Researchers studying sex differences in pain are strongly advised to consult a recent, comprehensive set of guidelines entitled “Strategies and Methods for Research on Sex Differences in Brain and Behavior” [17
]. In addition, methodological issues that are specific to pain research are discussed below.
It is generally agreed that the first stage of any sex difference study should be a comparison of gonadally intact adult females and males. In the absence of previous evidence for large menstrual/estrous cycle-related variations in the measure of interest, it is not absolutely necessary to test females in specific stages. However, a failure to observe sex differences in pain can be interpreted in multiple ways: (1) no sex difference exists; (2) the observed sex difference occurs only when females are in a particular stage of the menstrual or estrous cycle (e.g., [193
]), and the sex difference could not be discerned due to “averaging” across the cycle; (3) a sex difference in mechanism
exists even though the magnitude of the phenomenon under study is similar in males and females (e.g., [126
]); or (4) two sex-specific mechanisms exist that cancel each other out, resulting in no phenotypic sex difference [50
]. Thus, in the case of a negative finding when comparing gonadally intact females vs. males, interpretations 2, 3, and 4 should be acknowledged.
When designing and interpreting studies on sex differences in pain or nociception, it is important to recognize that there are sex differences in a number of physiological systems that may directly or indirectly affect measures of pain/analgesia. For example, adult male rodents have greater percentage of body fat than females [124
], while the opposite is true for humans [78
]. This sexual dimorphism can affect the distribution of highly lipophilic drugs, and therefore influence analgesic drug potency, efficacy, and duration of action. Other aspects of pharmacokinetics, including liver metabolism and membrane transport, may differ between the sexes [139
], possibly affecting analgesic potency, efficacy, and duration of action. Immune responses also differ between the sexes [21
], which may contribute to sex differences in response to chronic inflammatory and neuropathic pain. The activity level of female rodents varies dramatically across the estrous cycle [24
], which may introduce a confound on pain tests that allow subjects to locomote freely, such as the hot-plate test. Lastly, adverse effects of analgesics such as respiratory depression and nausea may occur differentially in females vs. males [32
], which is important to consider when comparing analgesic efficacy in females vs. males.
3.1. Are sex differences mediated by gonadal hormones?
As described by Becker et al. [17
], if a sex difference is observed in gonadally intact adults, a logical next step is to determine whether the sex difference can be attributed to the actions of gonadal steroid hormones, either activationally (in adulthood) or organizationally (during development). Organizational effects can only be practically examined in animal studies. Several recent reviews have described the multiple ways in which estrogens, progestins, and androgens may modulate nervous system function related to pain and analgesia [5
3.2. Testing across the estrous cycle: animal studies
The value of testing female rodents at different stages of the estrous cycle is debatable. In rodents, an influence of estrous cycle stage is not necessarily indicated by larger observed variance [142
]; therefore, large samples are typically required to adequately power studies examining females in specific stages of the estrous cycle. One of the most reliable and simple ways to determine stage of the estrous cycle in rodents, sampling vaginal cytology, requires that females be handled and probed daily for at least two cycles (approximately 8–10 days). Given that females and males often respond differently to acute stressors (e.g., [214
]), it is not clear whether daily handling of males adequately controls for this potential confound. Obtaining repeated vaginal samples in rodents also may affect sensitivity to drugs [209
]. An alternative to daily vaginal lavage is to measure motor activity (in the home cage), which peaks during proestrus [24
]. Because rodent estrous cycles differ substantially from primate menstrual cycles (), one cannot readily extrapolate a cycle effect in a rodent to one in a human. Therefore, although estrous cycle fluctuations in a pain/analgesia measure suggest that reproductive hormones modulate the effect of interest, this hypothesis is more directly and efficiently tested using a hormone depletion/replacement approach (see Section 3.4), which is readily accomplished in rodents. If females in different stages of the estrous cycle are tested, it is imperative that the investigator explicitly state how the stages – particularly the stage of most dramatic hormone change, proestrus – are defined, as there is more than one way to designate stage of cycle. One study has shown significant differences in analgesic sensitivity between females in “early” vs. “late” proestrus [19
Fig. 1 Patterns of estradiol, progesterone, and leuteinizing hormone (LH) in humans (a) and rats (b) during the reproductive cycle. Time unit of the x-axis in (a) is days; in (b), it is hours. Dark bars in (b) indicate dark period of the day/night cycle. Note (more ...)
3.3. Testing across the menstrual cycle and reproductive stages: human studies
In humans, while it may not always be important to test at different stages of the menstrual cycle, it is always important to consider whether such testing is appropriate. If menstrual cycle itself is not a factor to be evaluated, the investigator should plan to evaluate women in the same phase of their cycle. There is no one ideal phase of the cycle to choose. Both absolute and relative hormonal levels could influence pain. Times of rapid change in hormone levels (e.g., ovulation) or phases when subjective mood changes are thought to occur (e.g., pre-menstrual) may be of interest.
Start date of the last menstrual period can be reliably obtained from self-report, as can year and month of menarche [77
]. A number of methods, including analysis of daily vaginal secretions, basal body temperature, and ovulation kits can provide information on the occurrence and approximate timing of ovulation, although some methods are more reliable than others [17
] (p. 1663). Prediction of ovulation using urine-based home ovulation testing kits to assess leuteinizing hormone surge is a minimum standard in studies in which the menstrual cycle is of interest. These kits are relatively inexpensive, are easy for subjects to use, and have high sensitivity and specificity for detecting ovulation [138
]. Additional measurement of blood or salivary levels of mid-luteal progesterone can confirm ovulation [183
]. Although serum-based measurements are still the most widely used and considered the gold-standard for clinical research, saliva-based methods can be an alternative. Saliva-derived hormone measures reflect the fraction of the hormone that exerts biological effects [158
]. Furthermore, collection of saliva is non-invasive and pain-free; with appropriate instruction, the subject can collect saliva at home and store it in a home freezer without deterioration of the sample [66
]. A serious drawback is that salivary hormone levels may be below detection levels. Additionally, standardization across laboratories can be a problem. It should be noted that both serum and salivary measures of estradiol and progesterone exhibit great between-subject variability within the range of normal. However, because menstrual cycle length varies between and within women [198
], and hormone levels also vary from one day to the next in some phases of the cycle, if hormonal status is a critical variable in the research, it is best measured directly rather than inferred from self-report and ovulation measures of menstrual cycle phase.
The standard designations for stages of the menstrual cycle (menstrual, follicular, ovulatory, and luteal) are gynecological terms based on reproductive function. However, a woman’s actual hormone levels within a phase vary radically. Thus, researchers must consider whether their interest is in reproductive function or in the relationship of hormone levels to pain. If the interest is in hormone levels, the gynecological nomenclature (and the lack of standardization with which it is applied) represents an obstacle to progress in the field [165
]. We propose that use of the terms menstrual, follicular, ovulatory, and luteal be discouraged in human clinical pain studies, unless coupled with report of the actual days of the menstrual cycle, standardized to a 28-day cycle based on ovulation testing. One statistical standardization method is described by LeResche et al. [122
]; however, other reasonable standardization methods are possible, and reaching consensus on the issue of how to standardize menstrual cycles of various lengths would facilitate comparison across studies. For these reasons, hormone levels should be reported when available.
If the research question is specifically cycle
-related, as opposed to hormone-related, subjects with irregular cycles should be excluded. However, it is difficult to exclude irregularly cycling subjects based only on self-report; many subjects who describe their cycles as regular in fact have irregular cycles [198
]. Even monitoring cycles before study enrollment may not solve this problem because there are no agreed criteria for distinguishing between a regular and irregular cycle. If the research question concerns the relationship between pain and hormone levels (or hormone variability) and hormones are measured directly, including irregularly cycling subjects may not be a problem, provided sufficient hormonal variability occurs. However, this approach assumes that the neurohormonal processes related to pain are comparable for women with regular and irregular cycles, which may not be the case. Finally, it is important to record dysmenorrheal status, as dysmenorrheic women may differ from those without dysmenorrhea in their responses to (non-menstrual) pain stimuli, especially during the perimenstrual period (e.g., [15
Circadian rhythms have been documented for steroid hormone levels [26
], autonomic nervous system activity [27
], and drug absorption [36
]. These rhythms may be altered by menstrual cycle rhythms in women. For example, at the time of ovulation, decreased absorption of drugs such as aspirin and alcohol occurs and intestinal transit times are longer in the late luteal phase, during pregnancy and with hormone supplementation; these effects can influence drug onset time. With a few exceptions (e.g., [134
]), the effect of circadian rhythms on pain and analgesia is largely unexplored.
Inclusion of non-cycling subjects may be of interest at specific life stages (pre-puberty, pregnancy, menopause). The conventional standard is to define puberty according to Tanner stages [196
], based on clinical examination of development of secondary sex characteristics. Self-report measures with sufficient reliability and validity are available for use in studies where direct examination is not feasible [25
]. A woman is conventionally defined as “post-menopausal” one year after the last menstrual period, provided the cause of amenorrhea is not pregnancy, nursing, disease, or medical intervention [217
]. Date of menopause is the date of the last menstrual period. Standardized criteria for stages of the menopausal transition are also available [83
]. If pregnant women are studied, stage of pregnancy should be clearly noted, as pain responses are known to change dramatically over the course of pregnancy (e.g., [16
]). Reproductive history also may be important to document. For example, pain during breastfeeding in the first week postpartum is directly predicted by parity, with women who have given birth to more children experiencing more pain [87
]. Conversely, in the laboratory (and not in close proximity to birth), multiparous women have higher pain thresholds than nulliparous women [82
3.4. Hormone manipulations: animal studies
A useful approach for identifying effects of particular hormones on pain/analgesia in animals is gonadectomy with or without hormone replacement. However, it is important to keep in mind that hormone depletion via gonadectomy alters the physiological status of the animal in two significant ways. First, surgery can affect pain thresholds [93
] and sensitivity to analgesics. Second, gonadectomy disrupts the normal feedback loop that sex steroids exert on the anterior pituitary and hypothalamus, leaving both males and females in a prolonged state of elevated (e.g., gonadotropin-releasing hormone (GnRH), leuteinizing hormone) or depressed (e.g., prolactin) circulating hormones [1
]. The likelihood of unintended consequences due to altered hypothalamic/pituitary hormone secretion can be addressed by administering low (maintenance) doses of estradiol or testosterone rather than using gonadectomized animals with no hormone replacement. In the case of the female rat, daily s.c. injections of 1 μg estradiol benzoate are sufficient to reduce leuteinizing hormone and to elevate prolactin to diestrous levels without inducing receptive behavior [1
Although there are “chemical castration” alternatives to gonadectomy, these have drawbacks and have not been fully characterized in terms of their possible effects on pain/analgesia. For example, continuous GnRH can be used to shut down the hypothalamo–pituitary–gonadal axis in both males and females. However, elevated GnRH may interfere with some aspects of opioid analgesia [160
] and may cause long-term enhancement of glutamatergic post-synaptic activity [218
]. Inhibitors of aromatase, the enzyme necessary for the synthesis of estrogens from androgens, can be used to substantially reduce synthesis of estrogens, and 5-α-reductase inhibitors can be used to reduce dihydrotestosterone synthesis; however, these inhibitors may also reduce synthesis of other steroid hormones within these metabolic pathways. The androgen antagonist flutamide can be used to block the effects of androgens, but it can also affect aromatase activity that is induced by androgens [23
] and can act directly as an androgen agonist in some tissues [131
]. The estrogen receptor antagonist tamoxifen, like other selective estrogen receptor modulators, has been shown to act as a partial agonist in some peripheral tissues [92
]. ICI 182780 is a selective estrogen receptor antagonist, but it does not cross the blood–brain barrier [208
]. Ultimately, convergent approaches to manipulate gonadal hormones will provide the most definitive testing of hypotheses that gonadal hormones modulate pain/analgesia.
When using a gonadectomy/hormone replacement approach, several issues must be considered. First, an effect of gonadectomy (relative to gonadally intact controls) implicates a role for gonadal secretions in pain/analgesia, but does not implicate any specific hormone. For example, one cannot conclude that testosterone mediates a given outcome based only on a castration effect in a male. Rather, studies in which testosterone is replaced in castrated males are also necessary. Second, to study the origins of sex differences, it is important to make the sex hormone levels as similar as possible in males and females at the time of testing. That is, both sexes rather than just males should be tested with androgens, and both sexes rather than just females should be tested with estrogens. This approach enables one to assess whether the mechanisms in question are sexually differentiated. If both sexes respond similarly to the same hormone treatment – even if one sex does not normally experience that hormone state – it can be concluded that the mechanism is likely to be similar in the two sexes (that is, not sexually differentiated).
A third point regarding gonadectomy/hormone replacement approaches is that failure to obtain an ovariectomy effect in females does not necessarily mean that ovarian hormones do not modulate pain/analgesia, particularly if the comparison group of intact females is in unknown or low hormone stages (i.e., an ovariectomy effect in females may only be apparent if the intact control group is in proestrus or estrus, when gonadal hormones are at or near peak levels). Fourth, the timing of hormone administration may be critical: if hormone replacement begins weeks rather than days (or immediately) after gonadectomy, responsiveness to exogenous hormone administration may diminish, requiring higher doses or longer duration treatment to obtain comparable effects [38
]. Moreover, for estradiol, the interval between the last injection and behavioral testing as well as the duration of hormone exposure may dramatically influence the outcome [42
Another aspect of hormonal manipulations concerns the age at which they are done. In rodents, the effects of ovariectomy and estradiol replacement depend upon the age at which the surgery is done [33
]. Furthermore, there are important species differences in the pattern of hormonal changes during the progression through reproductive senescence (sometimes called “estropause” [33
]). Unlike mice and women, in whom both estradiol and progesterone levels fall to very low levels following reproductive senescence, when rats progress through estropause, their estradiol levels remain elevated [203
In the absence of relevant data to guide the design of a hormone replacement protocol, administration of a dose that has been shown to maintain or reinstate sexual behavior is a reasonable starting point. However, the dose ranges of hormone replacement relevant to the reproductive system may not be equivalent to ranges that modulate pain/analgesia. Even within the reproductive system, tissue sensitivities to estradiol vary widely [99
]. Thus, reasonable starting
doses for estradiol in ovariectomized rats would be in the range of 2 μg estradiol benzoate s.c., with testing beginning 24–48 h later; this dose/test interval has been shown to mimic proestrus–estrus, both behaviorally and in terms of plasma hormone levels [13
]. Likewise, 500 μg of s.c. progesterone is a reasonable starting point for progestin replacement, with testing beginning 4–6 h later [58
]. Of course, these testing intervals assume a nuclear steroid receptor-mediated (genomic) effect, whereas recent reports indicate that gonadal steroids may modulate pain thresholds at considerably shorter intervals [57
]. Thus, it is important to consider shorter treatment-test intervals as well. For in vitro preparations, 100–1000 × Kd
may be a reasonable starting dose to saturate a receptor [100
], followed by decreasing the dose until the effect of interest disappears.
No obvious rationale exists for examining the influence of progesterone alone in a gonadectomized animal (other than as a control for progesterone + other hormone combinations), because gonadectomized humans are not exposed to progesterone alone under any known clinical treatment condition. However, it would be informative to test progestins and estrogens alone and in combination in gonadally intact
animals, to model various clinical hormone treatments in humans (e.g., [48
3.5. Hormone manipulations: human studies
In human studies it is often impractical or unethical to manipulate hormones. However, women taking contraceptives for birth control or other reasons (and men taking androgens) can be studied. Additionally, short-term hormone administration is feasible (e.g., [187
]). Hormone replacement and supplementation therapies differ markedly in the particular androgens, estrogens, and progestins they contain; these should always
be clearly specified, as the various hormones likely exert different effects.
3.6. What types of pain tests are appropriate? Animal studies
The model system that is most appropriate is entirely hypothesis-driven. In animals, although acute pain tests such as the tail-flick test may not model clinical human pain per se, many advances have been made in understanding mechanisms of pain using such models. Face validity is less critical than using the model that is most appropriate to studying the specific mechanism of interest. In other words, mechanism “translates” better than phenomenology. A variety of approaches (e.g., visceral vs. cutaneous pain, acute vs. chronic [inflammatory or neuropathic] pain, behavioral vs. electrophysiological vs. molecular measures) are all worth pursuing, because they may reveal sex differences that have not yet been observed. Many existing pain models have not yet been applied to females, and animal models for female-specific pain syndromes are needed. In any model, the intensity of the noxious stimulus should be carefully considered. High-intensity stimuli are likely to result in a ceiling effect in which all subjects (or cells, etc.) respond at the limit of their capacity, thereby potentially obscuring individual differences (e.g., sex differences). Examination of more than one stimulus intensity is valuable in determining the generalizability of the sex difference.
As noted in reviews, sex differences in acute pain models have been observed, but are often protocol-, species-, and strain-dependent [140
]. Very few studies have investigated sex differences in chronic pain models such as nerve injury and persistent inflammation. Notable exceptions include nerve ligation [40
] and arthritis studies [4
]. This dearth of information makes it premature to suggest which protocols or animal species are most relevant to any particular human pain condition. One valuable approach involves parallel testing in the animal and human laboratory (e.g., [29
3.7. What types of pain tests are appropriate? Human studies
In contrast to the robust sex difference found in the prevalence and severity of many chronic pain conditions, reported sex differences in experimental pain responsiveness, while generally consistent in direction, are often subtle in magnitude and sometimes absent. Sex differences may be more consistently observed when using particular stimulus modalities or psychophysical protocols. Specifically, a meta-analysis reported that sex differences in threshold and tolerance measures were largest and most consistently found for pressure pain and electrical stimulation, while smallest and least consistently found for thermal pain stimuli [164
With respect to visceral pain, a lower pain threshold to esophageal distension was demonstrated in females [149
], whereas rectal stimulation studies showed no gender differences in mechanical thresholds [11
], or higher thresholds in healthy women [34
]. Additionally, women have shown larger referred pain areas to the mechanical and thermal stimulation of the esophagus [161
]. Characterizing sex differences in visceral pain, even with the limited number of studies published, is complicated. As with sex differences in somatic pain, detailed evaluations of visceral pain will be necessary to determine the mechanisms responsible.
Studies should include a range of stimulus intensities, as assessments of threshold or only mildly painful stimuli could fail to reveal differences that are manifest with more intensely noxious stimuli. Similarly, different physiological and psychological mechanisms operate in acute vs. sustained pain, such that sex/gender differences may be more prominent in tests of sustained pain.
Studies of experimentally induced hyperalgesia have demonstrated robust sex differences [29
], but not all aspects of pain show significant sex differences in these models. Such observations provide the opportunity to selectively evaluate different mechanisms of hyperalgesia, and identify those in which sex is a significant factor.
When testing pain sensitivity of persons with clinical pain, it may be informative to use quantitative tests that are closely related to and replicate the clinical pain, as well as stimuli that differ from it in location and/or quality. The decision on these options should be based on the questions posed. For instance, evaluation of pain sensitivity at asymptomatic body sites can reveal information about general pain-processing alterations, as has been shown for fibromyalgia [118
], various types of headaches [182
], and temporomandibular disorders (TMD) [175
]. Additionally, experimental provocation of a patient’s clinical pain allows separate assessment of that pathological condition, for instance evaluating pressure pain sensitivity of symptomatic muscles of fibromyalgia patients [188
], or rectosigmoid distension of irritable bowel syndrome patients [34
It is important to document as much of the experimental testing situation as possible. For example, experimenter (or clinician) sex should always be reported, as this variable may influence pain report in the laboratory (e.g., [14
]) and in the clinic [133
]. Other relevant factors to document include: (1) a detailed description of instructions to subjects and the training they received prior to data collection; (2) the history of subjects’ pain experiences; and (3) the history of subjects’ experience with similar experimental studies.
The goal of providing such information is to characterize the context of the experimental testing situation, and importantly, the subjects’ reaction to the context. While it is relatively simple to document the environmental features of the testing situation, it is more complicated – but likely more relevant – to document the subjects’ perception of and reaction to the testing environment. One example is the anxiety or stress that the subject feels in the testing environment. These states can vary substantially among subjects, and there is considerable evidence that stress or anxiety level influences pain perception, often in a sex-dependent manner [91
]. Accordingly, it is valuable to document both the environmental conditions and the subjective state of the participants at the time of evaluation. Decidedly relevant factors are identified in under “Context”.
Factors decidedly or likely relevant to sex differences in experimental studies of human pain
Many variables other than a person’s sex or gender account for individual differences in pain sensitivity. Thus, sex differences may be more prominent in one sample than another due to differences in other sample characteristics. It is important to document the characteristics of the study population as thoroughly as possible; a list of relevant characteristics is presented in under “Individual Factors.” Many of these factors are also relevant to clinical pain research.