Trends of HIV Mortality between 2001 and 2018: An Observational Analysis

Round 1

Reviewer 1 Report

In their manuscript, Jani et al describe mortality rates across 48 countries with complete data in the WHO Mortality Database across periods ranging from 2001-2018.

One of the major issues with studying mortality outcomes of chronic infections in national or worldwide public databases is that without knowing the incidence or yearly prevalence of the infection, the mortality rates are fairly meaningless to the population for which the estimate is useful. Here, the HIV prevalence varies substantially across countries, and it is no surprise that some countries have a higher burden of disease in the general population simply because the prevalence is higher. The discussion on this point is very limited.

Cross country comparisons are then problematic. The overall specificity of mortality to HIV-positive individuals is a real limitation, which makes other efforts, such as the Global Burden of Diseases study (GBD 2015 HIV Collaborators, Lancet HIV, 2016) or the WHO’s Global Health Observatory, much more informative. These results should be compared to these studies.

Another issue, at least at the country level, is that identifying HIV as the cause of death depends on whether the individual had been tested for HIV. There are likely differences in HIV testing rates across settings, hence there could be some misclassification of death. This potential for misclassification needs to be discussed. The authors never really go into detail on how the endpoint was adjudicated by the WHO. As the authors mentioned in the discussion, there are indeed differences in ART coverage between countries, yet these figures are never really relayed back to the endpoint of mortality.

There are several time-series models available that could be fit to give an effect estimate of the impact of the proportion undergoing ART, but this would require that the authors retrieve data from other sources. The description of the statistical methods used is wanting. It seems that LOWESS was only used to make plots and not to model changes across countries within the same region. “Visual inspection” is unclear. “Software requires continuous data throughout the observation period” can be remedied with a variety of methods. “…to avoid excess imputation” is odd to bring up when imputation was never discussed. How was “percent change” calculated? Would it not be more appropriate to use absolute change in mortality and not relative %, as higher % might be an artifact of very low rates of the baseline comparator? I would recommend that the authors read how others have used the WHO Mortality Database (particularly with respect to the LOESS functions: Ebmeier S et al, Lancet, 2017; Barco S et al, Lancet Res Med, 2020).

Finally, it is striking that sub-Saharan Africa, having the highest HIV prevalence and burden of HIV in terms of mortality worldwide, is only represented by South Africa and Mauritius. Would the same results be observed in western or eastern Africa?

Minor comments:

- ln 67. Methods. “completeness of data” is defined how?

- ln 78-79. The definition for estimated level of coverage is confusing. When was it reported in the Results?

- The “percent change in ASDR” concerns different time periods. Estimates need to be given in average yearly rates of change, or another method should be entertained.

- ln 267. “Injecting” should be used instead of “intravenous”.

- ln 269-70. Hepatitis B virus infection is probably more likely to cause liver-related death than HCV in sub-Saharan Africa and Southeast Asia.

Author Response

Reviewer-1

In their manuscript, Jani et al describe mortality rates across 48 countries with complete data in the WHO Mortality Database across periods ranging from 2001-2018.

One of the major issues with studying mortality outcomes of chronic infections in national or worldwide public databases is that without knowing the incidence or yearly prevalence of the infection, the mortality rates are fairly meaningless to the population for which the estimate is useful. Here, the HIV prevalence varies substantially across countries, and it is no surprise that some countries have a higher burden of disease in the general population simply because the prevalence is higher. The discussion on this point is very limited. Cross country comparisons are then problematic. The overall specificity of mortality to HIV-positive individuals is a real limitation, which makes other efforts, such as the Global Burden of Diseases study (GBD 2015 HIV Collaborators, Lancet HIV, 2016) or the WHO’s Global Health Observatory, much more informative. These results should be compared to these studies.

Epidemiological transition has brought reduction in the infectious diseases in many developed countries. However, diseases like AIDS, tuberculosis and influenza have emerged and re-emerged. Even though, incidence and prevalence are not available through WHO mortality database, it reports an actual data of mortality with no estimation which is the case in Global Burden of Disease database. Both the data-bases have their own benefits and limitations. With actual reporting, we have the real-world data although there is always a possibility of under-reporting. Estimation through GBD would help us in filling this gap. However, it is always an estimation and not the real-world data. We feel that the literature from different databases can help us in understanding these epidemiological trends in more detailed manner. We highly appreciate the suggestion of the reviewer for the comparison to previous studies and therefore, we have added few comparison in our discussion now. 

“Trends from the Global Burden of Disease (GBD) database have still shown high mortality in Sub-Saharan Africa and South Asia, as well as increasing mortality in South-East Asia. Apart from these regions, in our study we have also observed increasing ASDRs in Eastern Europe and central Asia region (Latvia, Lithuania, Estonia, Kyrgyztan, Georgia and, Armenia) as well as in Central Europe region (Czech Republic and Croatia) ¹⁸. This finding highlights the importance of harmonization of ‘real world’ data and the increasing need for independent multinational registries, in order to better understand health-system level differences in policy, clinical practice, and outcomes.”

Regarding effect of prevalence, we have added following sentences:

“The WHO Mortality Database does not have data pertaining to prevalence or incidence. Studies in the past using GBD database have shown that, in 2015, 75% of new cases were in Sub-Saharan Africa followed by South Asia (8.5%) and Southeast Asia (4.7%). The overall high prevalence and incidence can be affecting this high mortality substantially ¹⁸.”

Another issue, at least at the country level, is that identifying HIV as the cause of death depends on whether the individual had been tested for HIV. There are likely differences in HIV testing rates across settings, hence there could be some misclassification of death. This potential for misclassification needs to be discussed. The authors never really go into detail on how the endpoint was adjudicated by the WHO. As the authors mentioned in the discussion, there are indeed differences in ART coverage between countries, yet these figures are never really relayed back to the endpoint of mortality.

Following sentences added to the manuscript:

“ICD coding for HIV mortality presents a number of challenges. In order to achieve comparability over time, WHO has designed a standardized verbal autopsy instrument that can serve different purposes. Although, due to the current usage of database, this is one of the limitations ³⁸.”

There are several time-series models available that could be fit to give an effect estimate of the impact of the proportion undergoing ART, but this would require that the authors retrieve data from other sources. The description of the statistical methods used is wanting. It seems that LOWESS was only used to make plots and not to model changes across countries within the same region. “Visual inspection” is unclear. “Software requires continuous data throughout the observation period” can be remedied with a variety of methods. “…to avoid excess imputation” is odd to bring up when imputation was never discussed. How was “percent change” calculated? Would it not be more appropriate to use absolute change in mortality and not relative %, as higher % might be an artifact of very low rates of the baseline comparator? I would recommend that the authors read how others have used the WHO Mortality Database (particularly with respect to the LOESS functions: Ebmeier S et al, Lancet, 2017; Barco S et al, Lancet Res Med, 2020).

We have edited the methodology section and removed sentences that were not directly applicable to our statistics. We did not perform any imputation. For percentage change, we calculated as [(End ASDR – Start ASDR)/Start ASDR] *100 for each gender and country. We have also added absolute number in the table for the comparison with the relative change.

The updated methods paragraph is copied below:

“We computed male and female mortality rates and used a locally weighted scatter-plot smoothing (LOWESS) plot fit to the rates of male and female mortality using SAS v9.4 (SAS, Cary, NC) (Figure 1). LOWESS was only used to make plots for visualization of country-specific trends and not to model changes across the countries. Mortality data were missing in a small subset of countries in the WHO Mortality Database for one or more calendar years. If a country had more than three consecutive years of data missing during the observation period, it was excluded from the analysis. There were no other modifications to the data. Changes in ASDRs over the observation period are calculated as crude absolute differences between first and last data points for the earliest and most recent years available. Like our previous studies, percentage change was calculated as [(End ASDR – Start ASDR)/Start ASDR]*100 for each gender and country ^16,17. We have also provided absolute changes in the ASDR. This can serve as a comparator to the percentage changes especially for the countries with low baseline mortality.”

Finally, it is striking that sub-Saharan Africa, having the highest HIV prevalence and burden of HIV in terms of mortality worldwide, is only represented by South Africa and Mauritius. Would the same results be observed in western or eastern Africa?

We understand that South Africa and Mauritius are not entirely representative of the population of Sub-Saharan Africa. However, due to the availability and completeness of the data reported by WHO, the other countries did not meet the data quality and completeness criteria for inclusion in the study. Without analysis it would be difficult to comment if other regions in Africa have similar results. We have included this as a limitation in the manuscript.

Minor comments:

- ln 67. Methods. “Completeness of data” is defined how?

Further details about completeness of data is now added to the methods sections:

“The proportion of all deaths which are registered in the population covered by the vital registration system for a country (referred to as completeness) has been estimated by WHO for the latest available year. For Member States with incomplete vital registration systems, demographic techniques have been used by WHO to estimate the level of completeness of death recording for the specified population to allow the calculation of death rates.”

- ln 78-79. The definition for estimated level of coverage is confusing. When was it reported in the Results?

 

Further details about level of coverage is now added to the methods sections:

“The estimated level of coverage for deaths with a recorded cause for death is calculated by dividing the total deaths reported for a country-year from the vital registration system by the total estimated deaths for that year for the national population. The national population estimates used are those of the UN Population Division. Best estimates of death rates by age and sex, adjusted for incompleteness and incomplete coverage), are applied to the national population data to obtain total estimated deaths. WHO estimated coverage for a Member State may be less than 100% due to incompleteness of registration, or to coverage of only some parts of the national population, or to differences between the vital registration population and the UN estimated de-facto population.”

- The “percent change in ASDR” concerns different time periods. Estimates need to be given in average yearly rates of change, or another method should be entertained.

We calculated percentage change by [(End ASDR – Start ASDR)/Start ASDR]*100 for each gender and country. This equation is described in the methods section of the manuscript.

- ln 267. “Injecting” should be used instead of “intravenous”.

We have changed the word to “injecting”.

- ln 269-70. Hepatitis B virus infection is probably more likely to cause liver-related death than HCV in sub-Saharan Africa and Southeast Asia.

Following sentence added to the manuscript:

“Also, Hepatitis B virus co-infections are high in endemic regions (East and South-east Asia) or through medical or traditional scarification procedures in Africa²⁶.”

Author Response File: Author Response.docx

Reviewer 2 Report

Introduction:
Rationale / Justification
A paragraph needs to be added, why this study is important?

Objectives:
To describe current mortality rates for HIV is a broader statement, it will be better to divide objectives of the study into General objectives and specific objectives.

Methods:
When this study was conducted?

Discussion:
Countries with high treatment coverage have higher mortality rates, reasons need to be properly explained.

Weaknesses of study
Out of total194 countries, data of only 48 countries have been analyzed due to the non-availability of data.

Author Response

Reviewer-2

Introduction:
Rationale / Justification

A paragraph needs to be added, why this study is important?

Thank you for this suggestion. We have added the following paragraph to the manuscript relating to the study’s importance.

“In year 2014, the Joint United Nations Program on HIV and AIDS (UNAIDS) and partners set the ‘90-90-90 targets’; aiming to diagnose 90% of all HIV people, provide antiretroviral therapy (ART) for 90% of those diagnosed and achieve viral suppression for 90% of those treated, by 2020 ^12,13. This led to more widespread and earlier administration of ART. It is important to examine how the setting of these targets, and the subsequent efforts to meet them, have impacted global HIV mortality trends.”

Objectives:
To describe current mortality rates for HIV is a broader statement, it will be better to divide objectives of the study into General objectives and specific objectives.

We have divided objectives into primary and secondary. We have also updated it in the manuscript.

“The primary objective of this study was to describe current ASDRs for HIV and the secondary objectives were to compute the absolute and percentage changes in ASDRs categorized by country and gender through an observational analysis of HIV mortality data from 2001 to 2018 from the World Health Organization (WHO) Mortality Database. We also compared the mortality trends from HIV in 48 nations between 2001 and 2018.”

Methods:
When this study was conducted?

The data was extracted in October 2020 with available  data from the WHO Mortality Database from 2001 to 2018.

We have added “October 2020” as the month of data extraction.

Discussion:
Countries with high treatment coverage have higher mortality rates, reasons need to be properly explained.

Thank you for the suggestion. We have expanded on the reasons. One of the reason can be increasing drug resistance and lack of available testing for the same. Second reason can be emerging non-HIV related conditions in People living with HIV including malignancy and other infectious as well as non-communicable diseases.

“After more than a decade of ART scale-up and more than 19.5 million people on treatment, an increase in drug resistance is inevitable. Data from several national surveys from low-income and middle-income countries have shown increase prevalence of pretreatment HIV drug resistance to non-nucleoside reverse-transcriptase inhibitors above 10%. Drug resistance testing is largely not available in low- and middle-income countries and might be contributing to the high mortality despite good ART coverage²⁴ . Non-HIV-related conditions are also emerging as prominent health concerns in settings where ART is widely available. A prospective cohort study of PLHIV patients in Spain found that the most common non-AIDS events developing in these patients were psychiatric, hepatic, renal, cardiovascular, and malignant diseases.²⁵ Interestingly, the leading causes of non-AIDS-associated mortality in HIV-positive individuals were malignancy, most commonly lung, followed by liver disease, with which hepatitis C virus was frequently found associated²¹. Also, Hepatitis B virus co-infections are high in endemic regions (East and South-east Asia) or through medical or traditional scarification procedures in Africa²⁶. A large body of evidence indicates that HIV-positive individuals are above-average risk for cardiovascular disease ^{27, 28}. Co-infection with tuberculosis remains a major public health concern and is the main cause of high mortality in HIV globally, particularly, in resource-limited settings.³⁴ Older age, low CD4 count, and cigarette use are independent predictors of mortality in HIV-positive individuals ²⁹. Additional risk factors for non-AIDS events include a late stage at initial treatment, coinfection with hepatitis C virus, and injecting drug abuse ³⁰. This situation raises the question of how the health needs of HIV patients should be addressed beyond the access of ART. Given these trends, the management of HIV patients in low- and middle-income countries should focus on reasons of virologic failure in patients receiving ART and provide universal access to the health services of sufficient quality to be effective without any financial hardship on the patients to optimization timely identification and management of potential non-AIDS complications and comorbid conditions.”

Weaknesses of study

Out of total194 countries, data of only 48 countries have been analyzed due to the non-availability of data.

This is based on the inclusion criteria relating to availability and quality of data. A sentence highlighting this has been added to the limitations section of the study.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

I thank the authors for carefully answering my comments. The manuscript is clearer in many aspects, but there are still some issues that need to be reconciled:

- With regards to the comment on HIV prevalence, the added discussion points are sufficient. The estimated prevalence of HIV-positive individuals should be added to table 1 to make this point even clearer.

- The added text “WHO has designed a standardized verbal autopsy instrument that can serve different purposes.” is vague. Could the authors please provide statistics from validation studies? Or something more tangible? Particularly with respect to HIV? If such information is not available, then it needs to be made explicit.

- The statistical analysis section has been very much improved. The only issue is that the begin and start years for calculating the percent change in ASDR are not the same across countries, hence there could be some bias. The authors could simply conduct a sensitivity analysis showing what would happen when using the same start and end year across all counties (or a majority thereof). The limitation must be addressed.

- I am very pleased that absolute change is given in Table 1. However, these figures tell a different story. Some of the countries with massively large percent changes have absolute changes that are very much comparable to other countries (likely owing to the exceedingly small proportion of HIV-related deaths during the begin year). I was surprised to see absolutely no discussion this – and given the data, absolute differences would seem to be more meaningful (Rothman & Greenland, Modern Epidemiology). Please amend your results and discussion accordingly.

Minor comment:

- “through medical or traditional scarification procedures in Africa²⁶” Most HBV infections in SSA are from horizontal transmission during adolescence.

Author Response

Hiv revisions:

Thank the authors for carefully answering my comments. The manuscript is clearer in many aspects, but there are still some issues that need to be reconciled:

• With regards to the comment on HIV prevalence, the added discussion points are sufficient. The estimated prevalence of HIV-positive individuals should be added to table 1 to make this point even clearer.

o Response:            WHO mortality database does not have the data for HIV prevalence. We had highlighted this in our discussion.

• The added text “WHO has designed a standardized verbal autopsy instrument that can serve different purposes.” is vague. Could the authors please provide statistics from validation studies? Or something more tangible? Particularly with respect to HIV? If such information is not available, then it needs to be made explicit.
•  Response: We have tried explaining it with more details. The same reference on “Obtaining cause-of-death information for HIV/AIDS” has validation studies in Box 5. Details are explained as below in the manuscript
• “ICD coding for HIV mortality presents a number of challenges and it is one of the limitations of all database studies. However, to achieve comparability over time, WHO has tried to standardize ICD-10 coding for HIV with detailed guidelines. Although is-sues still exist when medical certification is not possible especially in sub-Saharan Africa as they occur at home without the presence of medical doctor. WHO has developed computerized coding of verbal autopsy to address this issue using algorithmic methods. Validation studies showed that based on the symptoms indicative of AIDS-defining illnesses from verbal autopsy, 89% of deaths among HIV-positive individuals are attributable to HIV 38.”
• The statistical analysis section has been very much improved. The only issue is that the begin and start years for calculating the percent change in ASDR are not the same across countries, hence there could be some bias. The authors could simply conduct a sensitivity analysis showing what would happen when using the same start and end year across all counties (or a majority thereof). The limitation must be addressed.

 Response: On request of reviewer, we did analysis of the data keeping 2007 as the start date and 2014 as the end data which would include majority of the countries. We had to exclude USA and Canada as their end dates were before that. However, the data generated would create lots of bias and would not give glimpse of the whole actual data. We have attached the table at the bottom. As this would be rather bias creating results, we have not added it to the manuscript.

However, we have highlighted in our limitation “Also, the start and end year of countries in our study varied based on the availability of the data.”

• I am very pleased that absolute change is given in Table 1. However, these figures tell a different story. Some of the countries with massively large percent changes have absolute changes that are very much comparable to other countries (likely owing to the exceedingly small proportion of HIV-related deaths during the begin year). I was surprised to see absolutely no discussion this – and given the data, absolute differences would seem to be more meaningful (Rothman & Greenland, Modern Epidemiology). Please amend your results and discussion accordingly.

 Response: Thanks for highlighting this part. We have now extensively added absolute change in results as well as discussion.     

Minor comment:

• “through medical or traditional scarification procedures in Africa26” Most HBV infections in SSA are from horizontal transmission during adolescence.

 Response: We have made the changes in the manuscript.

“Also, Hepatitis B virus co-infections are high in endemic regions (East and South-east Asia) 26”

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table with 2007 and 2014 as end dates.

Region	Country	Overall percentage change in mortality in males		Absolute changes in males		Overall percentage change in mortality in males		Absolute changes in males
		Till original end date	till 2014	Till original end date	till 2014	Till original end date	till 2014	Till original end date	till 2014
Europe	Israel	-15.487704	-47.5744	-0.8710228	-4.433223176	-58.44511	38.03298616	-1.5515038	0.598881863
	Armenia	163.612103	372.9322	8.69429131	13.7037181	141.859702	416.4545967	2.68540901	2.462382221
	Austria	-70.951157	-35.2418	-6.2979658	-3.567474209	-27.419324	-35.68442529	-0.6243851	-0.704756189
	Belgium	-52.49484	-42.7583	-4.2512488	-2.268456889	-73.478493	-27.53215009	-2.990902	-0.733661911
	Croatia	58.0304142	358.4954	1.11795453	1.286251058	30.8538892	167.9831859	0.113322	0.820553396
	Czech Republic	219.48253	354.8966	1.2450318	1.319071677	71.8918104	137.4748208	0.13263946	0.430620959
	Denmark	-77.044882	-67.3123	-6.5670572	-4.506275971	-49.902381	-28.5824473	-1.2639716	-0.598098111
	Estonia	672.974324	23.34052	30.3178296	10.44809687	5877.56501	-18.69218172	26.9468342	-3.7466784
	Finland	20.4165499	-72.4766	0.24769358	-2.147564674	-56.394189	267.7980926	-0.5329263	0.869948664
	France	-72.078316	-54.8002	-17.765674	-8.343797348	-64.338079	-52.11917974	-4.2297913	-2.552064024
	Germany	-50.007259	-21.5624	-4.05827	-1.505719388	-54.391227	-7.235007127	-1.555026	-0.121621567
	Hungary	-61.63948	62.94649	-0.8315576	0.920062069	202.779037	-46.58784365	0.66571424	-0.085026494
	Ireland	-57.282171	-33.0806	-2.4166132	-1.395599131	-49.09711	-40.53821319	-1.0342822	-0.853980074
	Italy	-73.087469	-38.9545	-21.709634	-8.634378772	-65.876717	-31.68357733	-5.3032354	-1.813336367
	Latvia	1010.39974	125.935	52.691637	38.44257484	2978.76295	226.577343	25.0937467	21.06271165
	Luxembourg	-6.6933227	-30.0032	-0.2420114	-3.640261633	-65.324394	71.96361032	-4.7687224	2.59241704
	Malta	104.843311	226.9867	9.9035685	6.529178756	NA	0	NA	0
	Netherlands	-81.363135	-52.4192	-9.1774405	-2.937934747	-78.81623	-27.14368883	-1.896482	-0.382183675
	Norway	-54.65209	16.25017	-1.5919129	0.369758084	-52.487327	40.19554806	-0.7554475	0.29546542
	Poland	-16.988135	-16.3808	-0.6362699	-0.854258278	-37.889664	57.96033824	-0.6202435	0.58229383
	Romania	-53.034392	49.76146	-15.058785	4.639599921	-72.095933	-11.5163712	-15.719514	-0.894786242
	Slovakia	-50.316827	-11.0797	-0.2968504	-0.036522549	NA	0	NA	0
	Slovenia	-74.3139	74.30681	-4.0526635	1.138381278	NA	-15.25812121	NA	-0.154280566
	Spain	-81.886941	-55.6628	-46.294251	-21.70892456	-79.504323	-42.16429409	-12.375399	-4.05710629
	Sweden	-79.216154	-79.1437	-4.2459366	-2.730988032	-25.830436	-72.84339115	-0.4029857	-1.216679299
	Switzerland	-79.812558	-50.7581	-16.86722	-4.397688068	-70.382455	-30.65581098	-5.8618442	-1.090489924
	United Kingdom	-27.052821	-34.4534	-1.2821637	-1.833181027	-33.998134	-59.18575408	-0.5577417	-1.52031717
	Turkey	60.7375928	50.09425	0.67695697	0.55833055	41.4937292	90.26707893	0.07033244	0.153003936
	Georgia	5392.65563	1802.003	22.4566197	17.87616985	1123.0204	651.6733187	5.28699864	3.067972726
	Kyrgyzstan	6998.62603	174.6448	29.2967188	23.17454332	962.019362	1795.702254	11.1708694	13.40965193
	Lithuania	436.498759	75.81398	9.79859263	5.109968787	576.777875	63.87111551	3.02837842	1.097819589
	Republic of Moldova	871.796197	33.45943	25.2571731	8.92386586	1467.70457	70.74757727	16.897346	5.588781757
	Serbia	-70.513106	11.46153	-7.1856834	0.286442763	-85.069813	-74.35344765	-3.3670369	-0.732947238
	Cyprus	19.5314302	-53.5807	0.41938218	-4.599745752	NA	-15.46863826	NA	-0.713063589
Western	Japan	-19.049726	-35.6009	-0.0810955	-0.267776783	-81.990697		-0.0432447
Pacific							-66.35441907		-0.061612907
	Malaysia	-7.9750421	-7.97504	-2.1266328	-2.126632796	156.711119	156.7111189	2.31193084	2.311930837
	Singapore	-19.736314	32.537	-2.8258231	4.658611157	50.1663088	-49.98741382	0.34365906	-0.342433559
	Australia	-65.482964	-41.0907	-6.4895631	-2.602764945	-93.802752	-89.59929366	-1.040554	-0.731103059
	New Zealand	-70.281577	5.776782	-3.898996	0.177032834	-64.49399	-80.75952302	-0.6024464	-1.392127231
Eastern	Egypt	-63.621749	145.0712	-0.2687721	0.10764541	114.383127		0.06181155
Mediterranean							1277.105491		0.382081925
	Bahrain	-49.946845	117.8704	-11.1005	4.832645427	-74.240858	-74.17940834	-4.6159756	-4.601178666
	Jordan	114.482031	-67.3907	0.81437703	-0.479389012	NA	0	NA	0
	Kuwait	33.3174935	-27.6171	0.12045845	-0.560839028	NA	8.614168518	NA	0.048932466
SouthEast Asia	Thailand	-69.968054	1.878263	-192.07258	1.825288073	-73.989224	-0.984658714	-118.2759	-0.55083728
Africa	Mauritius	436.026377	91.03361	109.680758	63.06474674	587.089308	110.4850474	26.6741821	15.75155015
	South Africa	56.641946	58.10283	168.964157	173.3219921	25.2653484	29.66264822	78.8817738	92.61072784

 

Author Response File: Author Response.docx

Round 3

Reviewer 1 Report

My previous comments have been fully addressed - job well done. I thank the authors for taking the time to add AC alongside PC. After seeing the analysis using uniform start and stop dates, I understand why the authors did not want to include it.

One small typo:

- ln 392. should be "... all databases studied."