Individuals in the simulation model were assumed to enter human immunodeficiency virus (HIV) care at age 40 years and to remain in the model until death or age 80 years. Current smokers continued smoking until the end of follow-up. Former smokers quit at age 40 years and remained abstinent. “Number at risk” tables are not needed here. The graphs represent model-generated outcomes, each line a simulated cohort of 1 million people entering HIV care at age 40 years. The number at risk are the number of people still alive at each time point, with no other form of censoring.
ART indicates antiretroviral therapy; LTFU, loss to follow-up. The intensity of smoking (number of cigarettes per day) was considered to be moderate (“average”). Current smokers continued smoking until the end of follow-up. Former smokers quit at age 40 years and remained abstinent. “Number at risk” tables are not needed here. The graphs represent model-generated outcomes, each line a simulated cohort of 1 million people entering HIV care at age 40 years. While some people are lost to follow-up from clinical care (in the sensitivity analyses that allow for loss to follow-up), the model still keeps track of them, recording when and how they die, so they are not lost from the model or from the results. The number at risk are the number of people still alive at each time point, with no other form of censoring.
eAppendix. Methods, Results, and References
eTable 1. Validation of Model-Generated Cumulative Lung Cancer Mortality by Age 75 for HIV-Uninfected Current Smokers: Comparison with Case-Control Studies from Western Europe
eTable 2. Validation of Model-Generated Cumulative Lung Cancer Mortality by Age 75 for HIV-Uninfected Former and Never Smokers: Comparison with a Case-Control Study from Western Europe
eTable 3. Cumulative Mortality by Age 80 from Various Causes for People Entering HIV Care at Age 40, by Smoking Exposure, Assuming Complete Antiretroviral Therapy Adherence and No Loss to Follow-up from HIV Care
eTable 4. Cumulative Mortality by Age 80 from Various Causes, by Age at Entry to HIV Care and Smoking Exposure
eTable 5. Cumulative Mortality by Age 80 from Various Causes for People Entering HIV Care at Age 40, by Smoking Exposure, Accounting for Incomplete Antiretroviral Therapy Adherence and Loss to Follow-up from HIV Care
eTable 6. Cumulative Lung Cancer Mortality by Age 80, by Smoking Exposure: Varying the HIV-Associated Independent Risk of Lung Cancer
eTable 7. Effect of Varying the Initial CD4 Cell Count on Model-Projected Cumulative Mortality by Age 80 from Various Causes, by Smoking Status
eTable 8. Model-Generated Cumulative Lung Cancer Mortality and Relative Risk Reduction Based on Age at Smoking Cessation, for Those Entering HIV Care at Age 40
eFigure 1. Overview of Flow Through CEPAC Monte Carlo Microsimulation Model
eFigure 2. Cumulative Mortality by Cause among Male Never Smokers Entering HIV Care at Age 40
Customize your JAMA Network experience by selecting one or more topics from the list below.
Reddy KP, Kong CY, Hyle EP, et al. Lung Cancer Mortality Associated With Smoking and Smoking Cessation Among People Living With HIV in the United States. JAMA Intern Med. 2017;177(11):1613–1621. doi:10.1001/jamainternmed.2017.4349
What is the risk of lung cancer death by smoking exposure for a person living with human immunodeficiency virus (HIV)?
In this microsimulation model–based analysis, cumulative lung cancer mortality by age 80 years for men and women who entered HIV care in the United States at age 40 years and continued to smoke an average number of cigarettes daily was 23.0% and 20.9%, respectively; if they quit smoking, the respective risks decreased to 6.1% and 5.2%. Those who continued to smoke were 6 to 13 times more likely to die from lung cancer than from traditional AIDS-related causes.
Smoking cessation should be a priority in the care of people living with HIV.
Lung cancer has become a leading cause of death among people living with human immunodeficiency virus (HIV) (PLWH). Over 40% of PLWH in the United States smoke cigarettes; HIV independently increases the risk of lung cancer.
To project cumulative lung cancer mortality by smoking exposure among PLWH in care.
Using a validated microsimulation model of HIV, we applied standard demographic data and recent HIV/AIDS epidemiology statistics with specific details on smoking exposure, combining smoking status (current, former, or never) and intensity (heavy, moderate, or light). We stratified reported mortality rates attributable to lung cancer and other non–AIDS-related causes by smoking exposure and accounted for an HIV-conferred independent risk of lung cancer. Lung cancer mortality risk ratios (vs never smokers) for male and female current moderate smokers were 23.6 and 24.2, respectively, and for those who quit smoking at age 40 years were 4.3 and 4.5. In sensitivity analyses, we accounted for nonadherence to antiretroviral therapy (ART) and for a range of HIV-conferred risks of death from lung cancer and from other non–AIDS-related diseases (eg, cardiovascular disease).
Main Outcomes and Measures
Cumulative lung cancer mortality by age 80 years (stratified by sex, age at entry to HIV care, and smoking exposure); total expected lung cancer deaths, accounting for nonadherence to ART.
Among 40-year-old men with HIV, estimated cumulative lung cancer mortality for heavy, moderate, and light smokers who continued to smoke was 28.9%, 23.0%, and 18.8%, respectively; for those who quit smoking at age 40 years, it was 7.9%, 6.1%, and 4.3%; and for never smokers, it was 1.6%. Among women, the corresponding mortality for current smokers was 27.8%, 20.9%, and 16.6%; for former smokers, it was 7.5%, 5.2%, and 3.7%; and for never smokers, it was 1.2%. ART-adherent individuals who continued to smoke were 6 to 13 times more likely to die from lung cancer than from traditional AIDS-related causes, depending on sex and smoking intensity. Due to greater AIDS-related mortality risks, individuals with incomplete ART adherence had higher overall mortality but lower lung cancer mortality. Applying model projections to the approximately 644 200 PLWH aged 20 to 64 in care in the United States, 59 900 (9.3%) are expected to die from lung cancer if smoking habits do not change.
Conclusions and Relevance
Those PLWH who adhere to ART but smoke are substantially more likely to die from lung cancer than from AIDS-related causes.
Antiretroviral therapy (ART) has dramatically improved the life expectancy of people living with human immunodeficiency virus (HIV) (PLWH), with a concomitant shift in morbidity and mortality from AIDS to non-AIDS diseases.1,2 Much of the non-AIDS disease burden is tobacco-related. Over 40% of PLWH in the United States smoke cigarettes, more than double the smoking prevalence in the general population.3-7 Among PLWH undergoing ART, smoking now reduces life expectancy more than HIV itself.8-10
Tobacco use and HIV together may accelerate the development of lung cancer.11-14 The risk of lung cancer is increased by the presence of HIV through mechanisms likely involving chronic inflammation, immunomodulation, and other infections.11,15-19 Lung cancer is the leading cause of cancer death among PLWH undergoing ART and is among the leading causes of death overall in this population.13,20
Despite the high smoking prevalence and the risk of lung cancer and other tobacco-related diseases, smoking cessation programs generally have not been successfully implemented in HIV care. As the population of PLWH in the United States ages, estimates of projected comorbidities can help guide where to direct attention and resources in HIV care. We sought to understand the likely impact of smoking cessation on lung cancer mortality among PLWH in HIV care in the United States. We compared the risk of lung cancer death against the risks of death from other causes as a function of smoking exposure.
We used the Cost-Effectiveness of Preventing AIDS Complications (CEPAC)-US model, a validated, widely published Monte Carlo microsimulation of HIV disease and treatment.10,21-23 After populating the model with published data, we projected lung cancer mortality among PLWH in HIV care according to smoking exposure. We defined smoking exposure by both smoking status (current, former, or never) and, for current and former smokers, intensity (heavy, moderate, or light) based on number of cigarettes per day. We accounted for competing risks of death from AIDS-related and non–AIDS-related causes, the latter stratified by smoking exposure. We modeled cohorts of 1 million men or women of a particular smoking exposure entering HIV care at a specific age (eg, 30, 40, or 50 years).
The primary outcome was cumulative lung cancer mortality by age 80 years. We combined the model-generated estimates with published epidemiologic data to project the total expected lung cancer deaths among PLWH in care in the United States.
For this analysis, an individual enters the model at the time of linkage to HIV care and is followed until death or age 80 years, whichever comes first. The model draws randomly from user-defined initial distributions of CD4 count and HIV RNA and tracks clinical outcomes as an individual transitions monthly through states of disease progression and treatment. Probabilities of transition between HIV-related health states depend on factors including current CD4 count and HIV RNA. All individuals are ART eligible, regardless of CD4 count.24 An individual may die from an AIDS-related cause (opportunistic infection or chronic AIDS-related disease), lung cancer, or another non–AIDS-related cause (eg, cardiovascular disease; eFigure 1 in the Supplement). Model details are available at http://www.massgeneral.org/mpec/cepac/.
We simulated cohorts stratified by sex, age at entry to HIV care, and smoking exposure (eg, current moderate smokers, former heavy smokers). We assumed no change in smoking status or intensity stratification over time. Former smokers are stratified by past smoking intensity. Within each cohort of former smokers, all individuals stop smoking at the same age (at entry to HIV care, except in sensitivity analysis, in which they quit later) and remain abstinent. We did not evaluate those who had quit smoking before entering HIV care.
We derived lung cancer mortality probabilities by sex, age, and smoking exposure. For current smokers, risks depend on smoking intensity. At the time of quitting smoking, risks decrease to those of former smokers and depend on age at cessation and prior smoking intensity.
We applied probabilities of non–AIDS-related death from all causes besides lung cancer (eg, other cancers, cardiovascular disease) by sex, age, and smoking exposure. Risks for current and former smokers are stratified in a similar manner to lung cancer risks. We additionally stratified former smokers’ risks by years since cessation, in accordance with available data.25
In the pre-ART era, smoking was not an important driver of mortality among PLWH.26 Untreated individuals were unlikely to survive long enough to develop lung cancer. Recognizing that smoking emerges as an important competing risk in the presence of virologic suppression, we assumed, in the base case, complete ART adherence and no loss to follow-up from HIV care. These in-care individuals are more likely to participate in a smoking cessation intervention. In sensitivity analysis, we relaxed this assumption, accounting for reported rates of ART nonadherence and loss to follow-up (Supplement).27 Additional specifications are described elsewhere.23,27
There are few published data on smoking-related mortality specific to PLWH. We therefore pursued 2 validation strategies. First, we compared our model-generated cumulative lung cancer mortality among HIV-uninfected people to results reported in general population studies in Western Europe.28,29 Second, we compared our model-generated cumulative lung cancer mortality among PLWH (accounting for ART nonadherence and loss to follow-up) with the modeled cumulative lung cancer incidence reported by the North American AIDS Cohort Collaboration on Research and Design (NA-ACCORD) (Supplement).30,31
We simulated PLWH reflecting those initiating HIV care in the United States25,32-39 (Table 1 and Supplement). In the base case, we assumed the same CD4 distribution regardless of cohort age at entry to care.
Smoking intensity for current and former smokers was based on number of cigarettes per day as derived by Rosenberg et al,25 who categorized US smokers into quintiles of cigarettes per day stratified by sex, age, and 5-year birth cohort, through the year 2000.25 Among current and former smokers, we considered the fifth, third, and first quintiles to be heavy, moderate, and light smokers. For example, among 40-year-old men, heavy, moderate, and light smokers consumed 35, 18, and 2 cigarettes per day, respectively (Table 1). These quantities changed with age according to published data (Supplement).25
To derive monthly lung cancer mortality by smoking exposure, we combined US general population data on lung cancer mortality rates in 2000 (to match the latest year of the lung cancer–deleted life tables25), lung cancer mortality risk ratios for current and former smokers (further stratified by smoking intensity) vs never smokers, and the proportions of current, former, and never smokers in the population (Supplement).25,36,40-42 Compared with never smokers, the lung cancer mortality risk ratio for male and female current moderate smokers was 23.6 and 24.2, respectively, and for those who quit at age 40 years, it was 4.3 and 4.5.
The derived smoking exposure-stratified lung cancer rates were not specific to PLWH. Therefore, in the base case, we accounted for an HIV-associated independent risk of lung cancer by multiplying the derived rates by 1.7. This was the multivariable-adjusted risk ratio (PLWH vs HIV-uninfected people) reported by the Veterans Aging Cohort Study,11 the largest study of the association between HIV and lung cancer.
We applied lung cancer–deleted, non–AIDS-related mortality probabilities (eg, from other cancers and cardiovascular disease) based on published lung cancer–deleted life tables through 2000 (Supplement).25 Mortality risks were stratified by sex, age, and smoking exposure. In the base case, we assumed no HIV-associated increase in these causes of death. We did not specify the cause of death within this group; estimates would have been inaccurate owing to the absence of many cause-deleted life tables.
To understand the robustness of our findings under parameter uncertainty, we performed sensitivity analyses with alternative assumptions and parameter estimates. These included (1) accounting for reported rates of ART nonadherence and loss to follow-up from HIV care, which increase AIDS-related death risk; (2) varying the HIV-associated risk ratio for lung cancer from 1.0 to 1.911; (3) varying initial CD4 count at entry to HIV care, which affects AIDS-related death risk; and (4) varying when former smokers quit: 10 or 20 years after entering HIV care, rather than at the time of entering care.
Our base case did not include a link between HIV and other non–AIDS-related causes of death (besides lung cancer). This reflected the inconsistency of reported smoking-adjusted, HIV-associated risks of death from these causes. However, we conducted an additional sensitivity analysis using a multiplicative risk factor of 1.7 to increase the independent effect of HIV on other non–AIDS-related mortality, including that from other cancers and cardiovascular disease. This value highlighted the reported HIV-conferred risk of myocardial infarction and matched the HIV-associated lung cancer risk ratio we applied in the base case.37-39,43
We derived the number of current, former, and never smokers among PLWH aged 20 to 64 years in HIV care in the United States by sex and 5-year age increment (the overall prevalence of current, former, and never smokers among PLWH was approximately 42%, 20%, and 37%, respectively, in 2009).3,31 Accounting for reported rates of ART nonadherence and loss to follow-up from care, we used model-generated cumulative lung cancer mortality for each sex, age, and smoking stratum to derive the expected total number of lung cancer deaths by age 80 years among PLWH in HIV care in the United States. To examine the potential impact of a smoking cessation intervention, we estimated the reduction in lung cancer deaths if 20% of current smokers in each sex and age cohort were to quit smoking (Supplement).
For HIV-uninfected people, model-generated cumulative lung cancer mortalities by age 75 years for current heavy smokers were 24.5% and 19.5% for men and women, respectively; these values are similar to those reported in the United Kingdom, 24.4% and 18.5% (eTables 1 and 2 in the Supplement).28 For 20-year-old PLWH in the United States, our model-generated cumulative lung cancer mortality by age 75 years was 5.0%, while the modeled cumulative lung cancer incidence reported by NA-ACCORD was 3.7% (Supplement).30
Among men, cumulative lung cancer mortality by age 80 years for heavy, moderate, and light current smokers entering HIV care at age 40 years was 28.9%, 23.0%, and 18.8%, respectively; for heavy, moderate, and light former smokers who quit smoking at age 40 years, it was 7.9%, 6.1%, and 4.3%; and for never smokers, it was 1.6%. Among women, the corresponding respective cumulative lung cancer mortality for heavy, moderate, and light current smokers was 27.8%, 20.9%, and 16.6%; for heavy, moderate, and light former smokers who quit at age 40 years, it was 7.5%, 5.2%, and 3.7%; for never smokers, it was 1.2% (Table 2 and Figure 1). Cumulative lung cancer mortality for ART-adherent current and never smokers varied little by age at entry to HIV care, with slightly higher results among those entering care at older ages owing to survival bias (Table 2).
Risks of mortality from lung cancer vs other causes varied by smoking exposure. For men who entered HIV care at age 40 years, adhered to the ART regimen, and continued to smoke at a moderate (“average”) level, cumulative mortality from lung cancer was 10 times that from AIDS-related causes (23.0% vs 2.3%; Figure 2A). The mortality for former smokers in this category is illustrated in Figure 2B, and for never smokers, it is illustrated in eFigure 2A in the Supplement. For women who entered HIV care at age 40 years, adhered to the ART regimen, and continued to smoke at a moderate (“average”) level, cumulative mortality from lung cancer was 8 times that from AIDS-related causes (20.9% vs 2.5%). Depending on sex and smoking intensity, ART-adherent smokers were 6 to 13 times more likely to die from lung cancer than from AIDS-related causes. For 40-year-old ART-adherent men who were current moderate smokers, the combined cumulative mortality from lung cancer and other non–AIDS-related causes—both of which were increased by smoking—was approximately 35 times that from AIDS-related causes (79.9% vs 2.3%); for women in this category, it was approximately 27 times higher (66.6% vs 2.5%) (eTable 3 in the Supplement). See eTable 4 in the Supplement for results for different ages at entry to HIV care.
For the sensitivity analyses described in the Methods section, we report instances where input data were most uncertain and/or where variation in the underlying parameter had a material impact on our findings. Results of other sensitivity analyses (including varying CD4 count at entry to HIV care and timing of smoking cessation) are reported in the Supplement.
When we accounted for reported rates of ART nonadherence and loss to follow-up from HIV care, cumulative lung cancer mortality among moderate smokers decreased, and cumulative AIDS-related mortality increased (Figure 2C and D). The results for never smokers are depicted in eFigure 2B in the Supplement. For men who were current heavy smokers, cumulative lung cancer mortality (22.7%) was similar to AIDS-related mortality (23.0%); for women, the corresponding results were 21.6% vs 25.5% (eTable 5 in the Supplement).
Model-generated cumulative mortality varied depending on the HIV-associated risk ratios applied for lung cancer and for other non–AIDS-related causes. For male and female current moderate smokers, cumulative lung cancer mortality varied from 10.5% to 23.0%, and cumulative other non–AIDS-related mortality varied from 45.7% to 78.0%. The lower numbers reflected scenarios assuming no direct influence of HIV on these causes of death (Table 3).
Applying sex- and age-specific, model-projected results to the approximately 644 200 PLWH aged 20 to 64 years in care in the United States (including current, former, and never smokers), 59 900 lung cancer deaths by age 80 years are expected (9.3% of this population) if smoking status does not change. If 20% of the 273 200 current smokers quit, their lung cancer risk would decrease to that of former smokers, and 6900 (11.5%) lung cancer deaths could be averted.
Using a microsimulation model, we found that ART-adherent PLWH in the United States who smoke cigarettes are 6 to 13 times more likely to die from lung cancer than from AIDS-related causes. Even when accounting for reported rates of ART nonadherence and loss to follow-up, we found that nearly 10% of PLWH initially linked to HIV care (including both smokers and nonsmokers) are expected to die from lung cancer if smoking habits do not change. Smoking cessation could substantially reduce lung cancer risk for an individual and avert many lung cancer deaths at the population level.
People with HIV whose virus is suppressed now have a life expectancy approaching that of HIV-uninfected people.44 However, life expectancy gaps persist, and smoking is a key driver of this difference.45 This smoking footprint will likely grow as PLWH age and develop lung cancer and other smoking-associated diseases. Lung cancer was already the leading cause of death in a study of PLWH in France in 2010; the smoking prevalence among PLWH is similar in France and the United States.3,20,46
In stratifying mortality risks by smoking status and intensity, our results can inform important conversations between clinicians and PLWH who smoke, helping both clinician and patient understand the patient’s risks of different diseases and the potential benefits of smoking cessation. Recognizing the increased risk of death from lung cancer vs that from traditionally feared AIDS-related causes may motivate a smoker to quit smoking,47 although this is not always the case. Our analyses also accounted for the smoking-conferred increase in risk of death from causes besides lung cancer and AIDS—this includes other cancers, other lung diseases, and cardiovascular disease. Our results provide evidence for HIV care programs and policymakers to include smoking cessation interventions as a key component of the comprehensive care of PLWH. Though smoking cessation is challenging, smoking prevalence among the US general population has decreased substantially in recent decades, from 42% in 1965 to 15% in 2015.7,48 While the current smoking prevalence among PLWH is much higher than that among HIV-uninfected people, a similar proportion of smokers in the 2 groups want to quit, offering hope for a potential decrease in smoking prevalence in PLWH as well.49
Perhaps counterintuitively, lung cancer risk is linked to adherence to HIV therapy. Those who are not ART-adherent are more likely to die of AIDS-related causes before developing lung cancer. Nonetheless, even when accounting for reported rates of ART nonadherence and loss to follow-up from HIV care, we found that for male heavy smokers, the risk of dying from lung cancer is similar to the risk of dying from AIDS-related causes.
Our results are subject to limitations inherent in the assumptions and simplifications of any model-based study. These include the following: (1) using data from the US general population in 2000 in the absence of smoking-stratified lung cancer data for PLWH; (2) uncertainty around the HIV-conferred risk of lung cancer, which could be lower or higher than what we applied in this analysis, and could vary by age11,13-17,50; (3) assuming no relationship between CD4 count and lung cancer risk, given conflicting reports affected by confounders such as smoking intensity15,19,51-54; (4) assuming that the HIV-associated increase in lung cancer mortality is similar to the increase in incidence, since median survival after lung cancer diagnosis is short11,40; (5) not accounting for disparities between PLWH and HIV-uninfected people in mortality after lung cancer diagnosis17,55-58; (6) not examining race: African American men are overrepresented among PLWH in the United States and may be more susceptible to lung cancer than people of other races31,59; (7) not differentiating by HIV transmission risk category: the national survey data for smoking prevalence among PLWH indicated little difference by sexual transmission risk category, but there were few injection drug users3; (8) not accounting for possibly higher average daily cigarette consumption among PLWH who smoke compared with HIV-uninfected people who smoke11; (9) assuming, in the base case, that smoking status does not change over time; and (10) not examining a potential impact on mortality from lung cancer screening with computed tomography.60,61 Estimating smoking-attributable mortality for various diseases would be informative, but the data for non-AIDS, non–lung cancer outcomes were too limited to do so. The smoking-conferred risk of death from smoking-related diseases may be even greater than that applied in our analyses, especially in PLWH.43 Though the model-generated estimates may be influenced by input parameter uncertainty, the magnitude of smoking-related harm with respect to lung cancer and other non–AIDS-related mortality, and the magnitude of the benefit from smoking cessation, remain robust, as shown in our sensitivity analysis.
In conclusion, there is a large expected burden of lung cancer among PLWH in the United States because (1) the smoking prevalence is very high in this population; (2) HIV itself increases the risk of lung cancer; and (3) PLWH are increasingly living long enough to develop lung cancer. For PLWH who adhere to ART, smoking is a much greater threat to their health than HIV itself. Clinicians caring for PLWH should offer guideline-based behavioral and pharmacologic treatments for tobacco use.62 Lung cancer is now a leading cause of death among PLWH, but smoking cessation can greatly reduce the risk. Lung cancer prevention, especially through smoking cessation, should be a priority in the comprehensive care of PLWH.
Accepted for Publication: July 5, 2017.
Corresponding Author: Krishna P. Reddy, MD, Medical Practice Evaluation Center, Massachusetts General Hospital, 50 Staniford St, 9th Flr, Boston, MA 02114 (firstname.lastname@example.org).
Published Online: September 18, 2017. doi:10.1001/jamainternmed.2017.4349
Author Contributions: Dr Reddy had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Reddy, Parker, Paltiel, Weinstein, Freedberg, Walensky.
Acquisition, analysis, or interpretation of data: Reddy, Kong, Hyle, Baggett, Huang, Parker, Paltiel, Losina, Freedberg, Walensky.
Drafting of the manuscript: Reddy.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Reddy, Kong, Huang, Parker, Losina, Weinstein.
Obtained funding: Reddy, Freedberg, Walensky.
Administrative, technical, or material support: Baggett, Paltiel, Freedberg, Walensky.
Supervision: Kong, Walensky.
Conflict of Interest Disclosures: None reported.
Funding/Support: This work was supported by awards from the National Institute on Drug Abuse (K01 DA042687, K23 DA034008, and R01 DA015612), the National Heart, Lung, and Blood Institute (T32 HL116275 and K01 HL123349), the National Cancer Institute (U01 CA199284), the National Institute of Mental Health (R01 MH105203), and the National Institute of Allergy and Infectious Diseases (R01 AI042006 and R37 AI093269) of the National Institutes of Health, and Massachusetts General Hospital (Steve and Deborah Gorlin Research Scholars Award to Dr Walensky).
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or Massachusetts General Hospital.
Additional Contributions: The authors thank Emily Martey, BA, Amy Zheng, BA, and Ethan Borre, BA (all of Massachusetts General Hospital), for their technical assistance. They received no compensation for their contributions beyond that received for their normal employment.