Defining infertility—a systematic review of prevalence studies

Abstract

BACKGROUND

Existing definitions of infertility lack uniformity, rendering comparisons in prevalence between countries or over time problematic. The absence of an agreed definition also compromises clinical management and undermines the impact of research findings. The aim of this study was to perform a systematic review of the literature to determine how infertility has been defined in prevalence studies and to come up with suggestions for a feasible and clinically relevant definition.

METHODS

MEDLINE, EMBASE, CINAHL and Cochrane Database of Systematic Reviews were searched for relevant population-based prevalence studies published between 1975 and 2010.

RESULTS

A total of 39 articles were included in the current review. The results highlight the heterogeneity of criteria used to define infertility and critical differences between demographic and epidemiological definitions. Demographers tend to define infertility as childlessness in a population of women of reproductive age, while the epidemiological definition is based on ‘trying for’ or ‘time to’ a pregnancy, generally in a population of women exposed to the risk of conception. There is considerable variation in terms of the duration of ‘trying for pregnancy’, the age of women sampled and their marital or cohabitation status. This leads to inconsistencies in determining the numerator and denominator used to calculate the prevalence of infertility.

CONCLUSIONS

There is a need for an agreed definition for infertility. We suggest a clinically relevant definition based on the duration of trying for pregnancy coupled with female age.

Introduction

Global estimates suggest that nearly 72.4 million couples experience fertility problems (Boivin et al., 2007). For couples and clinicians, a diagnosis of infertility signals the start of investigations and possible treatment. Agreement about the criteria used to make this diagnosis is also crucial to evaluating the effectiveness of alternative interventions for the condition. Accurate knowledge about the prevalence of infertility, awareness of secular trends and geographical differences are essential for providers of reproductive health care as well as policy-makers. Critical to establishing prevalence is a clear, unambiguous definition of the condition and an efficient instrument for making a diagnosis.

The Oxford English Dictionary defines ‘infertile’ as ‘not able to have babies or produce young’, which implies a state of sterility rather than ‘difficulty in conceiving’—which represents the view of many clinicians (http://www.oxfordadvancedlearnersdictionary.com/dictionary/infertility#infertile).

In defining infertility, the medical literature makes a distinction between a clinical/epidemiological versus a demographic approach (Table I). According to the International Committee for Monitoring Assisted Reproductive Technology and the World Health Organization, infertility is ‘a disease of the reproductive system defined by the failure to achieve a clinical pregnancy after 12 months or more of regular unprotected sexual intercourse’ (Zegers-Hochschild et al., 2009). Demographers, however, define infertility as the absence of a live birth in a sexually active non-contracepting woman (Larsen, 2005), and this approach is more in line with the public perception, as well as the dictionary definition, as couples seek live birth rather than conception. While most clinical definitions are predicated on lack of conception over 12–24 months, it is not uncommon for demographers to use a time horizon of 5 years. Lack of agreement on outcomes representing reproductive ‘success’ (e.g. pregnancy, ongoing pregnancy and live birth) has led to considerable variation in the estimated prevalence of infertility (Marchbanks et al., 1989; Schimdt and Munster, 1995; Larsen, 2005). While the need for an agreed definition has been highlighted in the past (Marchbanks et al., 1989; Schimdt and Munster, 1995; Habbema et al., 2004; Gnoth et al., 2005; Larsen, 2005), researchers and clinicians are yet to reach a consensus.

Infertility has been variably defined in studies aimed at determining its prevalence. Data in these studies have generally been self-reported and captured by means of survey questionnaires or life-event calendars. There is little consensus on a number of key factors, including (i) the appropriate method of survey; (ii) criteria for establishing the presence of infertility, i.e. length of time spent trying for a pregnancy, avoiding contraception, seeking medical help or formal diagnosis by a physician; (iii) the nature of the population of women sampled, i.e. appropriate age group, relationship status; (iv) the outcome measured i.e. pregnancy or live birth and (v) the outcome of interest, i.e. infertility/sterility/infecundity/subfecundity/childlessness.

Despite unease about the ambiguity of the term ‘infertility’, no previous systematic reviews on ways of defining this condition have been undertaken. Several new prevalence studies have been reported since the publication of a (non-systematic) review in 1995 (Schimdt and Munster, 1995). The objective of this study was to perform a systematic review of literature to determine the criteria used to define infertility in studies assessing the prevalence of this condition. In addition, we aimed to interpret our findings in the context of current clinical practice and suggest an optimal definition of infertility.

Methods

Meta-analysis of Observational Studies in Epidemiology guidelines for systematic reviews were followed (Stroup et al., 2000). (Checklist shown in Supplementary data Appendix.)

Search strategy and identification of studies

We aimed to identify population-based prevalence studies of infertility. Owing to the nature of the research question, we expected to identify observational studies; RCTs were not anticipated.

The following electronic databases were searched: The search strategy combined the following search terms: ‘infertility,’ ‘subfertility’, ‘subfecundity’, ‘sterility’, ‘infecundity’, ‘seeking medical help’, ‘physician diagnosis’, ‘childlessness’, ‘unresolved infertility’ and ‘prevalence’, ‘epidemiology’, ‘rate’, ‘ratio’, ‘definition’, ‘questionnaire stud*’, ‘survey*’, ‘epidemiological stud*’, ‘demographic stud*’, ‘prevalence stud*.’

1. MEDLINE (1975–2010),

2. EMBASE (1975–2010),

3. Cochrane library (1975–2010),

4. CINAHL (1975–2010).

Other resources

1. Websites of the World Health Organization and Population Council were also searched using the earlier-mentioned search terms.

2. Conference proceedings of the American Society of Reproductive Medicine (ASRM) and the European Society of Human Reproduction and Embryology were hand-searched from 1990 to 2010.

3. Bibliographies from relevant articles were hand-searched.

4. Personal communication with experts in the field.

Criteria for considering studies for this review

Inclusion criteria

Only published papers and reports based on population-based studies with a random selection of a representative sample surveyed by postal questionnaires or oral interviews were included. Where studies reported a number of reproductive health problems, only data relating to infertility were included.

Exclusion criteria

Clinic-based studies were excluded from this review. Although they offer an insight into the etiology of infertility, they are known not to be representative of the condition in the general population. The search was restricted to studies published in English as it was felt that nuances of language that may not be easy to translate could introduce an element of bias in defining infertility.

Outcome measures

The primary outcome was the set of criteria used to define infertility. Secondary outcomes included factors that determine variations in prevalence, including: nature of the population sampled, age group and denominator used to compute prevalence rates.

Selection of studies

The searches were independently carried out by S.G. and Z.P., who identified and reviewed suitable studies. Any disagreement was resolved by consensus and discussion with S.B. Contact with authors of primary studies was attempted where appropriate.

Data collection and analysis

Data extraction and management

Pre-designed data extraction forms were used to collect data. S.G. and Z.P. independently extracted the data that were validated by the second reviewer in cases of uncertainty.

Assessment of methodological quality of included studies

In the absence of standard criteria for the assessment of quality of prevalence studies, we developed an assessment tool based on the criteria proposed by Boyle (1998). The external validity of included studies was assessed by means of the adequacy of sampling methods, description of eligibility criteria, response rate and a priori calculation of sample size. A response rate of over 70% was considered to be satisfactory. Internal validity was evaluated on the basis of the nature of the measurement instrument and inclusion of confidence intervals around estimates.

Results

Description of studies

Our initial electronic search identified 29 657 studies. A search of the World Health Organization Website retrieved one relevant publication, and a hand search of cross references of relevant articles yielded a further 18 articles. Seventy were screened for eligibility. A total of 39 articles were finally included in the review (Fig. 1). Excluded studies (with reasons for their exclusion) are listed in Fig. 1. The methodological aspects of the included studies are summarized in Supplementary data, Table SI; 97.4% fulfilled at least three out of six quality assessment criteria, while 32/39 articles (82%) fulfilled four out of six criteria (Fig. 2).

Detailed descriptions of individual studies are included in Tables II and III. Eleven studies calculated the prevalence of infertility indirectly from survey or census data using a demographic approach (Ericksen and Brunette, 1996; Chandra and Stephen, 1998; Sundby et al., 1998; Larsen, 2000; Merlo and Rowland, 2000; Walraven et al., 2001; Larsen, 2003; Rutstein and Shah, 2004; Liu et al., 2005; Parr, 2005; Stephen and Chandra, 2006) (Table II). Two studies calculated ‘childlessness’ (Merlo and Rowland, 2000; Parr, 2005); six studies followed the demographic definition as mentioned in Table I (Ericksen and Brunette, 1996; Sundby et al., 1998; Larsen, 2000; Larsen, 2003; Rutstein and Shah, 2004; Liu et al., 2005) and three studies used a clinical definition to make their data comparable with epidemiological studies (Chandra and Stephen, 1998; Walraven et al., 2001; Stephen and Chandra, 2006). Considerable heterogeneity was evident in the definition used (duration of time to estimate infertility), age group of women included and denominator used to compute prevalence.

Twenty-eight regional epidemiological studies used the clinical definitions shown in Table I albeit with a degree of variability (Table III). Nineteen of these studies estimated current infertility (Rachootin and Olsen, 1982; Li, 1990; Adetoro and Ebomoyi, 1991; Bruckert, 1991; Hogberg et al., 1992; Webb and Holman, 1992; Gunnell and Ewings, 1994; Zargar et al., 1997; Philippov et al., 1998; Karmaus and Juul, 1999; Unisa, 1999; Che and Cleland, 2002; Geelhoed et al., 2002; Barden O'Fallon, 2005; Ahmadi Asr Badr et al., 2006; Kumar, 2007; Safarinejad, 2007; Bhattacharya et al., 2009; Herbert et al., 2009). Lifetime infertility was assessed in nine studies (Templeton et al., 1990; Schmidt and Munster, 1995; Sundby and Schei, 1996; Buckett and Bentick, 1997; Kuppers-Chinnow and Karmaus, 1997; Rostad et al., 2006; Oakley et al., 2008; Terava et al., 2008; Vahidi et al., 2009). The terms ‘infertility’, ‘subfecundity’, ‘subfertility’ and ‘childlessness’, ‘unresolved infertility’ were used almost synonymously and there was a significant disparity in terms of the duration of attempting pregnancy, age group of women included and denominator used to calculate prevalence rates.

Outcomes

Outcomes in demographic versus epidemiological studies have been described separately.

Demographic studies

Definition of infertility

Two studies estimated ‘lifetime childlessness’ (Merlo et al., 2000; Parr et al., 2005). One study calculated the prevalence of ‘impaired fecundity’, defined as ‘women of any marital status who are not surgically sterile with problems conceiving or carrying a pregnancy to term and unable to conceive after three years of unprotected intercourse’ (Chandra and Stephen, 1998). Of the six studies that followed the demographic definition, the duration used varied from 1 to 7 years.

The definition of infertility used by some studies included ‘absence of live birth despite being married for 5 or 7 years’ with no consideration of whether this was voluntary or involuntary (Larsen, 2000, 2003; Liu et al, 2005). Three studies used a similar definition with live birth as an outcome but did incorporate the presence or absence of any attempt to conceive (Ericksen and Brunette, 1996; Sundby et al., 1998; Rutstein and Shah, 2004). We identified three studies that used a definition similar to the ‘clinical’ definition ‘absence of pregnancy despite attempting conception for 1–3 years’ (Chandra and Stephen, 1998; Walraven et al., 2001; Stephen and Chandra, 2006).

Age group

The two studies estimating lifetime childlessness sampled women aged 40–49 years and 40–54 years. All other studies had a lower age limit of 15–20 years, whereas the upper age limit varied from 44, through 49, 54 up to 57 years.

Denominators used to compute rates of prevalence

Four studies were from developed countries and used ‘all women’ (Chandra and Stephen, 1998; Parr, 2005); ‘all married women’ (Stephen and Chandra, 2006) and ‘ever married women’ (Merlo and Rowland, 2000) as denominators. In the developing world, where universality of marriage is not uncommon, denominators in seven studies included ‘all women’ (Sundby et al., 1998; Rutstein and Shah, 2004), ‘married women’ (Walraven et al., 2001), ‘ever married women’ (Larsen, 2000), ‘married for 5 years’ (Larsen, 2003), ‘married for 7 years’ (Liu et al., 2005) and ‘couples exposed to conception for at least 5 years’ (Ericksen and Brunette, 1996).

Epidemiological studies

Definition

As Table III indicates, the main outcome in epidemiological studies is the absence of pregnancy, while descriptive terms used include ‘infertility’, ‘subfertility’, ‘subfecundity’, ‘childlessness’ and ‘unresolved infertility’. Attempting conception by means of regular unprotected intercourse has been variously defined as either ‘sexual activity without contraception’, ‘unsuccessful attempted conception’, ‘regular unprotected intercourse’, ‘trying for pregnancy’, ‘difficulty in conception’, and  ‘cohabitation and exposure to pregnancy’. Alternatively ‘time to pregnancy’ has been used to identify women with infertility.

Age group of sample

We found considerable disparity in the age group of women sampled; four studies collected data from women aged 15–49 years (Adetoro and Ebomoyi, 1991; Geelhoed et al., 2002; Ahmadi Asr Badr et al., 2006; Safarinejad, 2007) and three studies included women aged 16–44 years (Webb and Holman, 1992; Schmidt and Munster, 1995; Zargar et al., 1997), whereas the remaining studies included women aged between 18 and 69 years.

Denominator

Five studies used ‘women attempting conception’ as a denominator (Rachootin and Olsen, 1982; Schmidt and Munster, 1995; Oakley et al., 2008; Bhattacharya et al., 2009; Herbert et al., 2009). ‘Currently married women’ was used in seven studies (Li, 1990; Bruckert, 1991; Zargar et al., 1997; Unisa, 1999; Che and Cleland, 2002; Ahmadi Asr Badr et al., 2006; Kumar, 2007). Two studies (Safarinejad, 2007; Vahidi et al., 2009) had ‘ever married’ as a denominator in their respective studies, while one study (Rostad et al., 2006) opted for 'menopausal women’.

The remaining 13 studies included ‘all women’ in the denominator without any special consideration of their marital/cohabitation status or opportunity/wish to conceive.

Duration of trying for pregnancy

Duration of exposure to risk of pregnancy was reported inconsistently. Twelve studies used a duration of 12 months or more in order to define infertility, while five used a 24 month cut-off (Table III). Five studies reported multiple values for infertility prevalence based on alternative periods of time such as 12, 24 and 60 months. We found one study that had used 18 months as the defining duration (Adetoro and Ebomoyi, 1991).

Discussion

Principal findings

Our results highlight the fact that the absence of an established definition of infertility results in varying estimates of its prevalence within and between populations. There is considerable heterogeneity in terms of determining exposure and outcomes, the nature of the population sampled and the denominator used to calculate prevalence rates. Terms such as ‘infertility’, ‘subfertility’ and ‘subfecundity’ are used without being clearly defined, as are the terms ‘primary permanent infertility’, ‘unresolved infertility’, ‘childlessness’ and ‘primary unresolved infertility’. This makes it impossible to pool published data or make comparisons over time and across geographical regions. Thus, whether the wide variation in published rates of prevalence is real or spurious remains unclear.

Strengths and weaknesses

This, to our knowledge, is the first systematic review on this subject. An extensive literature search was conducted, minimizing the risk of missing information. We complied with existing guidelines for the reporting of systematic reviews and meta-analyses of observational data and were able to produce a comprehensive description of the characteristics of individual studies. Inconsistencies in terms of design, population and methods have made it difficult to interpret the results of individual studies or attempt any formal aggregation of data. Restricting our focus to English language papers may have resulted in exclusion of some relevant studies—although problems with translation of certain key words may have introduced further bias in terms of establishing definitions of infertility in the non-English literature. Despite highlighting the lack of agreement in defining infertility, the ideal definition still remains a debatable issue.

Quality of included studies

Formal evaluation of the quality of included studies was challenged by variations in design and the tools used to determine exposure and outcomes. We found few instruments specifically designed for prevalence studies and noted the limitations of standard scales used for assessing the quality of observational studies in this context. Most included studies met quality standards for half the accepted criteria, such as adequacy of sampling methods, description of eligibility criteria, response rate and a priori calculation of sample size, but very few fulfilled all of them.

Interpretation in the context of published literature

Infertility is a state ranging from near normal fecundity to an absolute inability to conceive. This breadth makes it debatable whether a single term can meaningfully cover the entire spectrum of the condition.

Alternative definitions of infertility can impact on estimated prevalence (Marchbanks et al., 1989; Larsen, 2005), which can vary between 6.8 and 38.6% in the same population depending on the criteria used. ‘Physician diagnosis’ yielded the lowest prevalence estimate, while an increased duration of 'trying’ from 12 months to 24 months reduced the prevalence by 12%. Larsen (2005) explored six alternative definitions of infertility—three based on self-reported data and three derived from birth history and time since last birth or marriage. The prevalence of infertility was significantly higher when defined by ‘unprotected intercourse for at least 2 years’ compared with ‘tried to conceive for at least 2 years’ suggesting that a sizeable number of women distinguish between having unprotected intercourse and ‘trying for a baby’. Larsen also stressed the importance of including ‘trying for pregnancy’ as the primary survey question in epidemiologic and demographic surveys. This distinction may also be reflected in the high proportion (33%) of women attending an antenatal clinic with either an unintended pregnancy or one about which they were ambivalent (Lakha and Glasier, 2006).

Data derived from demographic surveys or census data have their limitations: ‘absence of a live birth within a defined period of time’ seems not dissimilar to ‘primary infertility’ or ‘childlessness’ as described in epidemiological studies but is not clear whether the absence of live birth is voluntary or involuntary. An added drawback in the demographic definition is the lack of a specified time period, varying from 1 to 7 years in the studies assessed (Table II). Demographic surveys do, however, have the advantage of involving large population samples and repeated surveys allow comparisons across time and regions. The comparability of results from epidemiological and demographic studies has not been adequately researched. Dick et al. (2003) compared women's self-reported difficulty in conceiving against calendar derived and self-reported time taken to conceive and found that self-reported difficulty is useful in quantifying fertility problems. It was observed that only 66% of all women who were considered infertile using the calendar-derived time taken to conceive (Calender TTTC) reported a difficulty in conceiving. Alternatively, of the women who were reportedly fertile (according to Calender TTTC), 5% reported a difficulty in conceiving (Dick et al., 2003).

Interpretation of results in the literature needs to accommodate differences in the characteristics of the women sampled in different studies. The disadvantages of including adolescents or women >45 years of age have been highlighted (WHO, 1991). Inclusion of girls at the beginning of their reproductive lives can lead to an overestimate of infertility owing to the physiological barriers in conceiving (Vahidi et al., 2009). Recall bias may be an issue of concern in women >45 years of age, while the number of births that occur after the age of 40 years is increasing.

Marital and cohabitation status can also affect prevalence rates. Choice of marital status must be in concordance with the social structure of the region under study. It is helpful to be able to discriminate between voluntary and involuntary childlessness. Survey questions must identify couples at risk of pregnancy, and a duration of 12 months would be a more acceptable cut off for older women.

Clinical implications

The clinical utility of diagnosing infertility is to identify a group of couples who need further investigation and possible treatment. Clinicians need a definition sensitive enough to identify all couples who would benefit from investigation but specific enough to avoid over-investigation and premature treatment. It has been long recognized that key factors in determining the need for and urgency of active treatment include female age and duration (Collins et al., 1995). Cumulative probability of conception is known to decline with age (O'Connor et al., 1998; Dunson et al., 2002; Gnoth et al., 2005). The probability of conception wanes after the mid-30s and reduced fecundity with age is thought to be the outcome of reduced fecundability and fetal loss (O'Connor et al., 1998). Thus different age cohorts will have dissimilar probabilities of conception and fecundability. Thus, the duration of trying may need to be less in older women in order to provide an accurate prognosis that can inform the need for active management: conversely significant diagnosable pathology is more likely in a younger couple with the same duration of trying. It is therefore debatable whether a common duration of time of trying for pregnancy to meet an arbitrary clinical (distinct from epidemiological) definition of ‘infertility’ is applicable across all age groups, or whether an age-specific definition might be more appropriate.

Research implications

There is an urgent need for a consensus on criteria for defining infertility and a valid instrument to identify the condition. Further work needs to be carried out to refine the current policy based on duration of exposure in order to accommodate the diminishing risk of pregnancy in older women. There is, however, an age-related conflict between clinical and epidemiological priorities underlying this aspect of the condition.

Existing problems with current definitions of infertility and the way forward

Demographic and epidemiological definitions

The inherent inconsistency between demographic and epidemiological definitions of infertility is unlikely to be resolved, and both are relevant in their own contexts. The demographic definition provides important data on reproductive outcomes in large populations in exercises such as a national census, allows comparisons across geographical regions and alerts governments about critical changes in replacement fertility that can have major socio-political implications. The demographic approach is, however, unsuitable for clinical purposes as it fails to identify couples who need investigation and treatment in order to improve their chances of conception.

Current terminology used for epidemiological definitions

Infertility represents a spectrum ranging from near normal fertility to an absolute inability to conceive. Although it would appear useful to have specific terms that capture the mild as well as the severe end of the spectrum, this may not be practical. Habbema et al. (2004) have suggested a clinically meaningful definition of infertility based on three separate parameters: length of trying, diagnosis and prognosis. However, many prognostic indicators cannot be identified without investigation, and one objective of an accepted definition is to rationalize the use of investigations and interventions. The overriding importance of female age should however be recognized. Terms such as ‘infertility’, ‘subfertility’ and ‘subfecundity’, which are used interchangeably, could be replaced by a single universally acceptable descriptor.

Diagnosis on the basis of time to pregnancy

Time to pregnancy is commonly used to define infertility, but this approach has a number of limitations—the most important of which is its exclusion of women with unresolved infertility. Retrospective studies can be associated with significant recall bias, while prospective studies tend to include couples who are actively trying for pregnancy as opposed to those who are at risk. Defining ‘at risk’ couples can be problematic as it could lead to the exclusion of fertile couples with unintended pregnancies. Factors such as age, semen parameters and frequency of sexual intercourse can also affect time to pregnancy especially in retrospective studies on antenatal populations. Alternatives include the ‘current duration’ approach (Slama et al., 2006) where the population samples include couples who have yet to conceive but can state their duration of exposure to pregnancy.

The outcome of choice

In any definition of infertility, the outcome of choice should be ‘pregnancy’ rather than ‘live birth’. This is because the reasons for inability to carry a pregnancy to term are different from those that influence the ability to conceive, although it is important to recognize that it is a baby that couples seek rather than pregnancy.

Denominator

For reasons mentioned earlier, the population sampled should include women in the reproductive age group (e.g. 20–45 years) who are currently attempting (or ever attempted) conception.

Suggestion for a clinically meaningful definition of infertility

We would like to suggest a simpler, more pragmatic definition of infertility based on length of trying (or exposure to pregnancy) adjusted for female age (Fig. 3). Existing prognostic models (Eimers et al., 1994; Collins et al., 1995; Hunault et al., 2005) have underlined the importance of female age and duration as two key factors, easily elicited from a couple's history, which are strongly predictive of spontaneous conception. Thus older women should be regarded as infertile earlier, prompting appropriate investigation and treatment. This approach is already widely, if informally, used in clinical practice, with less advocacy of expectant management where the female partner is older (NICE, 2004). Figure 3 represents a tool that can be used to define infertility in couples, based on the duration of exposure to pregnancy (‘time of trying’) and age of the female partner. The curved action line is for representative purposes only, with its defining characteristics (slope, point of inflection) requiring data from future research.

Conclusions

This systematic review highlights the existing lack of, and need for, a consensus definition of infertility. It reinforces the case for a reliable and valid survey instrument for measuring infertility and emphasizes the impact of factors, such as age group of women sampled and their cohabitation/marital status, on prevalence rates. While a demographic definition may have its uses in the context of identifying fertility trends in a population, it is not fit for purpose in a clinical setting. In order to identify couples who would benefit from fertility investigations and treatment, we suggest that the definition of infertility should be based on an estimated chance of spontaneous conception that is driven by duration of trying for a pregnancy and the female partner's age.

Authors' roles

S.G. contributed to the conception and design, literature search, drafting the paper, analysis and interpretation of data; Z.P. was involved in literature search, interpretation of data, editing of draft paper and approval of final version; R.A.A. played a role in interpretation of results, revision of draft paper and approval of final version; S.B. initiated the project, contributed to conception and design, revised the manuscript critically for important intellectual content was involved in final approval of the version to be published.

References