after the exposure compared with the situation when the exposure was not present (counterfactual argument). The event may be a pregnancy defined by hormonal tests or ultrasound, embryonal loss or later pregnancy outcomes.
Figure 1 illustrates the identification of the case series within a group of infertile; all with at least two cycles and the event (conception) for the case series. If the cycle leading to the event and the preceding cycle are used as the two-time windows of interest, then only couples who had a discordant exposure history in these two time periods are informative.
|
In Figure 1, patient 1 is informative, indicating a negative effect of exposure on fecundity. Patient 2 is not informative (no event). Patient 3 is not informative; the third cycle led to a pregnancy following exposure but the second cycle (the referent cycle) was exposed as well. The fourth patient is informative, indicating a positive effect of exposure. The fifth patient is not informative, and the sixth patient provides evidence for a negative effect of the exposure.
The pregnancy odds ratio (OR) would simply be
 |
In order for selection bias to be active, recruitment of case series have to depend not only upon the event but the discordant pattern of exposure leading up to the event. This type of selection is unlikely if the number of allowed treatment cycles is low and within the normal treatment modality. To the best of our knowledge, this design has not been used yet in this setting.
Because the woman (the couple) is her own control, all time stable factors are adjusted for in the design. That will include genetic factors, many environmental exposures and even the underlying diseases and the treatment given to the disease, and the treatment does not vary over time.
By comparing one set of exposed infertile couples with another set of unexposed infertile couples in a standard follow-up study, a wider exposure variation will be available, and in principle all treated couples who meet the inclusion criteria based upon social and disease specific criteria could be enrolled. In this design, there is no requirement of a minimum number of treatment cycles or a specific set of cycle combinations, nor is it needed to require an exposure variation within the treated cycles for the cycles to be informative. In fact, stable exposure conditions, such as living close to power lines or radio transmitters, job exposures, or lifestyle habits like smoking are proper issues to study. Van Noord-Zaadstra was one of the first to use infertile couples in studying determinants of subfecundity. In her paper from 1991 (Van Noord-Zaadstra et al., 1991), she used women who were living with azoospermic males and were treated with donor semen for the first time to study the association between female age and female fecundity. She simply compared the probability of a successful pregnancy in treated cycles in different age groups.
In the proposed design, women exposed to the agent in question during their treated cycles are compared with other treated women, who are not exposed. In the absence of randomization, these two groups are made as comparable as possible by using proper exclusion/inclusion criteria and statistical modelling or stratification. The aim is to establish an unexposed treated group that would have the same outcome in their treated cycles as the exposed women would have had, had they not been exposed.
The between-couples design using IVF-treated patients has been used on several occasions to study smoking, alcohol, occupational hazards, age and twinning (Sills et al., 2000; Strom et al., 2000; Tielemans et al., 2001; Hjollund et al., 2004, 2005). Because it makes use of the entire data source and requires fewer restrictions on the cycle/exposure structure, it is more flexible than the case–crossover design. The price to pay for this flexibility is a higher potential for bias.
If data are analysed as a follow-up study, endpoints like a pregnancy, a spontaneous abortion, a birth or a birth with given characteristics are used. It is important to consider the timing of the exposure and of potential confounders in relation to the hypothesised ‘windows’ of causation, and the timing will differ for different endpoints.
|
Limitations in the cause-effect spectrum available for analyses |
|---|
|
TOPAbstractIntroductionRecruiting infertility patients...External validityDesign options for follow-up...Limitations in the cause-effect...Pros and consConclusionReferences |
|---|
The selection of oocytes and sperm cells in IVF makes it possible to study exposures resulting in effects that operate during fertilization, implantation or in early pregnancy. Selecting the oocytes and sperm cells with the best characteristics probably brings better control to the natural variation in sperms and oocytes available for the pregnancy attempt.
If it is possible to obtain 100 000 ‘normal’ sperm cells per oocyte, standard IVF is often performed. If there are fewer, many choose ICSI. In any case, sperm cells are selected, and the available number of oocytes may depend upon the FSH dose as well as other exposures. Only mature metaphase II oocytes will be fertilized and smoking, obesity, and other factors may have an influence on the process of maturation.
When preembryos are selected for transfer, a quality scoring is used. The number of embryos screened as well as the score value reflecting expected quality must therefore be adjusted for in the analysis. Without a doubt, additional diagnostic tools will be developed in the future. The better predictive value these tools will have, the better options we have for confounder adjustments.
The selected samples make it almost impossible to study exposures that operate before sperm ejaculation or ovulation unless the exposures result in effects on germ cells that are undetectable by the procedures used to choose germ cells. For instance, exposures to germ cell mutagenes or clastogenes might cause germ genomic damage or alterations of the germ cell chromatin structure that escapes recognition by the procedures used for preparation of specimens and that are not interfering with fertilizing capacity or growth of early stage embryo but do have effects on embryo implantation and survival. Exposures that impair the process before these events, but leave a few cells untouched and thus suitable for fertilization, could be overlooked because of this clinical selection.
IVF designs are not useful to detect exposures that cause subfecundity by reducing sperm count or by impairing sperm morphology or motility because some of these pathways are partly bypassed by the IVF technique. Similarly, in women, anovulation and tubal causes of subfecundity will not be easily detectable. This applies to within as well as between individual comparisons.
IVF may be useful for studies of exposures that interfere with fertilization (sperm–oocyte fusion), implantation and survival of the embryo—in particular early survival including male- and female-mediated embryonal loss. In this particular case, IVF data are of special value because of the ultrasounds, hormonal monitoring and so on, that are so hard to get in other settings. Thus IVF designs are not an alternative to population-based studies of subfecundity but may be a design option when some specific endpoints as fertilization, implantation and early embryo survival are of interest.
The ICSI method (and other more invasive methods) add important additional new procedures that may interfere with our ability to use data for making scientific inference. By injecting the sperm directly into the oocyte, biochemical consequences of natural penetration are bypassed and this may in itself add risk to the treatment. Such risk opens the door to causal relationships that may be irrelevant in normal procreation (Sills et al., 2000). If testicular spermatozoa are used, imprinting may be less complete, and the risk of structural or functional defects perhaps increased (Olsen et al., 1998), a situation that we would never observe in couples who conceive naturally.
|
Pros and cons |
|---|
|
TOPAbstractIntroductionRecruiting infertility patients...External validityDesign options for follow-up...Limitations in the cause-effect...Pros and consConclusionReferences |
|---|
The main advantage of using infertile couples in epidemiologic research is the existence of data and follow-up that would otherwise be too expensive to set up for research purposes alone in an unselected population. Furthermore, IVF patients often have a larger number of intermediate fertility data recorded, such as ovulation induction response, number of gametes available, fertilization rate, early implantation and so on. These endpoints provide useful design options to address different steps in the process towards a successful pregnancy.
The main disadvantage is that these monitoring systems are set up to monitor clinical success rather than to study avoidable causes of infertility. Furthermore, follow-up of treatment cycles may often be incomplete, if the couples dropout of treatment. Also, data may be missing or not be part of the routine recording system.
The case–crossover design offers a powerful design option, but only for a limited set of hypotheses. Exposure variation is needed during follow-up, and the exposure should have an immediate effect that is not carried over to the next cycle if exposure stops. Furthermore, the exposure should not have an inherent time-dependent variation. An exposure that changes in intensity over calendar time will produce biased results. In this situation, a sample from the base population is needed to adjust for the time trend (Greenland, 1996).
If exposure data cannot be built into the monitoring system to allow for prospectively collected data, the risk always exists that knowing the outcome of the treatment may interfere with how self-reported data are recorded retrospectively. That is probably of major concern for self-reported unhealthy life-style habits, but less of a problem for objective and well-recalled exposures, such as age, work tasks, long-lasting illnesses and some environmental exposures.
If the research question is of the type: ‘Will exposure to welding fumes increase the risk of early spontaneous abortion?’, we need not collect data on all in an IVF registry. We would rather select women who had an early abortion and try to reconstruct their welding history from the preconceptional time up to the time of abortion. We will do the same for a sample of women who reached the same stage in pregnancy without a miscarriage. We may select one or more controls per case, and it is possible to obtain almost the same statistical precisions as if we had collected welding information on all in the registry, but at much lower cost. A case–control study nested within a well-registered cohort is a powerful design that allows for a wide search of possible determinants of the disease, but the risk of making multiple comparisons should always be kept in mind.
If we ask a question of the type: ‘Does this exposure have an effect on any reproductive outcome?’, we need to reconstruct the entire follow-up study and look at all endpoints from lack of implantation, preclinical or clinical abortion, stillbirth, structural or functional defects and so on. Some women may, however, give up potentially harmful exposures, such as welding, if they feel that they can interfere with the success of the treatment. It may, therefore, lead to too little variation in the exposure distribution for meaningful analyses. Furthermore, use of IVF patients limits the spectrum of endpoints available for research.
|
Conclusion |
|---|
|
TOPAbstractIntroductionRecruiting infertility patients...External validityDesign options for follow-up...Limitations in the cause-effect...Pros and consConclusionReferences |
|---|
The use of infertile patients in aetiologic research has potential strengths and deserves more attention and closer scrutiny, although the selection of patients and the inference with treatment sets many limitations. If the ongoing clinical data collection could add exposure assessments (data or biological samples), this data source could be even more valuable in aetiologic research.
|
Acknowledgements |
|---|
|
TOPAbstractIntroductionRecruiting infertility patients...External validityDesign options for follow-up...Limitations in the cause-effect...Pros and consConclusionReferences |
|---|
The Danish Epidemiology Science Centre is supported financially by the Danish National Research Foundation and the Danish Medical Research Council (grant no. 22-02-0363).
|
References |
|---|
|
TOPAbstractIntroductionRecruiting infertility patients...External validityDesign options for follow-up...Limitations in the cause-effect...Pros and consConclusionReferences |
|---|
Andersen AN, Westergaard HB and Olsen J (1999) The Danish in vitro fertilisation (IVF) register. Dan Med Bull 46,357–360.[ISI][Medline]
Baird DD, Wilcox AJ and Weinberg CR (1986) Use of time to pregnancy to study environmental exposures. Am J Epidemiol 124,470–480.
Barlow DH (2003) The design, publication and interpretation of research in subfertility medicine: uncomfortable issues and challenges to be faced. Hum Reprod 18,899–901.
Bolumar F, Olsen J, Boldsen J and the European Study Group on Infertility and Subfecundity (1996) Smoking reduces fecundity: a European multicenter study on infertility and subfecundity. Am J Epidemiol 143,578–587.
Bonde JP, Hjollund NH, Jensen TK, Ernst E, Kolstad HA, Henriksen TB, Giwercman A, Skakkebaek NE, Andersen A and Olsen J (1998) A follow-up study of environmental and biologic determinants of fertility among 430 Danish first pregnancy planners: design and methods. Reprod Toxicol 12,19–27.[CrossRef][ISI][Medline]
Buck GM, Lynch CD, Stanford JB, Sweeney AM, Schieve LA, Rockett JC, Selevan SG and Schrader SM (2004) Prospective pregnancy study designs for assessing reproductive and developmental toxicants. Environ Health Perspect 112,79–86.[ISI][Medline]
Colombo B and Masarotto G (2000) Daily fecundability: first results from a new data base. Demogr Res 3,5.
Dew JE, Don RA, Hughes GJ, Johnson TC and Steigrad SJ (1998) The influence of advanced age on the outcome of assisted reproduction. J Assist Reprod Genet 15,210–214.[CrossRef][ISI][Medline]
Goldfarb JM, Austin C, Peskin B, Lisbona H, Desai N and Loret de Mola JR (2000) Fifteen years experience with an in-vitro fertilization surrogate gestational pregnancy programme. Hum Reprod 15,1075–1078.
Greenland S (1996) Confounding and exposure trends in case-crossover and case-time-control designs. Epidemiology 7,231–239.[ISI][Medline]
Hjollund NH, Bonde JP, Jensen TK, Ernst E, Henriksen TB, Kolstad HA, Giwercman A, Skakkebaek NE and Olsen J (1998a) Semen quality and sex hormones with reference to metal welding. Reprod Toxicol 12,91–95.[CrossRef][ISI][Medline]
Hjollund NH, Bonde JP, Jensen TK, Henriksen TB, Kolstad HA, Ernst E, Giwercman A, Pritzl G, Skakkebaek NE and Olsen J (1998b) A follow-up study of male exposure to welding and time to pregnancy. Reprod Toxicol 12,29–37.[CrossRef][ISI][Medline]
Hjollund NH, Bonde JP, Jensen TK, Henriksen TB, Andersson AM, Kolstad HA, Ernst E, Giwercman A, Skakkebaek NE and Olsen J (2000) Male-mediated spontaneous abortion among spouses of stainless steel welders. Scand J Work Environ Health 26,187–192.[ISI][Medline]
Hjollund NH, Bonde JP, Ernst E, Lindenberg S, Andersen AN and Olsen J (2004) Pesticide exposure in male farmers and survival of in vitro fertilized pregnancies. Hum Reprod 19,1331–1337.
Hjollund NH, Bonde JP, Ernst E, Lindenberg S, Andersen AN and Olsen J (2005) Spontaneous abortion in IVF couples – a role of male welding experience. Hum Reprod 20,1793–1797.
Jensen TK, Hjollund NHI, Henriksen TB, Scheike T, Kolstad H, Giwercman A, Ernst E, Bonde JP, Skakkebaek NE and Olsen J (1998) Does moderate alcohol consumption affect fertility? Follow up study among couples planning first pregnancy. BMJ 317,505–510.
Joffe M (2003) Infertility and environmental pollutants. Br Med Bull 68,47–70.
Juhl M, Olsen J, Nybo Andersen AM and Groenbaek M (2003) Intake of wine, beer and spirits and waiting time to pregnancy. Hum Reprod 18,1967–1971.
Maclure M (1991) The case-crossover design: a method for studying transient effects on the risk of acute events. Am J Epidemiol 133,144–153.
Olsen J (1994) Options in making use of pregnancy history in planning and analysing studies of reproductive failure. J Epidemiol Community Health 48,171–174.[Abstract]
Olsen J and Basso O (2004) Reproductive epidemiology. In Pigeot J and Ahrens W (eds), Handbook of Epidemiology. Springer Verlag, Germany, pp. 1043–1109.
Olsen J, Küppers-Chinnow M and Spinelli A (1996) Seeking medical help for subfecundity: a study based upon surveys in five European countries. Fertil Steril 66,95–100.[ISI][Medline]
Olsen J, Juul S and Basso O (1998) Measuring time to pregnancy. Methodological issues to consider. Hum Reprod 13,1751–1753.
Rachootin P and Olsen J (1982) Prevalence and socioeconomic correlates of subfecundity and spontaneous abortion in Denmark. Int J Epidemiol 11,245–249.
Rachootin P and Olsen J (1983) The risk of infertility and delayed conception associated with exposure in the Danish workplace. J Occup Med 25,394–402.[ISI][Medline]
Schenker JG and Ezra Y (1994) Complications of assisted reproductive techniques. Fertil Steril 61,411–422.[ISI][Medline]
Sharpe RM and Skakkebaek NE (1993) Are oestrogens involved in falling sperm counts and disorders in the male reproductive tract? Lancet 341,1392–1395.[CrossRef][ISI][Medline]
Sills ES, Tucker MJ and Palermo GD (2000) Assisted reproductive technologies and monozygous twins: implications for future study and clinical practice. Twin Res 3,217–223.[CrossRef][Medline]
Silver RI, Rodriguez R, Chang TSK and Gearhart JP (1999) In vitro fertilization is associated with an increased risk of hypospadias. J Urol 161,1954–1957.[CrossRef][ISI][Medline]
Steirteghem AV, Devroey P and Liebaers I (2002) Intracytoplasmic sperm injection. Mol Cell Endocrinol 186,199–203.[CrossRef][ISI][Medline]
Storgaard L, Bonde JP, Ernst E, Spano M, Andersen CY, Frydenberg M and Olsen J (2003) Does smoking during pregnancy affect sons’ sperm counts? Epidemiology 14,278–286.[CrossRef][ISI][Medline]
Strom CM, Levin R, Strom S, Masciangelo C, Kuliev A and Verlinsky Y (2000) Neonatal outcome of preimplantation genetic diagnosis by polar body removal: the first 109 infants. Pediatrics 106,650–653.
Tielemans E, Burdorf A, te Velde E, Weber R, van Kooij R and Heederik D (2001) Sources of bias in studies among infertility clients. Am J Epidemiol 156,86–92.
Van Noord-Zaadstra BM, Looman CWN, Alsbach H, Habbema JDF, te Velde ER and Karbaat J (1991) Delaying childbearing: effect of age on fecundity and outcome of pregnancy. BMJ 302,1361–1365.[ISI][Medline]
Wang JX, Norman RJ and Wilcox AJ (2004) Incidence of spontaneous abortion among pregnancies produced by assisted reproductive technology. Hum Reprod 19,272–277.
Weinberg CR, Baird DD and Wilcox AJ (1994) Sources of bias in studies of time to pregnancy. Stat Med 13,671–681.[ISI][Medline]
Wilcox AJ, Weinberg CR, O’Connor JF, Baird DD, Schlatterer JP, Canfield RE, Armstrong EG and Nisula BC (1988) Incidence of early loss of pregnancy. N Engl J Med 319,189–194.[Abstract]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||