Morphological and postural sexual dimorphism of the lumbar spine facilitates greater lordosis in females

Abstract

Previous work suggests females are evolutionarily adapted to have greater lumbar lordosis than males to aid in pregnancy load‐bearing, but no consensus exists. To explore further sex‐differences in the lumbar spine, and to understand contradictions in the literature, we conducted a cross‐sectional retrospective study of sex‐differences in lumbar spine morphology and sacral orientation. In addition, our sample includes data for separate standing and supine samples of males and females to examine potential sex‐differences in postural loading on lumbosacral morphology. We measured sagittal lumbosacral morphology on 200 radiographs. Measurements include: lumbar angle (L1–S1), lumbar vertebral body and disc wedging angles, sacral slope and pelvic incidence. Lumbar angle, representative of lordotic curvature between L1 and S1, was 7.3° greater in females than males, when standing. There were no significant sex‐differences in lumbar angle when supine. This difference in standing lumbar angle can be explained by greater lordotic wedging of the lumbar vertebrae (L1–L5) in females. Additionally, sacral slope was greater in females than males, when standing. There were no significant sex‐differences in pelvic incidence. Our results support that females have greater lumbar lordosis than males when standing, but not when supine – suggesting a potentially greater range of motion in the female spine. Furthermore, sex‐differences in the lumbar spine appear to be supported by postural differences in sacral‐orientation and morphological differences in the vertebral body wedging. A better understanding of sex‐differences in lumbosacral morphology may explain sex‐differences in spinal conditions, as well as promote necessary sex‐specific treatments.

Introduction

Lumbar lordosis (LL) refers to the curvature of the lumbar spine, expressed in humans as a response to bipedalism. Lordotic curvature is critical for biomechanical stability in the lumbar spine (Patwardhan et al. 1999), as it contributes to both the load‐bearing capacity and flexibility of the lumbar spine, which are important for activities of daily living (Sparrey et al. 2014). The lordotic curvature of the lumbar spine resembles a buckled (Euler) column in response to load‐bearing (Meakin et al. 1996) and promotes load‐sharing between the disc and facet joints (Sparrey et al. 2014).

Evolutionarily, LL is an adaptation aiding upright posture and bipedal locomotion (Lovejoy, 2005) by enabling sagittal balance of the upper body atop the hips (Been & Kalichman, 2013; Sparrey et al. 2014). Lordosis in humans develops following the onset of unassisted bipedal locomotion in juveniles (Abitbol, 1987; Cil et al. 2005; Shefi et al. 2013) and is argued to be a biomechanical response to gravitational load on the maturing vertebral column. Children who are paraplegic have spinal misalignment (Kilfoyle et al. 1965). In non‐human primates, the lumbar spine is not lordotically curved, as it is in humans. Lordosis occurs, however, in Japanese macaque monkeys trained for exclusive bipedal locomotion (Preuschoft et al. 1988). Chimpanzees are most closely related to humans and can only exhibit limited unassisted bipedal locomotion due in part to their flat and less mobile lumbar spines that do not allow the chimpanzee upper body to be positioned above their hips (Lovejoy, 2005; Lovejoy & McCollum, 2010). Humans with ‘flatback’, a pathologic flattening of the lumbar spine, exhibit impaired gait function (Sarwahi et al. 2002). LL is, then, load‐dependent and a critical adaptation enabling upright posture and bipedal locomotion in humans.

Additionally, LL allows an individual to maintain their posture and adjust the position of their center of mass (CoM) while carrying load (Meakin et al. 1996). Considering sex‐specific loading demands (e.g. pregnancy loads), females may be able to exhibit greater lordosis than males for them to accommodate increased upper body load and an anteriorly displaced CoM during pregnancy (Whitcome et al. 2007; Masharawi et al. 2010; Hay et al. 2015). Females increasingly extend their lumbar spines with developing pregnancy (Whitcome et al. 2007); however, if and to what extent the lumbar spine is sexually dimorphic in asymptomatic (and non‐pregnant) individuals is unclear. Characterizing the sex‐differences in LL may help explain (1) how sex impacts the wide range of normal LL (30–80° in standing asymptomatic adults; Been & Kalichman, 2013), and (2) sex‐differences in conditions pertaining to lumbar lordosis.

Despite extensive research, no consensus on whether sex‐differences in the lumbar spine exist has emerged (Been & Kalichman, 2013). As shown in Table 1, some studies report sex‐differences in LL, whereas others report no differences. The discrepancy in these results may be due to the variability in methodology and exclusion criteria used in the studies. Many of the clinically oriented studies were designed to examine sagittal balance and reported the effect of sex on LL as a secondary result, without controlling for study characteristics that may affect sex‐differences (e.g. including juveniles, age‐related degeneration, supine stance, skewed sex‐ratio in the sample). Only two previous studies have been explicitly designed to examine sex‐differences in lumbar morphology (Whitcome et al. 2007; Masharawi et al. 2010) but these studies reported significant sex‐differences in morphology of dry lumbar vertebrae from an osteological collection and, therefore, could not quantify the effect of the soft tissues (particularly the discs) on LL. In vivo sex‐differences in LL have recently been characterized using the shape of the vertebral canal (Hay et al. 2015), but the role of the discs remains unexplored. Furthermore, given the well‐established correlation between sacral orientation and lumbar lordosis (Legaye et al. 1998; Vialle et al. 2005; Boulay et al. 2006), sacral orientation parameters (e.g. pelvic incidence and sacral slope) should also differ by sex if sex‐differences in LL exist.

Table 1.

Review of imaging studies reporting on sex‐differences in in vivo lumbar lordosis. The literature is divided on whether or not significant sex‐differences in lumbar lordosis exist. In the table, differences are noted for those studies that found sex‐differences in lumbar lordosis. Italicized rows are studies comprising primarily juveniles. * Some overlap in study populations

Reported sex‐differences?	Research	n	Age	Method	Differences noted
No	Stagnara et al. (1982)	n = 100 (males, n = 57; females, n = 43)	20–29	Standing radiographs
	Voutsinas & MacEwen (1986)	n = 670 (males, n = 251; females, n = 419)	4–20	Standing radiographs
	Lin et al. (1992)	n = 149 (males, n = 76; females, n = 73)	18–76	Recumbant radiographs
	Jackson & McManus (1994)	n = 100 (males, n = 50; females, n = 50)	20–65*	Standing radiographs
	Korovessis et al. (1998)	n = 99 (males, n = 39; females, n = 79)	20–79	Standing radiographs
	Jackson et al. (1998)	n = 50 (males, n = 25; females, n = 25)	20–65*	Standing radiographs
	Korovessis et al. (1999)	n = 120 (males/females not reported)	20–79	Standing radiographs
	Jackson & Hales (2000)	n = 20 (males, n = 11; females, n = 9)	20–65*	Standing radiographs
	Cil et al. ( 2005 )	n = 151 (males, n = 79; females, n = 72)	3–15	Standing radiographs
	Boulay et al. ( 2006 )	n = 149 (males, n = 78; females, n = 71)	19–50	Standing radiographs
	Been et al. (2007)	n = 106 (males, n = 56; females, n = 50)	20–50	Standing radiographs
	Kalichman et al. (2011)	n = 191 (males, n = 104; females, n = 87)	40–80	Supine MRI
	Shefi et al. ( 2013 )	n = 210 (males, n = 173; n = 37)	2–20	Supine CT
Yes and no	Gelb et al. (1995)	n = 100 (males, n = 46; females, n = 54)	40–82	Standing radiographs	No differences in total LL or sacral inclination, but did find that females had significantly more segmental lordosis at L2–L3, L3–L4, and L4–L5.
	Cheng et al. (1998)	n = 340 (m, n = 142; f, n = 198)	54–71	Radiographs, stance unspecified	LA (between L1 and L5) was not different between sexes, but vertebral height ratios were significantly lower in males.
	Endo et al. (2012)	n = 50 (m, n = 25; f, n = 25)	23–47	Standing & sitting radiographs	LA (between L1 and L5) was significantly greater in females in sitting radiographs, but not when standing.
	Masharawi et al. ( 2012 )	n = 100 (m, n = 51; f, n = 49)	12–16	Supine MRI	3‐year follow up study in juveniles. Found LL to be significantly greater in females at the 12–13 age time point, but not different at the 15–16 age time point
Yes	Fernand & Fox (1985)	n = 938 (m, n = 418; f, n = 520)	17–84	Recumbant radiographs	LA (between L2 and S1): f = 47°, m = 43°, P < 0.01
	Amonoo‐Kuofi (1992)	n = 485 (m, n = 250; f, n = 235)	9–61	Recumbant radiographs	LA (between L1 and S1) was reported greater in females at every age group tested, no P‐values reported. (Age group averages for male LA were very low, ranging from 33 to 39°)
	Murrie et al. (2003)	n = 29 (m, n = 12; f, n = 17)	18–73	Supine MRI	LA (from L1/L2 disc space to L5/S1 disc space): females = 52, males = 48. (No P‐value provided for comparing LA between sexes in control population)
	Vialle et al. (2005)	n = 300 (m, n = 190; f, n = 110)	20–70	Standing radiographs	LA (between L1 and L5): females = 46°, males = 41°, P < 0.001
	Damasceno et al. (2006)	n = 350 (m, n = 143; f, n = 207)	18–50	Standing radiographs	LA (between L1 and S1): females = 62, males = 59, P < 0.01. This study also measured vertebral and disc wedging and found significantly more lordotic wedging in L2 and L4 for females.
	Oyakhire et al. (2013)	n = 300 (males, n = 144; females, n = 156)	18–77	Standing radiographs	LA (between L1 and S1): females = 50°, males = 47°, P < 0.05
	Hegazy & Hegazy (2014)	n = 93 (males, 46; females/males = 47)	25–57	Supine MRI	LA (between L1 and S1): f = 52°, m = 41°, P < 0.001.
	Zhu et al. (2014)	n = 260 (m, n = 104; females, n = 156)		Standing radiographs	LA (between L1 and S1): females = 49°, males = 43°, P = 0.03. No other measures (SS, PT, PI, thoracic kyphosis) were different between sex. VB and disc wedging not measured.

We designed a study that includes both males and females and aimed to characterize potential anatomical and postural sex‐differences in the lumbar spine. We aimed to clarify the conflicting results of prior work by eliminating age‐related degeneration and including both standing and supine individuals. We hypothesized (1) that females have greater LL than males, (2) that sex‐differences in LL are accentuated when standing, and (3) that pelvic parameters also exhibit sex‐differences. This study is the first to differentiate and quantify sexual dimorphism in the in vivo lumbar spine and pelvis in a controlled population.

Materials and methods

Study design

To determine the effect of sex on the degree of LL, we designed a cross‐sectional retrospective radiographic study that controlled several critical factors known to affect postural measurements. We excluded moderate to severe spinal degeneration (e.g. osteophytosis and disc degeneration) from our study sample. Although many studies of sagittal balance include degeneration as a normal feature of the aging spine (Been & Kalichman, 2013), we excluded any factor that could confound the measurement of lordosis. Both standing and supine radiographs were collected to compare sex‐differences in lordotic morphology under different postural loading conditions. In our study, ‘posture’ refers to standing or supine whole‐body orientations. Sex‐differences in LL were compared using relevant sagittal spine measures: vertebral wedging, disc wedging and sacral orientation.

Sample

Our sample included 200 adults ranging in age from 18 to 56 years with a mean and standard deviation of 32.9 ± 8.9 years. We collected lateral radiographs, either standing or supine, retrospectively from Harborview Medical Center in Seattle, Washington. Standing and supine radiographs are from separate subjects. Radiographs were taken as part of a routine medical examination designed to determine whether osseous injury had occurred after a traumatic event and we only collected those where the spinal exam concluded that no injury was present. Data were obtained for this study with human subjects approval (University of Washington #44154). Our exclusion criteria aimed to eliminate factors that may influence lumbar spine morphology and overall curvature, such as: moderate to severe degenerative disease (e.g. osteophytosis and disc space narrowing); congenital abnormalities (e.g. scoliosis, sacralized lumbar segments, and six lumbar vertebrae); fractures or other previous spine injuries; and previous spinal surgeries or devices. We also excluded radiographs when the vertebral endplates were unclear. We collected patient information for sex, ethnicity, radiographic orientation (standing or supine), and age at the time of radiography.

Angular measurements

On lateral radiographs, we measured lumbar angle, vertebral body wedging, intervertebral disc wedging, sacral slope and pelvic incidence (Fig. 1; variables defined in previous studies; Vialle et al. 2005; Been et al. 2011, 2010; O'Brien et al. 2004; De Carvalho et al. 2010). Lumbar angle (LA) is a way to assess LL and was defined as the Cobb angle between the L1 cranial endplate and S1 cranial endplate. Vertebral body and disc wedging angles were defined as the angle between the cranial and caudal endplates of the vertebrae and discs of the lumbar spine. Sacral slope was defined as the angle between two lines: (1) a line parallel with the S1 cranial endplate; and (2) a line parallel with the horizontal plane. Note, sacral slope changes as the posture of the body changes (between supine and standing). Pelvic incidence was defined as the angle between two lines: (1) a line perpendicular to the plane formed by the S1 cranial endplate and located at the endplate mid‐point; and (2) a line between the mid‐point of the S1 cranial endplate and mid‐point of a line connecting the central points of both acetabula. Pelvic incidence does not depend on postural orientation. Measurements were made using the OsiriX DICOM viewer.

Figure 1.

Radiographic measurements.

Statistics

The first author collected all radiographic measurements while blinded to patient information. Rater reliability was assessed via individual (single) intra‐ and inter‐class correlation coefficients (ICC) for lumbar angle and separate vertebral body and disc measurements. We selected a random subset of 30 radiographs that were measured by first author (J.F.B.) to assess intra‐rater reliability and an external rater to assess inter‐rater reliability. Student's t‐tests, power (1‐β), and effect size were calculated to assess the independent effect of sex on each morphological variable (lumbar angle, vertebral body and disc wedging angles, PI and SS). Multivariate linear regression analyses were used to assess the relationship between each morphological variable and sex while adjusting for the covariates age and ethnicity, which have been shown to have a relationship with lumbar curvature in prior reports (Been & Kalichman, 2013). Given the correlation between sagittal balance variables (LL, SS, PI), we did secondary multivariate regression analysis including SS and PI as additional covariates to understand how these sagittal balance variables may influence the relationship between sex and LA. For continuous covariates in the multivariate regression analyses, we used partial correlations for determining the correlation coefficient (r). In any test with a postural dependent variable (i.e. lumbar angle, disc wedging, sacral slope), separate analyses were done for standing and supine. A Bonferroni correction was applied to correct for multiple measures for the disc (five discs between L1 and S1, P = 0.05/5 = 0.01) and vertebral (L1–L5, P = 0.05/5 = 0.01) wedging angles. All statistical analyses were conducted using stata (Stata Corp, College Station, TX, USA).

Results

Of the 200 adults in our sample, 121 adults [females, n = 48, 34 ± 1.4 years (mean ± SE); males, n = 73, 34 ± 1.0 years] had standing radiographs, 75 adults (females, n = 39, 28 ± 1.2 years; males, n = 36, 32 ± 1.4 years) were supine, and four adults had unknown orientation. Of the 200 individuals in our sample: 36.5% self‐identified as Black, 32.5% Caucasian, 9% Hispanic, 6.5% Asian, 2% Native American; the remaining 13.5% did not disclose their ethnicity.

Lumbar lordosis

For LA measurements, ICC for intra‐rater reliability was 1.0 and inter‐rater reliability was 0.98. We found LA was 7.3° greater in females than males in the standing group [females, 60.3° ± 1.6° (mean ± SE); males, 53.0° ± 1.4°; P = 0.001; Table 2]. Conversely, LA was not significantly different between females and males while supine (females, 49.4° ± 1.5°; males, 46.5° ± 1.7°; P = 0.208). Overall, LA was 7.9° greater in standing than supine individuals (P < 0.001). Similarly, the covariate age was positively associated, albeit with a weak correlation, to LA when standing (r = 0.19, P = 0.024), but not when supine. Ethnicity was not associated with LA in either posture. See Table 3 for results from independent tests between LA and the covariates.

Table 2.

The t‐test results comparing angular measurements in the lumbar spine between sexes. Negative angles denote anterior wedging or kyphosis, positive angles denote dorsal wedging or lordosis. In addition to mean and standard error values, difference values are reported (i.e. females – males). P‐values indicate whether there are significant sex‐differences for each lumbosacral measure listed and significant values are defined by P < 0.05 and denoted by (*). Significance for the discs and vertebrae are adjusted by a Bonferroni correction [P = 0.05/(5 lumbar segments)] to be P < 0.01, therefore L3–L4 supine discs (0.017) are not significant based on the correction. Power (1‐β) and effect size (Cohen's d) are also reported

Level	Stance	Females (mean ± SE)	Males (mean ± SE)	Mean difference	P	1‐β	d
Lumbosacral measurement (°)
L1 vb	Standing + Supine	−2.44 ± 0.22	−4.06 ± 0.27	1.62	<0.001*	1	0.6
L2 vb	Standing + Supine	0.77 ± 0.24	−1.28 ± 0.26	2.05	<0.001*	1	0.8
L3 vb	Standing + Supine	2.09 ± 0.28	0.6 ± 0.23	1.48	<0.001*	1	0.6
L4 vb	Standing + Supine	4.26 ± 0.34	2.45 ± 0.24	1.80	<0.001*	1	0.6
L5 vb	Standing + Supine	10.48 ± 0.39	8.01 ± 0.32	2.47	<0.001*	1	0.7
L1–L2 disc	Standing	5.28 ± 0.40	5.51 ± 0.31	−0.22	0.654	0.1	−0.1
	Supine	2.3 ± 0.37	4.53 ± 0.33	−2.23	<0.001*	1	−1
L2–L3 disc	Standing	6.78 ± 0.35	6.76 ± 0.30	0.02	0.967	—	0.001
	Supine	2.81 ± 0.35	4.74 ± 0.35	−1.92	<0.001*	1	−0.9
L3–L4 disc	Standing	8.1 ± 0.34	8.11 ± 0.27	0.00	0.998	—	−0.001
	Supine	5.08 ± 0.48	6.68 ± 0.44	−1.59	0.017	0.8	−0.6
L4–L5 disc	Standing	10.89 ± 0.46	11.35 ± 0.41	−0.46	0.470	—	−0.14
	Supine	8.7 ± 0.46	9.72 ± 0.47	−1.02	0.126	0.3	−0.4
L5–S1 disc	Standing	15.08 ± 0.77	15.06 ± 0.58	0.02	0.981	—	0.004
	Supine	14.74 ± 0.88	16.32 ± 0.73	−1.58	0.175	0.2	−0.3
LA	Standing	60.34 ± 1.64	53.01 ± 1.44	7.32	0.001*	0.9	0.6
	Supine	49.39 ± 1.52	46.46 ± 1.74	2.93	0.208	0.3	0.3
SS	Standing	42.58 ± 1.37	38.87 ± 0.99	3.71	0.027*	0.6	0.4
	Supine	42.57 ± 1.50	42.19 ± 1.28	0.37	0.851	0.1	0.04
PI	Standing	54.84 ± 2.41	53.41 ± 1.46	1.43	0.595	0.1	0.1
	Supine	53.07 ± 2.94	57.66 ± 1.90	−4.59	0.181	0.3	−0.5

Table 3.

Univarate regression analyses comparing LA with separate variables: sex, SS, PI, age, and ethnicity. Correlation coefficients are provided for comparisons between LA and continuous variables. P‐values indicate whether there are significant sex‐differences for each relationship between LA and each variable listed and significant values are defined by P < 0.05 and denoted by (*)

	Standing		Supine
	Correlation (r)	P‐val	Correlation (r)	P‐val
E = β0 + β1*Sex	N/A	<0.001*	N/A	0.21
E = β0 + β1*PI	0.66	<0.001*	0.54	0.002*
E = β0 + β1*SS	0.85	<0.001*	0.85	<0.001*
E = β0 + β1*Age	0.15	0.11	0.01	0.94
E = β0 + β1*Ethnicity	N/A	0.76	N/A	0.97

Individual vertebrae and discs

For lumbar vertebral body and disc wedging measurements, the ICCs ranged from 0.88 to 0.94 for intra‐rater reliability and from 0.76 to 0.93 for inter‐rater reliability. We found that the vertebral bodies of all lumbar vertebra were significantly more wedged in females than in males (Fig. 2; Table 2; P ≤ 0.001). Vertebral body wedging angles included data from both standing and supine individuals, since individual vertebral body morphology is not dependent on postural orientation.

Figure 2.

Sex‐differences in vertebral body wedging for lumbar vertebral bodies. Means for vertebral body wedging angle (°) are plotted separately for females (solid black line) and males (dashed black line) with standard deviation bars. Trend in wedging patterns among lumbar vertebrae is similar between females and males, but the mean dorsoventral wedging for females is higher for each lumbar vertebrae. Vertebral body wedging is independent of posture and therefore, includes values from both standing and supine individuals.

Sex‐differences in disc wedging were tested separately for standing and supine subsamples because the disc is flexible and deforms in response to changes in load‐bearing and posture (Adams & Hutton, 1985; Meakin et al. 2009; Lord et al. 1997; demonstrated in Fig. 3). Disc wedging for each disc between L1 and S1 was not significantly different between standing females and males. Among supine individuals, males had significantly greater dorsal disc wedging than females for the two cranial‐most lumbar discs but the Bonferroni correction makes differences for L3–L4 non‐significant (Table 2).

Figure 3.

Examples of sex‐differences in lumbar curvature in standing and supine orientations. (A) Standing female, LA = 59.7°; (B) supine female, LA = 49.4°; (C) standing male, LA = 53.0°; (D) supine male, LA = 45.5°. Each of these were selected from our sample as examples depicting the mean values for standing and supine lordosis in females and males.

Sacral orientation

Pelvic measurements (SS, PI) were related to LA in both the standing and supine samples (Table 3). SS was measured in all of the standing (n = 121) and supine (n = 75) adults in our sample; however, PI was measured in 114 adults (standing, n = 81; supine, n = 33) because the acetabula were not visible in all radiographs. Given the significant sex‐differences in LA when standing, we tested whether sex‐differences in SS and PI exist. We found that (standing) SS had a significant relationship with sex (P = 0.027) but that PI did not (Table 2).

We explored further the relationships among LA, sex, SS and PI. SS and PI are themselves correlated (r ² = 0.53), so we included both in a multivariate regression analysis to determine whether their effects on LA are independent. In the multivariate regression analysis, SS remains strongly correlated with LA, but there is no longer a significant relationship between PI and LA (Table 4). For standing individuals, sex and sacral slope are significantly associated with LA. For supine individuals, only SS is significantly related to LA. This suggests that there are sex‐differences in postural orientation of the sacrum that relate to sex‐differences in standing LA.

Table 4.

Multivariate regression analyses for LA with covariates sex, SS, PI, age, and ethnicity. Partial correlation coefficients are provided for comparisons between LA and continuous variables. P‐values indicate whether there are significant sex‐differences for each relationship between LA and each variable listed and significant values are defined by P < 0.05 and denoted by (*)

Covariates	Partial correlation (r)	P‐value	Partial correlation (r)	P‐value
ELA = β0 + β1*Sex + β2*PI + β3*SS + β4*Age + β3*Ethnicity
Sex	N/A	0.02*	N/A	0.53
PI	0.11	0.37	−0.21	0.3
SS	0.79	<0.001*	0.75	<0.001*
Age	−0.03	0.79	−0.19	0.36
Ethnicity	N/A	0.97	N/A	0.58

Discussion

We hypothesized that females have greater lordosis than males, that these differences would be accentuated when standing (i.e. weight‐bearing) and that sacral orientation parameters predictive of LL would also exhibit sex‐differences. Our results indicate that sex‐differences in LL are present when standing, but not when supine. In the standing position, the lumbar spine assumes the shape of a buckled column (Meakin et al. 1996), demonstrating the functional role of curvature when loaded axially. In supine position, the lumbar spine appears to flatten when subjected to transverse loads. We found that in the standing sample, females, on average, have 7.3° greater lordosis than males, but that difference reduces to 2.9° (and is not significant) in the supine sample (Fig. 3). This result is further supported by sex‐differences in skeletal morphology (vertebral body wedging angles) and posture (sacral slope while standing).

Compared with previous studies, the unique value of this study is that we quantified sexual dimorphism in lumbar spines free of pathology and degeneration and explored the source of this variation quantitatively using characteristics which impact LL, including: the contributions of vertebral and disc wedging morphology to LL, the effect of body position on these measures, and the potential contribution of sacral orientation to sex‐differences in LL. Additionally, whereas many studies accept degeneration as a normal age‐related feature of the spine, we eliminated degenerated spines in order to investigate lumbar spine morphology of females and males without the lordosis‐damping effects of spine degeneration.

The primary limitation of this study was its retrospective design. Patient data, including body mass and pregnancy/parity history, were not uniformly available in the patient's medical record. Additionally, it would be valuable to have standing and supine radiographs from the same individuals. Future research should include a controlled prospective study where these data can be collected.

The ranges of LA, SS and PI observed in this study are in accordance with previously published results. Average standing L1–S1 Cobb angles have been reported as 51.3° ± 10.7° (Been et al. 2010) and 58.5° (Vialle et al. 2005; no reported standard deviation); we observed 55.9° ± 12.4°. Of the studies that did differentiate LA by sex, only one used methods that were directly comparable to those in our study (Damasceno et al. 2006). That study found that females had, on average, 2.7° greater LA than males when standing and that only L2 and L4 vertebral bodies were significantly more wedged. The differences between our results and theirs are unclear but may be due to differences between sample populations. Regardless, the sexual dimorphism in LA observed in this study was similarly present in standing radiographs, highlighting the effect of upright posture and load‐bearing on spinal curvature.

The correlations between pelvic parameters and LA are also well established (Schwab et al. 2009; Roussouly & Nnadi, 2010). The PI and SS values observed in this study are in agreement with those previously reported. However, sexual dimorphism in the relationship between sacral slope and LA has not been reported previously. We found that PI does not differ between sexes. Relating LL to a fixed measure such as pelvic incidence has value, but should be regarded with caution, as pelvic incidence does not include the significant effect of posture on normal lordosis in standing.

We found that standing females have, on average, 7.3° greater LL than males. In addition, each lumbar vertebral body was significantly more dorsally wedged in females and the sum of mean differences in vertebral wedging between females and males was 9.4° (Table 2). Two previous studies characterizing sex‐differences in dry lumbar vertebrae from an osteological collection (Whitcome et al. 2007; Masharawi et al. 2010) also found evidence of more vertebral wedging in particular lumbar vertebra of females: significant sex‐differences in vertebral wedging were found for L1–L4 in Whitcome et al. (2007) and in L1–L2 in Masharawi et al. (2010). Our data suggest that the amount of difference in vertebral wedging can account for the differences in LA. In supine patients, the average LA was significantly lower than in standing individuals and no sex‐differences were apparent. This suggests that changes in LL between standing and supine are sex‐dependent. We found the difference in LA between standing and supine was on average 10.9° in females (P < 0.001) and 6.5° in males (P = 0.007), which may imply differential lumbar range of motion between sexes. Future studies will be directed to understanding sex‐differences in the sagittal range of motion of the lumbar spine. Prior work has shown that there are sex‐differences in sagittal range of motion of the lumbar spine (Burton & Tillotson, 1988; Sullivan et al. 1994; Dreischarf et al. 2014), but determining how sex affects flexion and extension separately at the intervertebral level may provide insight into how vertebral wedging impacts intervertebral motion and whether females have a greater range of lumbar spine extension.

Our results for sacral orientation suggest that there are sex‐differences in the postural orientation of the sacrum that correspond to sex‐differences in LL, but not to the orientation of the sacrum within the pelvis. LA and SS differ between the supine and standing groups and between sexes, but PI does not. PI has been used to define surgical correction guidelines (Legaye et al. 1998; Le Huec et al. 2011; Lee et al. 2011; Roussouly & Pinheiro‐Franco, 2011; Bae et al. 2012; Berjano et al. 2014) and PI is correlated with LA and SS. It was therefore surprising that we found no sex‐differences in PI corresponding to sex‐differences in LL. Sex‐differences in SS, however, do exist. Taken together, this implies that females and males have a different pelvic‐orientation when standing that results in differences in SS that override the similarities in PI. In the few prior studies that reported non‐significant differences in PI between sexes (Peleg et al. 2007; Janssen et al. 2009), the discordance between no sex‐differences in PI and sex‐differences in LL was not investigated (note: Vialle et al. 2005 found significant sex‐differences in PI). Further research is necessary to investigate how sacral orientation and stance differences contribute to LL, as well as, how sacral orientation and stance may differ by sex.

Understanding the functional reason for greater lordosis in females than males is important for understanding developmental and biomechanical differences between the lumbar spines of females and males, as well as exploring potential trade‐offs that may accompany greater lumbar lordosis in females. Previous studies have suggested that sexual dimorphism in the lumbar spine is driven by the fetal load in pregnancy (Franklin & Conner‐Kerr, 1998; Whitcome et al. 2007). It was not possible to confirm this in our study, as parity history for our subjects was not available to us. Sex‐differences in vertebral wedging may imply, however, sex‐differences in loading, possibly beginning with sex‐differences in growth and development during skeletal maturation (Cil et al. 2005). Furthermore, the differences in mass distribution in the upper bodies of females and males or variations in spinal curvature in the cervical and thoracic spine may affect the degree of LL required to maintain stable posture (Hay et al. 2015).

Alternatively, sex‐differences in lordosis could merely be a by‐product of sexual dimorphism of the human pelvis (Tague, 1992; Rosenberg & Trevathan, 2002). Musculoskeletal differences of the hip may affect sacral orientation. Differences in lordosis may be primarily attributable to differences in sacral orientation due to pelvic dimorphism between females and males.

In conclusion, we found that LL is sexually dimorphic, with females having greater lordosis than males. This result, along with our findings that morphological and postural differences contribute to sex‐differences in LL, has functional and clinical implications for lumbosacral alignment. Understanding the variation in LL between sexes is a primary step toward explaining the wide variation in normal LL in adults. Additionally, sex‐differences in LL could indicate that females and males might be differently predisposed to certain lumbar conditions (e.g. risk of degenerative spondylolisthesis, which is five times higher for females than males; Sanderson & Fraser, 1996; Jacobsen et al. 2007). Finally, defining the quantitative sex‐differences in lumbar lordosis may promote the development of sex‐specific treatments for spinal maladies.

Conflict of interest

There are no conflicts of interest to declare.

Author contributions

J.F.B. collected and analyzed data and drafted the initial version of this manuscript. C.S.J. planned data analyses and edited this manuscript. E.B. contributed to study design and edited this manuscript. P.A.K. contributed to study design, planned data analysis and interpretation, and edited this manuscript.