Indirect fluorescence antibody test (IFA).

Two studies examined the diagnostic efficacy of IFA. The first, by Chamberlin et al [19], tested 33 cases (blood culture or smear positive) and 101 controls and found a sensitivity of 85% and a specificity of 92%. Four patients infected with other pathogens were tested for cross-reactivity (two with syphilis, two with cat-scratch disease, caused by Bartonella henselae) and one of each tested positive.

The second, by Knobloch et al [20], looked at 187 sera obtained from patients resident in an endemic area of the Northern district of Cajamarca and showed a much lower sensitivity of 45% when compared to ELISA as a reference standard.

PCR.

The only study found looking at PCR which used a reference standard was by Padilla et al. [21] All 10 smear-positive blood samples tested positive for B. bacilliformis and 5 additional malaria positive samples were correctly identified as being negative for Bartonella.

ELISA.

The larger of two ELISA evaluation studies [22] compared 27 lab-confirmed cases (by blood culture, smear or PCR) to 40 healthy controls who were not from, and had never visited endemic areas; and 10 sera of patients known to have different infections. Sensitivity using IgM was superior to IgG ELISA (85% vs 70%); specificity was 100% for both. One patient with Salmonella infection tested positive but none of the other pathogens cross-reacted with the test.

Eight of nine smear-positive patients tested by Maguiña [23] between 1969 and 1992 had positive ELISA.

Sonicated immunoblot.

Mallqui et al. [24] tested this technique on 42 confirmed Carrion's disease patients (blood smear or biopsy positive). Two methods of antigen preparation were tested; sonicated and glycine. For chronic disease the sensitivity was 94% using both, whereas for acute cases, sensitivity was 70% for sonicated and 30% for glycine. The specificity was quoted as 100% for both when sera of healthy volunteers were tested with immunoblot compared to blood smear. However, when testing sera known to be positive for brucellosis, C. psittaci and Coxiella burnettii, 34%, 5%, and 29% of the sera respectively cross-reacted and gave a false positive result.

Thin blood smear.

Despite being the oldest and most widely employed method of diagnosis of B. bacilliformis, only two published studies have looked at the sensitivity and specificity of this test. Ellis et al. [25] found that out of 11 PCR positive acute cases tested during the 1998 outbreak in the Urubamba region, only 4 were thin smear positive, giving a sensitivity of 36%, and a positive predictive value of 44%. A specificity of 96% was obtained as 125 individuals were thin smear negative out of 130 PCR negative individuals.

Pachas et al. [26] found that of 352 confirmed acute and chronic cases in Caraz (Ancash, Peru), 24% were thin smear positive.

Western blot.

Two studies by Maguiña et al. have looked at Western blot compared to reference standard. One [27] used a ‘clinical expert’ to make a clinical diagnosis in 11 chronic patients and the other [23] tested 9 culture-positive patients with acute disease. All of these tested positive for B. bacilliformis with Western blot.

Ciprofloxacin.

Studies of quinolone therapy have reached divergent conclusions.

Minnick [28] and Sobraques et al. [29] in separate studies published in 2008, found ciprofloxacin to have a minimum inhibitory concentration (MIC) of 0.25 but disagreed on whether this was high or low, with Sobraques [29] concluding that the organism was highly susceptible to ciprofloxacin and Minnick [28] discouraging its use, a suggestion purportedly later substantiated by the description of constitutive mutations in the quinolone-resistance-determining region of gyrase (gyrA) by Minnick [28] and Angelakis [30].

Biswas [31] and del Valle [32] found that B. bacilliformis can quickly become resistant to ciprofloxacin in vitro and therefore concluded that its use should be discouraged in the treatment of Carrion's Disease, whilst Rolain [33] concluded that there was ‘moderate evidence for the use of ciprofloxacin from opinions of respected authorities.’ This was based on unpublished reports.

Chloramphenicol.

Three in vivo studies have been carried out looking at the effectiveness of chloramphenicol, the first of which was carried out by Urteaga [34] in the 1950s and involved 19 cases. Although the results reported that the majority, 79% (15), made a rapid recovery, the study lacked details of sampling methods and doses used.

Chloramphenicol was the most used antibiotic in the acute phase in a cohort of 518 patients studied by Arroyo in Caraz [35]. The treatment resulted in clinical cure in 89%. However, none of the patients received the recommended loading dose of 50 mg/kg and 66.1% required a higher maintenance dose of 25 mg/kg and indeed 39.1% required a prolonged treatment course greater than 14 days.

A similar study by Maguiña [23] carried out in Lima found that 95.4% of acute patients who had received chloramphenicol, either alone or with another antibiotic, responded well though disaggregated data for mono- and poly-microbial treatment was not presented.

Rolain et al. [33] classify chloramphenicol as having good evidence for use from one or more well-designed clinical trials, though this is based on 2 observational studies [10], [23] which lacked in depth analysis.

Large blood transfusions.

One historical case-series by Hodgson [36] looked at the treatment of 2 acute cases with large blood transfusions. The patients survived but took 32 and 45 days to recover as well as 3.4 L and 8.15 L of blood respectively.

Rifampicin and streptomycin.

Arroyo [35], in his study of 518 cases, found that rifampicin was the most popular treatment for chronic disease leading to clinical cure in 93.1% of patients, though a large proportion (82%) required a prolonged treatment course of greater than 21 days.

Rolain [33] reported rifampicin as having ‘good evidence for use from one or more well-designed clinical trials’; this conclusion is based on one observational study carried out by Maguiña et al. [23] between 1969 and 1992 which found that 80% (37 of 46) of patients treated with rifampicin had a good response compared with 56% (5 of 9) who received streptomycin. From this data Rolain et al. maintained that the strength of evidence for streptomycin was the same as for rifampicin.

Rifampicin and azithromycin.

In 2003, a group of researchers from Caraz carried out a study [37] (as yet unpublished) comparing azithromycin, the current 1^st-line treatment, to rifampicin in 127 cases. The definition of cure used was reversion to negative blood cultures. Both antibiotics were equally efficacious and achieved a similar ‘cure time’ of 3 to 4 weeks.

Sultamicillin and deflazacort.

A single case report [38] was published after the successful treatment of a 12-year old girl with chronic verruga with 21 days of treatment (25 mg/kg of sultamicillin, 0.7 mg/kg of deflazacort).

Vector species.

A number of studies have attempted to implicate the Lutzomyia sandfly species in the transmission of Carrion's disease in different areas; most of them have involved CDC light trap collections during outbreaks. However, since few of them have identified B. Bacilliformis in the insect, the majority only present circumstantial evidence that these sandfly species could be the responsible vectors.

L. peruensis and, in smaller quantities, L. pescei were the only species found in a study carried out during an outbreak in Cusco [39]. Similar findings were reported by Ellis [25] who found only L. peruensis during an outbreak in the Cuzco region. In this second study, B. bacilliformis was identified using PCR in 2% of specimens collected (n = 2).

A further study carried out by Caceres [40] found a great abundance of Lu. robusta and Lu. maranonensis in intradomiciliary areas in the provinces of Jaen, San Ignacio and Utcubamba, however their possible role as vectors was not confirmed by PCR.

A different species, L. serrana was found to make up 93.6% of vectors collected in indoor CDC light traps and outdoor Shannon traps during an outbreak in Huamalíes, Huánuco in the high jungle area of Peru. However, once again carriage of B. bacilliformis by the sandflies could not be confirmed [41].

Vector characteristics.

One study looking at vectorial characteristics of L. verrucarum tried to ascertain whether vertical transmission of B. bacilliformis occurred between adults and their offspring [42]. The investigators were unable to make firm conclusions regarding this, but they did find that the lifespan and degree of oviposition in sandflies infected with low bacteraemic blood(3%) was much longer and higher respectively compared with those fed on high bacteraemic (80%) blood (p<0.001).

Control.

Studies looking at the control of Lutzomyia are scarce and none relate Lutzomyia control to Carrion's disease incidence. The first study was carried out in 1945 by Hertig and Fairchild [43] in the Rimac Valley, about 40 miles from Lima, where L. verrucarum and L. peruensis are abundant. Treatment of stone walls with DDT produced a marked reduction of sandflies, not quantified in the report, and treatment of stone walls combined with house spraying reduced sandflies to an ‘extremely low level.’

The only other study found was a report of the control programme in Caraz which took place between 2004 and 2008. The initiative involved indoor house spraying with residual insecticides [44]. The main conclusions drawn from this unpublished report were that intra-domiciliary sandfly populations remained remarkably low for over a year after spraying. This was unexpected as these insecticides were marketed to have an efficacy of 6 months. It is proposed that pyrethroids have additional irritant and repellent effects on the vectors that last longer than their known insecticide properties.

Incubation period.

In 1947 Ricketts [45] studied a group of patients who had recently visited endemic areas and proposed that their incubation periods, as determined by culture positivity, ranged from 20 to 100 days.

Immunisation.

In 1943 Howe and his team [46] carried out a case control study using military staff posted out to verruga zones to test a crude vaccine made of 4 inactivated strains of B. bacilliformis. Out of 22 vaccinated men, 55% had developed positive blood cultures by the end of their duty period. This was in contrast to the control group, 90% (9 of 10) of which developed positive cultures over a similar period of time.

Asymptomatic infection.

In 1997, Chamberlin [12] and her team in Caraz carried out a prospective cohort study which firstly used PCR to detect the number of asymptomatic bacteraemics in an endemic population; 0.5% of 555 were PCR positive. The study population was also tested for past infection using IFA and 45% had positive antibodies. The investigators then followed an abacteraemic cohort for 25 months monitoring them for clinical features of Carrion's disease. 127 cases were diagnosed in 25 months (12.7 per 100 person years) and cases were clustered, with 70% occurring in 18% of households.

Herrer, in the only other published epidemiological study done in 1953 [47], measured serial culture positivity in a group of students who had recently come to live in an endemic area. He found that 29% (4 of 14) had asymptomatic infection but that 2 failed to produce positive cultures on repeated testing. He concluded that this transient bacteraemia makes it less likely that asymptomatic carriers are a reservoir for infection.

The effect of ‘El Niño’.

Two studies have looked specifically at the effect of the ‘El Niño’ phenomenon on the number of cases of bartonellosis. One study [48] compared numbers of cases between pre-, post- and 1997–1998 ‘El Niño’ periods in the Cusco and Ancash areas and found that sea surface temperature correlated significantly with the number of cases of Carrion's disease.

The other study, by Chinga-Alayo et al. [13] described similar findings during the 1983–1988 event in Ancash with an almost 4-fold increase in monthly cases.

Reservoirs.

Due to the sporadic nature of Carrion's disease outbreaks, various candidate reservoirs have been postulated in a number of studies. Herrer [49], in 1953, proposed that euphorb plants, which are found in great abundance in verruga zones and yield a milky latex, may be a natural reservoir. However, in his study he failed to recover B. bacilliformis from the plants, infect seedlings or grow the microorganism in the presence of latex dilutions.

Birtles [50] investigated the possibility of intradomicilliary animals being reservoirs by testing 50 animals from the homes of 11 children who had recently had the illness. Bartonella-like organisms were isolated from 4 out of 9 small non-domesticated rodents trapped inside the houses but they were not confirmed to be B. bacilliformis.

Cooper et al. carried out two case-control studies [51], [52] in Zamora Chinchipe province, Ecuador, using questionnaires to determine whether there was an excess of dead or dying animals in case households compared to controls. Case households reported seeing significantly more dead or dying rodents [51] and chickens [52] than controls.

Acute disease.

Only studies looking at quinolones and chloramphenicol were found in the literature search but none comparing the two.

There has been much discussion recently about the use of quinolones in the management of acute Carrion's disease due to the publication of a number of experimental studies suggesting constitutive expression of gyrA mutations which confer quinolone resistance. However, in vivo studies looking at this class of antimicrobials are lacking and therefore it is difficult to draw meaningful conclusions from these findings.

Although these are in vitro studies, and therefore evidence level 5, some go on to claim that ciprofloxacin is inadequate [28], [30], [32] and should be removed from the current guidelines. One of these studies [32] goes on to give an example of an in vivo case of ciprofloxacin failure who died, however this patient was extremely sick on presentation and did not recover despite simultaneously receiving ciprofloxacin and ceftriaxone iv [53]. There is also some discrepancy between studies in acceptable MICs for antibiotics with some claiming that 0.25 makes ciprofloxacin inadequate [30] and others claiming that a MIC of 0.25 makes it highly effective [29].

The evidence behind the use of chloramphenicol is only slightly more established, with two level 4 observational studies [34], [35] claiming a 79% and 89% recovery rate respectively and a minireview [33] indicating that there is good evidence for its use based on two small observational studies.

Chronic disease.

The unpublished results of the study carried out by a team of Caraz researchers led by Jesus Gonzalez [37] are very compelling as this is the first study to compare two antibiotics and the only study to use microbiological cure as the end-point. Its results, which show that a weekly 1 g azithromycin dose is just as effective as daily rifampicin, has important implications. Not only could the weekly dose improve adherence but it would avoid the indiscriminant use of an anti-tuberculous drug in an area where multi-drug resistant (MDR) TB is a problem.

Rifampicin is now the most widely used antibiotic in the chronic phase and most patients have been found to respond well; 80% and 93.1% respectively in two studies one of which was a minireview and therefore level 3a evidence [23], [35].

Studies looking at less-conventional treatments; large blood transfusions for acute disease [36], and sultamicillin and deflazacort for verruga peruana [38], involved only two and one patient respectively and therefore their use in practice cannot be determined.