In this population-based study, the authors constructed age-specific and sex-specific hemoglobin percentiles from values reported for a defined healthy population in the Comprehensive National Nutrition Survey (CNNS). Age-specific and sex-specific 5th percentiles of hemoglobin derived for this healthy population was used as the study cut-off to define anemia. These were compared with the existing WHO cut-offs to assess significant differences between them at each year of age and sex for quantifying the prevalence of anemia in the entire CNNS sample. From the CNNS survey 41210 participants had a hemoglobin value, 8087 of whom were included in this study and comprised the primary analytical sample. Compared with existing WHO cut-offs, the study cut-offs for hemoglobin were lower at all ages, usually by 1-2 g/dL, but more so in children of both sexes aged 1-2 years and in girls aged 10 years or older. Anemia prevalence with the study cut-offs was 19·2 percentage points lower than with WHO cut-offs in the entire CNNS sample with valid hemoglobin values across all ages and sexes (10·8% with study cut-offs vs 30·0% with WHO cut-offs). The authors concluded that these findings support the re-examination of WHO hemoglobin cut-offs to define anemia.

Relevance: This analysis [1] was designed to identify the range of hemoglobin levels in normal infants, children, and adolescents in India, and thereby derive age and gender appropriate cut-off thresholds to define anemia. The justification was that currently used operational definitions of anemia are based on thresholds set by the World Health Organization over five decades back. The evidence-base for defining these thresholds had several lacunae, meriting a definitive study with robust methodology to address the issue [1]. Further, data from the National Family Health Survey-4 was insufficient to resolve the issue on account of missing information across the age spectrum.

Study methods: The analysis [1] was designed as a cross-sectional examination of hemoglobin level among a representative nation-wide sample of infants, children, and adolescents deemed to be, as near normal as feasible. Although defining an individual as normal or healthy (as opposed to asymptomatic) is extremely difficult, the investigators did this by systematically excluding participants (from the analysis), who could have a clinical condition (determined by measurement of biomarkers) that affected erythropoiesis and/or hemoglobin levels.

The sample population was derived from the Comprehensive National Nutrition Survey (CNNS), designed to analyze the nutritional status, prevalence of specific micronutrient deficiencies, and associated risk factors in a nationally representative cross-section of infants, children, adolescents, and their households [2,3]. CNNS was conducted in a systematic manner across all 30 Indian states [3], using a multi-stage sampling design method, in order to include a represen-tative sample of households and participants from birth through adolescence. Over 112000 such participants were included, and biological specimens, including blood samples, were obtained from more than 49000 participants older than 1 year [3].

For this analysis [1], there were 41210 participants >1year whose hemoglobin level was available from the CNNS. Among them, those with laboratory evidence of any condition that could affect hemoglobin content adversely, were systematically excluded to achieve a filtered group of ‘normal’ participants. The first step of filtering excluded participants across all ages with iron deficiency (measured by serum ferritin and transferrin receptor levels), folate or vitamin B12 deficiency (measured by the respective levels in serum), inflammation (defined by C-reactive protein), vitamin A deficiency, and two abnormal hemoglobin variants viz. HbS and HbA2. Those with hemoglobin value exceeding 5 standard deviations were also excluded, although the reasons were not given. The resultant pool of participants was used for the primary analysis. Among those older than 5 year, additional exclusions were made on the basis of biomarkers suggesting abnormal renal function (serum creatinine), dyslipidemia (serum cholesterol), and abnormal glycemic control (HbA1c). Three additional sub-groups were created by excluding those with i) low or missing serum albumin values (across all ages), ii) low or missing zinc levels (among those >5 year), and iii) evidence of stool parasitic infestation, or unavailable data for this (among those >5 year). The investigators then determined the range of hemoglobin values at age intervals of 1 year, and constructed percentile curves. The 5th percentile value was chosen as the threshold to define anemia. Data from participants in subgroups other than the primary analysis group were used for sensitivity analyses. Lastly, the investigators compared the age-wise prevalence of anemia by their definition against the thresholds prescribed by the World Health Organization.

Critical appraisal: Critical appraisal of the study methodology using criteria from tools designed for the purpose [4-7], is summarized in Box 1. Additional issues are highlighted below.

The authors reported that 8058 of 49486 (16.3%) blood samples obtained could not be analyzed because of insufficient volume or sample spoilage [1]. In a study with robust training of personnel, stringent sampling methods, meticulous storage and handling [3], loss of 1 in 6 samples appears to be disproportionately high.

It is unclear why the investigators chose the 5th percentile as the lower limit of the normal range. In Gaussian distributions, values below the 2.5th percentile mark (or two standard deviations) are generally considered abnormal. The authors mentioned that using this threshold would have resulted in a lower prevalence of anemia among the normal population, and a greater divergence from the WHO prevalence [1]. But this should not be a deterrent, considering the methodological limitations and biases in the studies from which the WHO thresholds were derived.

Although, the investigators did not present the age-wise range of hemoglobin values of the normal children and adolescents in the analysis, this can be indirectly inferred from the smoothed percentile curves in the publication [1]. Hemoglobin values in normal boys range from 9.0-13.5 g/dL at 1year, rising to 13.0-17.0 g/dL by 19 year. In normal girls, it ranges from 9.0-14.0 g/dL at 1y, with negligible increase beyond 5 year, remaining almost static in the range of 10.0-14.0 g/dL throughout childhood and adolescence. It appears that the 50th percentile value in boys rises steadily from about 11 g/dL at 1 year, to approximately 12.0 g/dL at 4 years, 12.5 g/dL at 10 years, and approximately 13.5 g/dL by 19 year of age. In contrast, the 50th percentile in girls, is around 11 g/dL at 1year, rises very slightly to around 12.0 g/dL by 6 year of age, and thereafter hovers around this level all through childhood and adolescence. The 5th percentile values paralleled the 50th percentile values. In boys, it was approximately 9.0 g/dL at 1 year, 10.0 g/dL at 5 year, 11.0 g/dL at 10 year, and just below 12.0 g/dL at 19 year. In girls, it was around 9.5 g/dL at 1 year, rising to just above 10.0 g/dL throughout childhood and adolescence. In both sexes, the values were 1.0 to 2.0 g/dL below the WHO thresholds used to define anemia.

Based on these thresholds, it will be interesting to see the hemoglobin values in children and adolescent with different types of anemia. This data is already available in the CNNS; one publication reported the prevalence of different types of anemia, but did not reveal the actual range of hemoglobin values [8]. It is essential to study the overlap in hemoglobin values between normal children, and those with different types of anemia.

The analysis [1] also has some additional interesting findings, not highlighted by the authors. For example, 13499 of 21586 children had to be excluded from the primary analysis on account of having laboratory markers of known causes of anemia. This translates to 62.5% of children across all age groups, suggesting that the prevalence of anemia (defined by robust laboratory biomarkers) is extremely high. Similarly, among those older than 5 years, 13407 of 19803 (67.7%) participants had to be excluded because of laboratory parameters confirming different types of anemia, or chronic conditions impacting hemoglobin. This suggest that two-thirds of children and adolescents have clinical conditions reducing hemoglobin. Likewise, 1203 of 5657 (21.3%) children >5 year old had lab confirmed zinc deficiency, and 767/4687 (16.4%) had stool infestation with parasites.

In the CNNS, 70% participants from 1-19 year had no anemia based on the WHO thresholds [1,8]. However, more than 60% participants had one or more laboratory markers of anemia (due to various causes), necessitating their exclusion from this analysis [1]. How to reconcile this difference? Three explanations are possible. One is that laboratory markers of anemia in children and adolescents, somehow do not go hand in hand with hemoglobin values, i.e., the markers could be far more sensitive than hemoglobin. The second is that the criteria for excluding >60% participants were ‘any abnormality’ in seven laboratory parameters, whereas the criteria for defining types of anemia used a combination of two parameters to define iron-deficiency anemia [8]. The third possibility could be that participants with abnormal vitamin A levels and abnormal hemoglobin variants were also excluded, whereas these two causes were not counted in the proportion with anemia. However, it seems unlikely that non-inclusion of these two causes could reduce the anemia prevalence by half.

In this analysis [1], the mean z-scores for anthropometric parameters were lower than 0, in fact closer to -1.0 [1]. Although this fits within the broad range of normal anthropometry, it suggests that the participants were thinner/smaller than expected. This raises the question whether the low(er) hemoglobin recorded in them could be a cause.

Interpretations and implications: Although the functional implications of low(er) hemoglobin levels in normal children and adolescents were not explored in this analysis [1], it necessitates re-thinking the clinical as well as public health consequences of anemia. If haemoglobin as low as 9.0 g/dL can be considered normal in infants, could it have any impact on growth, development, physical performance and cognition? Thresholds for initiating prophylactic and therapeutic micro-nutrient supplements in individual children would need revision. Blood transfusion thresholds in acute and chronic conditions may need to be re-examined. There could be public health implications in terms of modifications in focus, resource allocation, etc. One indirect silver lining could be that the thriving micronutrient supplement industry and the associated irrational prescription (and self-administration) of these products, may decline.

Conclusion: This analysis of data [1] from the CNNS [3] suggests that normal Indian infants, children, and adolescents have a wide range of hemoglobin across all ages. The lower limit in normal participants appears to be much lower than expected, calling for a re-look at the thresholds to define (and manage) anemia.

Funding: None; Competing interests: None stated.

Joseph L Mathew

1. Sachdev HS, Porwal A, Acharya R, et al. Hemoglobin thresholds to define anemia in a national sample of healthy children and adolescents aged 1-19 years in India: a population-based study. Lancet Glob Health. 2021;9:e822-31.

2. Population Council. India Comprehensive National Nutrition Survey. Accessed 13 June, 2021. Available from: https://www. popcouncil.org/research/india-comprehensive-national-nutrition-survey　

3. Ministry of Health and Family Welfare (MoHFW), Government of India, UNICEF and Population Council. 2019. Comprehensive National Nutrition Survey (CNNS) National Report. New Delhi. Accessed　13 June, 2021. Available from: https://www.popcouncil.org/uploads/pdfs/2019RH_CNNS report.pdf.　

4. The Joanna Briggs Institute Critical Appraisal tools for use in JBI Systematic Reviews. Checklist for Analytical Cross-Sectional Studies. Accessed 11 June, 2021. Available from: https://jbi.global/sites/default/files/2019-05/JBI_Critical_Appraisal -Checklist_for_Analytical_Cross_Sectional_Studies2017_0. pdf　

5. Moola S, Munn Z, Tufanaru C, et al. Chapter 7: Systematic reviews of etiology and risk . In: Aromataris E, Munn Z (Editors). Joanna Briggs Institute Reviewer’s Manual. The Joanna Briggs Institute, 2017. Accessed 11 June, 2021. Available from　https://reviewersmanual.joannabriggs.org/

6. Center for evidence-based management. Critical Appraisal of a Cross-Sectional Study (Survey). Accessed 11 June, 2021. Available from: https://cebma.org/wp-content/uploads/Critical-Appraisal-Questions-for-a-Cross-Sectional-Study-July-2014-1.pdf

7. Downes M. AXIS critical appraisal of cross-sectional studies. Accessed　11 June, 2021. Available from: http://www.cebm. net/wp-content/uploads/sites/3/2016/06/Evidence-Live-2016-MD.pdf

8. Sarna A, Porwal A, Ramesh S, et al. Characterisation of the types of anemia prevalent among children and adolescents aged 1-19 years in India: a population-based study. Lancet Child Adolesc Health. 2020; 4:515-25.

Anemia has continued to be a public health problem in India since many decades despite various interventions to reduce it at the national level. The prevalence of anemia among children 1-4 year, 5-19 year and among adolescents were reported as 40.5%, 23.4%, and 28.4% respectively. [1] The future scenario regarding anemia prevalence is also bleak in India and the neighboring South Asian Countries as the projected anemia prevalence among women in 2030 are 48%, 25%, and 32%. This means that India would still be falling short of anemia related nutrition targets of Sustainable Development Goals. [2]

The authors have diligently used the Comprehensive National Nutrition Survey data 2019 to answer a nagging public health nutrition question since many decades [3]. This study reports that that the newer cut offs suggested by this study lowers the prevalence of anemia among the entire CNNS sample by around 20 percentage points when compared to that with WHO cut offs. It also lowered the prevalence of anemia by 25·1 percentage points in ages 1-4 years, 15·3 percentage points in ages 5-9 years, 15·6 percentage points in ages 10-14 years, and 22·3 percentage points in ages 15-19 years. Even a slight change in cut-off will lead to a large impact on the absolute numbers for a populous country like ours. The indirect effect of such a change will also be on the possibility of increased resource allocation to various other public health problems. This study coerces the global health organizations as well as the pediatric associations across the world, by providing evidence from a large-scale study to rethink about hemoglobin cut offs, in India and for other countries [4]. Another study has also found similar results for anemia cut off among adult Indian population [5]. We need more studies primarily designed to find out hemoglobin cut offs for anemia in South Asian countries, including India. The public health impact of this study goes much beyond anemia. It also questions the need to re-examine the cut offs and the prevalence of other micro- and macro-nutrient deficiencies.

However, it would be pre-emptive to be celebrate the reduction in anemia using the lower cut offs in the study being discussed. We could be careful that the sustained efforts towards anemia reduction such as Anemia Mukt Bharat Abhiyaan could be slowed down with the present findings. Despite the lower prevalence of anemia, as per the new emerging cut offs from this study, we still have a long way to go to eliminate anemia among children, and adolescents in our country.

　Funding: None; Competing interests: None stated

Amir Maroof Khan

1. Sarna A, Porwal A, Ramesh S, et al. Characterisation of the types of anemia prevalent among children and adolescents aged 1-19 years in India: a population-based study. Lancet Child Adolesc Health. 2020;4:515-25.

2. Sharma D. Achieving sustainable development nutrition targets: the challenge for South Asia. J Glob Health. 2020;10:010303.

3. Ministry of Health and Family Welfare, UNICEF, Population Council. Comprehensive National Nutrition Survey (CNNS) National Report. 2019. Ministry of Health and Family Welfare, UNICEF, Population Council. Comprehensive National Nutrition Survey (CNNS) National Report. 2019. Accessed on July 18, 2021. Available from: http://www.nhm. gov.in/ New_Updates_2018/resources/CNNS_reports.zip

4. Sachdev HS, Porwal A, Acharya R, Ashraf S, Ramesh S, Khan N, et al. Hemoglobin thresholds to define anemia in a national sample of healthy children and adolescents aged 1-19 years in India: a population-based study. Lancet Glob Health. 2021;9:e822–31.

5. Varghese JS, Thomas T, Kurpad AV. Evaluation of hemoglobin cut-off for mild anemia in Asians - analysis of multiple rounds of two national nutrition surveys. Indian J Med Res. 2019;150:385-9.

Anemia continues to be a severe public health problem in India despite multipronged efforts through government programs for anemia control being in place for over five decades [1]. Like many developing countries, the World Health Organization (WHO) hemoglobin (Hb) cut offs are used in India to define anemia. These cut offs were proposed over half a century back in 1968. The cut offs were based on the studies performed predominantly on white populations in Europe and North America, which were reviewed by a group comprising clinical and public health experts [2]. Data from other ethnic groups/races/ countries was not available for review. Recently, the validity of these cut offs is being questioned, as they appear to be higher for population groups from low- and middle-income countries. In fact, these widely used cut offs are under review by WHO itself in view of genetic and racial variations and other emerging evidences [3, 4].

Sachdeva, et al. [4] have recently published age and sex-specific Hb percentiles, which are based on values reported for healthy population in the CNNS 2019. From the whole survey population, authors have excluded children with low serum ferritin, folate, vitamin B12, and retinol levels. Also excluded were children with evidence of inflammation; variant Hb and history of smoking. They considered age and sex-specific 5th percentiles of Hb derived for this stringently defined healthy population as the study cutoff to define anemia. The authors then compared these thresholds with existing WHO cutoffs for children at each year of age and sex for quantifying the prevalence of anemia in the whole CNNS sample [4].

Compared with existing WHO cut-offs, the study cut-offs for Hb were lower at all ages. Anemia prevalence with these cutoffs was 19·2 percentage points lower than with WHO cutoffs in the entire CNNS sample [4]. In a similar study, comparing Hb in adults of different racial/ethnic descent, the Hb cut-off for mild anemia in Asians was lower at 11.22 g/dL. Using the Hb cut-off derived in this study by Verghese, et al. [5] in place of the WHO cut-off of 12 g/dL results in a 17.9% point decrease in the prevalence of anemia in India. 　

If the cut offs proposed by these two studies are accepted, prevalence of anemia in India will be under 40% which is the cut off for defining anemia as a major public health problem! A good feeling indeed. However, it may lead to complacency in government efforts for anemia control as the figure for overall anemia prevalence will still be very high.

Firstly, as pointed out by Verghese et al. [5], this will result in decrease of prevalence of mild anemia. However, the grades of severity of anemia based on Hb will not change [6]. We have to be cognizant of the fact that micronutrients’ deficiencies may cause harm to developing brain and body in young infants even before anemia develops [7].

Most significant impact of this lowered cut off for defining anemia (and also labeling individuals with normal Hb) will be a change in Hb target while treating hematological conditions where therapy is aimed at keeping Hb normal for age. Most patients with transfusion dependent thalassemia (TDT) are managed on hyper transfusion regimen which aims to maintain a baseline hematocrit ‘as nearly normal as practicable’ [8]. With the lowered cut offs for defining normal or low Hb, a lowered cut off will be used for transfusion in TDT. For some conditions such as immune hemolytic anemia, an arbitrary target Hb of 10 gm/dl is advisable when tapering of therapy begins [9]. This may remain unchanged.

In conditions such as aplastic anemia and nutritional anemia also, transfusion practices may remain unchanged. Transfusion guidelines for anemia in children with severe malnutrition recommend transfusions at Hb of 4gm/dl or 4-6 gm/dL if patients have respiratory distress [10,11]. Similarly, most patients with aplastic anemia receive transfusions between Hb 6-8 gm/dL [12]. However, as aim of the therapy while treating deficiency anemia is to bring Hb to normal (and continuing therapy further for replenishing the store), the target Hb will be set lower.

Secondly, the hematology consultations are likely to decrease. Maximum hematology consultations-over a quarter- are for anemia which takes away a significant time of the hematologists [13,14]. However, the decrease in consultations may not be as pronounced as the decline in anemia prevalence may indicate as most consultations for anemia are for moderate or severe anemia [14]. Anemia is the commonest cause of donor deferral in our country. Various studies have shown that anemia accounts for 6.5% of all donor deferrals and over 50% of all deferrals [15-17].　 This is because the Hb cut off for donor suitability is 12.5 gm% for both male and female donors [17]. This cut off is likely to change in light of the lowered Hb cut off for defining anemia. This would result in more donor availability.

At the national level, the intervention required for alleviating anemia may become less intensive, require fewer financial inputs and will be more target-oriented due to lower prevalence.

Funding: None; Competing interests: None stated

Jagdish Chandra

Department of Pediatrics,
PGIMSR and ESIC Model Hospital, New Delhi.
Email: [email protected] 

1. Ministry of Health and Family Welfare, Government of India. International Institute of Population Sciences (IIPS). National Family Health Survey 4. Fact sheets [Internet]. [cited 2018 Jul 6]. Accessed July 1, 2021. Available from: http://rchiips.org/nfhs/factsheet_NFHS-4.shtml

2. Pasricha SR, Colman K, Centeno-Tablante E, et al. Revisiting WHO hemoglobin thresholds to define anemia in clinical medicine and public health. Lancet Haematol. 2018;5: e60–62.

3. Garcia-Casal MN, Pasricha SR, Sharma AJ, et al. Use and interpretation of hemoglobin concentrations for assessing anemia status in individuals and populations: results from a WHO technical meeting. Ann NY Acad Sci. 2019;1450: 5-14.

4. Sachdev HS, Porwal A, Acharya R, et al. Hemoglobin thresholds to define anemia in a national sample of healthy children and adolescents aged 1-19 years in India: a 5.　

5. Varghese JS, Thomas T, Kurpad AV. Evaluation of hemoglobin cut-off for mild anemia in Asians-analysis of multiple rounds of two national nutrition surveys. Indian J Med Res. 2019;150:385-89.

6. Vitamin and Mineral Nutrition Information System. Hemoglobin concentrations for the diagnosis of anemia and assessment of severity. WHO/NMH/NHD/MNM/11.1;World Health Organization, 2011.

7. Black MM. Micronutrient deficiencies and cognitive functioning. J Nutr. 2003;133:3927S-31S.

8. Piomelli S, Karpatkin MH, Arzanian M, et al. Hyper transfusion regimen in patients with Colley’s anemia. Ann NY Acad Sci. 1974;232:186-92.

9. Hill Q A, Stamps R, Massey E, et al. The diagnosis and management of primary autoimmune haemolytic anemia. British J Haematol. 2017;176:395-411.

10. Ashworth A, Khanum S, Jackson A, Schofield C. Guidelines for the Inpatient Treatment of Severely Malnourished Children. World Health Organization, 2003.

11. Bhatnagar S, Lodha R, Choudhury P, et al. IAP Guidelines 2006 for Hospital-Based Management of Severely Malnourished Children (adapted from WHO guidelines). Indian Pediatr.　2007;44:443-61.

12. Williams DA, Bennett C, Bertuch A, et al. Diagnosis and treatment of pediatric acquired aplastic anemia (AAA): an initial survey of the North American Pediatric Aplastic Anemia Consortium (NAPAAC). Pediatr Blood Cancer. 2014;61: 869-74.

13. Pai A, Kotak D, Facher N, et al. Development of a virtual benign hematology consultation service. 　Blood; 2018:132: 824.

14. Bluhm P, Eldem I, Abraham A, et al. Evaluation of pediatric hematology　referrals　at a tertiary university hospital in West Texas. J Pediatric Hematol Oncol. 2021 April 26.

15. Basavarajegowda A. Whole blood donor deferral causes in a tertiary care teaching hospital blood bank from south India. Hematol Transfus Int J. 2017;5:219-22.

16. Kandasamy D, Shastry S, Chenna D, et al. Blood donor deferral analysis in relation to the screening process: A single-center study from southern India with emphasis on high hemoglobin prevalence. J Blood Med. 2020;11:327-334.

17. Sundar P, Sangeetha S K, Seema D, et al. Pre-donation deferral of blood donors in South Indian set-up: An analysis. Asian J Transfus Sci. 2010;4:112-15.

Anemia is defined as hemoglobin level more than two standard deviations below the mean for the age and sex of the child. It results in decrease in the oxygen carrying capacity of the blood leading to several physiological changes. It affects not only the growth and development of the child but may also lead to several soft neurological signs. Although India has made considerable progress in terms of health indicators such as neonatal and infant mortality rate, maternal mortality etc, anemia has remained a significant health problem in children and adolescents [1]. The National Family Health Survey (NFHS-3) revealed that approximately 70% of Indian children had anemia. The scenario was only slightly better in National Family Health Survey (NFHS-4) with almost 60% children suffering from anemia [2]. The cut-offs used for defining anemia in Indian children has been based on the studies by World Health Organization (WHO) which were carried out almost 50 years ago on predominantly white population [3]. Previous analyses of data from National Health and Nutrition Examination Survey (NHANES) have shown that hemoglobin concentrations among healthy Asian, Hispanic and Black population were lower compared to the White population [4]. This variability has also been seen among different racial groups within the same country. These findings indicate that the cut-off for anemia needs evaluation in specific population groups [5]. This has significant implication for India where hemoglobin cut-offs determined in the local population could reduce the estimated prevalence of anemia. The present population-based study, which is representative of the healthy population of children and adults in India, supports the re-examination of WHO cut-offs to define anemia and seems suitable for national use [6].

Funding: None; Competing interests: None stated

Department of Pediatrics,
Institute of Medical Sciences,
Banaras Hindu University, Varanasi, India.
[email protected] 

1. Kapil U, Kapil R, Gupta A. Prevention and control of anemia amongst children and adolescent. Indian J Pediatr. 2019; 86: 523-31.

2. International Institute for Population Sciences. National Family Health Survey (NFHS-4); Fact Sheet. India: HPS: 2017

3. World Health Organization. Nutritional anemias. Report of a WHO group of experts. World Health Organization. Technical Report series. Vol. 405. Geneva. 1968.

4. Patel KV, Longo DI, Ershler WB, et al. Haemoglobin concentrations and the risk of death in older adults: Differences by race/ethnicity in the NHANES III follow up. Br J Haematol. 2009;145: 514-23.

5. Varghese JM, Thomas T, Kurpad AV. Evaluation of haemoglobin cut-off for mild anemia in Asians-analysis of multiple rounds of two national nutrition surveys. Indian J Med Res. 2019; 150: 385-89.

6. Sachdev HS, Porwal A, Acharya R, et al. Haemoglobin thresholds to define anaemia in a national　sample of healthy children and adolescents aged 1-19 years　in India: a popu-lation-based study. Lancet Glob Health. 2021; 9: e822-e31.