Recrudescence of onchocerciasis infection, simulation of risks, detection and control

WORLD HEALTH ORGANIZATION ORGANISATION MONDIALE DE LA SANTE ONCHOCERCIASIS CONTROL PROGRAMME IN WEST AFRICA PROGRAMME DE LUTTE CONTRE L'ONCHOCERCOSE EN AFRIQUE DE L'OUEST EXPERT ADVISORY COMMITTEE Thirteenth session Ouaeadoueou. 8-12 June 1992 ocP/EAcr3.9(ocP/Brs/MoD/ t /92) RECRUDESCENCE OF ONCHOCERCIASIS INFECTION, SIMULATION OF RISKS, DETECTION AND CONTROL WORLD HEALTH ORCANIZATION ONCHOCERCIASIS CONTROL PROCRAMME rN wEsT AFRTCA (OCP) RECRUDESCBNCE OF ONCHOCERCIASIS INFECTION simulation of risks, detection and control (ocP/Brs /MoDlt/e2) Introduction On the basis of simulations with the epidemiological simulation model ONCHOSIM it has been concluded that 14 years of vector control is sufficient to reduce the risk of recrudescence after cessation of control to below lVo even in the most afflicted areas. However, this only holds true for situations where vector control wu lNVo successful, and where there has been no invasion by infective flies or immigration of infected humans. Since it is a fact that, at least in some areas of OCP, these conditions will not be satisfied completely, the occurrence of recrudescence cannot be excluded. Model simulations have also shown that community-based treatment with mectizarL again under certain conditions, can be a powerfirl strategy for the control of recrudescence. The risk, detection and control of recrudescence is the theme of the present report. In the first chapter, a summary is given of the model-based analysis of strategies for the detection and control of recrudescence. The basic ideas will be giveq and it will be further demonstrated that arriving at an operational definition of recrudescence is complicated and depends on a lot of (basically unknown) factors. Part of this work has already been presented at EAC 1991. One of the recommendations of EAC 1991 was to extend the work of recrudescence with regard to the impact of human immigration. The phenomenon of human migration is introduced in chapter 2. I Simulation of strategies for detection and control of recrudescence 1.1 Statement of the problem The policy of OCP is to prevent recrudescence of infection after withdrawal of its larviciding campaigns. Nevertheless, the occurrence of recrudescence must be anticipated, and strategies must be designed to detect and control such renewed transmission and prevent the subsequent build-up of the prevalence and intensity of infection. Such strategies are neither simple nor straightforward. Model simulations have shown that under certain conditions community-based mectizan distribution can stop recrudescence. A first condition is that the recrudescence is detected timely. If mectizan strategy is started when the build-up of infection progressed too far it will not be able to control it, at least not with a time-limited (ocr/nrs/rioo/1/92) -l- -i - campaign. The success of such mectizan-based intervention further depends on the total duration of the campaign, the coverage of the population, and the frequency of applications. The question remains: what is a'timely'detection; does this only depend on the factors mentioned (duration, coverage, frequency)? The answer is: no. The problem is that it is likely that marked differences will exist in the dynamics of recrudescence. Sometimes the recrudescence will be slow, and will initially be invisible. In other situations it will proceed fast. In case of such fast recrudescence, the detection must be done early to guarantee a successful control by mectizan. In case of slowly progressing recrudescence, detection can occur late. This brings us to another consequence of this variability in recrudescence dynamics. It is obvious that a lot of money and scarce resources are wasted when a mectizan campaign is started without any need. This wilt however happen frequently if each isolated occurrence of (new) infection is interpreted as'recrudescence'. Hence, we need some criterion to discriminate benveen'real recrudescence' and 'false alarm,. An intuitive solution is to determine a threshold value for some epidemiological indicator: beyond this threshold we have recrudescence and we need to intervene, below it we need not do anything. This generates the following questions: which indicator is useful, what is an appropriate threshold, and - especially - when should we determine this epidemiological indicator in the community? 1.2 Basic ideas of post-larviciding surveillance The statement of the problem in the previous section reveals that there exist conflicting preferences with regard to detection and control. In order to be sure that recrudescence (including those cases which proceed fast) can be controlled by mectizan, early detection is required. In contrast, to be sure that real recrudescence can be discriminated from occasional infections, the epidemiologicat indicator should be measured much later. The solution to this problem is to use a longitudinal approach. Rather than makingjust one single epidemiological observation, regular post-larviciding surveys are required. The first surveys will lead to the detection and successful control of cases of fast recrudescence, later surveys will detect the slower cases of recrudescence, which can be controlled even after a considerable delay. The advantage of the longitudinal method 3in comparison with the single-survey cross-sectional method will be illustrated with a number of numerical examples. These examples have been generated by model simulations. Simulation of a fiercoura-like village The basic situation that has been simulated is a period of 12 years and 3 months of successful vector control in a formerly hyperendemic village, comparable to the village of Tiercoura (Burkina Faso, I-erab4 pre-control CMFL t70 mf/s). This situation has been simulated 600 times,lelding 107 cases of recrudescence (hence the risk is lSVo\. Of course, most likely, this basic situation does not have an equivalent in the real world. Many villages differ from Tiercoura (which is one of the highest endemic places in the OCP area) and larviciding will not already stop after such a period. However, although this artificial representation can be considered as a rather pessimistic view of reality, it nevertheless illustrates the problem very well, and allows us to make general conclusions. Single survey approach: prevalence of infection As an example, suppose that the (single) survey is planned 5 years after cessation of control, and that the prevalence of infection (positive skin-snip) in adults ( > 20) is used as an epidemiological indicator for recrudescence. We have found that, in order to detect all 107 cases of recrudescence with that single survey, we must use a threshold value of 2Vo.In other words, in all situations where the adult prevalence exceeds2Vo, mectizan-based intervention is started. The problem is that, with such a decision rule not only the cases of recrudescence are successfully controlled, but that intervention is also started in more than 65%o of the situations where no recrudescence occurs. If this single survey is done 8 years after cessation of control, then a prevalence of.2.SVo can be used to detect all recrudescence. Now in 45Vo of. the situations mectizan treatment is started without the need. However, a serious problem is that in a few cases the detection of recrudescence took place too late to be controlled by mectizan. This example stresses the need for a longitudinal multi-survey approach. Iangitudinal approach: incidence of infection Fig. 1 shows the step-wise detection of the events of recrudescence by means of a multi-survey approach. Now, the indicator for detection is the incidence of infection -4- among adultsl. Three different threshold values for this indicator have been used. The interval benveen surveys is 4 years, the first being done at the moment of cessation of control. In Fig. la the results are shown for a threshold value of l.75Vo. Hence, urs soon as at a given survey the incidence of infection (in comparison with the previous survey) exceeds a value of. L.7SVo, the community is treated. Control of recrudescence is done by annual mectizan treatment during 15 years, and with a constant mean coverage of 65Vo. The graph shows that at each survey a number of the 107 recrudescence cases are detected, and after about 30 years most of them have been detected and controlled. In all cases this recrudescence control has been successful, which can be concluded from the graph representing "treatment too late" which is zero all the time. Also the number of cases where control started unnecessarily ("no recrudescence") is negligible. Fig. 1b shows what would happen if a lower value for the incidence had been used as an indication of recrudescence (7.25Vo instead of l.75Vo). Now, most of the cases are detected earlier than in the previous example. But, as could be expected, there are a considerable number of situations where treatment was started without any need. A higher value for the incidence of infection is also sub-optimal. This is shown in Fig. 1c where the threshold value is increased from 1.75 to 2.SVo.Now, it not alone takes more time to detect most of the recrudescence cases, but there is also an unacceptable number of situations where the detection comes too late to successfully control the recrudescence. Hence, 1.75 seems the best value for this moment. However, we still cannot say that an annual incidence in adults of L.75Vo at a survey interval of 4 years is a good value under all circumstances. This is demonstrated in Fig. 1d. The only difference between Fig. 1a and Fig. 1d is that during the mectizan campaign for controlling recrudescence, the coverage of apptications decreases from65Vo at the first round of treatments to 35Vo at the last. Due to this less effective control of recrudescence, again in a considerable number of cases the detection comes too late. I Defined as: (100 x a) / (b x c) with: a = No. new positive adults b = Time (years) since previous survey c = No. negative adults at previous survey -5- oN !? ol @(\. ! ot o6r o 6r o ! o E& 8E 5 .T T SE aIo> o t E:988R88388eo o H gf; HHct o€27) -- EEE Erg r+t ooc ooaoEft-ooL H-o olb+.C oo EC(0 C o *.o o+.oo o) iE +(\r !? o (oN E:888R88888eo pcyrlr ru6pdurec ueccurrc l lo'oN E:g88R88e88eo E:g88R8888Reo pclistr rr0Fdurc uppeur.^l lo'oN @ GI ttN I o o| E RE E .T E SE oIo> t la : A 23 3 !tI .t T 5 EI 5o et t I I {) : A E 3 EI!I E 5 -E!c 6 I .ih ut ot A Af E ! flt E Ed EE 5o It t lO . Ag) 3 ! b3I d) E a E!c 6 =t oo -6- 1.3 A problem of optimization It is clear that many considerations enter the design of a stratery for post-larviciding surveillance. An optimal strategy has several characteristics. In the first place two potential failures must be minimized: Mectizan-based intervention is started in situations where the (newly) detected infections are not a sign of recrudescence and would have disappeared automatically. In addition, one would preferably detect recrudescence as soon as possible. Finally, the number of post- larviciding (and at a given moment also post-OCP) surveys should be minimized. It has already been mentioned that, to a certain extent, these requirements are conflicting, and therefore it is necessary to assign a relative importance to each of them. The problem of optimization is illustrated in Table 1. In the analysis presented there, an optimal strategy is defined as follows: 1. The risk of starting mectizan too late must be less than 1zo. 2. The number of unnecessary treatments must be minimized. 3. The post-larviciding period needed for detection of all recrudescence cases must be as short as possible. The order of importance is followed strictly. Hence condition 1 determines the margins left for conditions 2 and 3. Condition 2 must be satisfied to pay attenrion to condition 3. To keep it simple, in this definition we did not include the number of surveys, which of course should be taken into consideration when thinking about actual costs and resource allocation. Four different scenarios are assumed for the control of recrudescence: A. Annual mectizan treatment with a population coverage decreasing from 65Vo in the first year to only 35Vo at the last treatment. B. Annual mectizan treatment with a constant coverage of 65vo. C. Six-monthly mectizan treatment with a coverage decreasing from 65Vo to 35Vo. D. six-monthly mectizan treatment with a constant coverage of 65vo. In all scenarios the total duration of the mectizan campaign is 15 years. -? - Giventhe conditions (definitionof anoptimalstratery) and thescenarios fortreatment, Table 1 gives the optimal threshold values for the incidence of infection among adults (220) for several intervals benreen surveys. Like in the example of Fig. 1, the first survey is done just before cessation of vector control. For mectizan scenarios B,C, and D suivey intervals up to 8 years are possible. Scenario A permits intervals to only 4 years. l-onger interval will lead to more than lVo risk of failure, irrespective of the threshold value. For the different intervals and scenarios also the percentage of situations is given where treatment is started unnecessarily. The first general conclusion is that the optimal threshold decreases with increasing interval between surveys. The second conclusion is that higher thresholds are allowed if a better mectizan strategy can be assured. This second conclusion makes sense intuitively. If mectizan is more effective, the decision about the occurrence of recnrdescence can be postponed until the incidence has risen to a higher value and the difference with occasional new infections becomes more apparent. The best explanation of the relative high threshold values to be used at short intervals benveen surveys is that shortly after cessation of vector control a peak in the incidence of infection can be expected. The reason for this peak is a short period of renewed microfilarial production of parasites. By the end of larviciding the probability of harbouring more than one parasite has become low. The probability of having nrro worms of opposite sex (a condition for microfilarial output) is even lower. The short period during which the flies can transmit after cessation of control will in a few (especially highly exposed) persons cause pair formation of worms: those having only (a) female worm(s) can get a male worm and vice versa. However, since the original parasite population is dying rapidly, this period of increased mating chances is relatively short, explaining the rapid decrease in incidence short after the initial peak. In many cases this initial increase in incidence is not sufficient to cause recrudescence. However, if asurveyis doneat the momentof peak incidence, the threshold valueusedas indicator of recrudescence must be high in order to avoid mectizan treatment where this is not necessary. -8- Table 1. Optimal threshold values of the incidence of infection in adults for the detection and successful control of recrudescence. See text for description of mectizan treatment scenarios A,B,C and D. Interval (years) Optimal value for scenario o/o unnecessary treated for scenario: A B c o ABCD 1 2 3 4 5 6 7 8 3 2.25 1.75 0.75 n.a n.a n.a n.a 4,25 2.5 2.25 1.75 1.5 1.25 0.75 0.75 5.75 3.75 3 2 1.75 1.5 1.25 1.25 7 3.75 3 3 1.75 1.5 1.25 1.25 21.1 5.1 4.3 34.7'. n.a n.a n.a n.a 4.5 2.8 0.6 0.4 0.2 0.4 7.7 5.7 1.4 0 0 0.2 0 0 0 0 0 0 0 0 0 0 0 0 ' This unexpectedly high percentage has to do with the rigid delinition ol 'optirnal strategy'. Using 1o6 incidence at scenario A would lead to slightly more than 10,6 ol cases where treatment comes too late. Therefore, the value of 0.75o,6 was chosen, which leads to the very high percentage ol treated unnecessarily. Table 2. lmpact of several actions on three different aspects of the detection and control of recrudescence (-=improvement, +=deterioration). . on Action: v No. unnecessary Troated No.treat€d too late. Total period of det€ction lncreasing the interval between surveys + + Oecreasing the interval betwoen surveys + lncreasing the incidence threshold threshold e nc n ategy strategy nm€ct I1.4 Discussion Table 1 supports the statement made earlier that strategies for detection and control of recrudescence are not straightforward. On the basis of this table a tentative guideline has been formulated and presented to EAC 1991. This guideline reads as follows: An annual incidence among adults of 1 to 1.5 Vo determined at survey intervals of 3 to 4 years can be regarded as an early sign of recrudescence. The values in this guideline have been based on the last effective mectizan treatment strategy (scenario A). Assuming this most pessimistic scenario (is it really pessimistic?) gives rise to the rather low values of the incidence threshold, higher values or longer periods benveen surveys can be chosen if higher treatment coverage or frequency can be assured. Clearly, the different building blocks of what might be considered an "optimal stratery" are affected differentially by several factors. In Table 2 these factors (actions) are given together with their likely impact on: 1.. The total period required to detect all cases of recrudescence,2. The number of situations where treatment is started without any need, and 3. The number of treatment campaigns failures. Of course, a " - " does not always cause an increase in a given variable: If the expected number of unnecessary treatments is already zero, increasing the interval does not have any effect on this variable. Apart from the factors mentioned in Table 2there are a lot of additional points which need to be kept in mind and which complicate the process of decision making in actual field situations. Some of these complications could be systematically investigated with the model. Others are so uncertain and different from place to place that impact on decision making must be judged on the basis of expert opinion and intuition. The cause of recrudescence In the model-based analyses done so far, the cause of recrudescence is the existence of a residual reservoir of parasite infection which remains after a given period of 1007o successful vector control. Recolonizing flies become infected by this reservoir and in a number of situations this will lead to recrudescence. In order to arrive at a reasonable risk of recrudescence in the simulations, the prevalence of infection must still be fairly high (but falling rapidly). Since, however, in OCP-covered areas vector control will never be stopped if the prevalence of infection has not been reduced to low values near -10- zero, it is not likely that this type of recrudescence will be the rule. Instead, if it occurs, it will probably be caused by inadequate vector control (which will be responsible for a long "tail" in the epidemiological trends due to fresh infections), by re-invasion of infectious flies, or by immigration of infected persons after cessation of vector control. The impact of these alternative causes of recrudescence will be investigated. Especially, the combination of these alternatives with a lower risk of recrudescence will probably provide more realistic scenarios. The dynamics of recrudescence The results presented in this report are based on simulations of an area with extreme high pre'and post-control biting rates (Tiercoura). This implies that the dynamics of recrudescence will be profound. Even fairly low values of incidence may already be a warning of recrudescence, and higher values will cause a situation which cannot be controlled by mectizan. An extension of the present analysis could be the simulation of situations with more moderate levels of the fly biting rate (but still causing hyper-endemiciry in the absence of control) such as Folonzo, Awahikro, etc. It is to be expected that in such areas higher values of the incidence of infection can be tolerated before control must be launched. Apart from the endemicity of the village, the population size may also play a role. In the simulations the village size at the moment of cessation of control is about 300 individuals. The effects of alternative population sizes could be further investigated. Epidemiological surveillance At several points the simulation of epidemiological surveillance and the subsequent decision making may not be very realistic. The following assumptions have been used: ' The only epidemiolo$cal mec$ure of infection, used throughout the wlrcle post-control peiod b the skin-snip microfilaial count. Although the skin-snip has been a key tool in the epidemiological evaluation of vector control, it is questionable whether this diagnostic method wiltbe routinely used in the post-OCP period. Although taking a skin-snip is relatively painless, it is an invasive technique. It will, therefore, become difficult to find a sufficient number of volunteers in a situation where infections are rare and where the people no longer suffer from the disease. Immunodiagnosis will only be an alternative when simple (non-invasive) field tests become available, and when high specificity and acceptable sensitivity can be - l1 - assured. Other alternatives may be nodule palpation and screening for morbidity. However, both of these alternatives will not be very sensitive, and will provide an indication of recrudescence at a stage which no longer enables mectizan-based intervention. Some exploratory simulations have been done with an imaginary highly sensitive immunodiagnostic test, reacting with all O. volrnrlus related antibodies (originating from microfilariae, immature and mature male and female worms). Ageneral problem thatwas highlighted by these simulations is that, although the sensitivity is much higher than the skin-snip and infection can be detected at an earlier stage, it is still a problem to discriminate benveen real recrudescence and occasional infections. This will especially be a problem when antibodies against the mature parasite and microfilariae have a long half-life. tn that case the immunodiagnostic test will only detect real new infections in children born since the start of OCP. - All inhabitants of a village participate in surueillance This has never been the case in OCP conducted surveillances, and will not be the case in future. A coverage of less than l00Vo. with random probabilities of participation, will have implications for the incidence of infection. These implications could be investigated with the model. - A village is an isolued unit, spatial aspects (e.9. the epidemiological situation in sunoundingvillages) are not included in the model This has always been one of the limitations of the model, which is difficult to solve. Since recrudescence will not be a problem of a single village, but for a whole area close to a breeding site, it is likely that post-control surveillance will follow some kind of rotational schedule. For example, if a survey is done in a given village, then in the following year it will be done in a neighbouring village, and the next year again in another village. By the time a second (or third) survey is done in the first village, a lot of additional information is'available, which can be utilized in the process of decision making. -t2- ' A single epidemiological indicator is used: the incidence of infection among aduks. If this indicator is uceeded, then mectizan treatment is started immediuety. The concept of detection and control used in the simulations is rather rigid, not only because of the previous point (disregarding information of surrounding villages), but also because of the fixed period benueen surveys and the simple decision rule which relies on h single epidemiological indicator. Although by definition the incidence of infection is based on a pair of surveys, in fact only the results of the second survey play a role in the decision to start control or not. Undoubtedly, also the information (e.g. prevalence or intensity of infection) of the first survey of such a pair (which could be the very first survey after cessation of control) will be taken into account. A further simplicity of the present approach is related to the decision itself: mectizan treatment stafts immediately when the threshold is exceeded. A kind of nro step procedure would be much more suitable. For example, a given value of the incidence of infection is considered as a possible indication of recrudescence. Instead of blindly starting intervention, this indication is used as a reason to intensiff the surveillance schedule. If subsequent surveys (either in the same village or in surrounding villages) confirm the initial suspicion of recrudescence, then control could be implemented. ' The vaiation in the skin-snip count of one individual k descibed by a Poisson distibutiort Though this assumption cannot be rejected on the basis of available data, there is some indication that the skin-snip is more available. In case of higher variability, the probabiliry of false negative diagnosis increases (especially at low intensities). False negatives at a given survey may artificially increase the incidence between that survey and a subsequent one. The aspect of variability in skin-snip could be investigated with the model. 'The definition of an optimal strateg for decision and control of recnrdescence is largety dominated by the two failures: "mectizan campaign too late" and "unnecessary mectizan campaign". This is a complicated problem. In the examples shown so far, a very hierarchical concept of an "optimal strategy" is used. For example, the total number of surveys is not included in the definition. An important extension of the analysis could be the explicit consideration of relative costs: what are the costs of an epidemiological survey when compared with the costs of a mectizan treatment failure, etc. tI l3 The effect of mectizan treatment The present analysis stresses that mectizan treatment, though a very promising tool for morbidity control, will only succeed in controlling low-level transmission in an early stageof recrudescence. Of course, the exact implicationof thisgeneralstatementlargely depends on the assumptions about the power of the drug in reducing microfilarial levels in treated individuals and the effect of treatment on the productivity of adult worms. We have attempted to quantiry the model parameters related to mectizan treatment such thatthe modelgives aproperdescriptionof the resultsof theAsubende community trial. At present, the results up to the follow-up after the 2nd treatment have been included into the quantification. We can improve. the parameter quantification by including all data that are now available from the Asubende trial. Much more uncertainty will exist about the scenario of community treatment. We have shown the implications of a restricted number of possible scenarios. These scenarios differ according to the frequency and coverage of operations. One of the extensions of the present analysis could be the analysis of the impact of variations in the total duration of the mectizan campaigrs. Is a period of 5 years of six-monthly treatment with a high coverage as powerfrrl as our scenario A? 2 Human migration An earlier analysis show that if recrudescence occurs in the central OCP area, it will most likely be the result of human migration or invasion of infected flies from elsewhere. The potential consequences of migration of infected humans on an optimal situation of vector control (i.e. perfect vector control, no migratior\ and not much variation in exposure) are simulated. This optimal situation is referred to as "Standard" in the graph of Fig. 2 where it has been attempted to relate the risk of recrudescence to the prevalence of infection (rnf/s) at the moment of cessation of control. The situation that is simulated is a village like Tiercoura, which is relatively homogeneous with respect to human exposure to the flies, and where the epidemiological trends suggest highly successful vector control. In order to arrive at a sufficiently wide range of values for the prevalence of infection at the moment of stopping larviciding, different durations of control varying between 12 (high a-11 - prevalences at cessation, therefore high risk of recrudescence) and 15 years (low prevalences and therefore no recrudescence2l have been simulated whereby infected immigrants have been introduced. Fig.2 shows the results of simulations where, 1 month before cessation of vector control 4 immigrants enter the village, which have a skin-snip count of about 16 to 64 mf/s (village size at that moment is about 300 individuals). In this situation, recrudescence is still probable even at very low prevalences. Fig.3 shows the risk of recrudescence as a function of the duration of vector control, and from this, it becomes clear that human migration deserves separate attention, thus further studies are continuing. F19.2 lmpact ol lmmlgratlon on recrudcsccncc Simulatiorr c{ Tiercou.a lika yillae6 t 0.0 0.8 0.7 0.c 0.5 0..1 o.3 o.2 0.t 0 !II E f I E E f t a 12 16 20 21 2A Prwalorer ol nl ln tot l poputatbn Flg.3 lmpact ol lmmlgratlon on rccrude.cencc a lmmlerari! with rf loart rango t6 to a4 b.fo.tr il.dbl ol @fltfd \ \ \ \ \ 1.4\-F--------! l3 ta 15 l0 Yeart cl vadoa control 2 This only app-lies to these circumstances. In the other examples even at l5 years of vectorcontrol recrudescence may occur. 0 I E I e E €I a ht E E I t 0.9 0.8 o7 0.0 0.5 0.4 03 o2 0.r 0 M olJ a '..6r.d"n/- _aI StlndardlY -l / / t2 --7--. 5