The role of the Panton-Valentine leucocidin toxin in staphylococcal disease: a systematic review and meta-analysis.

BACKGROUND: Invasive community-onset staphylococcal disease has emerged worldwide associated with Panton-Valentine leucocidin (PVL) toxin. Whether PVL is pathogenic or an epidemiological marker is unclear. We investigate the role of PVL in disease, colonisation, and clinical outcome.
METHODS: We searched Medline and Embase for original research reporting the prevalence of PVL genes among Staphylococcus aureus pneumonia, bacteraemia, musculoskeletal infection, skin and soft-tissue infection, or colonisation published before Oct 1, 2011. We calculated odds ratios (ORs) to compare patients with PVL-positive colonisation and each infection relative to the odds of PVL-positive skin and soft-tissue infection. We did meta-analyses to estimate odds of infection or colonisation with a PVL-positive strain with fixed-effects or random-effects models, depending on the results of tests for heterogeneity.
RESULTS: Of 509 articles identified by our search strategy, 76 studies from 31 countries met our inclusion criteria. PVL strains are strongly associated with skin and soft-tissue infections, but are comparatively rare in pneumonia (OR 0·37, 95% CI 0·22-0·63), musculoskeletal infections (0·44, 0·19-0·99), bacteraemias (0·10, 0·06-0·18), and colonising strains (0·07, 0·01-0·31). PVL-positive skin and soft-tissue infections are more likely to be treated surgically than are PVL-negative infections, and children with PVL-positive musculoskeletal disease might have increased morbidity. For other forms of disease we identified no evidence that PVL affects outcome.
INTERPRETATION: PVL genes are consistently associated with skin and soft-tissue infections and are comparatively rare in invasive disease. This finding challenges the view that PVL mainly causes invasive disease with poor prognosis. Population-based studies are needed to define the role of PVL in mild, moderate, and severe disease and to inform control strategies. FUNDING: None.

Introduction

Staphylococcus aureus harmlessly colonises the skin and mucosa of about 30% of healthy adults and is the commonest cause of mild to moderate skin and soft-tissue infections, such as abscesses and wound infections. Such infections account for many consultations in primary care but rarely result in hospital admission or surgical treatment. The incidence of invasive community-onset staphylococcal infections such as pneumonia or osteomyelitis is low, but S aureus is a common cause of health-care-associated infections, with meticillin-resistant S aureus (MRSA) strains reported from hospitals and health-care facilities in most industrialised countries.

Panton-Valentine leucocidin (PVL) is a toxin composed of two components, LukS-PV and LukF-PV. These two components are secreted before they assemble into a pore-forming heptamer on neutrophil membranes, leading to neutrophil lysis. Epidemiological, historical, and biochemical research all point towards a role for PVL in pathogenesis but whether the toxin affects clinical presentation, disease severity, and outcome is unclear. The toxin has been linked to community-onset MRSA disease worldwide, but some community-acquired MRSA strains do not carry the PVL genes, particularly in Australia. Sequencing data suggest that circulating community-onset MRSA strains might be directly descended from a historical PVL-producing penicillin-resistant clone phage type 80/81 that circulated in the 1950s and 1960s and was highly virulent. Additionally, PVL has well established leucocidal properties, and causes dermonecrosis when injected into rabbits, but direct evidence from experimental studies with animal models that the PVL toxin causes invasive disease is scarce.

The molecular epidemiology and burden of community-onset staphylococcal disease is geographically disparate. In the USA, community-onset staphylococcal disease is endemic and is mainly meticillin-resistant associated with a PVL-producing clone, ST8/USA300. All USA300 isolates are sequence type 8 (ST8) but only a subset of ST8 isolates are USA300. Until about 2003, these infections were predominantly community-acquired, affecting people without risk factors for health-care-acquired infection, but USA300 has become a common cause of health-care-associated infection. In Europe, USA300 is rare, and community-onset MRSA is epidemiologically and clonally diverse. Many PVL strains are meticillin-sensitive S aureus (MSSA), and the commonest community-onset PVL-positive MRSA clone is ST80. PVL-positive community-onset MRSA strains are rarely reported in England and Ireland, but are common in parts of Europe, such as Greece, where PVL-positive MRSA circulates in both community and hospital settings. Different clones predominate in Australia where USA300 is rare and community MRSA and MSSA clones both have PVL genes. Although disease burden varies, clinical syndromes associated with community-onset staphylococcal disease are consistent across all regions. Skin and soft-tissue infections are common and often present as abscesses, furuncles, and carbuncles. A few patients develop necrotising pneumonia, sepsis syndrome, necrotising fasciitis, and musculoskeletal disease. Disease has been described in diverse populations including children, athletes, the military, men who have sex with men, and prisoners; a common feature is prolonged close contact, which is thought to play an important part in disease transmission.

Many studies have reported an association between PVL genes and invasive disease, implying that PVL is an epidemiological marker of a syndrome of severe infection. In some countries this notion has led to public health measures for individuals infected with PVL-producing strains. Importantly, aggressive clinical and public health responses, such as screening and decolonisation, might be more justifiable if PVL-positive disease is rare and mainly severe than if disease is common and not associated with severity or invasion. Modelling suggests decolonisation of patients with MRSA in settings such as intensive-care units is cost effective and improves health outcomes. By contrast, evidence from randomised trials that decolonisation reduces recurrence of community-onset S aureus skin infections is scarce, and no studies have addressed the cost effectiveness of disease control measures specifically for PVL-positive infection.

We review international evidence for the association between PVL genes, colonisation, disease, and outcome for S aureus pneumonia, bacteraemia, musculoskeletal infection, and skin and soft-tissue disease. In this Article the terms PVL-producing and PVL-carrying S aureus denote isolates of S aureus that contain the PVL genes. PVL-positive disease refers to S aureus pneumonia, bacteraemia, musculoskeletal infection, and skin and soft-tissue disease caused by strains containing the PVL genes. Invasive disease is defined as isolation of S aureus from a normally sterile body site. The term unselected isolates refers to studies that assessed prevalence of PVL genes among all S aureus specimens submitted to a laboratory in a defined time period, without additional exclusion criteria.

Methods

Search strategy and selection criteria

We searched Medline and Embase for articles reporting the proportion of isolates from S aureus pneumonias, bacteraemias, musculoskeletal infections, and skin and soft-tissue disease that produced PVL. English or French language articles published after 1946 (Medline) or 1980 (Embase) and before Oct 1, 2011, were eligible for inclusion. We used the following keywords in various combinations to search the scientific literature: “Staphylococcus aureus”, “meticillin resistance”, “community acquired”, “incidence”, “prevalence”, “risk factors”, “epidemiology”, “panton valentine”, and “leukocidin”. After peer review, we repeated the literature search in Medline to include articles reporting the proportion of isolates from S aureus colonisation that produced PVL. We used the same search terms as the previous search and hand-searched the reference lists of relevant articles. We excluded review articles; case reports; outbreaks; studies based at institutions, such as prisons or residential homes; studies in immunocompromised patients, such as those with HIV infection; and studies in high-risk settings, such as intensive-care units. We excluded small studies with fewer than ten isolates for pneumonia and musculoskeletal infection, or fewer than 50 isolates for bacteraemias, skin and soft-tissue disease, and colonisation. We excluded studies of bacterial clones, rather than unselected clinical isolates and colonisation studies exclusively reporting the prevalence of PVL genes among patients admitted to hospital.

Data extraction

Two independent researchers (LJS and EF) reviewed all abstracts to identify articles that required full-text review, with final decision reached through consensus. LJS and EF reviewed abstracts of articles against predefined inclusion criteria and all articles included were discussed with a third reviewer (ACH). We extracted data for the study setting, inclusion criteria, and onset of infection (hospital or community-onset) and entered them directly onto a database. LJS extracted clinical details for meticillin-resistance and infection type to classify studies into S aureus colonisation, bacteraemias, pneumonia, bone and joint infections, or skin and soft-tissue infections. We used information about sampling strategy to identify studies of unselected S aureus specimens submitted to a hospital laboratory in a defined time period. We extracted data for the prevalence of PVL strains for MRSA and MSSA isolates and used them to calculate the prevalence of PVL strains among each type of infection.

Statistical analysis

To examine the association between PVL genes, colonisation, and disease we calculated odds ratios (ORs) comparing the odds of infection with a PVL-positive strain for S aureus pneumonia, musculoskeletal infections, and bacteraemias, or colonisation relative to the odds of PVL-positive skin and soft-tissue disease. We calculated the odds of skin and soft-tissue infection relative to the odds of all other presentations of S aureus infection. We calculated the odds of furuncles or abscesses relative to the odds of any S aureus infection that did not present as an abscess or furuncle, specifying the comparator group. We used meta-analysis to obtain summary estimates of the odds of infection with a PVL-positive strain separately for S aureus colonisation, pneumonia, musculoskeletal disease, bacteraemia, and abscesses or furuncles and presented fixed-effects or random-effects models, depending on the results of tests for heterogeneity (p<0·05). We excluded studies from reference units and those with a comparator group of fewer than 30 isolates from meta-analysis.

To test the robustness of our estimates we used sensitivity analysis, excluding studies for which comparator isolates were obtained during a different time period or with a different sampling strategy. We identified a subgroup of studies that classified disease isolates and comparator skin and soft-tissue infections as either hospital-acquired or community-onset to assess the effect of selection bias with respect to type of acquisition of infection (appendix p 4 provides discussion of the limitations of this approach).

To investigate whether PVL strains are associated with poor outcome among cases of invasive disease we compared mortality, length of hospital stay, and the proportion requiring surgery for patients infected with PVL-positive and PVL-negative S aureus pneumonia, bacteraemia, musculoskeletal infections, and skin and soft-tissue infections. We calculated the proportion of PVL infections identified as ST8/USA300 to investigate whether this clone confounds the reported association between PVL and invasive disease. We did all analyses with STATA (version 12).

Role of the funding source

The funding source had no role in study design, data collection, data analysis, or data interpretation, or writing of the report. The corresponding author had full access to all data in the study and had final responsibility for the decision to submit for publication.

Results

After exclusion of duplicates, the Medline search identified 504 articles and we identified five more articles through hand searching reference lists (figure 1). We excluded 264 studies on review of the abstract. Of the remaining 245 articles, 76 met our inclusion criteria. The commonest reasons for exclusion were insufficient clinical information to determine prevalence for each disease category (77 studies) and small sample size (57 studies). We included 12 studies of pneumonia, 13 studies of musculoskeletal infections, 19 studies of bacteraemias, and 41 studies of skin and soft-tissue disease from 31 countries (some studies included more than one type of disease).

Figure 1

Study selection

For the association between PVL and pneumonia, data were available for 11 studies (appendix pp 1, 5). We compared odds of infection with a PVL-positive strain between pneumonia and skin and soft-tissue disease for six studies and summarised them with fixed-effects meta-analysis (test for heterogeneity p=0·77; figure 2). Overall, individuals with pneumonia were less likely to be infected with a PVL-positive strain than were those with skin and soft-tissue disease (figure 2). In a subset of four studies with similar sampling strategies for identification of pneumonia and comparator isolates, patients with pneumonia were less likely to be infected with a PVL-positive strain than were individuals with skin and soft-tissue disease (pooled OR 0·28, 95% CI 0·14–0·55). The estimate from the only study directly comparing community-onset pneumonia to community-onset skin and soft-tissue disease (0·26, 0·08–0·82) was much the same as that obtained from the subset of studies, suggesting selection bias relating to infection-onset did not have a major effect on our results.

Figure 2

Fixed-effects meta-analysis comparing the odds of infection with Panton-Valentine leucocidin-producing strain for Staphylococcus aureus pneumonia versus skin and soft-tissue disease

OR=odds ratio

Data were available for 13 studies of musculoskeletal disease (appendix pp 1, 6). We calculated ORs comparing musculoskeletal and skin and soft-tissue infections for six studies and summarised them with random-effects meta-analysis (test for heterogeneity p=0·01; figure 3). Overall, patients with musculoskeletal infections were less likely to be infected with a PVL-positive strain than were individuals with skin and soft-tissue disease (figure 3). These estimates are skewed by data from Senegal, which showed high prevalence of PVL genes among S aureus myositis and osteomyelitis compared with skin and soft-tissue infections (OR 2·37, 95% CI 0·76–7·42). One possible explanation is a difference in the proportion of infections that are routinely sampled in Africa. Exclusion of this study and restriction of analysis to studies that had much the same sampling frame for disease and comparator isolates reduced the odds of infection with a PVL-positive strain for patients with musculoskeletal disease relative to individuals with skin and soft-tissue infection (0·34, 0·20–0·57) and substantially reduced between-study heterogeneity (p=0·43). These estimates might have been subject to selection bias in terms of whether infection was community-acquired or hospital-acquired because the only study directly comparing community-onset musculoskeletal infections to community-onset skin and soft-tissue infections reported a lower OR of 0·20 (0·09–0·45).

Figure 3

Random-effects meta-analysis comparing the odds of infection with a Panton-Valentine leucocidin-producing strain for Staphylococcus aureus musculoskeletal disease versus skin and soft-tissue infection

OR=odds ratio.

19 studies included data for bloodstream infections (appendix pp 1–2, 7–8). We calculated ORs comparing bacteraemias and skin and soft-tissue infections in four studies and summarised them with fixed-effects meta-analysis (test for heterogeneity p=0·65; figure 4). Compared with patients with skin and soft-tissue disease, those with bacteraemia were less likely to be infected with a PVL-positive strain (figure 4). All these studies had a homogeneous study design, obtaining S aureus specimens submitted to one hospital laboratory for both blood and skin and soft-tissue isolates. Only one study recruited community-onset bacteraemias and community-onset skin and soft-tissue infections, reporting a similar estimate to that obtained through meta-analysis (OR 0·16, 95% CI 0·07–0·38), suggesting that selection bias in terms of hospital-onset or community-onset of infection did not affect our estimates.

Figure 4

Fixed-effects meta-analysis comparing the odds of infection with a Panton-Valentine leucocidin-producing strain for Staphylococcus aureus bacteraemia versus skin and soft-tissue disease

OR=odds ratio.

41 studies of skin and soft-tissue infections reported the prevalence of PVL strains among S aureus (appendix pp 2–3, 9–13). We did a fixed-effects meta-analysis for four of five studies that included data on abscesses or furuncles (test for heterogeneity p=0·57; figure 5). We excluded one study because most comparator isolates were derived from bacteraemias in Europe, which are very rarely caused by PVL-positive strains. Overall the odds of PVL-positive infection were ten-times greater in patients with abscesses or furuncles than in those with other presentations of S aureus disease (pooled OR 10·51, 95% CI 7·42–14·88).

Figure 5

Fixed-effects meta-analysis comparing the odds of infection with a Panton-Valentine leucocidin-producing strain for Staphylococcus aureus abscesses or furuncles to all other presentations of S aureus disease

OR=odds ratio.

21 studies reported prevalence of PVL genes among S aureus colonisation (appendix pp 3–4, 14–15). We did random-effects meta-analysis for five studies (test for heterogeneity p<0·0001), having excluded one study because the comparator population was substantially different from all other included studies. Pooled OR was 0·07 (95% CI 0·01–0·31; figure 6), suggesting that PVL strains are rare in colonising isolates compared with isolates causing skin and soft-tissue infections. Although these studies have similar designs, each compares specimens from healthy people to isolates derived from the local hospital laboratory. The comparator group will differ depending on the type of hospital, which might explain some of the heterogeneity between estimates.

Figure 6

Random-effects meta-analysis comparing the odds of colonisation with Panton-Valentine leucocidin-producing Staphylococcus aureus versus Panton-Valentine leucocidin-positive skin and soft-tissue infection

OR=odds ratio.

For the association between PVL and clinical outcome of invasive disease, data were available for seven studies of S aureus pneumonia (table). In adults we identified little evidence that infection with a PVL-positive strain was associated with poor outcome. In the four predominantly small studies with outcome data from Thailand, Australia, the USA, and Singapore, PVL strains were not associated with increased mortality, prolonged hospital stay, or increased disease severity assessed by the Acute Physiology and Chronic Health Evaluation II score for intensive-care mortality. A large case series from France reported a high mortality rate (28 of 50) among cases of PVL-associated pneumonia, but without a comparator population of PVL-negative cases we cannot infer whether PVL contributed to pathogenicity. One study from the USA reported higher rates of surgery in children infected with PVL-positive USA300 than in those with PVL-negative disease (p=0·03), but because most infections were caused by USA300 we cannot determine the effect of PVL independent of USA300.

Table

Outcome studies

	Location	Dates	Study setting	Inclusion criteria	Isolates (n)	Association between PVL and clinical outcome	Evidence for role of PVL in pathogenicity	PVL strains identified as ST8/USA300; n/N (%)
Pneumonia
Carillo-Marquez and colleagues60	USA	2001–09	Texas children's hospital, Houston	Children admitted to hospital diagnosed with SA pneumonia (not ventilated)	110	One patient with PVL died; seven patients developed sequelae; 103 of 117 (88%) deemed cured or improved; 12 of 21 (57%) non-USA300 infections required surgery vs 71 of 89 (80%) for USA300 infections; p=0·03	Uncertain, confounded by USA300	75/82 (92%) USA300 MRSA; 14/28 (50%) USA300 MSSA
Geng and colleagues61	China	2006–08	Eight regional hospitals	Children admitted to hospital with a clinical diagnosis of CA MRSA pneumonia	55	One patient with PVL died; no difference in proportions with necrotising pneumonia according to PVL status	No	NA
Gillet and colleagues24	France	1986–2005	Cases referred to the French staphylococcal reference unit from France (32 cases) and elsewhere	Patients with clinical signs of pneumonia and PVL-SA culture from respiratory tract	50	Mortality in 28 of 50; acute respiratory distress syndrome in 24 of 47; leucopenia and airway haemorrhage associated with mortality	Uncertain, no comparator	NA
Hsu and colleagues43	Singapore	2004	Singapore general hospital	SA specimens from adults	30	Mortality attributable to infection in one of 19 PVL-positive cases and three of 76 PVL-negative cases	No	NA
Nickerson and colleagues45	Thailand	2006–07	Sappasithiprasong hospital	Patients admitted to hospital with pneumonia	11	PVL status did not affect outcome; four of five (80%) mortality for PVL-positive vs of five six (83%) for PVL-negative	No	NA
Peyrani and colleagues62	USA	2008–10	Four academic medical centres in Kentucky, Michigan, Ohio, and Florida	Isolates from patients with hospital-acquired or ventilator-assisted pneumonia	109	Mean severity score (APACHE II) similar in PVL-positive and PVL-negative cases; no difference in 28 day mortality (eight of 25 [32%] PVL-positive vs 24 of 63 [38%] PVL-negative, p=0·59) or mean length of hospital stay (PVL-positive 20·4 days [SD 8·1] vs PVL-negative 21·0 days [7·6], p=0·74)	No	29/29 USA300 (100%)
Wehrhahn and colleagues46	Australia		Royal Perth hospital and Fremantle hospital	Patients admitted to hospital with community-onset pneumonia	22	Median hospital stay 16 days for PVL-positive and PVL-negative cases, p=0·70; 30 day mortality two of five (40%) for PVL-positive and six of 17 (35%) for PVL-negative cases, p=1·00	No	ST8-MRSA-IV less than 5%*
Bone, muscle, and joint infections
Abdel-Haq and colleagues50	USA	2006–07	Michigan hospital, Detroit	Children admitted to hospital with bone and joint infections	22	17 of 20 PVL-positive cases and two of two PVL-negative cases needed surgery	No	20/20 (100%) USA300 MRSA
Dailiana and colleagues63	Greece	2003–06	Thessalia hospital	Consecutive patients with purulent musculoskeletal infections affecting the arm and shoulder attending hospital	81	32 of 70 (46%) PVL-positive cases admitted to hospital vs 78 of 157 (50%) PVL-negative cases	No	0
Dohin and colleages32	France	2001–05	Edouard Herriot hospital, Lyon national reference unit	Cases: children with CA PVL-positive SA bone and joint infections identified from the national register; controls: random selection of children with PVL-negative SA bone and joint infections	14	Median hospital stay of 45 days for PVL-positive cases vs 13 days for PVL-negative cases. Median follow-up till total cure 15·5 months (IQR 6–60) for cases vs 3 months (1–79) for controls	Yes	NA
Martinez-Aguilar and colleagues64	USA	2000–02	Texas children's hospital, Houston	Patients with osteomyelitis, septic arthritis, or pyomyositis	56	Ten of 33 (30%) PVL-positive cases had complications such as chronic osteomyelitis and deep-vein thrombosis vs none of 23 PVL-negative cases, p=0·002	Yes	NA
Pannaraj and colleagues31	USA	2000–05	Texas children's hospital, Houston	Children admitted to hospital with pyomyositis or myositis	24	Muscle drainage required for 14 of 17 (82%) PVL-positive cases and two of seven (29%) PVL-negative cases, p=0·02	Yes	13/13 (100) USA300 MRSA; 1/9 (11) USA300 MSSA
Bacteraemias
Seybold and colleagues19	USA	2004	Grady Memorial hospital, Atlanta	Clinical consecutive MRSA BSI isolates	116	Crude in-hospital mortality three of 39 with PVL-positive USA300 vs 22 of 77 for PVL-negative strains	No	39/39 (100%) USA300
Wehrhahn and colleagues46	Australia	Not provided	Royal Perth hospital and Fremantle hospital	Consecutive patients with community-onset BSI	66	30 day mortality one of ten (10%) for PVL-positive cases vs nine of 56 (16%) for PVL-negative cases, p=0·62	No	ST8-MRSA-IV less than 5%†
Skin and soft-tissue infections
Gubbay and colleagues65	Australia	2001–02	Westmead hospital, Sydney	Children attending hospital with gentamicin-sensitive MRSA infection	59	PVL not associated with need to admit to hospital; 27 of 68 (40%) PVL-positive cases needed surgery vs four of 32 (13%) PVL-negative cases	Yes	0
Jahamy and colleagues66	USA	2005–06	St John hospital, Michigan	Isolates from inpatients with soft-tissue foci of infection	194	78 of 96 (81%) PVL-positive cases needed surgery vs 52 of 98 (53%) PVL-negative cases, p<0·001	Yes	NA
Kaltsas and colleagues51	USA	Not provided	Montefiore medical centre	Isolates from patients with skin or soft-tissue infection	101	Longer median length of stay for PVL-negative cases than for PVL-positive cases; eight of 44 (18%) PVL-negative cases developed osteomyelitis vs three of 53 (6%) PVL-positive cases; 40 of 54 (74%) PVL-positive cases needed surgery vs 28 of 47 (60%) PVL-negative cases; in multivariate analysis PVL-positive MSSA four times more likely to need surgery (OR 4·7, 95% CI 1·14–19·7)	Yes	7/10 (70) USA300 MRSA
Kanerva and colleagues67	Finland	2004–06	Reference laboratory, Helsinki	Epidemiologically defined clinical CA-MRSA isolates referred to the reference laboratory	156	41 of 72 (57%) PVL cases needed surgery vs 27 of 84 (32%) for PVL-negative, p=0·003; 11 of 82 (13%) PVL-positive cases admitted to hospital vs 19 of 109 (17%) PVL-negative cases; patients infected with PVL-positive strains more likely to develop disease (82 of 90 vs 109 of 208, p<0·01)‡	Yes	33/90 (37%) ST8 MRSA-IV3
Munckhof and colleagues68	Australia	2004–05	Eight hospitals in southwest Queensland	Consecutive clinical isolates of non-multiresistant MRSA, multiresistant MRSA, and MSSA from patients attending hospital	384	SA was nine times more likely to be reason for admission when associated with PVL (OR 9·4, p<0·001); PVL positive cases more likely to need surgery (OR 8·2, p<0·001)	Yes	NA
Muttaiyah and colleagues44	New Zealand	2008	Auckland hospital	Clinical MSSA isolates	234	PVL-positive MSSA cases were 7·4-times more likely to need surgery (95% CI 4·1–13·3)	Yes	NA
Tong and colleagues53	Australia	2006	Royal Darwin hospital, Northern Territory	SA isolates defined as non-multiresistant MRSA, MRSA, and MSSA	419	MRSA: compared with PVL-negative cases, PVL-positive cases were more likely to need surgery (OR 5·0, p<0·001) and had shorter median length of stay (2·5 days vs 5 days, p=0·001); MSSA: PVL-positive cases had shorter length of stay (2 days vs 4 days, p<0·001); PVL-positive cases had increased risk of surgery (OR 6·8, p<0·001)	Yes	NA

PVL=Panton-Valentine leucocidin. ST8=sequence type 8. SA=Staphylococcus aureus. MRSA=meticillin-resistant SA. MSSA=meticillin-sensitive SA. NA= not available. CA=community-acquired. APACHE=Acute Physiology and Chronic Health Evaluation. BSI=bloodstream infection. OR=odds ratio.

For entire study, data unavailable for pneumonia alone.

For entire study, data unavailable for bacteraemia alone.

For entire study, data unavailable for skin and soft-issue infections alone.

Four of five studies of musculoskeletal infection that included outcome data were done in children (table). Compared with children infected with PVL-negative strains, those with PVL-positive musculoskeletal infection reported higher rates of surgical drainage (14 of 17 PVL-positive vs two of seven PVL-negative, p=0·02), complications such as chronic osteomyelitis and deep vein thrombosis (ten of 33 PVL-positive cases vs none of 23 PVL-negative cases, p=0·002), and lengthier hospital stay (median 45 days vs 13 days, p<0·001). Two of these studies included molecular typing and all PVL-positive MRSA isolates were USA300. Conversely one small study reported similar rates of surgery among children infected with S aureus, irrespective of whether the strain encoded the PVL gene. In adults the one study with outcome data reported much the same rates of hospitalisation among individuals with musculoskeletal infection irrespective of whether they were infected with PVL-positive or PVL-negative S aureus (46% [32 of 70] vs 50% [78 of 157]).

We did not identify evidence to support a link between PVL and poor outcome with bacteraemia on the basis of data from two studies from Australia and the USA (table). Compared with patients with PVL-negative bacteraemias, those infected with PVL-positive strains had similar 30 day mortality (16% [nine of 56] vs 10% [one of ten], p=0·62), and reduced crude in-hospital mortality (29% [22 of 77] vs 8% [three of 39]), although this result might be indicative of underlying population differences in terms of age and comorbidities, because elderly people are more likely to be infected with PVL-negative MRSA than are younger adults.

All seven outcome studies provided good evidence that compared with individuals with PVL-negative skin and soft-tissue infections, those infected with a PVL-positive strain were more likely to require surgery. This association was independent of meticillin resistance and USA300 (table). We identified some evidence that compared with patients with PVL-negative disease, patients with PVL-positive disease had a shorter hospital stay, both for individuals infected with PVL-MRSA (median 2·5 days vs 5 days, p=0·001) and PVL-MSSA strains (median 2 days vs 4 days, p<0·001), although these analyses did not adjust for potential confounding factors such as age.

Discussion

PVL is strongly associated with skin and soft-tissue disease and is comparatively less common in colonisation and in other forms of invasive disease, such as pneumonia, musculoskeletal disease, and bacteraemia. Results from observational studies suggest that infection with a PVL-positive strain does not predict poor clinical outcome for staphylococcal pneumonia, musculoskeletal disease, or bacteraemia in adults, but patients with PVL-positive skin and soft-tissue disease are more likely to require surgical treatment (appendix p 4 provides a discussion of limitations of methods used in these studies). In children with musculoskeletal disease we identified some evidence that infection with a PVL-positive strain is associated with surgical treatment, lengthy hospital stay, and chronic osteomyelitis. As far as we are aware, this is the first review to systematically assess the association between PVL genes, colonisation, and S aureus disease and to investigate whether PVL predicts clinical outcome.

Although strains carrying PVL genes are commonly identified in invasive staphylococcal disease, direct evidence increasingly suggests that PVL is not the main determinant of severity or outcome. Clinical studies from Australia have directly compared outcome for PVL-positive and PVL-negative strains for a range of invasive infections, consistently reporting outcome to be independent of PVL, with the exception of increased surgical treatment for skin and soft-tissue infections. Results of two studies of clinical outcome among patients with hospital-acquired pneumonia showed similar clinical outcome and mortality among patients irrespective of PVL, even after adjustment for potential confounders. In-vitro studies have failed to correlate the amount of PVL toxin produced by different S aureus strains with the severity of clinical disease, and experimental studies of animal models of skin and soft-tissue infection and pneumonia have not convincingly shown PVL to have an effect on the development of disease independent of bacterial strain. Much of the work linking PVL genes to invasive staphylococcal disease is derived from the USA where most community-onset staphylococcal disease and even some hospital-acquired strains are meticillin-resistant USA300. This situation contrasts with Europe where USA300 is rare and a high proportion of PVL-positive strains are meticillin-sensitive, and Australia where half of community-MRSA clones are PVL-positive, and MSSA strains contribute a large burden of PVL-positive disease. The epidemiological association between invasive disease and PVL genes might be confounded by USA300, and this potential confounding effect has become more apparent as investigators have examined the role of PVL in settings where USA300 is not common.

We report that PVL-producing strains are mainly associated with disease rather than colonisation. Because S aureus colonises a third of the healthy population, high rates of colonisation with PVL-producing strains might be expected in regions where disease is prevalent. This theory is not borne out by data from the USA, where community-acquired PVL-positive MRSA infection is common but most individuals are colonised with PVL-negative MSSA strains. This finding suggests that disease can occur in the absence of nasal colonisation, and evidence increasingly suggests that non-nasal sites such as the throat and inguinal areas might be important colonisation sites for PVL-positive S aureus. Results of a large study of S aureus transmission in the household suggest interplay between colonisation and disease-causing isolates is complex and might depend on bacterial strain. More than a quarter of 350 infected index case patients in this study were colonised with a strain type discordant from their infecting isolate and only 12% were colonised with a concordant strain. The relation between bacterial strain and disease potential was further investigated in a study from China, in which investigators sequenced carriage and disease isolates from children, concluding that disease potential of a strain carrying PVL genes depended partly on the bacterial genetic background. Screening of patients for PVL carriage might not accurately assess their risk of disease, and if colonising and disease-causing isolates are systematically different, eradication of a strain type rarely associated with disease could render a patient vulnerable to recolonisation with a more pathogenic strain.

Evidence that PVL is associated with skin and soft-tissue infections is strong and independent of strain type among both MRSA and MSSA strains. Compared with PVL-negative strains, PVL-positive strains are more likely to be truly community-acquired, occurring in individuals who have not had contact with health care. When MRSA first emerged in the community in the early 1990s, an epidemiological definition of community-acquisition was developed on the basis of risk factors for health-care use to distinguish between hospital-acquired and community-acquired MRSA. Community strains then started to enter hospitals in the USA, blurring the distinction between hospital and community-acquisition and limiting the usefulness of a definition based on health-care contact. Genotypic definitions focusing on molecular markers such as PVL, or susceptibility to specific antibiotics can imperfectly discriminate between health-care or community-acquisition in a local setting, but DNA sequencing techniques are required to reliably compare strains between regions or laboratories. Increasingly, so-called community strains cause a spectrum of nosocomial disease and genotypic definitions cannot reliably predict clinical presentation. In a retrospective study of 160 patients infected with USA300 about 20% had invasive disease, mainly among those with health-care-associated infection. Such patients were older and had more comorbidities compared with those with community-acquired S aureus, suggesting characteristics of the patients rather than the bacterial strain are a major determinant of disease presentation and outcome.

Individuals with PVL-positive skin and soft-tissue infections are more likely to require surgery than are those with PVL-negative infection, but we identified little data suggesting that they have a worse clinical outcome, provided that they receive appropriate surgical treatment and the correct antibiotics, although this finding might simply be indicative of the younger age and scarcity of comorbidities among this group. Patients with PVL-positive skin and soft-tissue infections might have more recurrent infections than those with PVL-negative disease, but most studies do not collect detailed data for outcomes. A clinical syndrome is clearly associated with PVL-positive skin and soft-tissue infections, but whether this syndrome warrants specific clinical and public health action is uncertain.

The decision to test for the PVL toxin is triggered by the presence of either invasive or recurrent (mainly skin) disease, as an instrument to guide clinical and public health measures. If PVL is assumed to be rare and mainly associated with invasive disease then aggressive antibiotic treatment combined with a low threshold for surgical intervention and a so-called search and destroy policy around the patient with screening and decolonisation of close contacts is potentially justified. Conversely if infection with a PVL strain is not exclusively associated with invasive disease and does not predict poor clinical outcome then aggressive clinical and public health actions might do more harm than good. Decolonisation has been shown to improve clinical outcome in high-risk settings such as renal units, intensive care, and surgery, but we are unaware of any data for the clinical or cost effectiveness of community-based screening and decolonisation for PVL strains.

The approach to management and control of PVL-positive disease contrasts between the USA and Canada and the UK. In the USA, where PVL-positive MRSA is common, toxin testing is not routinely recommended. Patients with skin and soft-tissue infections are advised on simple hygiene measures to restrict the spread of infection and public health measures, such as contact screening and decolonisation, are only required when a defined outbreak occurs outside the household. Much the same guidance is in place in Canada. In England where a high proportion of PVL cases are associated with MSSA strains, PVL toxin testing is recommended for any patient who presents either with community-associated invasive disease or with recurrent skin and soft-tissue infection caused by S aureus, such as boils, abscesses, or eyelid infections. All patients with recurrent infection or community-onset invasive disease should be tested for PVL, and when positive, be offered decolonisation after appropriate treatment, irrespective of clinical presentation. Household contacts might be offered screening or decolonisation, or both, after risk assessment. This policy is based on the assumption that decolonisation can reduce risk of further disease, but because PVL strains are common and usually not highly pathogenic this policy might not be justified. Moreover we identified little evidence that decolonisation is effective for individuals with PVL-positive disease, and treatment could risk elimination of a colonising strain that is rarely associated with disease.

Estimation of the burden of PVL-associated disease in the community is challenging because surveillance systems rely on specimens sent to reference units and mainly focus on MRSA, and research studies predominantly analyse retrospective collections of clinical isolates. Both methods are biased towards severe disease because minor skin and soft-tissue infections are not sampled in routine clinical practice and a high proportion of PVL strains are meticillin-sensitive. These methods tend to provide a false impression of the prevalence of severe disease because they represent the so-called tip of the clinical iceberg of S aureus infection (figure 7). Without population-based studies we cannot know what absolute proportion of PVL-positive specimens are associated with colonisation and mild, moderate, and invasive disease. More research is needed to test the hypothesis that PVL strains are also associated with minor skin and soft-tissue infections that are undiagnosed and treated uneventfully in the community without recourse to secondary care. If this hypothesis is true, it strongly questions the need for a search and destroy policy around individuals who are infected with PVL-positive strains.

Figure 7

Clinical iceberg of Panton-Valentine leucocidin-associated infection

More research is needed to understand the interplay between host factors, colonisation, bacterial virulence, and PVL to identify individuals at risk of invasive disease with a poor prognosis. This research in turn should guide a proportionate and evidence-based approach to clinical and public health control measures.