NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Forum on Drug Discovery, Development, and Translation; Board on Health Sciences Policy; Institute of Medicine. The Global Crisis of Drug-Resistant Tuberculosis and Leadership of China and the BRICS: Challenges and Opportunities: Summary of a Joint Workshop by the Institute of Medicine and the Institute of Microbiology, Chinese Academy of Sciences. Washington (DC): National Academies Press (US); 2014 Feb 28.

Cover of The Global Crisis of Drug-Resistant Tuberculosis and Leadership of China and the BRICS

The Global Crisis of Drug-Resistant Tuberculosis and Leadership of China and the BRICS: Challenges and Opportunities: Summary of a Joint Workshop by the Institute of Medicine and the Institute of Microbiology, Chinese Academy of Sciences.

Show details

8Rapid Diagnostic Technologies: Status and Limitations

Key Messagesa

  • The rapid detection of drug resistance using molecular diagnostics poses challenges that vary depending on technological and genetic factors.
  • Sensitivity and specificity vary by test and patient group and can be critical determinants of whether a diagnostic test is appropriate in a given setting.
  • Susceptibility testing of some FLDs and most SLDs can be difficult to standardize, and results often vary across high-quality laboratories.
  • Decentralized testing can reduce turnaround time, lower initial default rates, improve communication, and require less reliance on logistics. On the other hand, centralized testing can use sophisticated technology, encompass a broad range of tests, and provide better quality assurance.

Identified by individual speakers.

As emphasized by the speakers on infection control, whose presentations were summarized in the preceding chapter, rapid diagnostic tests are essential to starting MDR TB patients on effective treatment regimens quickly and stopping the spread of the disease. Seven speakers at the workshop examined the state of rapid diagnostic technologies. These technologies can enable more rapid detection of resistance, but they need to be applied within strong systems that can identify potentially drug-resistant patients, test them quickly, and start them on effective therapies.


The basis for drug resistance in TB is the acquisition of chromosomal mutations, termed genotypic resistance by Mark Nicol, Wernher and Beit Professor and Head of the Department of Medical Microbiology, University of Cape Town and National Health Laboratory Service of South Africa. In some cases, those mutations give rise to phenotypic resistance—the relative inability of M.tb. to grow in the presence of an antibiotic. In turn, phenotypic resistance can be associated with clinical resistance, which is a failure of the patient to respond to antibiotic therapy.

Genotypic resistance can be detected by molecular testing using GeneXpert, LPA testing, sequencing, or some other means. Phenotypic resistance is detected by culture-based methods—the failure of the organism to grow in the presence of antibiotic. Clinical resistance to a single drug is usually undetected (because combination therapy is used), but may result in treatment failure or relapse.

Genotypic susceptibility testing is straightforward for some drugs. With rifampicin, for example, almost all resistance is associated with mutations in a very small region, just 81 base pairs, of the rpoB gene. Screening that region of the gene can identify rifampicin-resistant strains rapidly and accurately. However, a major problem with genotypic resistance is that the relationship between the genotype and the phenotype may not be clear, especially for some drugs. With some second-line injectable drugs, for example, the drug-resistant population has a higher frequency of certain mutations, but these mutations may also be found in some drug-susceptible strains. A given mutation is not always associated with resistance, whether because of the genetic background of the strain or some other factor. In addition, many of the mutations giving rise to resistance to some drugs remain unknown.

Mutations leading to resistance also demonstrate geographic variability. In Russia, for example, a high proportion of kanamycin resistance is associated with a particular mutation, but that is not the case in South Africa. Thus, a test that works well in South Africa may not work at all or may perform poorly in Russia.

People also can be infected with more than one strain of M.tb. Whereas phenotypic tests are quite effective in detecting mixed populations of drug-sensitive and drug-resistant bacilli, most genotypic methods are relatively insensitive for this purpose.

Phenotypic tests also pose a variety of problems. They are slow, they rely on culturing the organism, and they can be complex to apply for some drugs. For example, a patient may have low- or high-level resistance, so testing a drug at a single concentration may fail to identify low-level resistance. Furthermore, the resistance levels conferred by different mutations can overlap, thus complicating the identification of a critical concentration for testing at which susceptible strains will not grow but resistant strains will (Böttger, 2011). Biosafety is also a concern when working with large populations of drug-resistant bacilli. Finally, the main problem with clinical resistance is that it is detected only when treatment has failed.

Nicol spoke about LPA testing and GeneXpert testing, which have been implemented sequentially in South Africa as part of large rollout programs. Both techniques have been largely automated, although they still require well-trained personnel. For example, the LPA requires well-designed laboratories and stringent laboratory control to prevent contamination. It has been used successfully in well-run reference or large regional laboratories, said Nicol, but it does not perform as well when the testing is decentralized. “In South Africa one of the mistakes we made was to try to decentralize this assay too far, when in fact it needed to be kept in the centralized facilities for quality-assurance purposes,” Nicol said.

The introduction of LPA testing in South Africa had positive although incremental effects on treatment delays, Nicol reported. Data gathered from a low-income housing area on the outskirts of Cape Town showed that its use reduced the median time to treatment initiation for smear-negative patients from 88 to 69 days and for smear-positive patients from 59 to 37 days. The problem in this part of South Africa is that only 30 percent of cases are smear-positive because of high rates of HIV infection, so even with LPA testing, many patients are started on treatment too late. For example, a survey of 73 patients from January 2008 to June 2009 showed that more than half had died while waiting for TB treatment, with a median time of death of 25 days from sputum sampling. “Clearly we needed to do better,” said Nicol.

GeneXpert, which uses overlapping DNA probes to detect the presence of mutations, also is highly automated and is relatively easy to use. In evaluation studies, it had high sensitivity and specificity for detection of rifampicin resistance. When it was used for patient management, however, its specificity was somewhat lower because some patients were found to be rifampicin-resistant with GeneXpert but rifampicin-sensitive with the LPA. The problem is that rifampicin-resistant TB is relatively uncommon, which means that GeneXpert will produce a fairly high proportion of false positives among those cases identified as having rifampicin resistance. Redesign of the test has since improved its specificity.

With GeneXpert, the time between collection of a sputum sample and diagnosis of DR TB was less than 24 hours in demonstration studies in Cape Town. For an LPA performed on a culture sample, the time was around 3 weeks; for an LPA performed directly on a smear-positive sputum sample, it was about 1 week.

In an unpublished study comparing GeneXpert with smear microscopy plus a culture if the smear was negative and a patient was HIV positive, 80 percent of patients with positive GeneXpert results were started on treatment within 1 week. In the routine arm on the study, only about 50 percent of patients were being treated by day 30. However, a high initial default rate, which has been a problem in South Africa, occurred among patients in both arms of the study. Even in the GeneXpert arm, almost 20 percent of patients with a positive test had not been started on TB treatment after 2 months. “You can have the best diagnostic in the world, but unless you have a program that is able to link results … to appropriate care, you're wasting your time,” said Nicol.

Shortly after the WHO recommendation on diagnostics was promulgated, the minister of health in South Africa made the bold decision to implement GeneXpert testing as a replacement for smear testing. This was a correct and well-informed decision, said Nicol, but the strategy has been complicated by problems with specificity. First, GeneXpert is performed on all patients, not just those suspected of having DR TB. If the GeneXpert results are positive and patients are resistant to rifampicin, they are referred for MDR treatment immediately rather than waiting for confirmatory testing by LPA. This approach again raises the issue of false positives, but Nicol argued that such patients are relatively few in number and can be addressed through other means.

Through the third quarter of 2012, almost 1.5 million GeneXpert cartridges had been procured under concessional pricing, almost half of which came from the South African public sector. At that point in time, the country had more than 100 sites for GeneXpert testing, which had detected about 7,800 rifampicin-resistant cases. The results of this testing have provided an accurate picture of the rifampicin resistance rate in South Africa and, by implication, of the rates of MDR TB. The results also have revealed striking variation among provinces, relative differences that have remained stable since the test was introduced.

The use of GeneXpert in South Africa has reduced to 12 days the gap between collection of the first sputum sample and initiation of treatment for MDR TB. Nicol called this “a tremendous advance” that will make “an enormous difference in terms of infection-control issues” and lead to a reduction in the transmission of MDR TB.

The GeneXpert-driven algorithm still has several important limitations. It does not include routine isoniazid susceptibility testing, although the significance of this omission is subject to debate. The South African algorithm is complex, especially because it calls for collecting multiple samples from patients. GeneXpert also is more costly than culture-based methods, but it compares favorably with some of the noncommercial phenotypic tests if one considers the infrastructure costs associated with culture-based methods.

The greatest bottleneck currently is the delay in second-line DST. Once patients have been identified as resistant to rifampicin, there is a considerable delay in culture-based susceptibility testing for SLDs. Resistance can be amplified if such patients are given inappropriate or suboptimal MDR TB treatment.

Technologies are being developed to overcome this problem, including microarray systems, phage-based assays, and new phenotypic assays. Nicol focused on three of these new technologies:


A second-line LPA targets fluoroquinolones, ethambutol, and injectables. The problem with this test is that its performance varies in different parts of the world depending on the prevalence of resistance mutations. Its sensitivity and specificity also vary for SLDs. Thus, the test can rule in but does not do well at ruling out XDR TB. Also, it can identify a proportion of patients with XDR TB, but that proportion varies geographically.


An XDR GeneXpert assay is being developed that targets the same genes as the second-line LPA. However, a decision has not yet been made on whether to continue its development.


Various kinds of targeted and whole-genome sequencing can identify specific mutations, can distinguish missense from silent mutations, and may detect mixed allelic variants. However, sequencing requires well-trained staff and costly equipment and poses a bioinformatics challenge when used in routine diagnostics.

A major tension in South Africa is between centralized and decentralized susceptibility testing. There is pressure to decentralize testing, which can reduce turnaround time, lower initial default rates, improve communication, and require less reliance on logistics. On the other hand, centralized testing can use sophisticated technology; allow for a broader range of tests; and provide better quality assurance, which can make it more cost-effective.


Rapid Diagnoses Using the Simultaneous Amplification Test2

Jin Chen, Chair, Department of Clinical Laboratory Science; and Deputy Director, Shanghai Key Laboratory of Tuberculosis, Tongji University School of Medicine, Shanghai Pulmonary Disease Hospital, described a new technology used for rapid TB diagnostics known as the simultaneous amplification test (SAT). An isothermal RNA amplification assay for M.tb., it has been used in many Chinese TB clinical laboratories with good results. Its target for amplification is the 16S rRNA in the bacterium, which enables the test to perform live bacterial detection. The test has a high sensitivity because one cell has thousands of 16S rRNAs. It has a simple protocol because RNA purification is not required. It also has low cross-contamination, unlike PCR amplification, and can perform rapid amplification under isothermal conditions in just 40 minutes. The total procedure time is about 2 hours once the sputum has been processed, and the results clearly indicate whether a sample is positive or negative.

Based on data collected in his laboratory using 177 sputum samples from TB patients and 67 sputum samples from other lung disease patients, Jin Chen concluded that the SAT is more sensitive and specific than the MGIT 960 mycobacteria testing instrument. For example, the SAT is much better than the MGIT 960 at detecting false negatives.

In a test of rifampicin-sensitive and –resistant and isoniazid-sensitive and -resistant strains, the SAT was clearly able to detect drug resistance. Simple, rapid, cost-effective, and reliable, the technology is suitable for resource-limited laboratories and has already been approved by the State Food and Drug Administration of China.

Using GeneXpert to Detect Drug Resistance3

Yao-Ju Tan, Director, Clinical Laboratory Department, Guangzhou Chest Hospital, described an evaluation of the Xpert MTB/RIF test and the effects of gene mutations in DR TB. Xpert MTB/RIF is an automated molecular test for M.tb. and resistance to rifampicin that can be performed in less than 2 hours. It uses real-time PCR and requires minimal skill and training, according to Yao-Ju Tan. Its performance has been evaluated in a number of countries, including Azerbaijan, India, Peru, and South Africa.

To evaluate its performance in China, sputum samples from 613 patients were analyzed both with the Xpert MTB/RIF test and by conventional DST. The test's sensitivity for detecting rifampicin resistance was 90.9 percent, its specificity was 80.7 percent, and the concordance rate was 86 percent. The positive predictive value was 83.7 percent, and the negative predictive value was 88.8 percent; however, the sensitivity for detecting resistance in smear-negative and culture-positive sputum was only 46.8 percent.

The performance of the test also was evaluated using bronchoalveolar lavage fluid (BLF) from 90 patients. In this case, the test's sensitivity for detecting rifampicin resistance was 90.2 percent, its specificity was 71.9 percent, and the concordance rate was 82 percent. The positive predictive value was 80.4 percent, and the negative predictive value was 85.2 percent. The sensitivity for detecting resistance in smear-negative and culture-positive BLF was higher than with sputum, at 76.5 percent.

The sensitivity for detecting M.tb. was 98 percent in smear-positive and culture-positive sputum and 51.6 percent in smear-negative and culture-positive sputum—more sensitive than smear microscopy. The sensitivity was 98.9 percent in smear-negative and culture-negative sputum. Using BLF, the sensitivity for detecting M.tb. was 97.6 percent in smear-positive and culture-positive samples and 82.4 percent in smear-negative and culture-positive samples—also more sensitive than smear microscopy. The specificity was 93.8 percent in smear-negative and culture-positive samples.

Yao-Ju Tan also discussed the characteristics of mutations in the rpoB gene. Some rifampicin-resistant clinical strains are resistant to rifabutin, which WHO has recommended for the treatment of MDR TB. Sequencing of the rpoB gene suggested that these different resistance patterns are due to mutations in different positions or in different combinations in the gene. In particular, the beginning region of the gene may confer both rifampicin and rifabutin resistance, which may make it possible to discriminate rifabutinsensitive cases and treat them accordingly.

Finally, Yao-Ju Tan spoke about mutations in the pncA gene, which encodes an enzyme that activates the anti-TB drug pyrazinamide. Mutations in this gene result in reduced or lost activity of the enzyme and are considered the primary mechanism of pyrazinamide resistance in M.tb. Also, mutations in the rpsA gene, which encodes a ribosomal protein, can confer increased pyrazinamide resistance. In 120 clinical isolates from the Guangzhou Chest Hospital, the mutations in pncA and rpsA were detected by DNA sequencing. The greatest number of mutations occurred in pncA in pyrazinamide-resistant isolates. However, 29 pyrazinamide-resistant isolates had no mutations in the two genes.

Yao-Ju Tan concluded by noting that the mutation frequency of pncA in pyrazinamide-resistant isolates in China is much lower than that seen in Western countries. This may be the case because most isolates in China have low-level resistance to pyrazinamide.

Use of the LPA in China4

Hairong Huang, Deputy Director, National Clinical Laboratory on Tuberculosis, Beijing Chest Hospital, described the use of LPA testing in China. There, two LPA techniques are commonly used: the Hain test from Germany and a domestically produced test called the CapitalBio Microarray. The two tests are similar, both relying on PCR amplification, hybridization, and machine determination of results, although they use different probes.

The Hain test, which is endorsed by WHO for the diagnosis of DR TB, produces similar results in different countries for sensitivity to rifampicin resistance. But its sensitivity for isoniazid resistance varies among countries because the probe covers only some gene mutations.

In 2009, data on the Hain test became available from China, showing that the test's sensitivity is 88 percent for rifampicin resistance but only 80 percent for isoniazid resistance. This finding makes sense, said Huang, because mutations affecting isoniazid resistance in China are different from those in other countries.

Huang also presented unpublished data from four city-level laboratories in four provinces concerning the test's sensitivity for rifampicin resistance as compared with traditional DST. In these cases, the test's sensitivity was relatively low, ranging from 85 percent to 59 percent, although the specificity remained high.

Huang attributed the difference in sensitivities for rifampicin resistance to the difference between research and actual practice. A given test depends not just on technology but also on the skills and performance of the technicians administering it. Even a good technique can produce unimpressive results. These results can complicate the treatment of patients, said Huang. Because the positive predictive values of LPA tests are relatively low, physicians may not know the best way to treat patients. For new cases of MDR TB, for example, the positive predictive value can be as low as 50 percent, which “is really a disaster for a doctor,” said Huang. “You cannot make decisions.”

Huang urged laboratories to perform traditional DST as well as LPA testing. Doing so means more work for technicians, but is necessary to obtain accurate results until more reliable methods are available. Huang also expressed concern about the long-term sustainability of LPA testing in terms of having the necessary reagents and always being able to cover its costs.

Evaluation of the GenoType MTBDRplus5

Wei Ge, who presented a talk prepared by Haiying Wang, Shandong Provincial Chest Hospital, described an evaluation of the LPA GenoType MTBDRplus. This test was approved by WHO in 2008 for identifying members of the M.tb. complex and detecting drug resistance. Using cultivated samples of pulmonary smear-positive patient material, it can identify resistance to rifampicin and isoniazid in about 5 hours.

In the evaluation, 305 stored AFB (acid-fast bacilli)-positive specimens and 118 isolates were submitted to both GenoType MTBDRplus testing and traditional DST. For the clinical isolates, the sensitivity and specificity were 100 percent for rifampicin resistance and 88.7 percent and 87.0 percent, respectively, for isoniazid resistance. For the AFB-positive specimens, the sensitivity and specificity for rifampicin resistance were 91.2 percent and 95.4 percent, respectively, and for isoniazid resistance were 85.7 percent and 92.1 percent, respectively.

Ge concluded by noting that the GenoType MTBDRplus assay has been validated as a rapid and reliable first-line diagnostic test on isolates or AFB-positive specimens for isoniazid and rifampicin resistance in Shandong. Future work includes rechecking all isolates and specimens for which discordant results were obtained with GenoType MTBDRplus and conventional DST.

Use of the GeneChip in China6

Pang Yu, Associate Professor, National Tuberculosis Reference Laboratory, National Center for Tuberculosis Control and Prevention, China CDC, described an evaluation of the use of GeneChip for detecting DR TB in China. GeneChip offers a set of analytical platforms that provide rapid, affordable, and substantial information at the DNA or RNA level. GeneChip enables the analysis of thousands of genes simultaneously in a parallel manner across samples. It can be used to scan the gene transcriptional profiles, discover new genes, sequence DNA, and analyze mutations.

The Chinese company CapitalBio developed GeneChip for TB diagnosis. The test is based on detecting the most commonly found mutations in the rpoB, katG, and inhA genes and makes it possible to obtain drug susceptibility results for rifampicin and isoniazid in 8 hours.

Comparison of the results of DST and the GeneChip tests in 330 M.tb. isolates and 129 sputum samples produced concordance rates of 91.8 percent for the isolates and 94.6 percent for the sputum samples for rifampicin resistance, said Yu. For isoniazid resistance, the concordance was 70.2 percent for the isolates and 78.1 percent for the sputum samples.

Evaluation of GeneChip for samples from 2,247 smear-positive patients in four cities in different provinces found a sensitivity of 87.6 percent for rifampicin resistance and 80.3 percent for isoniazid resistance, using DST as the gold standard. When GeneChip was used to diagnose MDR TB, the sensitivity was about 73 percent. The positive predictive values for these three outcomes were 84.2 percent, 74.0 percent, and 77.8 percent, respectively.

From these results, Yu concluded that the efficacy of GeneChip is similar to that of imported products, and GeneChip has high acceptability in the hospitals where it is used because of better biosafety. GeneChip can be used for MDR TB diagnosis in China at the city, provincial, and national levels, Yu said.


Many different mutations can lead to the same phenotype of drug resistance, observed Megan B. Murray, Professor, Department of Global Health and Social Medicine, Harvard Medical School. For example, numerous mutations in a number of different genes are associated with isoniazid resistance. In contrast, drug-sensitive TB has very low genetic diversity compared with other bacteria.

Murray and her colleagues have been involved in an effort to develop the TB Drug Resistance Mutation Database, which lists all the mutations found in clinical studies to be associated with resistance and evaluates the strength of the evidence for calling a particular genetic change a drug-resistance mutation. The database contains high-confidence mutations that have been reported repeatedly in large and well-conducted studies, along with a large collection of mutations for which less confidence is warranted because they have not been reported very often or they were reported in lower-quality studies.

Many of the mutations in the database are plausible drug-resistance mutations in that they change molecules in locations that would be expected to affect their function. For example, many mutations that create rifampicin resistance change the rpoB gene in regions that affect the receptor-binding site for rifampicin (although some of the mutations are in locations that appear to be unrelated to the binding site). Similarly, mutations known to cause isoniazid resistance occur in regions involved in the activation and binding of the drug, although mutations occur throughout the molecule involved in activating isoniazid, suggesting that anything that impairs the function of that protein could reduce activation.

With funding from BMGF, the mutation database project launched an initiative to identify mutations and their frequency in drug-resistance genes in M.tb. The project developed an archive of 1,800 well-characterized strains from 9 countries and 6 contributing laboratories, sequenced 28 genes and promoters, and created a public database. In addition, a machine-learning approach was used to identify an optimal set of single nucleotide polymorphisms for diagnosing resistance by drug. Selecting a set of mutations that optimizes sensitivity and specificity, this approach identified 24 mutations, all in rpoB, that yield a sensitivity of about 93 percent and a specificity of 92 percent. The approach also identified sets of mutations yielding good sensitivity for isoniazid, although the sensitivity was not as good for pyrazinamide and ethionamide, nor was it high for either ciprofloxacin or levofloxacin.

Murray identified four gaps in this analysis, which she labeled the “four Es”—errors, epistasis, efflux, and exotic mutations:


Repeated studies comparing the results of DST for FLDs and SLDs across high-quality laboratories rarely produce perfectly concordant results. For pyrazinamide, for example, DST is often inaccurate because the pH of the media can inhibit growth, as can the size of the innocula. But errors can produce a loss of sensitivity in molecular diagnostics, said Murray.


Epistasis, or the interaction of multiple mutations, also can be an important factor. On many occasions, and especially with isoniazid, two mutations, each of which causes low-level isoniazid resistance, together cause relatively high isoniazid resistance. If the partner mutation is not detected, identifying a single mutation can lead to an inaccurate result. Furthermore, multiple mutations can produce a stepwise increase in resistance (Meacci et al., 2005).


Efflux is a mechanism used by bacteria to extrude toxic substances. Exposing M.tb. to one anti-TB drug can cause greater expression of efflux pumps, which can increase resistance to other drugs (Louw et al., 2011). This finding will be explored further in the next few years, Murray said.


As an example of exotic or rare mutations, Murray cited rifampicin-resistant isolates that do not have known rpoB mutations in the rifampicin resistance-determining regions (Siu et al., 2011); rather, they have mutations in a different part of the gene that affects resistance. These variants are rare, so they do not make a large contribution to overall resistance levels, and they may exact a fitness cost that does not occur with the common mutations. However, some studies have found that rare mutations with fitness costs can subsequently be overcome by compensatory mutations (Pym et al., 2002; Comas et al., 2011).

Genes associated with mutations conferring resistance or a growth or fitness advantage in the presence of a drug are likely to be under strong evolutionary positive selection. To explore this process, Murray and her colleagues sequenced multiple strains of M.tb. from around the world, half of which were drug-resistant and half of which were drug-sensitive. Strains that had acquired stepwise resistance to individual drugs over time were oversampled. Using a newly developed technique for detecting selection in the M.tb. genome, the researchers identified a set of genes that are involved in the evolution of resistance. Many of these genes have functions that remain unknown, but most with known functions encode cell wall components or are known to be involved in the cell wall's permeability. Some of these mutants may confer low-level resistance, some may be compensatory mutations correcting for a fitness loss from an earlier drug-resistance mutation, and some may be correcting for metabolic changes. “We don't know,” said Murray. “It opens up a new avenue for research into drug resistance. It also opens up the idea that what we have considered a very black-andwhite, simple process—that organisms acquire drug-resistance mutations and then either are or are not resistant—is a simplification of what actually is going on. There are probably many steps to the acquisition of resistance, and much more to be learned.”



This section is based on the presentation by Mark Nicol, Wernher and Beit Professor and Head of the Department of Medical Microbiology, University of Cape Town and National Health Laboratory Service of South Africa.


This subsection is based on the presentation by Jin Chen, Chair, Department of Clinical Laboratory Science; and Deputy Director, Shanghai Key Laboratory of Tuberculosis, Tongji University School of Medicine, Shanghai Pulmonary Disease Hospital.


This subsection is based on the presentation by Yao-Ju Tan, Director, Clinical Laboratory Department, Guangzhou Chest Hospital.


This subsection is based on the presentation by Hairong Huang, Deputy Director, National Clinical Laboratory on Tuberculosis, Beijing Chest Hospital.


This subsection is based on the presentation by Wei Ge and Haiying Wang, Shandong Provincial Chest Hospital.


This subsection is based on the presentation by Pang Yu, Associate Professor, National Tuberculosis Reference Laboratory, National Center for Tuberculosis Control and Prevention, China CDC.


This section is based on the presentation by Megan B. Murray, Professor, Department of Global Health and Social Medicine, Harvard Medical School.

Copyright 2014 by the National Academy of Sciences. All rights reserved.
Bookshelf ID: NBK195973


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (2.4M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...