Determinants of relapse periodicity in Plasmodium vivax malaria
© White; licensee BioMed Central Ltd. 2011
Received: 3 August 2011
Accepted: 11 October 2011
Published: 11 October 2011
Skip to main content
© White; licensee BioMed Central Ltd. 2011
Received: 3 August 2011
Accepted: 11 October 2011
Published: 11 October 2011
Plasmodium vivax is a major cause of febrile illness in endemic areas of Asia, Central and South America, and the horn of Africa. Plasmodium vivax infections are characterized by relapses of malaria arising from persistent liver stages of the parasite (hypnozoites) which can be prevented only by 8-aminoquinoline anti-malarials. Tropical P. vivax relapses at three week intervals if rapidly eliminated anti-malarials are given for treatment, whereas in temperate regions and parts of the sub-tropics P. vivax infections are characterized either by a long incubation or a long-latency period between illness and relapse - in both cases approximating 8-10 months. The epidemiology of the different relapse phenotypes has not been defined adequately despite obvious relevance to malaria control and elimination. The number of sporozoites inoculated by the anopheline mosquito is an important determinant of both the timing and the number of relapses. The intervals between relapses display a remarkable periodicity which has not been explained. Evidence is presented that the proportion of patients who have successive relapses is relatively constant and that the factor which activates hypnozoites and leads to regular interval relapse in vivax malaria is the systemic febrile illness itself. It is proposed that in endemic areas a large proportion of the population harbours latent hypnozoites which can be activated by a systemic illness such as vivax or falciparum malaria. This explains the high rates of vivax following falciparum malaria, the high proportion of heterologous genotypes in relapses, the higher rates of relapse in people living in endemic areas compared with artificial infection studies, and, by facilitating recombination between different genotypes, contributes to P. vivax genetic diversity particularly in low transmission settings. Long-latency P. vivax phenotypes may be more widespread and more prevalent than currently thought. These observations have important implications for the assessment of radical treatment efficacy and for malaria control and elimination.
In endemic areas of Asia, Oceania, Central and South America, and in the horn of Africa Plasmodium vivax malaria is a major cause of morbidity. It is an important contributor to early pregnancy loss and reduced birth weight which increases mortality in infancy [1, 2]. Plasmodium vivax is a sophisticated and resilient malaria parasite which was once prevalent over much of the inhabited world. It has receded from North America, Europe and Russia, but in the tropics vivax malaria remains a major cause of childhood illness. In most endemic areas, P. vivax cohabits with Plasmodium falciparum. Mixed infections with the two species are common. P. vivax is more difficult to control and eliminate than P. falciparum because of its tendency to relapse after resolution of the primary infection. In endemic areas relapse of vivax malaria is a major cause of malaria in young children, and an important source of malaria transmission. Relapse also occurs in Plasmodium ovale infections and in several of the simian malarias, notably Plasmodium cynomolgi, which has often been used as an animal model of vivax malaria. The factors which control relapse and determine their remarkable periodicity are not known.
During the malaria therapy experience it became clear that both the incubation period and the number of relapses were determined by the numbers of sporozoites inoculated [25–28, 38]. The Dutch had shown first , and others later confirmed, that low sporozoite inocula often resulted in an extended incubation period of 7-10 months. The more sporozoites that were inoculated, the more likely was an early infection (incubation period two weeks), and the more relapses that followed - provided that prompt anti-malarial treatment (quinine) was given each time. Later clinical and experimental studies, reported after the Second World War, were to confirm these observations. In order to ensure that there was a short incubation period preceding the malaria illness malaria therapy infections were produced typically by the bites of 5-10 infected mosquitoes, and either no treatment or partial suppressive treatment was given. Previous theories that seasonal influences were important determinants of relapse were largely rejected as the intervals from primary illness to relapse in neurosyphilis patients were generally similar whichever month the infection started [22–26].
Between the 1920s to the 1940s the long-latency infection was regarded as the "usual" P. vivax phenotype. Throughout the endemic areas of Europe vivax malaria peaked in the late spring and early summer (largely from inoculations the previous year) [20, 25]. In southern Europe there was often a bimodal pattern with a late summer peak of falciparum malaria . The epidemiological studies of malaria epidemics in Sind (now Pakistan) and Ceylon (Sri Lanka) followed a similar pattern [19, 25, 39]. This suggested that long-latency P. vivax was also responsible for the peak of vivax malaria cases which occurred the year after the falciparum malaria epidemics in these two tropical areas (although other interpretations are also possible). During the Second World War observations on Allied soldiers fighting in North Africa, Italy, the Caucasus, and Greece, and further observations in Japanese occupation forces in China, all pointed clearly to long-latency P. vivax with similar illness patterns to the Madagascar and St Elizabeth strains [40–45]. In contrast the soldiers on both sides fighting in the Indo-Burman and South Pacific campaigns encountered vivax malaria with a very different relapse pattern. Relapses were frequent and the relapse rate was very high -in some companies all soldiers were infected and all relapsed. Infections occurred at three week intervals if quinine was given, and seven week intervals following mepacrine treatment [42–49]. Multiple relapses were very common and there was no evidence of long latency. The "type strain" for this tropical frequent relapse P. vivax phenotype was the "Chesson strain" isolated from a soldier of that name who had acquired the infection in New Guinea [50–52]. In volunteers infected with the Chesson strain 80% of relapses occurred within 30 days of initial treatment with quinine.
In 1902, three years before his death, the eminent protozoologist Fritz Schaudinn reported that he had observed direct infection of erythrocytes by sporozoites . There was therefore no need to postulate a tissue stage of malaria. By the 1930s this theory was largely discredited as others had tried and failed to reproduce the observations, and by then tissue stages of bird malarias had been demonstrated unequivocally. The existence of a tissue stage of the malaria life cycle in humans was considered sufficiently likely that the Malaria Commission of the League of Nations in 1933 suggested that the sporozoite in human malaria went on to divide in cells of the endothelial system as did Haemoproteus in birds. In 1937 James and Tate showed that the exoerythrocytic development of Plasmodium.gallinaceum took place in the brain capillaries of chickens . After the Second World War, Fairley's brilliant work at the Cairns experimental station showed that sporozoites were cleared from the blood within one hour of mosquito feeding on volunteers, and in the case of P. vivax, parasites only returned to the blood one week later [55–57]. Several researchers had noted earlier that in the latent or inter-relapse period even transfusion of as much as 500 mL blood could not transmit P. vivax to a volunteer recipient, whereas if P. falciparum recurred -it could always be transmitted beforehand by blood transfusion. It was clear then that there was a pre-erythrocytic tissue stage which preceded the blood stage infection, and also that subsequent relapses originated from an exoerythrocytic stage - but where was it? SP James, the eminent British malariologist, was so convinced that there must be an exoerythrocytic stage in the primate malarias that he told the young PCC Garnham to stay in East Africa until he had found it -and so he did! In 1947, Garnham identified the pre-erythrocytic development of Hepatocystis (then Plasmodium) kochi in the hepatocytes of African monkeys . Shortly afterwards in England definitive studies by Shortt and Garnham identified the site of pre-erythrocytic development in primate malarias as the liver, first in P.cynomolgi infected Rhesus monkeys [59, 60], and then in a heroic experiment with P. vivax in a very heavily infected volunteer who underwent open liver biopsy [61–63]. This classic work still did not identify the persistent stage, although later primate work suggested that relapses might arise from arrested development of hepatic pre-erythrocytic schizonts [63, 64]. Forty years later Krotoski, working with Garnham and colleagues at Imperial College, finally identified the dormant stages or "hypnozoites" of P.cynomolgi and P. vivax responsible for relapses in the liver [65–69]. Although parasite bodies, which are probably hypnozoites, have since been demonstrated in liver cell cultures  remarkably little is known of their biology. The term relapse is now used specifically to describe recurrences of malaria derived from persistent liver stages of the parasite (hypnozoites) whereas recrudescence refers to a recurrence of malaria derived from persistence of the blood stage infection. The relapse arises after the "awakening" of these hypnozoites and the subsequent intrahepatic schizogony followed by blood stage multiplication. The question remained unanswered as to how the hypnozoites were woken, and what determined their remarkable periodicity.
Following the Second World War artificial infection studies were conducted in several locations to study different anti-malarial treatment regimens. The administration of an effective schizontocidal drug allowed better definition of relapse periodicity than the malaria therapy studies (where the objective had been to produce a sustained high fever). In the more prevalent tropical strains of P. vivax treated in soldiers fighting in the Indo-Burmese and South-West Pacific campaigns in the Second World War, the relapses were documented to occur at intervals of 3 to 4 weeks (as in the Chesson strain) [42–52, 76]. These observations were similar to those in the seminal chemotherapy studies of Sinton and colleagues in Kasauli, Himachal Pradesh, India in the 1920s and 1930s [32, 77, 78]. Interest in the long-latency phenotype revived during the early 1950s as P. vivax malaria which had a long-latency was an important problem in soldiers fighting in the Korean war . During this period patients who had received very heavy inoculations, typically soldiers from the Second World War, would return to clinics for many years complaining of recurrences of malaria. The long latencies and the long relapse intervals in some infections and the multiple relapses in others despite anti-malarial treatment gave rise to the old saw that "you never got rid of malaria".
The second factor accounting for lengthening inter-relapse intervals in artificial infections was the acquisition of blood stage immunity against the single infecting genotype. In the studies of the St Elizabeth strain there was clear evidence that late relapses were attenuated if there was an early infection, but this did not affect the interval to latency (Figure 11), whereas with increasing size of sporozoite inoculum there was a corresponding shortening of the interval. These observations suggest that inoculum size is more important than immunity in determining the duration of latency or the inter relapse interval - at least for the first relapses with genetically homologous parasites . In Schmidt's primate studies (Figure 9) the lengthening of inter-relapse intervals was less prominent than in the human investigations [51, 52, 71, 82, 83] and increasing inter-relapse intervals were seen only after 12 or so relapses , although in these experiments the inocula were very large, and all animals received chloroquine treatment which was given early (on the second day of patency). Schmidt also observed in the Rhesus monkeys infected with P.cynomolgi , as Coatney had done earlier in volunteers infected with the Chesson strain of P. vivax , that occasionally a very long interval would follow a series of short intervals. Thus the steadily increasing intervals between the homologous strain relapses result from both the "running out of hypnozoites" with successive relapses, which results in reversion to the mean intervals associated with single hypnozoite activation, together with slower asexual growth rates resulting from the acquisition of asexual stage immunity. Blood stage immunity against homologous strains of P. vivax, which persists for many months, was a consistent observation in malaria therapy and challenge studies [34, 84]. Boyd noted that if the initial infection was allowed to run its natural course, then relapse did not occur and reinfection with the homologous strain was not possible .
Two other key observations were made during this early era of malaria therapy and drug evaluation which were not satisfactorily explained until decades later. It was noted that haemolytic reactions occurred sporadically with plasmoquine in patients of Asian, African or South European descent, but were uncommon in Caucasians originating further north . This was explained later by the epidemiology of glucose-6-phosphate dehydrogenase deficiency. In the Southern United States it proved difficult of impossible to infect patents or volunteers of West African descent with P. vivax . This was later shown to reflect the absence of the Duffy blood group receptor for P. vivax invasion of erythrocytes in this population.
Artificial infections provided invaluable information but they differed from natural infections in several important respects . The infections were in non-immune adults whereas the burden of vivax disease in endemic areas was in children. Adults in the malaria endemic areas have usually developed significant immunity to a broad range of local parasites which controls symptoms and reduces parasite densities. The artificial infections in the majority of volunteer studies and in malaria therapy followed the bites of 5-10 infected anopheline mosquitoes selected for maximal infectivity based on salivary gland sporozoite loads in sibling mosquitoes. The timing of inoculation in malaria therapy and experimental studies to coincide with maximum infectivity contrasts with the natural setting where anopheline mosquitoes display a wide range of infectivities depending on sporozoite age and other factors. The inocula in artificial infections were therefore usually "supranormal". This resulted in reliable infections but did not bring out the important stochastic component of P. vivax epidemiology resulting from low sporozoite inocula in areas of low seasonal transmission. Median sporozoite inocula in natural infections are estimated to be less than 10 sporozoites [90–92]. If Garnham's estimates (50:50 ratio of immediately developing parasites to hypnozoites) are correct this corresponds to a median of 5 hypnozoites in tropical P. vivax infections. The "strains" of P. vivax used in malaria therapy were likely to have been of a single (albeit evolving) genotype or very closely related interbreeding genotypes which were passaged through a very large number of patients over many years. Even if these infections originated as a mixed genotype infections in the donor patient it is likely that with multiple passage in malaria therapy the "strains" became purified through successive interbreeding to a very closely related group of genotypes. In contrast multiple unrelated genotype infections are common in natural infections.
The malaria therapy patients usually had neurosyphilis and were often very frail. Overall the mortality associated with P. vivax malaria therapy was approximately 7% (and was 10% with P.malariae infections -generally regarded as the mildest human malaria) . This high mortality reflects the underlying condition, although it was undoubtedly contributed to by the infection as well. The objective of treatment in natural infections is cure, but in malaria therapy quinine was used to "damp down" the more severe infections, not to eliminate them. Reinfection with a different genotype was usual in endemic areas but in malaria therapy this was undertaken only occasionally if the first infection was insufficient, and in volunteer studies was performed to demonstrate the "strain specificity" of the immune response.
The proportion of P. vivax infections which relapse is often thought of as an intrinsic property of the malaria parasites which varies considerably by geographic region. Tropical "strains" relapsed more than temperate "strains". But the relapse proportion is also clearly a function of the sporozoite inoculum and immunity (if any). As described earlier if artificial sporozoite induced infections were allowed to continue for weeks until self-termination the relapse did not usually occur and reinoculation was usually unsuccessful [34, 84, 93] (Figure 5). In this setting the probability of relapse with a presumed single genotype depended on the duration of preceding illness.
Relationship between the proportion of patients relapsing with vivax malaria and total number of relapses experienced
Proportion of incident P. vivax infections followed by ≥1 relapse (%)
Mean number of relapses per incident infection
Prospective studies from India over the past 25 years have recorded relapse rates following chloroquine treatment of between 8.6% (Orissa) and 8.9% (Madhya Pradesh) and 40.1% (Delhi) [97–99]. In Mumbai 19 out of 150 patients with vivax malaria and treated with chloroquine only and who were followed for one year had a relapse (17 within six months) . Reinfection was considered unlikely. Higher rates were reported from the Delhi area  where the authors concluded "Based on the foregoing epidemiologic features, three distinct relapse patterns were observed in the present study, and it can be concluded that the P vivax population in northern India is polymorphic. Group I is the tropical or Chesson strain type of frequent relapsing P. vivax with a short period of latency between the primary attack and the first relapse, which is similar to other Southeast Asian strains such as those from Thailand and Vietnam. Group III is the temperate or St. Elizabeth-type strain that has a long period of latency between the primary attack and the first relapse. Group II is intermediate between these two types" . This intermediate group, if it exists, has not been well characterized. The ratio of short to long-latency relapse phenotypes in Aligarh, Uttar Pradesh was estimated as 4:1 . Overall relapse rates from India have been relatively low by comparison with South-East Asia [97–107]. Data from travellers returning from the Indian sub-continent have suggested that long incubation-period P. vivax may be common in the Punjab, and this is consistent with evidence in the past from Northern India, Pakistan and Afghanistan [108, 109] and also the early observations of Fearnside, Yorke, and MacFie with artificial infections of Indian origin [12, 22, 23]. There are few data on temporal patterns of relapse in travellers in recent years, although studies in Eritrean immigrants to Israel, Turkish immigrants to Germany, and US soldiers returning from Somalia suggest the presence of long-latency phenotypes in the countries of origin [110–112]. Most travellers receive radical treatment for vivax malaria in temperate countries, which may be more effective against the long incubation or long-latency infections than against the tropical frequent relapse phenotypes.
The tropical frequent relapse phenotype was documented by Sinton and colleagues in British soldiers stationed in India in the 1920s and 1930s, and was observed in soldiers fighting in North-East India and Burma, and in prisoners of war in Thailand . More recently Luxemburger et al studied 342 children with acute vivax malaria treated with chloroquine on the north-western border of Thailand in 1995 and 1996 . Reappearance of P. vivax occurred in one patient on day 21 and in 8 by day 28, giving a 28-day cure rate of 97% [95% confidence interval (CI) 95-99%]. By day 63, the relapse/re-infection rate was 63% (95% CI 57-69%). Most reappearances of parasitaemia (85%; 121/143) were symptomatic. Silachamroon et al studied adult patients in Thailand (infections from Burma, Thailand, or Cambodia) with acute vivax malaria who were treated with either 5 days (N = 157) or 7 days (N = 159) of artesunate monotherapy . Relapse rates within 28 days were 52.2% and 47.8% respectively. The timing of the relapses suggested that very few if any relapses emerged from the liver before the eighth day after starting anti-malarial treatment. In Papua Indonesia the relapse rate estimated at six weeks following artemether-lumefantrine treatment was 38% (the total number may well have been higher because of later emerging relapses) . In French Guiana the relapse rate in children was 70% (nearly all relapses occurred within three months) but both recrudescence and reinfection could not be excluded . Although South American P. vivax is generally regarded as "tropical frequent relapsing" in phenotype, a recent report from Rio de Janeiro that six of 80 travellers presenting with vivax malaria (who had returned from the Amazon region and not received chemoprophylaxis) had an incubation period of between three to 12 months, and another from Brasilia describing long-latency in three patients, suggest that long-latency forms may coexist with frequent relapse phenotypes in Brazil [115, 116].
Before the Second World War most malaria infections were treated with quinine. Early relapses of P. vivax were observed to occur approximately three to four weeks after starting quinine treatment. The 8-aminoquinoline plasmoquine (plasmochin, pamaquine) was evaluated clinically shortly after its discovery in 1924 in Germany. Sinton and colleagues in India soon provided evidence that plasmoquine synergized with quinine in the treatment of acute vivax malaria and also reduced the rate of recurrence (mainly relapse) [85, 86]. "Sinton's regimen" of one week's quinine plus plasmoquine was endorsed by the League of Nations and generally replaced the two-month regimens previously in vogue . In the 1930s the newly introduced sulphonamides were shown to have activity against malaria, but were more effective against P. falciparum than P. vivax . The discovery of mepacrine (quinacrine, atebrine) in 1932 and its subsequent introduction provided a simpler, somewhat better tolerated, treatment although there was a pharmacokinetic interaction with plasmoquine which resulted in increased plasmoquine concentrations and oxidative toxicity, and precluded using the drugs simultaneously . Mepacrine was very slowly eliminated and extensively distributed . It was an effective treatment of vivax malaria. Mepacrine treatment markedly delayed the early relapse of tropical P. vivax which, usually then presented six to eight weeks after the primary infection rather than three weeks later, as was usual following quinine. When mepacrine was used as a prophylactic it suppressed infections for at least one month after stopping the drug [47, 119, 120]. Thereafter relapses followed soon afterwards in tropical areas where frequent relapse "strains" were prevalent, and many months later where long-latency "strains" were prevalent.
Radical treatment pharmacodynamics; Quantitative considerations
Consider two groups of patients who represent two extremes
Without radical treatment group A has an 80% relapse rate (eg some soldiers who fought in the Pacific in the Second World War)
Without radical treatment group B has a 20% relapse rate (eg some soldiers who fought in the Korean War)
Assuming a fixed fractional proportion of relapses and no acquisition of immunity, then the total number of relapses/100 patients is
group A = 395,
group B = 24.
These numbers represent the minimum number of viable activatable hypnozoites (VAH) i.e. there are 16 times more in group A compared to group B. It is likely that the distribution of VAH is random among the patients and therefore conforms to a Poisson distribution.
If primaquine at a dose of 0.25 mg base/kg (15 mg adult dose) reduces the number of viable activatable hypnozoites (VAH) by 90%, and there is no difference in susceptibility between the groups, and this effect is consistent across all patients then the post treatment number of VAH is
group A = 39 or 40
group B = 2 or 3.
Thus we would expect 13 to 20 times more relapses in group A compared to group B.
This hypothetical example simply points out that the apparent differences in primaquine "resistance" may reflect differences in the biology of the parasite rather than drug susceptibility per se.
The mid 1950s saw a decline in clinical research on vivax malaria. Meanwhile most countries adopted the 15 mg base/day primaquine regimen usually given for 14 days. Five day regimens of primaquine (total adult dose 75 mg base) were recommended in the Indian subcontinent, although there was no evidence they were effective [97–105, 133]. Baird has recently pointed out that the much improved tolerability of primaquine when taken with food was not emphasized sufficiently at this time and thus later dosage recommendations may have been limited by perceived or observed poor tolerability . In the recent reawakening of interest in malaria it has been suggested that resistance to the radical curative activity of primaquine has emerged  - but it is not at all clear if there has been any significant change in susceptibility. As noted earlier, because the tropical phenotype is usually associated with a greater number of relapses than the temperate phenotype, then it requires a greater proportional reduction in activatable hypnozoites to prevent all relapses. More clinical pharmacology evidence on these important points is needed.
Various theories to explain the remarkable periodicity of Plasmodium vivax infections have been proposed  including reinfection of liver cells from released merozoites, intrinsic differences in latency periods of the inoculated sporozoites (tachyzoites, bradyzoites), and activation of dormant parasites by external stresses or seasonal stimuli [141, 142]. In bird malarias there is reinfection of tissues from blood stage parasites. Following their seminal discovery of the pre-erythrocytic developmental stage in the liver, Shortt and Garnham initially suggested that this might also occur in the primate malarias [59–61]. However there is no convincing evidence to support this theory, and most are now agreed that reinfection from the blood stage back to the liver to produce secondary tissue stages does not take place in the primate malarias.
The temperate zone P. vivax usually had an incubation period of 8-9 months so emergence of an infection acquired in late summer or autumn could coincide with vector emergence in the following late spring or summer. Further south in temperate climes P. vivax infections had a primary illness two to three weeks after mosquito inoculation but the first relapse occurred 8-9 months later. Although the interval ("latency") between the primary infection and first relapse for this "Madagascar phenotype" was long (8-10 months), the subsequent inter-relapse intervals were short (Figure 7). In fact they were similar to the intervals from primary infection to early relapse which occurred in some Madagascar/St Elizabeth phenotypes and all Chesson phenotypes. This observation sits uncomfortably with a simple preprogrammed biological clock conjecture in which each sporozoite has a programmed latency. Lysenko et al pointed out that if the inoculated sporozoites were indeed a mixture of short and long-latency "zoites" then long latencies would only be evident if the sporozoite inocula were very small (otherwise there would almost invariably be one or more tachyzoites in the inoculum, and its emergence and subsequent treatment would obscure any later emergence of bradyzoites) . Under this conjecture then perhaps failure to relapse early simply implies that long-latency (bradyzoites) only were inoculated. Much of the theorizing preceded discovery in 1982 of the dormant or persistent liver stage-now called the hypnozoite [65–69]. After the discovery of the hypnozoite the biological clock model was refined. The generally accepted theory (as described, and disputed, by Schmidt) has been that "sporozoite populations of all strains of P. vivax, P. ovale, and relapsing simian malarias such as P. cynomolgi, include a subpopulation that completes development promptly and is responsible for the early primary attack, and a group of subpopulations that undergo partial development to the resting hypnozoite stage. Subsets of these dormant pre-erythrocytic stages are preprogrammed to resume development at different times and, via this built-in time clock, evoke the sequence of relapses that characterize sporozoite-induced infections with the aforementioned plasmodia" . As discussed previously it is difficult to understand how multiple relapses could occur at regular intervals with generally small inocula (median ~ five hypnozoites) and a simple clock mechanism. This remarkable efficiency suggests that some activation or feedback mechanism must operate in addition. A recent suggestion is that mosquito bites themselves might provide the trigger - perhaps by parasite sensing of a mosquito protein . This is difficult to reconcile with the similarity of latency periods in indigenous peoples living in malaria endemic areas and the latency periods observed in malaria therapy patients and returned soldiers (i.e. away from seasonal mosquitoes and independent of season). There are relatively few recent data on relapse patterns in frequent relapse "tropical" vivax malaria, but the available evidence confirms the remarkable periodicity documented in the volunteer studies with the "Chesson strain". Most informative are studies in which anti-malarials such as quinine or the artemisinin derivatives have been used as these drugs are eliminated rapidly and do not suppress or delay the emergence of subsequent relapse [107, 144, 145]. Much of the early volunteer data with tropical "strains" was confounded partly by slowly eliminated drug effects and inoculations of unnaturally large numbers of sporozoites. Coatney realized this and so, in his classic series of 204 sporozoite-induced infections with the Chesson strain , he made detailed longitudinal observations following a single infected mosquito bite (Figure 8). The number of relapses varied considerably. In the seven volunteers who were not reinfected the median (range) number of relapses following a single bite was five (one to nine) and 11 of the 39 relapses in this group (28%) occurred more than six months after the initial infection. The interval from one relapse to the next was remarkably similar but overall the inter-relapse intervals gradually lengthened. Importantly there could then be very long intervals between relapses (four relapses occurred with preceding latencies exceeding six months. The maximum documented interval was 397 days). This proves that long-latency does occur with the tropical frequent relapse phenotype.
These patterns of relapse are also illustrated well in the primate model (Figure 9); the Rhesus monkey infected with P.cynomolgi (the primate malaria "equivalent" of P. vivax) . In Schmidt's detailed longitudinal series  he noted that "The mean days separating each of the first four attacks (primary and first three relapses) were essentially identical in infections treated with chloroquine in combination with potentially curative agents, varying from 18.8 to 21.8 days from onset of patency in infections induced with the smaller inoculum and from 18.4 to 22.1 days in infections larger inoculums". In these monkey experiments where different sporozoite inocula were evaluated, it was noted that, although there was no clear difference in the number of relapses between monkeys inoculated with 5 × 102 sporozoites up to 5 × 106 sporozoites, "the intervals between relapses were related to size of inoculum, being distinctly shorter in monkeys inoculated with 5 × 106 sporozoites than in those challenged with 5 × 102 sporozoites, with recipients of 5 × 104 sporozoites occupying an intermediary position". Taken together with the absence of any relapses following an inoculum of only 5 sporozoites in three monkeys this argues for activation of a proportion of hypnozoites per relapse, and is consistent with the earlier observations in soldiers and malaria therapy patients of a fixed fraction of relapses following each illness episode in vivax malaria (Figure 16). It is unfortunate that inocula between 5 and 500 sporozoites were not studied in the primate model.
Any theory seeking to explain the remarkable biology of Plasmodium vivax relapse must accommodate the following
1. Relapses show remarkable periodicity.
2. Early relapses reach patency around three weeks after starting treatment which suggests emergence from the liver at least one week earlier.
3. Not all P. vivax primary infections are followed by a relapse. In Thailand approximately 50% of infections are followed by a subsequent relapse within 28 days if a rapidly eliminated anti-malarial drug (artesunate) is given for treatment of the primary infection and primaquine is not given . Elsewhere the probability of relapse generally varies between 20% and 80%. Animal experiments, the malaria therapy experience, and volunteer studies all suggest this proportion is a function of sporozoite inoculum.
4. Multiple relapses are common, particularly in young children, even though sporozoite inocula are thought to be relatively small (median 6-10 sporozoites). It is not uncommon in tropical areas for children to have four to six relapses at 4-6 week intervals and sometimes more following an incident infection. Even larger numbers of relapses were observed in soldiers following intense exposure and in Rhesus monkeys receiving very large sporozoite inocula. Importantly the fraction of people experiencing a relapse after each illness episode in a particular location appears constant
5. In long-latency phenotypes there is commonly a period of 8-9 months either before the first symptomatic infection, or between the first symptomatic infection and the first relapse. This long-latency interval appears to be normally distributed (mode 28 weeks for the Madagascar strain [24, 25] (Figure 2). Sometimes there are several short interval relapses followed by a long interval. Conversely long latencies may also occur after multiple relapses in the tropical frequent relapse phenotype  (Figure 8).
6. if there are further relapses after the long latent period then they occur frequently with short intervals which are very similar to those observed in the tropical "strains".
7. The relapses in clinical studies conducted in endemic areas are commonly with a genotype which is different to that identified in the primary infection (48% in Columbian isolates, 55% in Indian isolates, 61% in Thai and Burmese isolates, and 71% in East Timor isolates) [146–148].
8. A remarkably high proportion of acute infections with Plasmodium falciparum are followed by an episode of P. vivax infection. The proportion is currently 30% in Thailand [136–138] and 50% in Myanmar . The intervals between the acute P. falciparum malaria illness and the subsequent P. vivax malaria are similar to those between acute P. vivax malaria and the subsequent P. vivax relapse. The epidemiological characteristics suggest that these are all relapses (Figure 20).
It is also interesting and perhaps relevant that in endemic areas, despite often low seasonal transmission, P. vivax maintains a high degree of genetic diversity
Swellengrebel and de Buck  noted that relapse rates were higher in naturally acquired P. vivax infections (up to 50%) than in mosquito-borne infections with the local strains in neurosyphilis patients (despite larger inocula in the latter). They ascribed this to immunity, and this was undoubtedly a contributor, but another explanation is possible. That is that in endemic areas a high proportion of the population have latent P. vivax hypnozoites which can be activated by a sufficient stimulus.
Four lines of evidence support this "Activation of latent hypnozoites" (ALH) hypothesis.
Mixed blood stage infections with P. falciparum and P. vivax are underestimated by microscopy, but, even with sensitive PCR techniques, this proportion is insufficient to explain the 30% to 50% of patients in SE Asia who experience vivax malaria following falciparum malaria . The interval between the primary P. falciparum infection and the subsequent P. vivax malaria strongly suggests this is a relapse (Figure 20). Persuasive supporting evidence that these are P. vivax relapses and not simultaneously acquired infections comes from entomology studies in which single anopheline mosquitoes have been examined for both species . If these mixed species infections resulted from simultaneous inoculation then 30 to 50% of anopheline vectors carrying one species should also carry the other. In fact finding vectors with both P. falciparum and P. vivax sporozoites is relatively uncommon (overall in Asia 6.6% of P. falciparum sporozoite positive mosquitoes (17 of 258) also contained P. vivax). In the published literature not one of the 45 individually examined malaria positive wild anopheline vectors trapped in Thailand contained both malaria species . This is also supported by the rarity of finding patients or healthy subjects with gametocytes of both species in the blood at the same time. Development rates in the mosquito are also slightly different (P. vivax being more rapid). Although space-time clustering of infections may occur in low transmission areas it is implausible that over 20% of P. falciparum inoculations would be associated with a separate P. vivax inoculation within one or two days, particularly when individuals receive on average less than one infectious bite per year. There is other supporting anecdotal evidence from travellers and from soldiers with brief periods of exposure in SE Asia. These groups have much lower rates of P. vivax following P. falciparum. The most plausible explanation for these findings is that the majority of these P. vivax episodes arise from hypnozoites which were latent in the liver of the patient at the time of acquiring P. falciparum (ALH). The remarkable similarity of both the timing of the P. vivax recurrences and their variance strongly suggest that latent P. vivax hypnozoites are activated by acute falciparum malaria.
If P. falciparum malaria activates latent P. vivax hypnozoites then there is no reason why P. vivax malaria should not do the same. This would explain satisfactorily the finding of heterologous genotypes in one half to two thirds of P. vivax relapses [146–148]. In these cases where the original genotype was not detected in the malaria recurrence then either the inoculated infection did not relapse, or its hypnozoite(s) were activated but their progeny were outcompeted by the earlier activation or more rapid growth of the progeny of the activated latent hypnozoite(s). In approximately one third of P. vivax relapses studied in South East Asia, the malaria parasites isolated are either identical or closely genetically related to the primary infection. Relatedness could occur if there is little genetic diversity in the area where P. vivax malaria was acquired, or could result from recombination in the anopheline vector with the production of genetically related sibling sporozoites (i.e. simultaneous inoculation). If the "cross" took place several cycles of infection previously then subsequent infections may contain highly related parasites through successive interbreeding between related siblings. More work on this important area is needed. However in the majority of relapses the parasites are clearly genetically unrelated. In the one third of patients in whom the relapse is homologous or highly related with the primary infection, either there were no latent hypnozoites (as in volunteer studies), or the homologous infection's hypnozoites' progeny won the race to patency against the heterologous hypnozoites' progeny. It is evident then that close "races" between different genotypes to reach patency commonly result in gametocyte genotype mixtures in relapses (which may then recombine in the mosquito). Further support for the ALH hypothesis comes from observations in mothers and their infants living in a malaria endemic area on the north-west border of Thailand. The relapses of vivax malaria in the babies' mothers were usually genetically different to those which caused the primary infection, as in other patients studied in this area, whereas the relapses which followed the first P. vivax infection of life in their babies were usually of the same genotypes as those which caused the initial infection . Obviously the infants could not have any previously acquired latent hypnozoites in their livers to be activated by the illness.
Higher rates of vivax relapse in indigenous compared with artificial infection would also be explained by the ALH hypothesis. The incidence and number of relapses depends on the number of sporozoites inoculated. If all relapses derived from the inoculated infection then artificial infections which follow inoculations with 5-10 times more sporozoites should have higher, not lower rates of relapse. Relapse rates were particularly high in soldiers who were immunologically naïve and underwent intense exposure. If all relapses derived from the most recent inoculum then there should be no relationship between intensity of exposure and number of relapses.
Four of seven volunteers receiving a single infected mosquito bite in Coatney's series had relapses of the Chesson strain of P. vivax with variable but long latency periods - all exceeding six months after their preceding relapse . It is not uncommon to encounter patients returning from malaria endemic areas where tropical phenotypes are prevalent with relapses more than 3 months after either a primary infection or return to the non-endemic area (of course these might also be long-latency phenotypes particularly if the interval is eight -10 months). This proves that some hypnozoites of frequent relapse phenotypes can remain dormant for long periods.
If acute malaria activates latent hypnozoites and thereby causes vivax malaria relapses then it seems that a significant systemic illness is necessary for this activation. In the first half of the twentieth century, it was widely believed that a variety of external stresses could bring on a relapse of malaria . By analogy, relapses in the bird haemosporidian parasite Leukocytozoon are provoked by the stresses of egg laying and the exhaustion of long migratory flights . It was even taught in some textbooks of tropical medicine that Cinchona alkaloids should be given before surgery to pre-empt a relapse of malaria. However formal studies to provoke relapses of vivax malaria, which included forced route marches, simulated altitude, and induced hypoxia, did not yield convincing results [153, 154]. Nevertheless it is interesting that soldiers in field hospitals in the Mediterranean region between 1940 and 1945 (who had acquired P. vivax in North Africa or Italy) were apparently considerably more likely to experience vivax malaria (presumably a relapse) if they had pneumonia or hepatitis compared with trench foot [155, 156]. In our own series of children seen every day for over 18 months on the Thai-Burmese border during a study of the SPf66 malaria vaccine P. vivax episodes were associated with malaria, but not with minor illnesses (S Lee; personal communication), so it seems that a substantial systemic stimulus is required.
For temperate strains of P. vivax, as suggested by several investigators previously, there are clearly at least two populations of hypnozoites, one becoming activatable early (as for tropical P. vivax) and another remaining latent and not immediately activatable. The study of Cooper et al, in which blood induced homologous (St Elizabeth strain) infections were induced 120 days after infectious mosquito bites, is informative in this respect . The blood inoculations reliably gave rise to symptomatic malaria but critically did not affect the timing of the subsequent relapse. This suggests that the hypnozoites (half way through their sleep) were absolutely refractory at this time. It is likely that once hypnozoites do become activatable, that there is a background relatively low rate of spontaneous activation in order to account for the distribution of latencies. Thus for the long-latency P. vivax the first relapse after the long latent period is usually spontaneous, but the illness then activates further hypnozoites, accounting for the subsequent short inter-relapse intervals. If this is correct it follows that first relapses which occur with a long-latency interval (i.e. 8-10 months after the primary infection) should usually be of a similar genotype to the initial infection, whereas relapses at other times could be heterologous. If there are subsequent relapses i.e. which follow the first relapse after the latent period (i.e. with a short periodicity) then these could be genetically heterologous.
Plasmodium vivax in temperate zones has clearly evolved to adapt to the long winters across the Northern hemisphere [17, 19, 27–29, 79, 158]. It is interesting to speculate that the large proportion of sporozoites dedicated to latency, and the relatively low number of relapses compared with the tropical strains may reflect the high natural wastage of hypnozoites in hepatocytes which die before the hypnozoites become activatable. In experimental P. cynomolgi bastianelli infection in the Rhesus monkey Garnham observed a ten-fold reduction in the number of hypnozoites in serial liver biopsies over a nine month period, but it is not clear how much of this reduction resulted from activation and how much was from cell death .
The hypnozoite can be considered as an unactivated sporozoite. If the duration of pre-erythrocytic development of the liver stage is similar for sporozoites and hypnozoites, as seems likely, this puts activation at the time or shortly after presentation with acute malaria (of any species). If the ALH hypothesis is correct then the key biological difference between frequent relapse and long-latency P. vivax phenotypes lies in the temporal distribution of susceptibility to activation among the sporozoites. This can then be subdivided into the proportion of sporozoites which activate immediately and the subsequent temporal distribution of susceptibility to activation among the remaining hypnozoites. The genetic basis for this regulation may be difficult to find, as it could well reflect subtle quantitative rather qualitative differences. Recent studies in murine malaria indicate a central role for iron sequestration in controlling pre-erythrocytic development through malaria illness induction of the iron regulatory hormone hepcidin [159, 160]. It is possible that iron availability also plays an important regulatory role in controlling hypnozoite activation and pre-erythrocytic development. One possible mechanism contributing to regular short interval (three weeks) relapses is a malaria illness related temporary halt to liver-stage development (mediated by reduced iron availability), which is then lifted with clinical recovery. Some of these hypotheses may be testable in the ex-vivo hepatocyte culture systems .
In summary this ALH hypothesis proposes that there is indeed a biological clock (which is most evident in temperate strains) determining latency in Plasmodium vivax. This clock (which could be an intrinsic parasite clock, or could reflect a host-parasite interaction) determines the length of the interval before the hypnozoite becomes susceptible to activation, but that there is a separate sensing mechanism which determines whether or not activation does occur. The trigger could be either a positive activation stimulus or removal of inhibition. This mechanism may activate spontaneously once the hypnozoite has become susceptible, and spontaneous activation presumably usually explains the first relapse after a long latent period, but once susceptible activation is much more likely with an external systemic illness such as malaria. Activation must involve signaling via the infected hepatocyte (which is very sensitive to systemic inflammatory responses). Importantly the individual probability of activation for each hypnozoite is low, allowing accumulation of latent but "activatable" hypnozoites after each sporozoite inoculum. This implies that people living in vivax endemic areas commonly harbour latent but "activatable" hypnozoites. In endemic areas of South-East Asia, if patients who acquire falciparum malaria are representative of the population, this figure is approximately one quarter to one third. The periodicity of P. vivax relapses is derived from the sequential activation of hypnozoites by illness. Mathematical modeling will provide valuable insights into which activation-inhibition model best fits the observed malaria therapy, volunteer, and epidemiological information.
Slowly eliminated drugs interefere with the resistance protection provided by relapse. Chloroquine has a slow multiphasic elimination phase and suppresses the relapse of sensitive parasites for approximately two weeks (from two to six weeks). Mefloquine has a different profile of elimination, and provides longer suppression of vivax relapse. The residual drug levels reduce potential interference with resistance emergence by relapse providing greater suppression of the sensitive compared with the resistant parasites and they therefore reduce the delay on resistance emergence. In summary early relapse provides a brake on the emergence and spread of low-grade resistance in Plasmodium vivax by pre-empting recrudescence.
Activation of hypnozoites from different preceding inoculations will commonly result in two or more genotypes reaching patent parasitaemia at similar times. As gametocytogenesis in P. vivax occurs simultaneously with asexual stage development this provides an effective method of increasing the likelihood that a mosquito will ingest gametocytes of different genotypes, thereby facilitating meiotic recombination between genetically unrelated P. vivax parasites. This must be an important contributor to the relatively high degree of genetic diversity in P. vivax often found in areas with very low seasonal transmission.
A substantial proportion of the population in P. vivax endemic areas harbours latent but activatable hypnozoites. Some of these may derive from inoculations which were not followed by any illness.
If relapse rates exceed 50%, then relapse becomes the predominant cause of vivax malaria.
Spontaneous or activated relapse followed by asymptomatic parasitaemia may be an important source of P. vivax transmission.
Reducing P. vivax transmission will have a smaller than currently predicted effect on the incidence and prevalence of vivax malaria initially.
Reducing P.falciparum transmission may reduce the incidence of P.vivax infections and reduce P.vivax transmission. However any effect on the incidence of clinical disease would probably be delayed because reducing falciparum malaria will also reveal vivax malaria by lifting suppression in mixed infections, and a reduction in vivax incidence will reduce immunity. A single radical treatment for all malaria infections may be justified in areas where both parasites are prevalent (i.e. ACT + radical primaquine regimen) .
It is very difficult to exclude the presence of long-latency P. vivax phenotypes in studies conducted in vivax endemic areas. Long-latency phenotypes may be prevalent over a much wider area of the tropics than currently thought.
Assessment of the efficacy of interventions requires characterization of the prevalent relapse phenotypes.
Assessments of radical curative activity where long-latency phenotypes are prevalent require one year's follow-up. Genotyping should assist in assessing long-latency relapse.
The proportion of genotypically different (heterologous) relapses will fall as transmission intensity falls
In therapeutic assessments a follow-up period of six months or less may miss a significant proportion of the relapses. If the majority of relapses in endemic areas derive from heterologous latent hypnozoites then malaria control interventions which are effective will not prevent relapses emerging for months or years, although their number will reduce as the reduced transmission will result in less illness and therefore less hypnozoite activation. Mass drug administration with radical curative regimens (currently primaquine is the only option) would be the only way to eliminate this reservoir of infection quickly. Epidemiological assessments in older children and adults in endemic areas may underestimate the burden of vivax malaria as partial immunity (and premunition) will ameliorate disease severity and may lead to reduced activation of relapses. This would result in relatively low relapse rates. The proportion of acute falciparum malaria infections, which are followed by P. vivax may be a better indicator of the prevalence of latent hypnozoite carriage.
It is possible that much of the variance in responses to primaquine is explained by differences in rates and burdens of latent hypnozoite carriage and degree of immunity and not variation in drug susceptibility. The same factors which affect therapeutic responses in the blood stage infection appear to affect the responses to hypnozoitocidal treatment i.e. organism load and immunity. We have tended to consider 8-aminoquinoline efficacy only from the perspective of the drug, and we do not take into account either organism load, organism phenotype, or host immunity. Taking a quantitative approach to assessing 8-aminoquinoline radical curative activity based on hypnozoite burdens may be a valuable approach. Patients with very large liver burdens of hypnozoites from either a very heavy inoculation or multiple inoculations and little or no immunity (such as soldiers) would be expected to have a larger number of relapses than travellers who have a brief period of exposure. Studies of radical curative activity in soldiers may not be comparable to studies in travelers. Studies in adults may not be comparable to studies in children. The apparent radical curative activity of primaquine would be expected to improve as malaria transmission falls. Lower dose regimens may have useful efficacy in this context. Long term follow up (minimum one year) of well characterized patients with parasite genotyping in low transmission settings should help to dissect the contributions of pre-existing versus recently inoculated hypnozoites to relapse. These should be accompanied by entomology and genotyping studies in wild-trapped anopheline vectors to determine mixed genotype and genetic recombination rates.
It is interesting to speculate on the role of relapse in enhancing the opportunities for recombination and how, in the pre-anti-malarial era, relapse would have contributed to persistence of the untreated infection. Rechallenge experiments showed that a "strain specific" immunity developed after protracted symptomatic infection with that P. vivax strain. But this did not always prevent subsequent parasitaemia, particularly if rechallenge was many years later . Despite considerable investment in a P. vivax vaccine there is relatively little information how immunity to vivax malaria is acquired and maintained in the context of frequent infection and relapse with different genotypes . It is generally recognized that asymptomatic P. vivax infections are common in endemic areas but their overall contribution to P. vivax transmission in malaria-endemic areas, and the importance of relapse in maintaining these asymptomatic infections has not been characterized adequately. However the malaria therapy experience suggested that asymptomatic infections were a very important source of infection . An effective vaccine which did not confer long-lasting immunity and required frequent boosting might provide a selective pressure towards longer latency.
There are major questions about the basic biology and the epidemiology of Plasmodium vivax relapse which first need recognizing, and then need answering, if we are to address seriously controlling and eliminating this important malaria parasite .
I am very grateful the advice and data given by my colleagues in South East Asia, which have contributed significantly to the development of this theory, to Professor PCC Garnham (1901 to 1994) for early discussions on the mechanism of relapse, to Professor Piet Kager for introducing me to the brilliant work of the early Dutch malariologists, and to Katherine Battle and Simon Hay for information on relapse studies. I am a Wellcome Trust Principal Fellow.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.