Prevalence, risk factors and associations of primary Raynaud's phenomenon: systematic review and meta-analysis of observational studies.

OBJECTIVE: To systematically review the literature with regard to the prevalence, incidence, risk factors and associations of primary Raynaud's phenomenon (PRP).
METHOD: A systematic review of the literature of observational studies for PRP was undertaken using five electronic databases. Any studies reporting prevalence, incidence and risk factors of PRP were collected. Relative risk or OR and 95% CI were extracted or calculated to present the association between risk factors and PRP. Random effects model was used to pool the results.
RESULTS: 33 articles assessing a total of 33,733 participants were included in this analysis (2 cohort, 17 cross-sectional and 14 case-control studies). The pooled prevalence of PRP was 4.85% (95% CI 2.08% to 8.71%) in the general population. The pooled annual incidence of PRP was 0.25% (95% CI 0.19% to 0.32%). Risk factors and associations for PRP included female gender (OR=1.65, 95% CI 1.42 to 1.91), family history (OR=16.6, 95% CI 7.44 to 36.8), smoking (OR=1.27, 95% CI 1.06 to 1.53), manual occupation (OR=2.66 95% CI 1.73 to 4.08), migraine (OR=4.02, 95% CI 2.62 to 6.17), cardiovascular disease (OR=1.69, 95% CI 1.22 to 2.34) and marital status (married, OR=0.60, 95% CI 0.43 to 0.83). The definition of PRP varied considerably between studies.
CONCLUSIONS: This is the first systematic review of the prevalence, incidence, risk factors and associations of PRP. Further study using uniform strict criteria for the condition is required to confirm these findings, particularly the possible association with cardiovascular disease. Published by the BMJ Publishing Group Limited. For permission to use (where not already granted under a licence) please go to http://group.bmj.com/group/rights-licensing/permissions.

This is the first meta-analysis of the literature for the global epidemiology of primary Raynaud's phenomenon (PRP).

The prevalence and incidence of PRP in different countries were estimated. Female gender, positive family history, smoking and migraines were found to be the major risk factors for PRP.

The lack of original data restricted an adequate estimation of the age effect on PRP.

Different definitions of PRP handicapped a comparison between countries.

Introduction

In the 19th century, Maurice Raynaud first described Raynaud's phenomenon (RP) as an episodic, symmetrical, vasospastic disorder resulting in classic triphasic colour change, trophic changes limited to the skin and uncomfortable sensory symptoms of the extremities in the absence of arterial occlusion.1 Further criteria have been suggested to distinguish primary RP (PRP) from secondary RP, which include detail regarding symptom duration, negative autoimmune serology, normal serum inflammatory markers and capillaroscopy and the clinical absence of any underlying disease.1–4 Use of colour charts to aid diagnosis has also been used.4 5 Despite this, there is no unifying definition that is used worldwide for PRP.

There have been a number of studies performed in various countries reporting the prevalence of RP. The reported prevalence ranges from less than 1% (in men) and up to 20% (in women) depending on definitions and population selected.6 In contrast, few studies have examined the incidence of PRP, and the true burden of PRP in the general population remains unclear. PRP is thought to be more common in women, particularly when it develops at a young age.7 There are also reports of a hereditary component and links with other vasospastic conditions such as migraine.8–11 It is uncertain whether other comorbidities or risk factors particularly related to vascular diseases such as ischaemic heart disease and/or smoking have an association with PRP.

The primary objective of this study was to perform a systematic review of observational studies to summarise the literature with regard to the prevalence, incidence and risk factors/associations of PRP. The secondary objective was to examine the current definitions used to define PRP worldwide.

Methods

Literature search—data sources and search strategy

A comprehensive systematic literature search was undertaken in June 2011 and rerun in October 2014 using five databases: MEDLINE, EMBASE, CINAHL, AMED and PubMed. The search terms for “Raynauds” or “Raynauds disease” were combined with the terms “epidemiology”, “prevalence”, “risk” or “incidence” to generate the citations (see online supplementary appendix 1 for full details of search strategy). “Cross sectional”, “case-control” or “cohort” studies and “systematic review” were also applied for types of studies.

Abstracts were reviewed and the full papers were sought where abstracts were felt to be relevant. Any duplicate articles were excluded (figure 1 and appendix 1 and 2). Where there was difficulty in article retrieval, the authors were contacted via email. The literature search and abstract review was completed by RG and validated by WZ. Reference lists of the review articles were also examined for relevant studies.

Figure 1

Flow chart diagram showing results of systematic literature search.

Review (219), Case report (34), Letter (14), Webpage (5), Animal studies unrelated to Raynaud's phenomenon (RP) (4), Connective tissue disease (602), Vibration induced disease (169), Diagnosis/investigation of RP (11), Secondary progression from primary RP (4), Drugs related to RP (53), RA/inflammatory arthritis (21), Other musculoskeletal (37), Fibromyalgia (9), Genetics not related to RP (11), Autoantibodies (13), Treatment involving sympathectomy (45), Ophthalmology studies (14), Psychiatric conditions (11), Haematological disease (14), Infectious disease (24), cardiovascular disease (64), Respiratory disease (17), Gastrointestinal disease (27), Renal disease (27), Dermatological (30), Endocrine disease (31), Neurological disease (43), Cancer (106), Drugs not related to RP (63), Vascular intervention (64), Laser Doppler flowmetry (8), Breast implant rupture (4), Others: including motor vehicles, dentists, seafarers, fishermen, aircrafts, breast feeding (17)

Letter to editor (1)

Review (5)

Diagnosis/Investigation of RP (14)

Secondary RP (39)

Treatment RP (1)

Same patient sample as is already included in another study used (3)

Case only study (1)

Inclusion and exclusion criteria

Inclusion criteria: studies reporting the prevalence and/or incidence of PRP; studies reporting potential risk factors associated with PRP; studies reporting human data on PRP in people of any age; studies in any language (4 articles required translation—1 Japanese, 1 Turkish, 1 French, 1 Italian).

Exclusion criteria (figure 1): studies assessing treatment of PRP; studies involving participants with RP secondary to other diseases; studies assessing RP in a specific occupation, for example, people using vibration tools; unpublished material, case reports, editorials, letters or reviews.

Data extraction and quality assessment

Study characteristics including age range, gender ratio and total number of participants in the study were documented. The study design, country, setting (ie, hospital or community based) were also assessed and noted. If more than one article used the same study population, the article where the data were felt to be presented most clearly was used in the study. The definition of PRP and instruments used to confirm the condition were also documented. The number of cases of PRP out of the number of people studied in a certain time in the general population was documented as unadjusted crude prevalence. Incidence figures were documented if the number of new cases of PRP in the population at risk studied over a given period of time was stated.

Individual OR, relative risk (RR) or HR and their 95% CI were extracted or calculated for the following:

Constitutional: age, gender

Environmental: employment, education, marital status and sex hormone medications

Genetic: family history in 1st degree relatives

Associations: smoking, alcohol, cardiovascular disease (CVD) and migraine

All studies were reviewed by RG to assess study quality and for data extraction and were validated by WZ. An independent reviewer (RK) assessed a random selection of articles to ensure quality of data extraction. Study quality was assessed according to study design (cohort, cross-sectional and case–control), setting (community or hospital), sample size, case definition, exposure definition, confounding factors and adjustment. Quality scoring for studies was not performed as it is not possible or fair to assign equal weight to different quality aspects related to the study. However, current consensus standards of reporting meta-analysis of observational studies in epidemiology12 were followed, and subgroup/sensitivity analysis was undertaken to examine the changes of the estimate according to different quality aspects.

Statistical analysis

Individual data for prevalence and incidence were derived from the original report either directly or indirectly from the information provided in each study. The pooled proportion was calculated as the back transform of the weighted mean of the transformed proportion, using inverse arcsine variance weights for the fixed effects model and DerSimonian and Laird13 weights for the random effects model.14 Cumulative incidence and 95% CI were transformed into incidence rate data (ie, incidence per 100 person-years) and pooled incidence rate was estimated. Individual data for OR, RR and HR were pooled to present the overall relative risk of all observational studies, as well as separately for each specific risk measure or study design as appropriate. Random effects mode was used to pool the data.15 Heterogeneity was examined using Forest plots, Cochran Q tests and I2 statistic as a measure for inconsistency due to chance.16 17 Publication bias was assessed using funnel plots and Eggers test or the Harbord test if the number of studies included in the meta-analysis was too small (≤4).18 All analyses were undertaken using StatsDirect V.2.7.9.

Results

Study characteristics

In total, 2378 citations were found in the initial literature search. All 467 duplicates were removed and 1878 citations were excluded as they did not meet the inclusion criteria (figure 1). The final number of studies available for analysis was 33 (33 733 participants). There were two cohort studies19 20 (1 632 participants), 17 cross-sectional studies6 21–36 (25 797 participants) and 14 case–control studies7–11 37–45 (6 304 participants; table 1). Data for incidence and prevalence were taken from cohort and cross-sectional studies, respectively. Data for risk factors were taken from all studies as long as the results were reported.

Table 1

Characteristics of studies

	Cohort	Cross-sectional	Case–control	All studies
Number of studies	2	17	14	33
Number of participants	1632	25 797	6304	33 733
Age	18–81	12–84	16–79	12–84
Setting
Community based	2	14	5	21
Hospital based	0	2	9	11
Community and hospital	0	1	0	1
Region of study
USA	1	2	7	10
Europe	1	11	6	18
France and USA	0	1	0	1
Japan	0	2	0	2
New Zealand	0	1	0	1
Israel	0	0	1	1
Diagnosis*
Questionnaire	0	5	2	7
Questionnaire+examination	2	9	12	23

*Diagnosis was classified as questionnaire based or questionnaire and examination based. The former includes phone survey, postal questionnaire and face-to-face interview whereas the latter includes clinical examination, blood testing including serology, use of colour chart/photographs and capillaroscopy in addition to the questionnaire. Colour chart/photographs were used in 12 studies (cohort 2, cross-sectional 7, case–control 3) and capillaroscopy was used in 10 studies (cohort 1, cross-sectional 2, case–control 7).

Age ranges across different study designs were as follows: case–control (16–79 years), cohort (18–81 years), cross-sectional (12–84 years). Sixty-seven per cent of the studies involved participants recruited in a community-based setting. The majority of studies were conducted in Europe (18)6 9 19 22 24 26 27 29 30 32 33 35–37 39 41–43 or the USA (10),7 8 10 11 20 21 23 38 40 44 however, other countries of origin included Japan (2),25 28 New Zealand (1)34 and Israel (1).45 One comparison study included participants from the USA and France.31

Participants were surveyed by means of phone, face-to-face interview and/or postal questionnaire. Twenty-six studies included a physical examination that also included blood testing (including serology), nailfold capillaroscopy and use of colour chart/photographs (table 1).

Ten studies used specific criteria to define PRP (3 studies8 21 32 Allen and Brown;1 39 36 37 LeRoy and Medsger2; 46 22 33 34 UK Scleroderma Study Group4). The remaining studies used a combination of cold sensitivity, varying degrees of colour change and sensory symptoms via questionnaire or interview to define PRP. Colour charts or photographs to indicate colour change were used in 12 studies6 7 9 10 19 20 25 28 29 31 33 35 and nailfold capillaroscopy was performed as part of the examination in 10 studies.8 9 19 31 36 37 39 41–43 In 15 studies, blood testing including serology and/or inflammatory markers was performed.6 8 9 11 24 30 33 36 37 39–44

Studies with clear definition of PRP or clear exclusion criteria for secondary RP were categorised as ‘definite PRP’ in this study. Studies with less clear definition of PRP were categorised as ‘possible PRP’. Studies with clear definition of secondary RP were excluded.

Prevalence of PRP

The overall prevalence for definite PRP varied from 1.6% to 7.2% in six cross-sectional studies in the general population (women: 2.1–15.8% and men: 0.8–6.5%).21 23 25 27 29 33 The pooled prevalence was 4.85% (95% CI 2.08% to 8.71%; figure 2), with 5.74% (95% CI 2.74% to 9.75%) in women and 4.12% (95% CI 1.60% to 7.74%) in men. We used the Harbord test to detect publication bias (1.59, 92.5% CI −21.6 to 24.8; p=0.87). The overall prevalence for possible PRP ranges from 3.98% to 12.7% (women: 4.5–17.9% and men: 3.4–7.2%) in three cross-sectional studies.26 34 35 The prevalence in specific populations varies depending on the studies (table 2).

Figure 2

Forest plot showing the pooled prevalence of definite primary Raynaud’s phenomenon for five general population studies.

Table 2

Prevalence of primary Raynaud's phenomenon in 17 studies

First author	Country	Setting	Samplesize	Age mean(SD)/range	Female (%)	Prevalence
						Overall (%)	Female (%)	Male (%)
Prevalence of definite primary Raynaud's phenomenon in general population studies
Brand (1997)	Boston, USA	Com	4182	51.8	52.2	7.2	7.8	6.5
Fraenkel (1999)	Boston, USA	Com	1525	53.9	52.5	7.8	9.6	5.8
Harada (1991)	Ehime, Japan	Hosp	3873	20–70	51.6	1.6	2.1	1.1
Ivorra (2001)	Valencia, Spain	Com	276	54.4	74.3	3.3	3.4	2.8
Maricq (1997)*	South Carolina, USA	Com	2086/432	18+	NS	NS	3.4	0.8
	Toulon, France	Com	1998/189	18+	NS	NS	11.4	2.8
	Nyons, France	Com	1996/345	18+	NS	NS	5.8	6.2
	Grenoble, France	Com	2069/272	18+	NS	NS	12.1	6.4
	Tarentaise, France	Com	2000/296	18+	NS	NS	15.8	6.3
Onbasi (2005)	Van, Turkey	Com	768	29.2 (10.4)	46.6	5.9	7.0	4.9
Prevalence of possible primary Raynaud's phenomenon in general population studies
Heslop (1983)	Southampton, UK	Com	450	20–59	50.9	12.7	17.9	7.2
Purdie (2009)	New Zealand	Com	234	18+	56.8	11.5	17.3	4.0
Sahin (2003)	Van, Turkey	Hosp	251	28.9	53.4	3.98	4.5	3.4
Prevalence of primary Raynaud's phenomenon in single gender only population studies
Leppert (1987)	Vasteras, Sweden	Com	2705	18–59	100	–	15.6	–
Olsen (1978)	Copenhagen, Denmark	Com	67	21–50	100	–	22.4	–
Tzilalis (2011)	Athens, Greece	Com	3912	18–28	0	–	–	0.18
Prevalence of primary Raynaud's phenomenon in studies using hospital personnel
Cakir (2008)	Edirne, Turkey	Com	1414	27.2	59.3	3.6	4.8	1.9
Gallo (1994)	Milan, Italy	Com	1920	15–84	68	4.2	4.5	3.9
Iwata (1987)	Japan	C&H	1470	18–59	56.8	4.8	6.5	2.5
Voulgari (2000)	Ioannina, Greece	Com	500	33.7 (6.2)	77.8	5.2	6.4	0.9
Prevalence of primary Raynaud's phenomenon in studies assessing children
Jones (2003)	Manchester, UK	Com	716	12–15	50.8	14.9	17.6	12.2

*This study involved two stage sampling.

C&H, Community and Hospital; Com, Community; Hosp, Hospital; NS, not significant.

In six studies assessing the general population we found the lowest prevalence of PRP in Japan, with an overall prevalence of 1.6 (2.1% in women, 1.1% in men).25 Highest overall prevalence figures were found in the USA with a median prevalence of 7.5% (7.8% in women, 5.8% in men).21 23 A study from France also showed high prevalence figures of 11.75% in women and 6.3% in men (median values;31 table 3).

Table 3

Regional variation of prevalence of primary Raynaud's phenomenon for general population studies including prevalence rates for males and females

Country	Number of studies	Number of participants	Prevalence (%)
			Male	Female	Total
USA*	3	6 139	5.8	7.8	7.5
France*	1	1 102	6.3	11.75	–
Spain	1	276	2.8	3.4	3.3
Turkey	1	768	4.9	7.0	5.9
Japan	1	3 873	1.1	2.1	1.6
Total	6	12 158

*Median values calculated for prevalence. The US gender figures include data from Maricq et al31 (France and the USA). Total US prevalence figure includes data from two US-only studies.

Five studies reported prevalence of PRP by age.6 28–30 34 Three did not find any age-related prevalence.6 28 30 Purdie et al34 reported a higher prevalence of PRP in younger compared to older age groups, whereas Fraenkel et al38 reported higher prevalence in older age groups in men (adjusted OR=2.3, 95% CI 1.0 to 5.2 highest vs lowest tertile) but not in women (adjusted OR=0.9, 95% CI 0.4 to 1.6). Jones et al29 also showed a slight increase in prevalence by age in yearly increments between ages 12 and 15 years.

Incidence of primary Raynauds phenomenon

Only two studies reported incidence rates.19 20 Carpentier et al reported an annual incidence rate of 0.25% (95% CI 0.17% to 0.33%), with 0.24% in women and 0.26% in men and Suter et al reported a 7-year incidence of 1.87% (2.2% in women and 1.5% in men), which was converted to an annual incidence rate of 0.26% (95% CI 0.17% to 0.39%). The pooled annual incidence rate of these two studies was therefore 0.25% (95% CI 0.19% to 0.32%).

Risk factors and associations

In 18 studies (23 197 participants), there was a positive association between female gender and PRP (OR=1.65, 95% CI 1.42 to 1.91).6 7 9 20–29 33–35 37 44 Family history, assessed in two studies looking at first-degree relatives, also had a positive significant association with PRP (OR=16.6, 95% CI 7.44 to 36.8).8 9 No significant association was found with education beyond primary school age6 37 (table 4). Manual occupation (not including vibration tool use) had an OR of 2.66 (95% CI 1.73 to 4.08) in one study of 3873 participants.25 In four studies, being married was associated with a lower risk of PRP with OR of 0.60 (95% CI 0.43 to 0.83) compared with being single/divorced/widowed.6–7 23 37 Smoking was found to have an association in nine studies giving a pooled OR of 1.27 (95% CI 1.06 to 1.53).7 9–11 20 22 23 32 44 Alcohol use,23 44 participants with positive Helicobacter pylori investigations39 41 and those with conditions such as diabetes,23 hypertension11 23 and hypercholesterolaemia23 did not have a significant association with PRP.

Table 4

Risk factors of Raynaud's phenomenon

Risk factor	Number of studies (Number of participants)	Pooled OR	95% CI	I2% (95% CI)*	p (heterogeneity)
Female	18 (23 197)	1.65	1.42 to 1.91	17.2 (0 to 53)	0.25
Family history of RP	2 (421)	16.6	7.44 to 36.8	–	0.34
Marital status†	4 (2 650)	0.60	0.43 to 0.83	16.9 (0 to 73)	0.31
Education‡	2 (891)	1.52	0.89 to 2.59	–	0.24
Manual occupation	1 (3 873)	2.66	1.73 to 4.08	–	–
Smoking	9 (8 501)	1.27	1.06 to 1.53	6.2 (0.8 to 57.1)	0.38
Alcohol	2 (4 967)	0.33	0.02 to 5.37	–	<0.0001
Migraine§	6 (2 595)	4.02	2.62 to 6.17	35.9 (0 to 73.6)	0.17
Diabetes	1 (1 525)	0.51	0.2 to 1.27	–	–
Hypertension	2 (1 711)	1.00	0.67 to 1.48	–	0.46
Hypercholesterolaemia	1 (1 525)	0.86	0.53 to 1.40	–	–
Coronary heart disease	1 (81)	0.58	0.1 to 3.31	–	–
Cardiovascular disease¶	1 (3 442)	1.69	1.22 to 2.34	–	–
Helicobacter pylori**	2 (265)	0.91	0.51 to 1.63	–	0.07
CP	2 (268)	0.69	0.34 to 1.38	–	0.88
Oestrogen replacement therapy††	2 (1 242)	2.34	1.42 to 3.84	–	0.81

Bold typeface indicates statistically significant results. *The I2 values are stated where more than three studies were assessed.

†Marital status references used are single/separated/widowed/divorced apart from Fraenkel et al38 and Keil et al,7 where references used are widowed/separated/divorced.

‡References for education used are primary school37 and <12 years education.6

§O'Keeffe11 did not report whether their calculation for OR was adjusted/unadjusted. All other reported calculations for OR are unadjusted.

¶Cardiovascular disease includes history of angina, myocardial infarction, coronary insufficiency, intermittent claudication, congestive cardiac failure, stroke and transient ischaemic attack.

**Positive investigation for H. pylori uses urea breath test41 and serology.39

††Adjusted for age, BMI, alcohol, cigarettes and B adrenoreceptor antagonists in the study by Fraenkel et al.38

BMI, body mass index; CP, contraceptive pill; RP, Raynaud's phenomenon.

Migraine had a positive significant association with a pooled OR of 4.02 (95% CI 2.62 to 6.17) in six studies.6 9–11 22 45 One study of 3442 participants reported a positive association of CVD with PRP with an OR of 1.69 (95% CI 1.22 to 2.34).44 CVD in this study included a history of ischaemic heart disease, intermittent claudication, congestive cardiac failure and cerebrovascular disease. A single study using 81 participants did not show a positive significant association of coronary heart disease with PRP.40

A positive association was found in participants taking oestrogen replacement therapy alone in two studies with an OR of 2.34 (95% CI 1.42 to 3.84).23 38 However, no significant association was found in combined oestrogen and progesterone replacement therapy38 in postmenopausal women or in contraceptive pill (CP)10 11 use in two other studies. A study by Smyth et al 9 looking at allele frequencies of known polymorphisms of candidate vasoactive mediator genes (eNOS, BKRG, ET01 and ETA receptor genes) did not show any association. Shemirani et al42 looked at clotting factors in participants with PRP and found a significant association with methyltetrahydrofolate reductase C677T mutations (OR=0.4, 95% CI 0.2 to 0.9) but no difference in other thrombosis-associated alleles (FVLeiden, prothrombin G20210 A).

Discussion

This is the first meta-analysis of the literature for the prevalence, incidence, risk factors and associations of PRP. Overall, the pooled mean prevalence of PRP in the general population was 4.85% (95% CI 2.08% to 8.71%; figure 2) and the mean incidence was 0.25% (95% CI 0.17% to 0.33%) per annum.19 Major risk factors/associations of PRP include female gender, family history of PRP, migraine, smoking, CVD, manual occupation, oestrogren replacement therapy and possibly, marital status (table 4).

Variations in prevalence were observed between countries (table 3), though this could reflect use of different diagnostic criteria rather than real differences in prevalence. The heterogeneity of prevalence figures may also reflect the differences in the way the studies were conducted, the selection of participants (eg, age and gender) and the disease definition. All studies (except for Maricq et al31) demonstrate a higher prevalence of PRP in women. This may be due to a relationship with female hormones as two studies found an association between PRP and use of oestrogen replacement therapy alone,23 38 although no association was found between combined oestrogen and progesterone replacement or the CP.10–11 38 In contrast, prevalence of PRP does not increase with age in five published studies with a wide age range of participants from 12 up to 84 years.6 28–30 34 This accords with the clinical observation that PRP usually starts in teenage years and that later development, which is far less common, is characteristic of secondary RP. While the former may be driven predominantly by genetic risk factors, later onset ‘primary’ Raynaud's may be predominantly influenced by environmental exposures such as vascular microtrauma from manual usage and vibrating tools. In terms of other environmental factors we did find a weak negative association between marital status and PRP with an OR of 0.60 (95% CI 0.43 to 0.83) in those that are married versus single/separated/widowed/divorced.6–7 23 37 However, there is no plausible biological explanation for this and the reported data may not be free from confounding bias.

The association of CVD and autoimmune disease is well documented and thought to be due to accelerated atherosclerosis as a result of chronic inflammation, treatment such as glucocorticoids as well as the traditional risk factors for CVD.46–52 A link between CVD and PRP has been shown in only one study44 and the reason for this association is not known. It is unlikely to be due to an inflammatory process or related to medication, and with PRP having predominance for the female population and onset at a young age, it is not clear if traditional cardiovascular risk factors play a part. However, smoking was found to have a positive association with PRP in our study (OR=1.27, 95% CI 1.06 to 1.53). It is well known that smoking is one of the three (smoking, hypertension and hyperlipidaemia) main risk factors for cardiovascular and cerebrovascular disease.53–59 Smoking may have the same risk factor for PRP and CVD. Whether smoking causes PRP first and then CVD is an interesting question that deserves further research. More interestingly, we found a very strong association between migraine and PRP (OR=4.02, 95% CI 2.62 to 6.17). It has been previously shown that migraine is due to a cascade of vascular and neural events.46 47 However, a review by Rosamund suggested that migraine was not shown to be linked with coronary heart disease but possibly shares a common underlying pathophysiology with RA and other vasospastic disorders such as variant angina.60 61 It is thought there may be other factors that could affect the underlying mecanism for these vasospastic conditions as episodes occur at different times with differing precipitants.62 63 Further study may help clarify whether PRP is a benign vasospastic disorder or whether there is underlying pathology affecting the vascular wall associated with traditional risk factors seen in CVD.

There are a number of caveats to this study. Firstly, it was striking that there was no uniform definition for diagnosis of PRP. Only 39% of studies looking at prevalence had a precise definition for PRP, thereby reducing the number of studies we used to assess pooled prevalence. It is possible that the variation in definition of PRP together with the way participants were recruited and assessed may have led to underestimation or overestimation of the true prevalence of PRP in the general population. We feel that an amalgamation of the generally more commonly used definitions would ensure that the diagnosis is clear by assessing symptoms, using a colour chart or photographs for confirmation of colour change and carefully exclude underlying conditions including checking for digital infarcts/ulceration, nailfold capillaroscopy, and assessing autoimmune screen and inflammatory markers. Secondly, as our objective was to specifically examine the epidemiology of PRP, a large proportion of studies were excluded because they focused on secondary Raynaud's phenomenon, especially related to connective tissue diseases and vibration white finger. In addition, we also excluded studies that looked at investigation or treatment of PRP. This left only a small number of studies to assess. From the studies included, there was a great deal of variation in the population of participants used. Nine of the 17 studies used investigated participants in the general population, whereas the remainder examined specific populations such as single gender, children or hospital/medical personnel. Furthermore, there was considerable variation in the risk factors addressed in each study and this may have affected the significance and association, or lack of association between the risk factors and PRP. We tried to extract as many risk factors from each study as possible to use in our analysis. In the future, a larger multinational population study may help us to get a better understanding of the disease. This would be particularly useful if standardised criteria were used to include participants in the studies, using strict definition for PRP (as mentioned previously), and data were collected in a similar fashion assessing a wide variety of possible risk factors (particularly related to CVD and vasospastic disorders) for more accurate data analysis.

Conclusion

This first systematic review summarises the burden of PRP in the general population using published literature. It is not a rare condition (prevalence 4.85% and annual incidence 0.25%). It starts at a young age, is more common in women, and associates with a family history and with smoking. In addition, people with PRP are four times more likely to have migraine than those without this condition.