Tuberculosis Detection by Giant African Pouched Rats
Total Page：16
File Type：pdf, Size：1020Kb
The Behavior Analyst 2011, 34, 47–54 No. 1 (Spring) Tuberculosis Detection by Giant African Pouched Rats Alan Poling Western Michigan University and Anti-Persoonsmijnen Ontmijnende Product Ontwikkeling Bart Weetjens, Christophe Cox, and Negussie Beyene Anti-Persoonsmijnen Ontmijnende Product Ontwikkeling Amy Durgin and Amanda Mahoney Western Michigan University and Anti-Persoonsmijnen Ontmijnende Product Ontwikkeling In recent years, operant discrimination training procedures have been used to teach giant African pouched rats to detect tuberculosis (TB) in human sputum samples. This article summarizes how the rats are trained and used operationally, as well as their performance in studies published to date. Available data suggest that pouched rats, which can evaluate many samples quickly, are sufficiently accurate in detecting TB to merit further investigation as a diagnostic tool. Key words: behavior analysis, discrimination training, tuberculosis, pouched rats, transla- tional research The article by Critchfield (2011) articles describing how these rats are and the accompanying commentaries trained and used have recently ap- in this issue of The Behavior Analyst peared in behavior-analytic journals highlight the potential importance of (Poling, Weetjens, Cox, Beyene, & translational research in behavior Sully, 2010, 2011; Poling, Weetjens, analysis. Also in this issue, Jones Cox, Mgode, et al., 2010;). Anti- (2011) provides an informative illus- Persoonsmijnen Ontmijnende Prod- tration of this type of research, uct Ontwikkeling (Anti-Personnel specifically the use of behavior-ana- Landmine Detection Product Devel- lytic strategies to teach dogs to detect opment, APOPO), a nonprofit hu- explosives and to evaluate their manitarian organization located in performance in doing so. Jones fo- Morogoro, Tanzania, is responsible cused on scent detection of remote for the use of rats as mine-detection explosives, in which samples that animals. APOPO is also exploring may or may not contain explosives other humanitarian applications of are collected in one area and ana- scent-detecting rats. One that shows lyzed by animals located elsewhere. considerable promise is using the As he notes, dogs also are used animals to detect tuberculosis (TB) operationally for the direct detection in humans. The purpose of the of land mines and other explosive present article is to provide a sum- remnants of war and are sometimes mary of APOPO’s work in this area. used to detect cancers in humans (e.g., Moser & McCulloch, 2010). An Introduction to TB Like dogs, giant African pouched Tuberculosis is an infection, often rats (Cricetomys gambianus) are used but not always of the lungs, caused operationally to detect land mines; by Mycobacterium tuberculosis.As data reported by the World Health Address correspondence to Alan Poling, Department of Psychology, Western Michigan Organization (WHO, 2008, 2009a, University, Kalamazoo, Michigan 49008 (e- 2009b) make clear, it is a serious mail: [email protected]). and common disease. According to 47 48 ALAN POLING et al. the WHO, TB was responsible for the common problem in everyday appli- deaths of 1,770,000 people in 2007. cations in the developing world (Bur- About 2,000,000,000 people currently gess et al., 2001; Hargreaves et al., are infected with TB, and roughly 1 2001; Steingart et al., 2006). For in 10 of them will become seriously ill example, a recent study in Nigeria with the disease. Most people infect- revealed a sensitivity of only 23% ed with TB live in developing coun- (Ani et al., 2009). That is, fewer than tries. The disease is especially com- one in four people with active TB mon in sub-Saharan Africa, where it were detected. is the primary cause of death in people with HIV. Despite this, in Pouched Rats as TB Detectors 2008 only 1% of people with HIV were tested for TB. A major reason In the hope of developing a viable for the scarcity of TB testing in alternative to or adjunct for micros- people with HIV, and in the popula- copy, APOPO is exploring the use of tion at large, is the unavailability of a African giant pouched rats (Cricet- cheap, fast, and accurate test. omys gambianus) to detect the pres- The purpose of any TB detection ence of TB. These large and long- system is to classify people into two lived rats, which are native to much discrete categories: those who have of Africa and have an excellent sense the disease and those who do not. of smell, detect TB by sniffing The sensitivity of a given test refers to sputum samples. They are trained to its ability to do the former, whereas respond consistently in one way its specificity refers to its ability to do (pause) if the sample contains the the latter. At present, the technique TB bacillus (is positive) and respond most widely used to diagnose TB in in another way (not pause) if the the developing world is sputum sample does not contain the bacillus smear microscopy (hereafter micros- (i.e., is negative). Each rat can test copy) (Dye, Watt, Bleed, Hosseini, & hundreds of samples each day, allow- Raviglione, 2005; Steingart et al., ing inexpensive testing. 2006), a technique in use for decades APOPO’s goal is to produce rats and still recommended for low- and that consistently emit an easily ob- middle-income countries (e.g., Cor- served indicator response when they bett et al., 2003; Dye et al., 2005). In smell a sputum sample that contains microscopy, trained technicians view M. tuberculosis and not to emit this slides (typically magnified 100 times) response at other times. In essence, that contain sputum samples stained this is a signal-detection task (Green by the Ziehl-Neelsen (or less often the & Swets, 1966), in which the odor Kinyoun) method (see Steingart et uniquely associated with M. tubercu- al., 2006), which makes Mycobacte- losis constitutes a signal and all other rium tuberculosis, an acid-fast bacil- odors constitute noise. The rats’ task lus, fairly easy to see and count. is to respond only to the signal. This Microscopy has very good specificity, is accomplished by training the rats characteristically above 90% in pub- in an operant stimulus discrimination lished studies (Steingart et al., 2006). task in which the designated indica- The sensitivity of the method, which tor response, pausing for at least 5 s is comprehensively reviewed by Stein- at a hole immediately above a spu- gart et al., varies widely in published tum sample, is reinforced when the studies, ranging from roughly 20% to sputum sample is known to contain 80%. Strategies for improving sensi- M. tuberculosis but not when the tivity have been developed, such as sample does not contain the bacillus. increasing the time that microscopists Such differential reinforcement estab- view slides (e.g., Ramsay et al., 2009), lishes the odor of TB as a discrimi- but low and variable sensitivity is a native stimulus (SD) that reliably RATS DETECT TB 49 evokes the operant response (paus- emphasize that an evidence-based ing), which rarely occurs in its approached to TB diagnosis is es- absence. sential (Pai, Ramsay, & O’Brien, In signal-detection terminology, 2008). That is, high-quality research, emitting the indicator (operant) typically reported in peer-refereed response when the signal (SD)is journal articles, must document the present on a given trial is termed a value of new diagnostic techniques hit and emitting that response when before they are widely adopted. the signal is not present is termed a APOPO recently has published three false alarm. Indicating that the signal studies describing the use of rats as is not present on a given trial, either TB detectors. by withholding the response indicat- In the first study, Weetjens, ing a signal (as in our procedure) or Mgode, Machang’u, et al. (2009) by emitting another response (as in described a procedure in which procedures used by others), is termed pouched rats were trained in a a correct rejection if the signal is not rectilinear cage to sniff each of 10 present and a miss if the signal is holes. A small pot containing a present. Hits and correct rejections sputum sample taken from a patient are correct responses, whereas false at a direct observation of treatment alarms and misses are incorrect. In short course (DOTS) TB center in operational TB systems it is highly Dar es Salaam, Tanzania, was placed desirable to have a high rate of hits immediately below each hole. DOTS and a relatively low rate of false centers are open to all citizens with alarms. respiratory difficulties and are in- By dividing total hits by total hits tended to detect and provide treat- plus total misses and multiplying the ment for TB. The rats were trained result by 100%, one can produce a by operant conditioning to pause for quantitative (i.e., percentage) mea- at least 5 s at holes where the sputum sure of the sensitivity of the rats, or of sample was positive for TB (as any other TB diagnostic. In essence, confirmed by other methods) but as noted previously, sensitivity sum- not to pause at holes where the marizes how good the test is at sputum sample was negative. In detecting patients with TB. Specific- essence, the training procedure in- ity, in contrast, refers to how good volved differential reinforcement, ar- the test is at detecting patients who ranged so that rats (which were do not have the disease. A percentage mildly food deprived) received food measure of sensitivity can be deter- (bananas mixed with food pellets, mined by dividing total correct rejec- both mashed) only when they paused tions by total correct rejections plus for the required time (5 s) at holes total false alarms and multiplying by with positive samples. The training is 100%. As noted previously, an ideal equivalent to that used to establish diagnostic has high sensitivity and the odor of 2,4,6-trinitrotoluene high specificity.