- Email: [email protected]
Role of concurrency in generalised HIV epidemics – Authors' reply
Correspondence
We declare that we have no conflicts of interest.
*Stéphane Helleringer, Hans-Peter Kohler [email protected] Department of Population and Family Health, Mailman School of Public Health, Columbia University, New York, NY 10032, USA (SH); and University of Pennsylvania, Philadelphia, PA, USA (HPK) 1
2
3
Tanser F, Barnighausen T, Hund L, Garnett GP, McGrath N, Newell ML. Effect of concurrent sexual partnerships on rate of new HIV infections in a high-prevalence, rural South African population: a cohort study. Lancet 2011; 378: 247–55. Maughan-Brown B, Venkataramani AS. Measuring concurrent partnerships: potential for underestimation in UNAIDS recommended method. AIDS 2011; 25: 1549–51. Helleringer S, Kohler HP, Kalilani-Phiri L, Mkandawire J, Armbruster B. The reliability of sexual partnership histories: implications for the measurement of partnership concurrency during surveys. AIDS 2011; 25: 503–11.
www.thelancet.com Vol 378 November 26, 2011
4
5
Carroll RJ, Ruppert D, Stefanski LA. Measurement error in nonlinear models. Boca Raton: Chapman & Hall/CRC, 1998. Phillips AE, Gomez GB, Boily MC, Garnett GP. A systematic review and meta-analysis of quantitative interviewing tools to investigate self-reported HIV and STI associated behaviours in low- and middle-income countries. Int J Epidemiol 2010; 39: 1541–55.
Authors’ reply In response to our population-based study in which we followed up more than 7000 HIV-negative women over 5 years and failed to find evidence that concurrent sexual partnerships are an important driver of HIV incidence, James Shelton, Martina Morris, and Helen Epstein cite studies on HIV incidence in stable concordant HIVnegative partnerships as “direct evidence” for the concurrency hypothesis. However, these studies do not test the concurrency hypothesis (that concurrent sexual partnerships increase the rate of spread of HIV in a population) because they lack a meaningful counterfactual—ie, HIV incidence in people with the same total number of partners over the observation period but in serially monogamous partnerships. Furthermore, Morris and Epstein’s claim that the concurrency hypothesis has already been tested and shown to hold runs counter to ongoing public debates on the topic,1,2 other empirical data,3–5 and the results of a recent systematic review.6 The claim is also inconsistent with Morris and Epstein’s call for a randomised controlled trial to test the hypothesis, which would not be ethically permissible if it had indeed already been shown that the hypothesis held true. Morris and Epstein write that the coefficient on men’s partners does not change after controlling for men’s HIV prevalence and that therefore either “there is something wrong” with our study or that “men’s previous lifetime partners affect a woman’s future incidence, even in a neighbourhood with no HIV infected men”. This argument is erroneous for three reasons. First, the premise is not based
on our study. Nowhere in the article do we examine how the magnitude and significance of the coefficient on men’s partners change when we additionally control for HIV prevalence. For this letter, we have added HIV prevalence to the regression with the men’s partners variable and, as expected, the magnitude of the coefficient on the partners variable is reduced (by 11%) and the standard error increased (by 5%). Second, our study does not include any neighbourhoods with zero HIV-infected men, as Morris and Epstein imply. Third, as discussed in our article, men’s previous lifetime partners proxy for the rate of partner turnover over the duration of the study, which affects women’s future HIV incidence over and above HIV prevalence. Shelton, Morris, and Epstein, and Chris Kenyon and Bob Colebunders, argue that community-level concurrency variables would be insensitive to detecting a relation to HIV incidence because of their ecological nature. In our analysis, the community-level variables men’s lifetime partners and HIV prevalence have the expected large and significant effects on HIV acquisition, whereas the adjusted hazard ratio of the concurrency variable is very close to 1 (0·99, p=0·730). This finding makes it unlikely that we failed to detect a concurrency effect on HIV acquisition owing to insensitivity in ecological measurement, because such an insensitivity would affect all community variables. Further, Morris and Epstein have themselves used community-level concurrency variables in their studies, arguing that “ecological analysis is an appropriate method” to examine the hypothesis.7 Stéphane Helleringer and Hans-Peter Kohler suggest that the concurrency variable could be more affected by measurement error than the lifetime partners variable. However, this suggestion remains unsupported by evidence, since neither of the two studies Helleringer and Kohler cite directly compared measurement error in the two variables.
Giacomo Pirozzi/Panos
instruments are used, men can both “swagger” during surveys, but also fail to report some of their sexual partnerships.2 In a detailed study during which we traced the partners of roughly 400 men residing in a small Malawian community,3 we found that men’s reports of their concurrent partnerships were highly unreliable, leading to large biases in the estimated prevalence of concurrency. Such measurement error in an exposure variable is known to substantially bias relative risk estimates downwards.4 Whether measurement error would similarly attenuate the association between lifetime partners and HIV incidence is far from certain. In a systematic review of survey administration techniques, Phillips and colleagues5 found that lifetime partnerships were more consistently reported than very recent (probably concurrent) partnerships. Measurement error thus disproportionately affects the association between recent concurrency and HIV incidence. Tanser and colleagues’ approach does not constitute a valid test of the hypothesis that concurrency drives HIV transmission. Improved measurement of concurrent partnerships and corrections for measurement errors4 are needed.
1845
Jenny Matthews/Panos
Correspondence
Published Online October 5, 2011 DOI:10.1016/S01406736(11)61549-3
Morris and Epstein write that empirical evidence is lacking that respondents in our study preferentially chose partners from their immediate neighbourhoods. This is incorrect. As stated in the paper, our data show that 61% of women reported at least one partnership with a man who lived in the same isigodi (a Zulu term for a small neighbourhood community headed by a local chief). Morris and Epstein further claim that the data in our study have “serious problems” and extract other data from our past publications to back up this claim.8 However, they extract the wrong values and fail to consider the relevance of these cross-sectional data for the study at hand. From the extracted data, Morris and Epstein deduce that our contact rate was “low”. In fact, the contact rate for this longitudinal study (ie, the proportion contacted of those eligible to be contacted by the HIV surveillance at least twice over the observation period) was 99·7%. The overall participation rate in the HIV cohort is thus correctly reported in our article and does not relate to any of the values Morris and Epstein calculate. As we state in our article, “in any observational study involving the collection of data for sexual behaviour and HIV acquisition, the possibility of bias affecting the results must be considered” and can never be completely ruled out. However, our work provides the most rigorous test of the concurrency hypothesis in a general population in sub-Saharan Africa to date. We fully agree with Shelton that it is important to raise awareness that “having multiple partners puts one’s family at risk”. Indeed, our results suggest that this is precisely what a clear, well constructed message around multiple partnering would help achieve. We declare that we have no conflicts of interest.
*Frank Tanser, Till Bärnighausen, Lauren Hund, Geoffrey P Garnett, Nuala McGrath, Marie-Louise Newell [email protected]
1846
Africa Centre for Health and Population Studies, University of KwaZulu-Natal, PO Box 198, Mtubatuba 3935, South Africa (FT, TB, NM, MLN); Department of Global Health and Population, Harvard School of Public Health, Boston, MA, USA (TB); Department of Biostatistics, Harvard School of Public Health, Boston, MA, USA (LH); Department of Infectious Disease Epidemiology, Imperial College, London, UK (GPG); Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London, UK (NM); and Centre for Paediatric Epidemiology and Biostatistics, University College London Institute of Child Health, London, UK (MLN) 1
2
3
4
5
6
7
8
Lurie MN, Rosenthal S. Concurrent partnerships as a driver of the HIV epidemic in sub-Saharan Africa? The evidence is limited. AIDS Behav 2010; 14: 17–24. Mah TL, Shelton JD. Concurrency revisited: increasing and compelling epidemiological evidence. J Int AIDS Soc 2011; 14: 33. Lagarde E, Auvert B, Carael M, et al. Concurrent sexual partnerships and HIV prevalence in five urban communities of sub-Saharan Africa. AIDS 2001; 15: 877–84. Mishra V, Bignami-Van Assche S. Concurrent sexual partnerships and HIV infection: evidence from national population-based surveys. Calverton: Macro International, 2009. Maher D, Waswa L, Karabarinde A, Baisley K. Concurrent sexual partnerships and associated factors: a cross-sectional population-based survey in a rural community in Africa with a generalised HIV epidemic. BMC Public Health 2011; 11: 651. Sawers L, Stillwaggon E. Concurrent sexual partnerships do not explain the HIV epidemics in Africa: a systematic review of the evidence. J Int AIDS Soc 2010; 13: 34. Morris M, Epstein H, Wawer M. Timing is everything: international variations in historical sexual partnership concurrency and HIV prevalence. PLoS One 2010; 5: e14092. Tanser F, Hosegood V, Bärnighausen T, et al. Cohort profile: Africa Centre Demographic Information System (ACDIS) and population-based HIV survey. Int J Epidemiol 2008; 37: 956–62.
Department of Error Boyce N. Alice Roberts: the skull beneath the skin. Lancet 2011; 378: 1132—In this Profile (Sept 24), the URL for Alice Roberts’s website should be http://www.alice-roberts.co.uk. This correction has been made to the online version as of Oct 5. Nunn A, Barnes A, Cornwall A, Rana A, Mena L. Addressing Mississippi’s HIV/AIDS crisis. Lancet 2011; 378: 1217—In this Correspondence (Oct 1), Aadia Rana of the Department of Medicine, Warren Alpert Medical School of Brown University, Providence, RI, USA, should have been listed as an author, after Alexandra Cornwall. This correction has been made to the online version as of Nov 25. Vuylsteke A, Sharples L, Charman G, et al. Circulatory arrest versus cerebral perfusion during pulmonary endarterectomy surgery (PEACOG): a randomised controlled trial. Lancet 2011; 378: 1379–87—In the Summary of this Article (Oct 15), the last line of the Findings should have read “At 12 weeks, two patients had died (one in each group).” This correction has been made to the online version as of Nov 25. Cavo M, Tacchetti P, Patriarca F, et al, for the GIMEMA Italian Myeloma Network. Bortezomib with thalidomide plus dexamethasone compared with thalidomide plus dexamethasone as induction therapy before, and consolidation therapy after, double autologous stem-cell transplantation in newly diagnosed multiple myeloma: a randomised phase 3 study. Lancet 2010; 376: 2075–85—In this Article (Dec 18/25, 2010), the first two sentences of the fifth paragraph of the Results should have read: “Incorporation of VTD induction and consolidation therapy into double autologous stem-cell transplantation overcame the adverse effect of t(4;14) on progression-free survival: at 3 years, 69% of patients with the abnormality were free from progression, relapse, or death compared with 74% of those without (p=0·66). By contrast, the adverse relation of t(4;14) with progression-free survival was retained in patients randomly assigned to the TD group: at 3 years, 37% of patients with the abnormality were free from progression, relapse, or death compared with 63% of those without (p=0·0131).” These corrections have been made to the online version as of Nov 25.
www.thelancet.com Vol 378 November 26, 2011
We declare that we have no conflicts of interest.
*Stéphane Helleringer, Hans-Peter Kohler [email protected] Department of Population and Family Health, Mailman School of Public Health, Columbia University, New York, NY 10032, USA (SH); and University of Pennsylvania, Philadelphia, PA, USA (HPK) 1
2
3
Tanser F, Barnighausen T, Hund L, Garnett GP, McGrath N, Newell ML. Effect of concurrent sexual partnerships on rate of new HIV infections in a high-prevalence, rural South African population: a cohort study. Lancet 2011; 378: 247–55. Maughan-Brown B, Venkataramani AS. Measuring concurrent partnerships: potential for underestimation in UNAIDS recommended method. AIDS 2011; 25: 1549–51. Helleringer S, Kohler HP, Kalilani-Phiri L, Mkandawire J, Armbruster B. The reliability of sexual partnership histories: implications for the measurement of partnership concurrency during surveys. AIDS 2011; 25: 503–11.
www.thelancet.com Vol 378 November 26, 2011
4
5
Carroll RJ, Ruppert D, Stefanski LA. Measurement error in nonlinear models. Boca Raton: Chapman & Hall/CRC, 1998. Phillips AE, Gomez GB, Boily MC, Garnett GP. A systematic review and meta-analysis of quantitative interviewing tools to investigate self-reported HIV and STI associated behaviours in low- and middle-income countries. Int J Epidemiol 2010; 39: 1541–55.
Authors’ reply In response to our population-based study in which we followed up more than 7000 HIV-negative women over 5 years and failed to find evidence that concurrent sexual partnerships are an important driver of HIV incidence, James Shelton, Martina Morris, and Helen Epstein cite studies on HIV incidence in stable concordant HIVnegative partnerships as “direct evidence” for the concurrency hypothesis. However, these studies do not test the concurrency hypothesis (that concurrent sexual partnerships increase the rate of spread of HIV in a population) because they lack a meaningful counterfactual—ie, HIV incidence in people with the same total number of partners over the observation period but in serially monogamous partnerships. Furthermore, Morris and Epstein’s claim that the concurrency hypothesis has already been tested and shown to hold runs counter to ongoing public debates on the topic,1,2 other empirical data,3–5 and the results of a recent systematic review.6 The claim is also inconsistent with Morris and Epstein’s call for a randomised controlled trial to test the hypothesis, which would not be ethically permissible if it had indeed already been shown that the hypothesis held true. Morris and Epstein write that the coefficient on men’s partners does not change after controlling for men’s HIV prevalence and that therefore either “there is something wrong” with our study or that “men’s previous lifetime partners affect a woman’s future incidence, even in a neighbourhood with no HIV infected men”. This argument is erroneous for three reasons. First, the premise is not based
on our study. Nowhere in the article do we examine how the magnitude and significance of the coefficient on men’s partners change when we additionally control for HIV prevalence. For this letter, we have added HIV prevalence to the regression with the men’s partners variable and, as expected, the magnitude of the coefficient on the partners variable is reduced (by 11%) and the standard error increased (by 5%). Second, our study does not include any neighbourhoods with zero HIV-infected men, as Morris and Epstein imply. Third, as discussed in our article, men’s previous lifetime partners proxy for the rate of partner turnover over the duration of the study, which affects women’s future HIV incidence over and above HIV prevalence. Shelton, Morris, and Epstein, and Chris Kenyon and Bob Colebunders, argue that community-level concurrency variables would be insensitive to detecting a relation to HIV incidence because of their ecological nature. In our analysis, the community-level variables men’s lifetime partners and HIV prevalence have the expected large and significant effects on HIV acquisition, whereas the adjusted hazard ratio of the concurrency variable is very close to 1 (0·99, p=0·730). This finding makes it unlikely that we failed to detect a concurrency effect on HIV acquisition owing to insensitivity in ecological measurement, because such an insensitivity would affect all community variables. Further, Morris and Epstein have themselves used community-level concurrency variables in their studies, arguing that “ecological analysis is an appropriate method” to examine the hypothesis.7 Stéphane Helleringer and Hans-Peter Kohler suggest that the concurrency variable could be more affected by measurement error than the lifetime partners variable. However, this suggestion remains unsupported by evidence, since neither of the two studies Helleringer and Kohler cite directly compared measurement error in the two variables.
Giacomo Pirozzi/Panos
instruments are used, men can both “swagger” during surveys, but also fail to report some of their sexual partnerships.2 In a detailed study during which we traced the partners of roughly 400 men residing in a small Malawian community,3 we found that men’s reports of their concurrent partnerships were highly unreliable, leading to large biases in the estimated prevalence of concurrency. Such measurement error in an exposure variable is known to substantially bias relative risk estimates downwards.4 Whether measurement error would similarly attenuate the association between lifetime partners and HIV incidence is far from certain. In a systematic review of survey administration techniques, Phillips and colleagues5 found that lifetime partnerships were more consistently reported than very recent (probably concurrent) partnerships. Measurement error thus disproportionately affects the association between recent concurrency and HIV incidence. Tanser and colleagues’ approach does not constitute a valid test of the hypothesis that concurrency drives HIV transmission. Improved measurement of concurrent partnerships and corrections for measurement errors4 are needed.
1845
Jenny Matthews/Panos
Correspondence
Published Online October 5, 2011 DOI:10.1016/S01406736(11)61549-3
Morris and Epstein write that empirical evidence is lacking that respondents in our study preferentially chose partners from their immediate neighbourhoods. This is incorrect. As stated in the paper, our data show that 61% of women reported at least one partnership with a man who lived in the same isigodi (a Zulu term for a small neighbourhood community headed by a local chief). Morris and Epstein further claim that the data in our study have “serious problems” and extract other data from our past publications to back up this claim.8 However, they extract the wrong values and fail to consider the relevance of these cross-sectional data for the study at hand. From the extracted data, Morris and Epstein deduce that our contact rate was “low”. In fact, the contact rate for this longitudinal study (ie, the proportion contacted of those eligible to be contacted by the HIV surveillance at least twice over the observation period) was 99·7%. The overall participation rate in the HIV cohort is thus correctly reported in our article and does not relate to any of the values Morris and Epstein calculate. As we state in our article, “in any observational study involving the collection of data for sexual behaviour and HIV acquisition, the possibility of bias affecting the results must be considered” and can never be completely ruled out. However, our work provides the most rigorous test of the concurrency hypothesis in a general population in sub-Saharan Africa to date. We fully agree with Shelton that it is important to raise awareness that “having multiple partners puts one’s family at risk”. Indeed, our results suggest that this is precisely what a clear, well constructed message around multiple partnering would help achieve. We declare that we have no conflicts of interest.
*Frank Tanser, Till Bärnighausen, Lauren Hund, Geoffrey P Garnett, Nuala McGrath, Marie-Louise Newell [email protected]
1846
Africa Centre for Health and Population Studies, University of KwaZulu-Natal, PO Box 198, Mtubatuba 3935, South Africa (FT, TB, NM, MLN); Department of Global Health and Population, Harvard School of Public Health, Boston, MA, USA (TB); Department of Biostatistics, Harvard School of Public Health, Boston, MA, USA (LH); Department of Infectious Disease Epidemiology, Imperial College, London, UK (GPG); Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London, UK (NM); and Centre for Paediatric Epidemiology and Biostatistics, University College London Institute of Child Health, London, UK (MLN) 1
2
3
4
5
6
7
8
Lurie MN, Rosenthal S. Concurrent partnerships as a driver of the HIV epidemic in sub-Saharan Africa? The evidence is limited. AIDS Behav 2010; 14: 17–24. Mah TL, Shelton JD. Concurrency revisited: increasing and compelling epidemiological evidence. J Int AIDS Soc 2011; 14: 33. Lagarde E, Auvert B, Carael M, et al. Concurrent sexual partnerships and HIV prevalence in five urban communities of sub-Saharan Africa. AIDS 2001; 15: 877–84. Mishra V, Bignami-Van Assche S. Concurrent sexual partnerships and HIV infection: evidence from national population-based surveys. Calverton: Macro International, 2009. Maher D, Waswa L, Karabarinde A, Baisley K. Concurrent sexual partnerships and associated factors: a cross-sectional population-based survey in a rural community in Africa with a generalised HIV epidemic. BMC Public Health 2011; 11: 651. Sawers L, Stillwaggon E. Concurrent sexual partnerships do not explain the HIV epidemics in Africa: a systematic review of the evidence. J Int AIDS Soc 2010; 13: 34. Morris M, Epstein H, Wawer M. Timing is everything: international variations in historical sexual partnership concurrency and HIV prevalence. PLoS One 2010; 5: e14092. Tanser F, Hosegood V, Bärnighausen T, et al. Cohort profile: Africa Centre Demographic Information System (ACDIS) and population-based HIV survey. Int J Epidemiol 2008; 37: 956–62.
Department of Error Boyce N. Alice Roberts: the skull beneath the skin. Lancet 2011; 378: 1132—In this Profile (Sept 24), the URL for Alice Roberts’s website should be http://www.alice-roberts.co.uk. This correction has been made to the online version as of Oct 5. Nunn A, Barnes A, Cornwall A, Rana A, Mena L. Addressing Mississippi’s HIV/AIDS crisis. Lancet 2011; 378: 1217—In this Correspondence (Oct 1), Aadia Rana of the Department of Medicine, Warren Alpert Medical School of Brown University, Providence, RI, USA, should have been listed as an author, after Alexandra Cornwall. This correction has been made to the online version as of Nov 25. Vuylsteke A, Sharples L, Charman G, et al. Circulatory arrest versus cerebral perfusion during pulmonary endarterectomy surgery (PEACOG): a randomised controlled trial. Lancet 2011; 378: 1379–87—In the Summary of this Article (Oct 15), the last line of the Findings should have read “At 12 weeks, two patients had died (one in each group).” This correction has been made to the online version as of Nov 25. Cavo M, Tacchetti P, Patriarca F, et al, for the GIMEMA Italian Myeloma Network. Bortezomib with thalidomide plus dexamethasone compared with thalidomide plus dexamethasone as induction therapy before, and consolidation therapy after, double autologous stem-cell transplantation in newly diagnosed multiple myeloma: a randomised phase 3 study. Lancet 2010; 376: 2075–85—In this Article (Dec 18/25, 2010), the first two sentences of the fifth paragraph of the Results should have read: “Incorporation of VTD induction and consolidation therapy into double autologous stem-cell transplantation overcame the adverse effect of t(4;14) on progression-free survival: at 3 years, 69% of patients with the abnormality were free from progression, relapse, or death compared with 74% of those without (p=0·66). By contrast, the adverse relation of t(4;14) with progression-free survival was retained in patients randomly assigned to the TD group: at 3 years, 37% of patients with the abnormality were free from progression, relapse, or death compared with 63% of those without (p=0·0131).” These corrections have been made to the online version as of Nov 25.
www.thelancet.com Vol 378 November 26, 2011