Author’s response to reviews Title: An outline of anemia among adolescent girls in Bangladesh: findings from a cross- sectional study Authors:

Minor Comments. 1. Lines 44-48. Suggest rewriting don't include repeat the covariates for which you report the AOR in the text here. Were any of the co-variates included in models "a priori"? Response: Rephrased the lines as per the suggestion and it appears: “After controlling for relevant covariates in multiple logistic regression model....” 2. Line 47 Define "poverty" in the text as 1.54 is the AOR of the bottom wealth quintile compared to the richest quintile. Response: Rephrased the lines as per the suggestion and it appears as: “......non-pregnancy (AOR: 6.10, 95% CI = 2.70-13.78, p-value <0.001), and households with bottom wealth quintile (AOR: 1.54, 95%CI = 1.03-2.30, p-value = 0.037) were identified as significant risk demographic factors of anemia among adolescent girls of Bangladesh.” Introduction: Minor Comments: 1. Line 65. Suggest rewording to "though anemia has multifaceted etiology, the most common contributing factor is iron deficiency". Otherwise it reads as if you are suggesting that iron deficiency is the distal cause of all anemia, even if other factors are also involved. Response: We changed the lines as suggested. Now we rewrite as: “Though anemia has multifaceted etiology, the most common contributing factor is iron deficiency” 2. Line 72 SCD is most common in Africa. Are any of these inherited disorders common in Bangladesh? Response: Yes, Thalassemia is common in Bangladesh. According to World health Organization (WHO), there are about 3% beta-thalassemia carrier and about 4% Hb E/beta-thalassemia carrier in Bangladesh. [13] Palit S, Bhuiyan RH, Aklima J, Emran TB, Dash R. A study of the prevalence of thalassemia and its correlation with liver function test in different age and sex group in the Chittagong district of Bangladesh. Journal of basic and clinical pharmacy. 2012 Sep;3(4):352. 3. Lines 77-81. Are there any studies that suggest that cognitive defects in iron deficient adolescents or adults can be reversed by iron supplementation/improvement in anemia? Response: Yes, we found a meta-analysis on the role of iron supplementation on cognition. We added a line within this paragraph as follows: “Several studies reported that iron supplementation among anemic adolescents and women had a role in cognition.” Falkingham, M., Abdelhamid, A., Curtis, P., Fairweather-Tait, S., Dye, L., & Hooper, L. (2010). The effects of oral iron supplementation on cognition in older children and adults: a systematic review and meta-analysis. Nutrition journal, 9(1), 4. Methods: Minor Comments: 1. Lines 113-114 The text reads to suggest that all the adolescents in the survey were mothers of children under 5, revise to improve clarity. Response: We have amended the following lines to clarify your queries: “In the base survey, the sample was drawn from 21 equally divided clusters encompassing the entire country, of which 7 clusters were randomly selected for the present study. Thus, we opt to select all the adolescent girls (1467) from the selected households of 7 clusters as participants of the present study. However, we were unable to reach 10% of them due to unavailability and 1,314 adolescent girls were finally selected.” 2. Line 122 what version of ODK was used? What is meant by "7" android? Response: We corrected the line as per your suggestion as follows: “The entire data was collected electronically using ODK (Open Data Kit version 2.0)....” 3. Lines 125-140 Was ODK being used offline during data collection or online. If offline, how often was data uploaded? Where was he ODK server hosted and who maintained it? Can you give any specific examples of het the QC checks and data validation? Response: We address your queries and added a line to the text as follows: “The ODK was used in offline during the data collection and the questionnaires were uploaded every evening. The ODK server was hosted at the Head Office and was maintained by an experienced ICT person. To ensure accuracy of information of such a large dataset, a number of quality control measures were undertaken at different stages of the data collection procedure. For example, a researcher statistician was position at the head office to check the live data and to provide feedback thereby.” 4. Lines 151 finger prick is "capillary" blood not venous. Response: Line is revised as suggested and rewrite as “Capillary blood was collected from all participants through pricking the fingertips by lancet” 5. Line 163 "stadiometer"? Make? Or were these made locally by local artisans? If so How were different ones compared to ensure measurements were similar across research teams? Response: Yes, it was locally prepared and all the machines were checked several times and with similar subjects to validate the machine before using. 6. Lines 174-180 were any variables included a-priori in multivar models? What criteria was used for including variables in multivar model, what strategy was used in model building? Response: No, we did not incorporate any variable as priori in multivar model. We described the model building strategy described by Akaike (1973) where the multiple logistic regression model with smallest Akaike Information Criteria (AIC) value considered as the best model. 7. Lines 180-184 not necessary to show the equation for multivariable logistic regression. Response: We omit the regression equation. Results: Major Comment 1. Tables 2. It is more appropriate to present the percentages as column percentages. For example, present the proportion of girls who were currently married for the anemic and nonanemic groups = 251/678 = 37% vs. 252/636 = 39%. Response: We thought that since our outcome of interest is anemia it is more meaningful to present particular characteristics among anemic adolescent girls. We also excluded the nonanemic cases from the table to make it more clarified. Results: Minor Comments: 1. Why was pregnancy set as the baseline value for comparison in logistic regression? Pregnancy was quite rare and it would be more normal to use non-pregnancy as the baseline and compare the effect of pregnancy compared to non-pregnancy. Response: We revised the reference category and now set non-pregnancy as the baseline value while running the logistic regression. 2. Table 2 the numbers of currently pregnant in the non-anemic group add up to more than the total number in this group. Response: The value was rechecked and corrected accordingly. 3. Table 2 the starts indicating the range of p-values are not required as the actual values are presented. Note do not report p=0.000, report the actual value or <0.001 or <0.0001 as per author guidelines. Response: We rewrite all p-values where p-value less than 0.001 were written as <0.001. 4. Lines 227-231 rewrite the text more concisely. Don't repeat all the results in the text that are in the tables. Response: Lines 227-243 were rewritten: “In simple logistic regression analyses, malnutrition and pregnancy status were significantly associated with anemia among adolescent girls (Table 3). The crude odds ratio (COR) for malnutrition status with anemia is found as 1.5 with p-value 0.052 where the 95% confidence interval was found as (1.0, 2.2). Likewise, the COR for non-pregnancy compared to pregnancy status with anemia was found 0.2 with p-value <0.001 where the 95% confidence interval was (0.1-0.4). However, after adjusting for potential covariates such as age, wealth index, marital status, regional effect and food security in multiple logistic model, malnutrition remained as an independent risk factors for anemia. Indeed, a malnourished adolescent girl had 40% more chance of anemia than a non-malnourished girl. It was also found that a pregnant adolescent girl had 5 times less chance of anemia compare to a pregnant girl. On the other hand, adolescent girls hailing from poorest or second poorest households were 50% more prone to become anemic compared to those from richest households. Besides, other covariates such as marital status, region and household food security status did not show any significant association with anemia status among adolescent girls.” 5. Lines 234-5 interpreting an OR of 1.42 as increasing the risk by 42% is not quite correct. This is only true for risk ratios. The odds are increased by 42%. Response: The interpretation was revised and written as: “a malnourished adolescent girl had 40% more chance of anemia than a non-malnourished girl” 6. Table 3 and results text for multivariable logistic regression results see previous comments about model building and how and what co-variates were included in the final model. No results presented for age, literacy? I would suggest age should be included "a priori" and as a continuous variable in years. Response: Now we considered the age variable for adjustment. Discussion: Minor Comments: 1. Line 252 "the prevalence of all form (sic) of anemia were found (51.6%), most of them (45.97%) were mildly anemic ..." rewrite as: Response: Rewritten as “While, the prevalence of all form of anemia were found 51.6%, most of them (46%) were mildly anemic and few (5.4%) were moderate with only 0.23% severely anemic” 2. "mild anaemia was the most common at 46% of the study population, with only 5.4% moderately anemic and 0.23% severely anemic. Can you see the difference in these sentences? The first suggests that only 46% of the anemic cases were mildly anemic! Response: Rewritten as suggested as: “mild anemia was the most common at 46% of the study population, with only 5.4% moderately anemic and 0.23% severely anemia” 3. Line 258 – in the DHS anemia data quoted – who were the study population? Response: Now the information about the study population has been added in the lines: “Prevalence of anemia reported in our study was similarly high to the most recent Bangladesh Demographic and Health Survey (BDHS) where the prevalence of any form anemia were 48.6%, while mild, moderate and severe anemia were 39.2%, 9.4% and 0.0% respectively among adolescent