Search strategy and selection criteria

The electronic databases PubMed, Web of Knowledge, Scopus, and the World Health Organization’s WHOLIS and regional office databases were searched for articles documenting historical malaria resurgence events using the search terms “malaria” and either “resurg*”, “reemerg*”, or “re-emerg*” (wild-card operators were used to ensure that the search would identify “resurging”, “resurgence”, and any other form of the word). The searches, conducted on Aug 1, 2011, included references published on any date up until the day of the search and included those published in English, French, or Spanish. All records resulting from these searches were screened, and full-text articles were assessed if the reference appeared to describe or allude to a resurgence event. In addition, the reference lists of all articles for which the full text was reviewed were hand-searched, and the full text of those references that appeared relevant to malaria resurgence were retrieved.

Full-text articles were read to evaluate whether they included mention of any resurgence event. Although the term “resurgence” is sometimes used in a general, non-specific way to refer to any increase in malaria, this review defined a resurgence event more narrowly as:

An increasing trend in malaria incidence or prevalence following suppression achieved through implementation of control efforts.

Accordingly, any report of an increase in malaria incidence or prevalence in assessed articles was included in analysis if it appeared to a) involve an increase over a period of more than a single year or transmission season (i.e., there was a upward “trend” over time and not just a single aberrant season), and b) occur in a region where endemic malaria had previously been reported but where transmission had subsequently been suppressed to some degree through anti-malarial interventions. Any reference to such an event, whether national or subnational, was recorded, regardless of article type or quality.

Evaluation of causes

After compiling all resurgence events in the identified citations that met these criteria, the same articles were reviewed for suggested causes of those resurgence events. Additionally, to identify potential causes suggested elsewhere, the same databases as above were searched for the name of the country involved and “malaria.” Search results were limited to articles published within a few years after the start of the resurgence event. The full-text of all results of these searches that appeared likely to discuss potential causes of resurgence was read, and the reference lists of these articles were hand-searched for relevant sources.

A data extraction form was used by two reviewers to classify all suggested causes for resurgence into categories. Three overarching categories were used to classify causes: 1) weakening of the malaria programme, 2) increasing intrinsic potential for malaria transmission, and 3) technical problems such as insecticide or drug resistance. All suggested causes for resurgence were recorded from each article, regardless of article type or quality of evidence. Additionally, however, each suggested cause was classified in regard to the degree of supporting evidence into one of two levels: a) assertions, without quantitative analysis or detailed argument for why that factor was a cause of resurgence, or b) evidence-based claims, where in-depth qualitative or quantitative analysis was used to provide evidence in support of the factor as a cause. Suggested causes of resurgence for each documented resurgence event were independently evaluated for category and level of evidence by the two reviewers, and disagreements between the reviewers were resolved by consensus.

Reported causes of resurgence

Suggested causes of resurgence fell into all three of the general categories. These categories – which were not mutually exclusive – included weakening of the malaria control programme (68/75 = 91%), increases in the intrinsic potential for malaria transmission (44/75 = 59%), and technical problems including drug and insecticide resistance (24/75 = 32%). Subcategories of each are described below.

Only 45 of the 273 (16%) suggested causes for resurgence events identified by the review were classified by reviewers as presenting in-depth qualitative or quantitative analysis to support the assertion. Of these 45 suggested causes, 27 (60%) implicated weakening of malaria programmes, 15 (33%) increases in malaria potential, and 3 (9%) technical problems such as resistance.

Weakening of the malaria control programme

Programmatic weakening was attributed to a variety of causes (which are not mutually exclusive), including funding shortages (37/68 = 54%), complacency and other issues with poor execution (32/68 = 47%), war or disaster (17/68 = 25%), purposeful cessation of control activities (17/68 = 25%), community non-cooperation (7/68 = 10%), or unknown or unstated factors (7/68 = 10%). The effects of programmatic weakening are illustrated by the increases in malaria that accompanied the scaling down of indoor residual spraying (IRS) in much of Latin America (Figure2).

Funding issues were the single most commonly cited reason for resurgence, mentioned in 37/75 (49%) events. Many of these involved time-limited bilateral commitments that funded interventions too costly to continue once the funding period had ended. For example, USAID provided DDT to India for an eradication programme at the end of the 1950s, which led to a very large reduction in the malaria burden, from an estimated 100 million annual cases in the early 20^th century to about 100 thousand cases in 1965[18]. When the USAID commitment ended, however, India proved unable to procure or produce the necessary insecticide to continue the programme, with over a 30% shortfall in 1965–66. Insufficient DDT was likely a key factor resulting in a resurgence of malaria to a peak of 6 million cases by 1976[18] (Figure3). In the western Kenyan highlands, three WHO-supported sprayings of dieldrin reduced malaria prevalence to 0.5-2.0%, after which the cost of malaria control was transferred to the local government in 1957. Since the cost of spraying was equivalent to the entire health budget for the district, the programme was terminated, and a “striking increase” in malaria had occurred by 1959[19].

Reasons for funding reductions or cessation were not clear for all events, but in several, donors appear to have reallocated funding specifically because successful reductions in malaria burden had occurred. In Thailand, for example, bilateral assistance for malaria control was phased out in 1970 following reductions of the slide positivity rate (SPR) to 1.7% in 1969; within six years the SPR had risen to 8.0%[20]. In Swaziland, the reduction of malaria to the point that it was no longer perceived as a public health problem in the 1950s resulted in significant cut-backs in funding to the WHO-funded control programme, including reductions of the staff from 36 to only seven at the end of the 1960s; significant malaria epidemics involving thousands of cases followed[21].

The reliance of malaria programmes on a few major donors has meant that any change in donor priorities may put continued suppression of malaria at risk. A US-led campaign in Monrovia, Liberia, caused hospital admissions at the public hospital to decrease by about 95% between 1945 and 1947. Thereafter, the programme was deemed too expensive, the budget was cut by 80% in 1948, and by 1950 an assessment concluded that control measures were no longer having any impact on transmission[22]. In Zanzibar, a USAID project in the 1980s was terminated, despite having about $US 1 million in undisbursed funds, due to the perception that the project was a failure[23], and malaria rates on the island of Pemba rose from 23.2% in 1989[23] to over 60% in 1994[24]. In Ethiopia, funding from USAID and WHO was halted in 1974 following the overthrow of the government by a military regime[25]. DDT application to households plummeted from a 1974 peak of 117,040 houses to only 8,139 houses in 1985. Incidence increased from 1.1 cases per 1,000 person-years in 1980 to 65.9 cases per 1,000 person-years in 1989[25]. Similarly, in Indonesia, a DDT programme protecting 17 million people by 1959 was scaled back following withdrawal of assistance from the USA during a tumultuous political period in the early 1960s[26], and malaria increased from <6,000 cases in 1963 to 346,000 in 1973[20].

In 32/75 events (43%), the weakening of operations for reasons other than funding shortfalls was blamed for subsequent resurgence. In several examples, this weakening was attributed to a sense of complacency within the programme or government resulting from the perception that malaria was no longer a threat. In these examples, commentators do not suggest that insufficient resources were available, nor that programmes were purposefully halted; instead they indicate that the programmes failed to operate sufficiently well despite the apparent availability of resources to do so. In Mauritius, for example, successful certification of elimination was said to have led to a laxness in control: regular testing for malaria was halted and vector control was scaled back, providing an ideal environment for malaria to return following the trigger of a natural disaster[27].

In 17/75 events (41%), war, strife, or natural disaster disrupted programme operations and prevented continued suppression of malaria. For example, in the wake of the dissolution of the Soviet Union, war and strife damaged control efforts throughout the region even as the Soviet support for interventions vanished. The Nagorno-Karabakh civil war in Azerbaijan in the early 1990s interrupted control efforts[28], while war in Tajikistan similarly contributed to disruption and resurgence of malaria[29] (Figure4). In Myanmar, the national malaria control programme lowered incidence from 217 per 1,000 in 1950 to 65 per 1,000 in 1957 (with a prevalence of only 0.11%)[20]. Troubles beset the programme in the 1960s and 1970s, however, and malaria metrics crept upwards; following rebellion and chaos in 1988, slide prevalence doubled from 7.3% in 1987 to 14.0% in 1992, and then continued increasing to 46.4% in 2010[30]. In China, malaria was reduced from a reported 6.8 million cases in 1954 to 1.58 million in 1959, but following natural disasters, 10 million cases were reported in 1960[31].

In 17/75 events (23%), malaria interventions were purposefully halted, often because they were intended as time-limited experimental pilots rather than ongoing programmes, especially in sub-Saharan Africa (Figure5). In Central Liberia, for example, WHO and UNICEF started funding one of the first pilot projects intended to examine whether eradication was possible in Africa in 1953; cessation of WHO involvement in 1961 led to rapid deterioration[22]. In Pare and Taveta on the Kenyan-Tanzanian border, malaria was greatly reduced from 1956–59, but it resurged to pre-intervention levels within three years after the pilot ended[32]. In Yaounde, Cameroon, pilot spraying began in 1954 and produced “excellent results”[33], but cessation of spraying in 1960 resulted in complete recovery of the vector[33] with subsequent increases in malaria[34]. And in Garki, Nigeria, prevalence resurged rapidly from <5% prevalence back to a baseline of around 50% following the cessation of the intervention[35].

Purposeful decisions to halt successful malaria programmes have also occurred outside of experimental situations. In Zanzibar, a WHO-supported IRS programme reduced malaria prevalence from 76% in 1957 to <5% in 1967[9], after which Sheik Karume suspended the programme due to his stated belief that Africans were “malaria-proof”[36]. Resurgence rapidly followed cessation of the spraying, and by 1973 prevalence had returned to 54% on the island of Unguja[37]. Elsewhere, countries followed WHO’s eradication guidance and withdrew all vector control measures from regions where malaria had apparently been interrupted[38], in some cases despite lacking sufficiently strong surveillance to maintain elimination in the absence of vector control. In Sri Lanka, for example, a highly successful IRS campaign reduced malaria from 2.8 million cases in 1946 to only 17 (11 of which were imported) in 1963[20]. Cessation of spraying following the subsequent move from attack to consolidation in 1964[39], exacerbated by weakened surveillance and increased population movement for mining and agriculture[26], led to a very large increase in malaria incidence, from 150 cases in 1964 to 538,000 in 1969[20]. In Madagascar, an eradication campaign reduced malaria to very low incidence in a few residual foci by 1960. By 1979, even prophylaxis and treatment centers were closed, and without the checks of either IRS or chemotherapy, the malaria incidence began to rise at a rapid rate[40]. Reintroduction of DDT spraying brought malaria back under control[40].

In 7/75 (9%) of events, resurgence was attributed to decreases in community acceptance or participation in malaria programmes. In French Guiana, it was suggested that resistance of domestic pests, such as cockroaches, to insecticide spraying caused the population to lose confidence in the effectiveness of IRS[41]. Declining compliance with spray campaigns may have contributed to resurgence in Kisumu, Kenya[42], São Tomé and Príncipe[43], and Sabah, Malaysia[26], among others.

Increases in malaria potential

Increases in intrinsic malaria potential were attributed to a variety of causes (which are not mutually exclusive), including movement of humans or mosquitoes (32/44 = 73%), development and land-use changes (19/44 = 43%), climate or weather (11/44 = 25%), war and civil strife (8/44 = 18%), and worsening of socioeconomic conditions (5/44 = 11%).

The most common rationale for why increased transmission potential may have contributed to resurgence involved the movement of humans and mosquitoes and the parasites they carry, cited in 32/75 (43%) events. In Brazil, for example, nearly one million immigrants moved into the Amazon region during the 1970s seeking new farmland and attracted by gold mining in the region. The increase in susceptible individuals in the receptive region, combined with a possible influx of infections from endemic Bolivia, may have contributed to the rapid rise in malaria incidence in the late 1970s and early 1980s[44]. In Thailand at the end of the 1960s, surveillance struggled to detect cases among a mobile population that traveled back and forth from endemic areas[20], while Afghanistan’s nomadic population created similar challenges in the early 1970s[26]. The migration of farmhands from neighbouring endemic countries to Belize in the 1970s fueled transmission[45], while movement of infected mosquitoes across the Demilitarized Zone between the Koreas is believed to have sparked resurgence of malaria in Republic of Korea 14 years after elimination[46]. Movement of refugees and soldiers from wars across countries and decades has been implicated as a cause of malaria resurgence, including in Spain[47], Bangladesh[48], Vietnam[49], Sri Lanka[50], Sudan[51], and Azerbaijan[28].

Nineteen of the 75 events (25%) were at least partially attributed to changes in development or industry, including agricultural development, creation of dams or highways, or other land-use changes. For example, in the USSR, new irrigation and construction of hydroelectric power stations may have increased breeding sites for mosquitoes in the 1960s and 1970s[52]. In Swaziland, development of sugar plantations in the receptive lowveld of the country involved bringing large numbers of potentially infected Mozambican workers into a region where increased agriculture had increased the potential for malaria transmission[53]. Similarly, in Costa Rica, development of the banana industry exacerbated malaria by moving workers from endemic areas into regions with increased suitability for vector breeding[5], while simultaneously reducing the coverage of malaria control programmes to protect them[54].

Eleven of the 75 resurgence events (15%) were attributed at least in part to climate or weather. In Debre Zeit, Ethiopia, malaria rates rebounded in the 1980s after having been successfully suppressed with extensive use of DDT. During this period of increasing malaria rates, positive correlations existed between temperatures and monthly malaria incidence[25]. In Kenya’s western highlands, warming temperatures may have played a role in an increase in epidemic malaria during the 1990s[55]. However, the importance of such factors is controversial given the many other important changes occurring in the region during that time period, including the emergence of chloroquine resistance[56, 57]. Increasing temperatures were also suggested as a cause of resurgence in northern Pakistan[58].

Eight of the 75 events (11%) involved increases in transmission potential attributed to war or strife, while another five (7%) were said to be related to a worsening of socioeconomic circumstances. For example, war in Bangladesh in the early 1970s uprooted millions of people and destroyed homes, and, even if the malaria programme had been able to continue its efforts, the lack of stable communities would have made continued control extremely difficult[48]. In 1969, before war began, fewer than 100,000 cases of malaria were reported, but a few years after the war, annual incidence had tripled[20].

Technical problems

Malaria resurgence was attributed primarily to two types of technical problems: vector (14/23 = 61%) and drug (15/23 = 65%) resistance. In 14 of the 75 events (19%), vector resistance to insecticides was suggested as a cause of resurgence. For example, in Nicaragua, an eradication programme failed in 1960 when resistance to dieldrin and DDT was coupled with a shortage of funds, with malaria returning to original levels[26]; DDT was among the cheapest pesticides available, and switching to new insecticides involved an increase in costs at a time when funds were scarce[59]. In Afghanistan, “already deficient operations” were unable to cope with the development of DDT resistance despite an infusion of funding from the USSR in the early 1970s. Malaria incidence there increased from <20,000 cases in 1970 to 127,000 in 1976[26], and the subsequent invasion of the country by the USSR finished off the malaria control programme[60]. Increasing DDT resistance of mosquitoes in Java may have been an important contributor of resurgence through 1973[26]. In the USSR, apparent behavioural changes in the vector population were stated to have negatively impacted the programme in the early 1960s[52].

Drug resistance was implicated as a cause of resurgence in 15/75 events (20%). For example, quantitative analysis of factors contributing to resurgence of malaria in South Africa in the 1990s demonstrated that drug resistance was associated with malaria incidence[61], probably because infections that were not effectively treated remained to contribute towards onwards transmission. Similarly, resistance to chloroquine has been suggested as a likely candidate for explaining increases in malaria in the western Kenyan highlands[56, 62, 63] and northern Pakistan[64].