<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.1 20151215//EN" "https://jats.nlm.nih.gov/publishing/1.1/JATS-journalpublishing1.dtd">
<article xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" article-type="letter" dtd-version="1.1" specific-use="sps-1.9" xml:lang="en">
  <front>
    <journal-meta>
      <journal-id journal-id-type="nlm-ta">Arq Bras Cardiol</journal-id>
      <journal-id journal-id-type="publisher-id">abc</journal-id>
      <journal-title-group>
        <journal-title>Arquivos Brasileiros de Cardiologia</journal-title>
        <abbrev-journal-title abbrev-type="publisher">Arq. Bras. Cardiol.</abbrev-journal-title>
      </journal-title-group>
      <issn pub-type="ppub">0066-782X</issn>
      <issn pub-type="epub">1678-4170</issn>
      <publisher>
        <publisher-name>Sociedade Brasileira de Cardiologia - SBC</publisher-name>
      </publisher>
    </journal-meta>
    <article-meta>
      <article-id pub-id-type="publisher-id" specific-use="scielo-v3">7FkWWZzNXPjD6b5VvYPpyjp</article-id>
      <article-id pub-id-type="publisher-id" specific-use="scielo-v2">S0066-782X2019001201155</article-id>
      <article-id pub-id-type="doi">10.36660/abc.20190601</article-id>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Letter to the Editor</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>Longitudinal Data and Correlated Measures Bias: The Alternative of Mixed Models</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-9679-5287</contrib-id>
          <name>
            <surname>Lopes</surname>
            <given-names>Johnnatas Mikael</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
          <xref ref-type="corresp" rid="c1"/>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Guedes</surname>
            <given-names>Marcello Barbosa O.G.</given-names>
          </name>
          <xref ref-type="aff" rid="aff2">2</xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Cavalcanti</surname>
            <given-names>Rafael Limeira</given-names>
          </name>
          <xref ref-type="aff" rid="aff3">3</xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Souza</surname>
            <given-names>Clecio Gabriel de</given-names>
          </name>
          <xref ref-type="aff" rid="aff2">2</xref>
        </contrib>
      </contrib-group>
      <aff id="aff1">
        <label>1</label>
        <institution content-type="orgname">Universidade Federal do Vale do S&#227;o Francisco</institution>
        <addr-line>
          <city>Paulo Afonso</city>
          <state>BA</state>
        </addr-line>
        <country country="BR">Brazil</country>
        <institution content-type="original">Universidade Federal do Vale do S&#227;o Francisco UNIVASF - Colegiado de Medicina, Paulo Afonso, BA - Brazil</institution>
      </aff>
      <aff id="aff2">
        <label>2</label>
        <institution content-type="orgname">Universidade Federal do Rio Grande do Norte</institution>
        <addr-line>
          <city>Natal</city>
          <state>RN</state>
        </addr-line>
        <country country="BR">Brazil</country>
        <institution content-type="original">Universidade Federal do Rio Grande do Norte UFRN - Fisioterapia, Natal, RN - Brazil</institution>
      </aff>
      <aff id="aff3">
        <label>3</label>
        <institution content-type="orgname">Centro Universit&#225;rio Maur&#237;cio de Nassau</institution>
        <addr-line>
          <city>Natal</city>
          <state>RN</state>
        </addr-line>
        <country country="BR">Brazil</country>
        <institution content-type="original">Centro Universit&#225;rio Maur&#237;cio de Nassau UNINASSAU - Fisioterapia, Natal, RN - Brazil</institution>
      </aff>
      <author-notes>
        <corresp id="c1"><bold>Mailing Address: Johnnatas Mikael Lopes,</bold> UNIVASF - Colegiado de Medicina - Centro de Forma&#231;&#227;o Profissional de Paulo Afonso (CFPPA) - Rua da Aurora, S/N, Quadra 27, Lote 3. Postal Code 48607-190, Bairro General Dutra Paulo Afonso, BA - Brazil E-mail: <email>johnnataslopes2@gmail.com</email>
				</corresp>
      </author-notes>
      <pub-date date-type="pub" publication-format="electronic">
        <day>2</day>
        <month>12</month>
        <year>2019</year>
      </pub-date>
      <pub-date date-type="collection" publication-format="electronic">
        <month>12</month>
        <year>2019</year>
      </pub-date>
      <volume>113</volume>
      <issue>6</issue>
      <fpage>1155</fpage>
      <lpage>1157</lpage>
      <history>
        <date date-type="received">
          <day>02</day>
          <month>09</month>
          <year>2019</year>
        </date>
        <date date-type="rev-recd">
          <day>04</day>
          <month>09</month>
          <year>2019</year>
        </date>
        <date date-type="accepted">
          <day>04</day>
          <month>09</month>
          <year>2019</year>
        </date>
      </history>
      <permissions>
        <license license-type="open-access" xlink:href="https://creativecommons.org/licenses/by/4.0/" xml:lang="en">
          <license-p>This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.</license-p>
        </license>
      </permissions>
      <kwd-group xml:lang="en">
        <title>Keywords</title>
        <kwd>Cohort Studies</kwd>
        <kwd>Longitudinal Studies</kwd>
        <kwd>Cross-Section,Studies</kwd>
        <kwd>Epidemiology</kwd>
        <kwd>Biostatistics</kwd>
      </kwd-group>
    </article-meta>
  </front>
  <body>
    <p>Longitudinal studies have two important data typologies: single outcomes or repeated measures.<xref ref-type="bibr" rid="B1">1</xref> Single outcome, such as death or disease onset, should have a different data treatment than those studies with repeated measures outcome. But, they have in common the detection of changes over time and the contributing factors for this change. Cohort differs from cross-sectional studies that desire only variables relationshiP , without causal effect.</p>
    <p>Fernandes et al.<xref ref-type="bibr" rid="B2">2</xref> wrote an article entitled <italic>The RelationshiP between Lifestyle and Costs Related to Medicine Use in Adults</italic>, published in this journal, volume 112, number 6, 2019, and they used behavioral independent variables to estimate their effects on drug costs outcome, collected as repeated measurements in a prospective cohort design.</p>
    <p>The aim of this exposition is to show that, probably, there was a mistake in the Fernandes et al.<xref ref-type="bibr" rid="B2">2</xref> data analysis, which compromises the causality inferences due to the great possibility of the estimates&#8217; accuracy to be mistaken.</p>
    <p>Let's get to the facts. Considering the prospective cohort design with repeated measures, there is a hierarchical structure in the outcome data due to their clustering in the same participant after various measures. Data cluster leave to the model error, that is the difference between what was predicted by the model and the actual measurement, of the same participant, at different times, to be correlated.<xref ref-type="bibr" rid="B3">3</xref> This is a condition for not using multiple linear regression (MLR) which assumes the independence of the model error given by the assumption that the distribution of each participant is equal. MLR does not extract from the data which is variability within the individual from variability between individuals (population).<xref ref-type="bibr" rid="B3">3</xref>
		</p>
    <p>Using RLM in repeated measures generates regression coefficients with standard errors biased. This requires covariance matrix application that will produce more reliable estimates, in others words, narrower confidence intervals from Mixed Effects Models.<xref ref-type="bibr" rid="B4">4</xref> This is the best alternative to verify changes over time or the conditioners effects on repeated measures outcomes in longitudinal studies, controlling for individual effects.</p>
    <p>There is greater variability between individuals than within individuals, mainly due to biological and social conditioning differences, it&#8217;s observed that drug costs will be more correlated over time in the same individual than among participants. To think that this distribution is the same among the participants ignores theoretical assumption in the social determination on people's behavior.<xref ref-type="bibr" rid="B5">5</xref>
		</p>
    <p>Build distinct MLRs (A, B, C and D), see Fernandes et al.<xref ref-type="bibr" rid="B2">2</xref>, does not control this covariance effect, and therefore may be producing coefficients with confidence intervals biased in independent variables and can not detect the rate of change from basal either.<xref ref-type="bibr" rid="B3">3</xref> In addition, with mixed models it would also be possible to take advantage of measurements that were measured on lost participants, increasing modeling sensitivity.<xref ref-type="bibr" rid="B4">4</xref>
		</p>
    <p>From another perspective, the objective of the research being to estimate the interrelation of drug cost and behavioral habits, without establishing causality, would only require a cross-sectional design of the participants with the collection of outcome data and independent variables at a single moment. Thus, the basal regression model would be sufficient to estimate gross and adjusted associations.<xref ref-type="bibr" rid="B1">1</xref>
		</p>
    <p>Thus, the use of RLM should be restricted to cross-sectional research designs and longitudinal studies with repeated measures outcomes need to differentiate the individual effect of the population effect in the identification of temporal changes and their conditioning. Possibly, the findings of Fernandes el al.<xref ref-type="bibr" rid="B2">2</xref> should be based regarding their conclusions about the inverse relationshiP between alcohol use and drug costs or the statistically non-significant relationships with body fat, gender and smoking status that have great impact on other health situations, especially chronic diseases.</p>
  </body>
  <back>
    <ref-list>
      <title>References</title>
      <ref id="B1">
        <label>1</label>
        <mixed-citation>1. Medronho RA, Bloch KV, Raggio LR, Werneck GL. Epidemiologia. 2&#170; ed. S&#227;o Paulo: Atheneu; 2009. 685 P .</mixed-citation>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Medronho</surname>
              <given-names>RA</given-names>
            </name>
            <name>
              <surname>Bloch</surname>
              <given-names>KV</given-names>
            </name>
            <name>
              <surname>Raggio</surname>
              <given-names>LR</given-names>
            </name>
            <name>
              <surname>Werneck</surname>
              <given-names>GL</given-names>
            </name>
          </person-group>
          <source>Epidemiologia</source>
          <year>2009</year>
          <edition>2&#170; ed</edition>
          <publisher-loc>S&#227;o Paulo</publisher-loc>
          <publisher-name>Atheneu</publisher-name>
        </element-citation>
      </ref>
      <ref id="B2">
        <label>2</label>
        <mixed-citation>2. Fernandes RA, Mantovani AM, Sanches Codogno J, Camilo Turi-Lynch B, Pokhrel S, Anokye N. Rela&#231;&#227;o entre Estilo de Vida e Custos Relacionados ao Uso de Medicamentos em Adultos. Arq Bras Cardiol. 2018;11(2):749-55.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Fernandes</surname>
              <given-names>RA</given-names>
            </name>
            <name>
              <surname>Mantovani</surname>
              <given-names>AM</given-names>
            </name>
            <name>
              <surname>Sanches Codogno</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Camilo Turi-Lynch</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Pokhrel</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Anokye</surname>
              <given-names>N</given-names>
            </name>
          </person-group>
          <article-title>Rela&#231;&#227;o entre Estilo de Vida e Custos Relacionados ao Uso de Medicamentos em Adultos</article-title>
          <source>Arq Bras Cardiol</source>
          <year>2018</year>
          <volume>11</volume>
          <issue>2</issue>
          <fpage>749</fpage>
          <lpage>755</lpage>
        </element-citation>
      </ref>
      <ref id="B3">
        <label>3</label>
        <mixed-citation>3. Fausto MA, Carneiro M, Antunes CMDF, Pinto JA, Colosimo EA. O modelo de regress&#227;o linear misto para dados longitudinais: Uma aplica&#231;&#227;o na an&#225;lise de dados antropom&#233;tricos desbalanceados. Cad Saude Publica. 2008;24(3):513-24.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Fausto</surname>
              <given-names>MA</given-names>
            </name>
            <name>
              <surname>Carneiro</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Antunes</surname>
              <given-names>CMDF</given-names>
            </name>
            <name>
              <surname>Pinto</surname>
              <given-names>JA</given-names>
            </name>
            <name>
              <surname>Colosimo</surname>
              <given-names>EA</given-names>
            </name>
          </person-group>
          <article-title>O modelo de regress&#227;o linear misto para dados longitudinais: Uma aplica&#231;&#227;o na an&#225;lise de dados antropom&#233;tricos desbalanceados</article-title>
          <source>Cad Saude Publica</source>
          <year>2008</year>
          <volume>24</volume>
          <issue>3</issue>
          <fpage>513</fpage>
          <lpage>524</lpage>
        </element-citation>
      </ref>
      <ref id="B4">
        <label>4</label>
        <mixed-citation>4. Helena Constantino Spyrides M, Jos&#233; Struchiner C, Tereza Serrano Barbosa Gilberto Kac M. An&#225;lise de Dados com Medidas Repetidas. In: Kac G, Sichieri R, Gigante D Epidemiologia nutricional. Rio de Janeiro: Editora FIOCRUZ/Atheneu; 2007.P . 245-60.</mixed-citation>
        <element-citation publication-type="book">
          <source>Kac G, Sichieri R, Gigante D Epidemiologia nutricional</source>
          <chapter-title>An&#225;lise de Dados com Medidas Repetidas</chapter-title>
          <year>2007</year>
          <publisher-loc>Rio de Janeiro</publisher-loc>
          <publisher-name>Editora FIOCRUZ/Atheneu</publisher-name>
        </element-citation>
      </ref>
      <ref id="B5">
        <label>5</label>
        <mixed-citation>5. Garbois JA, Sodre F, Dalbello-Araujo M. Da no&#231;&#227;o de determina&#231;&#227;o social &#224; de determinantes sociais da sa&#250;de. Sa&#250;de Debate. 2017;41(112):63-76.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Garbois</surname>
              <given-names>JA</given-names>
            </name>
            <name>
              <surname>Sodre</surname>
              <given-names>F</given-names>
            </name>
            <name>
              <surname>Dalbello-Araujo</surname>
              <given-names>M</given-names>
            </name>
          </person-group>
          <article-title>Da no&#231;&#227;o de determina&#231;&#227;o social &#224; de determinantes sociais da sa&#250;de</article-title>
          <source>Sa&#250;de Debate</source>
          <year>2017</year>
          <volume>41</volume>
          <issue>112</issue>
          <fpage>63</fpage>
          <lpage>76</lpage>
        </element-citation>
      </ref>
    </ref-list>
  </back>
  <sub-article article-type="reply" id="s1" xml:lang="en">
    <front-stub>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Reply</subject>
        </subj-group>
      </article-categories>
      <contrib-group>
        <contrib contrib-type="author">
          <name>
            <surname>Fernandes</surname>
            <given-names>R&#244;mulo Ara&#250;jo</given-names>
          </name>
        </contrib>
      </contrib-group>
    </front-stub>
    <body>
      <p>I appreciate the opportunity to answer the questions concerning our manuscript recently published in Arquivos Brasileiros de Cardiologia.<xref ref-type="bibr" rid="B6">1</xref> Academic discussion is always healthy and welcome.</p>
      <p>Firstly, thank you for your interest in our study. The question raised refers to the use of linear regression in the treatment of data from a prospective cohort with repeated measures, which is believed to have caused mistaken estimates (mixed linear regression is suggested instead). Linear regression is debated as it fails to detect intra-individual variability properly, as it focuses on variability between individuals. From a theoretical point of view, this statement is correct, but it does not reflect the way the data were analyzed in the study.</p>
      <p>The dependent variable of the study was defined as &#8220;drug spending over 12 months.&#8221; In the study, we did not analyze the history of drug spending over the year<xref ref-type="bibr" rid="B7">2</xref> (and how this history would be affected by behavioral variables), nor did we seek to identify the relationshiP between changes compared to baseline (for dependent and independent variables). We did try to analyze the relationshiP of behavioral variables with the final amount spent over the year.</p>
      <p>In fact, this dependent variable is unusual in its construction, as it was longitudinally designed (expenditures on drugs computed over 12 months), but treated cross-sectionally (total amount spent over 12 months). The total amount of drug spending reflects a cross-sectional construct, although its construction considers the 12 months of follow-uP . This particularity of the dependent variable, added to the fact that the behavioral variables were collected at only two moments (baseline and at the end of 12 months), led us to create the four models proposed in the study, which characterize a cross-sectional view of the problem (especially models A [baseline data] and B [at the end of 12 months]). Unfortunately, the monthly assessment of behavioral variables was not an available methodological option.</p>
      <p>In an ideal model, the dependent variable and the independent variables should be collected monthly, allowing to identify the impact of changes on behavioral variables on changes in drug spending history over the year. However, I repeat, this was not the purpose of the study.<xref ref-type="bibr" rid="B6">1</xref> For this type of analysis, specific structural equation modelling (latent growth curve analysis) would be more suitable (even more so than mixed linear regression), as they would make it possible to analyze the direct impact of changes on independent variable (slope)over changes observed on dependent variable (slope).<xref ref-type="bibr" rid="B8">3</xref> The &#8220;impact&#8221; measures generated by the model are easily interpreted, as they can be expressed as correlation coefficients, which additionally provide effect-size measurements.<xref ref-type="bibr" rid="B9">4</xref>
			</p>
      <p>Additionally, the dependent variable as it was presented (cross-sectionally, with spending accruing over follow-uP time) was necessary due to the particularities observed in its structure. Unlike other variables usually measured in different areas of health sciences (height, blood pressure, lipid profile components), which do not have zero value, drug spending occurs irregularly, reflecting the high occurrence of zero values (that is, spending can be reported in the first month of collection, then no spending can be reported over the subsequent months). Against this background, analyses considering the month-to-month variable would be problematic. Likewise, the issue of intra-individual variability needs to be considered with caution in this study because drug spending in the previous month does not recur in the following month, unlike what was observed for variables like height<xref ref-type="bibr" rid="B10">5</xref> which, even without any gain, the amount of the previous month will repeat in the following month.</p>
      <p>Finally, the absence of significant relationships for obesity and smoking is not surprising in this study, as the sample is relatively young, without the presence of chronic diseases and low occurrence of smoking.</p>
    </body>
    <back>
      <ref-list>
        <title>References</title>
        <ref id="B6">
          <label>1</label>
          <mixed-citation>1. Fernandes RA, Mantovani AM, Sanches Codogno J, Camilo Turi-Lynch B, Pokhrel S, Anokye N. Rela&#231;&#227;o entre Estilo de Vida e Custos Relacionados ao Uso de Medicamentos em Adultos. Arq Bras Cardiol. 2018;11(2):749-55.</mixed-citation>
          <element-citation publication-type="book">
            <person-group person-group-type="author">
              <name>
                <surname>Fernandes</surname>
                <given-names>RA</given-names>
              </name>
              <name>
                <surname>Mantovani</surname>
                <given-names>AM</given-names>
              </name>
              <name>
                <surname>Sanches Codogno</surname>
                <given-names>J</given-names>
              </name>
              <name>
                <surname>Camilo Turi-Lynch</surname>
                <given-names>B</given-names>
              </name>
              <name>
                <surname>Pokhrel</surname>
                <given-names>S</given-names>
              </name>
              <name>
                <surname>Anokye</surname>
                <given-names>N</given-names>
              </name>
            </person-group>
            <chapter-title>Rela&#231;&#227;o entre Estilo de Vida e Custos Relacionados ao Uso de Medicamentos em Adultos</chapter-title>
            <source>Arq Bras Cardiol</source>
            <year>2018</year>
            <volume>11</volume>
            <issue>2</issue>
            <fpage>749</fpage>
            <lpage>755</lpage>
          </element-citation>
        </ref>
        <ref id="B7">
          <label>2</label>
          <mixed-citation>2. Lemes &#205;R, Fernandes RA, Turi-Lynch BC, Codogno JS, de Morais LC, Koyama KAK, Monteiro HL. Metabolic Syndrome, Physical Activity, and Medication-Related Expenditures: A Longitudinal Analysis. J Phys Act Health. 2019;16(10):830-5.</mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <name>
                <surname>Lemes</surname>
                <given-names>&#205;R</given-names>
              </name>
              <name>
                <surname>Fernandes</surname>
                <given-names>RA</given-names>
              </name>
              <name>
                <surname>Turi-Lynch</surname>
                <given-names>BC</given-names>
              </name>
              <name>
                <surname>Codogno</surname>
                <given-names>JS</given-names>
              </name>
              <name>
                <surname>de Morais</surname>
                <given-names>LC</given-names>
              </name>
              <name>
                <surname>Koyama</surname>
                <given-names>KAK</given-names>
              </name>
              <name>
                <surname>Monteiro</surname>
                <given-names>HL</given-names>
              </name>
            </person-group>
            <article-title>Metabolic Syndrome, Physical Activity, and Medication-Related Expenditures: A Longitudinal Analysis</article-title>
            <source>J Phys Act Health</source>
            <year>2019</year>
            <volume>16</volume>
            <issue>10</issue>
            <fpage>830</fpage>
            <lpage>835</lpage>
          </element-citation>
        </ref>
        <ref id="B8">
          <label>3</label>
          <mixed-citation>3. Brailean A, Aartsen MJ, Muniz-Terrera G, Prince M, Prina AM, Comijs HC, Huisman M, Beekman A. Longitudinal associations between late-life depression dimensions and cognitive functioning: a cross-domain latente growth curve analysis. Psychol Med. 2017;47(4):690-702.</mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <name>
                <surname>Brailean</surname>
                <given-names>A</given-names>
              </name>
              <name>
                <surname>Aartsen</surname>
                <given-names>MJ</given-names>
              </name>
              <name>
                <surname>Muniz-Terrera</surname>
                <given-names>G</given-names>
              </name>
              <name>
                <surname>Prince</surname>
                <given-names>M</given-names>
              </name>
              <name>
                <surname>Prina</surname>
                <given-names>AM</given-names>
              </name>
              <name>
                <surname>Comijs</surname>
                <given-names>HC</given-names>
              </name>
              <name>
                <surname>Huisman</surname>
                <given-names>M</given-names>
              </name>
              <name>
                <surname>Beekman</surname>
                <given-names>A</given-names>
              </name>
            </person-group>
            <article-title>Longitudinal associations between late-life depression dimensions and cognitive functioning: a cross-domain latente growth curve analysis</article-title>
            <source>Psychol Med</source>
            <year>2017</year>
            <volume>47</volume>
            <issue>4</issue>
            <fpage>690</fpage>
            <lpage>702</lpage>
          </element-citation>
        </ref>
        <ref id="B9">
          <label>4</label>
          <mixed-citation>4. Maher JM, Markey JC, Ebert-May D. The other half of the story: effect size analysis in quantitative research. CBE Life Sci Educ. 2013; 12(3):345-51.</mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <name>
                <surname>Maher</surname>
                <given-names>JM</given-names>
              </name>
              <name>
                <surname>Markey</surname>
                <given-names>JC</given-names>
              </name>
              <name>
                <surname>Ebert-May</surname>
                <given-names>D</given-names>
              </name>
            </person-group>
            <article-title>The other half of the story: effect size analysis in quantitative research</article-title>
            <source>CBE Life Sci Educ</source>
            <year>2013</year>
            <volume>12</volume>
            <issue>3</issue>
            <fpage>345</fpage>
            <lpage>351</lpage>
          </element-citation>
        </ref>
        <ref id="B10">
          <label>5</label>
          <mixed-citation>5. Fausto MA, Carneiro M, Antunes CM, Pinto JA, Colosimo EA. Mixed linear regression model for longitudinal data: application to an unbalanced anthropometric data set. Cad Saude Publica. 2008;24(3):513-24.</mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <name>
                <surname>Fausto</surname>
                <given-names>MA</given-names>
              </name>
              <name>
                <surname>Carneiro</surname>
                <given-names>M</given-names>
              </name>
              <name>
                <surname>Antunes</surname>
                <given-names>CM</given-names>
              </name>
              <name>
                <surname>Pinto</surname>
                <given-names>JA</given-names>
              </name>
              <name>
                <surname>Colosimo</surname>
                <given-names>EA</given-names>
              </name>
            </person-group>
            <article-title>Mixed linear regression model for longitudinal data: application to an unbalanced anthropometric data set</article-title>
            <source>Cad Saude Publica</source>
            <year>2008</year>
            <volume>24</volume>
            <issue>3</issue>
            <fpage>513</fpage>
            <lpage>524</lpage>
          </element-citation>
        </ref>
      </ref-list>
    </back>
  </sub-article>
  <sub-article article-type="translation" id="s2" xml:lang="pt">
    <front-stub>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Carta ao Editor</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>Dados Longitudinais e o Vi&#233;s da Correla&#231;&#227;o de Medidas: A Alternativa dos Modelos Mistos</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-9679-5287 </contrib-id>
          <name>
            <surname>Lopes</surname>
            <given-names>Johnnatas Mikael</given-names>
          </name>
          <xref ref-type="aff" rid="aff4">1</xref>
          <xref ref-type="corresp" rid="c2"/>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Guedes</surname>
            <given-names>Marcello Barbosa O.G.</given-names>
          </name>
          <xref ref-type="aff" rid="aff5">2</xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Cavalcanti</surname>
            <given-names>Rafael Limeira</given-names>
          </name>
          <xref ref-type="aff" rid="aff6">3</xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Souza</surname>
            <given-names>Clecio Gabriel de</given-names>
          </name>
          <xref ref-type="aff" rid="aff5">2</xref>
        </contrib>
      </contrib-group>
      <aff id="aff4">
        <label>1</label>
        <institution content-type="original">Universidade Federal do Vale do S&#227;o Francisco UNIVASF - Colegiado de Medicina, Paulo Afonso, BA - Brasil</institution>
      </aff>
      <aff id="aff5">
        <label>2</label>
        <institution content-type="original">Universidade Federal do Rio Grande do Norte UFRN - Fisioterapia, Natal, RN - Brasil</institution>
      </aff>
      <aff id="aff6">
        <label>3</label>
        <institution content-type="original">Centro Universit&#225;rio Maur&#237;cio de Nassau UNINASSAU - Fisioterapia, Natal, RN - Brasil</institution>
      </aff>
      <author-notes>
        <corresp id="c2"><bold>Correspond&#234;ncia: Johnnatas Mikael Lopes,</bold> UNIVASF - Colegiado de Medicina - Centro de Forma&#231;&#227;o Profissional de Paulo Afonso (CFPPA) - Rua da Aurora, S/N, Quadra 27, Lote 3. CEP 48607-190, Bairro General Dutra Paulo Afonso, BA - Brasil E-mail: <email>johnnataslopes2@gmail.com</email>
				</corresp>
      </author-notes>
      <kwd-group xml:lang="pt">
        <title>Palavras-chave</title>
        <kwd>Estudos de Coortes</kwd>
        <kwd>Estudos Longitudinais</kwd>
        <kwd>Estudos Transversais</kwd>
        <kwd>Epidemiologia</kwd>
        <kwd>Bioestat&#237;stica</kwd>
      </kwd-group>
    </front-stub>
    <body>
      <p>Estudos longitudinais t&#234;m duas tipologias importantes de dados: desfechos &#250;nicos ou medidas repetidas.<xref ref-type="bibr" rid="B1">1</xref> Coortes de desfecho &#250;nico, como a morte ou surgimento de doen&#231;a, devem ter um tratamento de dados diferente daqueles estudos com desfecho de medidas repetidas. Mas todos eles t&#234;m em comum a detec&#231;&#227;o de mudan&#231;as ao longo do tempo e os fatores contribuintes para esta altera&#231;&#227;o. Isto difere dos estudos transversais que buscam apenas rela&#231;&#227;o de vari&#225;veis, sem o necess&#225;rio efeito de causalidade.</p>
      <p>O estudo de Fernandes et al.,<xref ref-type="bibr" rid="B2">2</xref> intitulado de <italic>Rela&#231;&#227;o entre Estilo de Vida e Custos Relacionados ao Uso de Medicamentos em Adultos,</italic> publicado neste peri&#243;dico no volume 112, n&#250;mero 6 em 2019, utilizou vari&#225;veis independentes de comportamento para estimar os efeitos sobre o desfecho de custos com medicamentos, coletado na forma de medidas repetidas em um desenho de coorte prospectiva.</p>
      <p>O que se pretende com esta exposi&#231;&#227;o &#233; evidenciar que, provavelmente, ocorreu um equ&#237;voco na an&#225;lise de dados do trabalho de Fernandes et al.,<xref ref-type="bibr" rid="B2">2</xref> que compromete as infer&#234;ncias de causalidade devido &#224; grande possibilidade de precis&#227;o das estimativas apresentadas estarem equivocadas.</p>
      <p>Vamos aos fatos. Considerando o desenho de coorte prospectiva com medidas repetidas constata-se a exist&#234;ncia de estrutura hier&#225;rquica nos dados do desfecho devido ao aninhamento deles nas diversas medidas do mesmo participante. O aninhamento de dados faz com que o erro, isto &#233;, o que foi predito pelo modelo e a real aferi&#231;&#227;o das medidas, do mesmo participante em momentos distintos sejam correlacionados.<xref ref-type="bibr" rid="B3">3</xref> Isto &#233; uma condi&#231;&#227;o para n&#227;o uso da regress&#227;o linear m&#250;ltipla (RLM) a qual tem como pressuposto a independ&#234;ncia do erro dada pela assun&#231;&#227;o de que a distribui&#231;&#227;o de cada participante &#233; igual. A RLM n&#227;o extraidos dados aquilo que &#233; variabilidade dentro do indiv&#237;duo da variabilidade entre indiv&#237;duos (popula&#231;&#227;o).<xref ref-type="bibr" rid="B3">3</xref>
			</p>
      <p>Usar RLM em medidas repetidas gera erros-padr&#227;o dos coeficientes de regress&#227;o viciados. Exige-se nesta situa&#231;&#227;o a aplica&#231;&#227;o de uma matriz de covari&#226;ncia que produzir&#225; estimativas mais confi&#225;veis, ou seja, intervalos de confian&#231;a mais estreitos a partir de modelos de efeitos mistos.<xref ref-type="bibr" rid="B4">4</xref> Esta &#233; a melhor alternativa para se verificar mudan&#231;as ao longo do tempo ou o efeito de condicionantes no desfecho em estudos longitudinais, controlando os efeitos individuais.</p>
      <p>Como existe uma variabilidade maior entre indiv&#237;duos que intra-indiv&#237;duo, devido principalmente &#224;s diferen&#231;as biol&#243;gicas e de condicionantes sociais, observa-se que os custos com medicamentos ir&#227;o ser mais correlacionados ao longo do tempo no mesmo indiv&#237;duo que entre os participantes. Pensar que essa distribui&#231;&#227;o &#233; igual nos participantes do estudo, ignora o pressuposto te&#243;rico da determina&#231;&#227;o social em sa&#250;de no comportamento das pessoas.<xref ref-type="bibr" rid="B5">5</xref>
			</p>
      <p>A cria&#231;&#227;o de RLM distintas (A, B, C e D), ver Fernandes et al.,<xref ref-type="bibr" rid="B2">2</xref> n&#227;o controla esse efeito de covari&#226;ncia e, portanto, podem estar produzindo coeficientes com intervalos de confian&#231;a enviesados para as vari&#225;veis independentes e n&#227;o consegue tamb&#233;m detectar a taxa de mudan&#231;a em rela&#231;&#227;o ao basal.<xref ref-type="bibr" rid="B3">3</xref> Al&#233;m disso, com os modelos mistos seria poss&#237;vel tamb&#233;m aproveitar medidas que foram aferidas em participantes perdidos, aumentando a sensibilidade da modelagem.<xref ref-type="bibr" rid="B4">4</xref>
			</p>
      <p>Em outra perspectiva, sendo o objetivo da pesquisa estimar a inter-rela&#231;&#227;o do custo com medicamentos e os h&#225;bitos comportamentais, sem o estabelecimento de causalidade, seria necess&#225;rio apenas um desenho transversal dos participantes com a coleta de dados do desfecho e vari&#225;veis independentes em um &#250;nico momento. Assim, o modelo de regress&#227;o basal seria suficiente para estimar associa&#231;&#245;es brutas e ajustadas.<xref ref-type="bibr" rid="B1">1</xref>
			</p>
      <p>Dessa forma, o uso de RLM deve ficar restrito aos desenhos transversais de pesquisa e que estudos longitudinais necessitam diferenciar o efeito individual do efeito populacional na identifica&#231;&#227;o das mudan&#231;as temporais e os condicionantes da mesma. Possivelmente, os achados de Fernandes et al.,<xref ref-type="bibr" rid="B2">2</xref> devem estar enviesados quanto &#224;s conclus&#245;es sobre a rela&#231;&#227;o inversa do uso de &#225;lcool com os custos dos medicamentos ou as rela&#231;&#245;es n&#227;o-significativas do ponto de vista estat&#237;stico com a gordura corporal, g&#234;nero e tabagismo que t&#234;m grande impacto em outras situa&#231;&#245;es de sa&#250;de, principalmente cr&#244;nicas.</p>
    </body>
  </sub-article>
  <sub-article article-type="reply" id="s3" xml:lang="pt">
    <front-stub>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Carta-resposta</subject>
        </subj-group>
      </article-categories>
      <contrib-group>
        <contrib contrib-type="author">
          <name>
            <surname>Fernandes</surname>
            <given-names>R&#244;mulo Ara&#250;jo</given-names>
          </name>
        </contrib>
      </contrib-group>
    </front-stub>
    <body>
      <p>Agrade&#231;o a oportunidade de resposta aos questionamentos relacionados ao nosso manuscrito recentemente publicado no Arquivos Brasileiros de Cardiologia.<xref ref-type="bibr" rid="B6">1</xref> A discuss&#227;o acad&#234;mica sempre &#233; saud&#225;vel e bem-vinda.</p>
      <p>Inicialmente, agrade&#231;o o interesse em nosso trabalho. O questionamento levantado refere-se ao uso da regress&#227;o linear no tratamento de dados provenientes de uma coorte prospectiva com medidas repetidas, fato que teria levado a gera&#231;&#227;o de estimativas equivocadas (faz-se a sugest&#227;o do uso da regress&#227;o linear mista). A cr&#237;tica sustenta-se na fragilidade da regress&#227;o linear em detectar a variabilidade intra indiv&#237;duo, ao passo que o modelo centra na variabilidade entre os indiv&#237;duos. Do ponto de vista te&#243;rico, o coment&#225;rio est&#225; correto, por&#233;m n&#227;o reflete a forma como os dados foram analisados neste estudo.</p>
      <p>A vari&#225;vel dependente deste estudo foi definida como &#8220;custos com medicamentos acumulados ao longo de 12 meses&#8221;. Neste estudo, n&#227;o se tentou analisar a trajet&#243;ria dos custos com medicamento ao longo do ano<xref ref-type="bibr" rid="B7">2</xref> (e como estas trajet&#243;rias seriam afetadas pelas vari&#225;veis comportamentais) e, tamb&#233;m, n&#227;o se buscou identificar a rela&#231;&#227;o entre mudan&#231;as frente &#224; linha de base (para vari&#225;veis dependentes e independentes). Houve sim, a tentativa de analisar a rela&#231;&#227;o das vari&#225;veis comportamentais com o montante final acumulado ao longo do ano.</p>
      <p>De fato, esta vari&#225;vel dependente &#233; inusitada em sua forma de constru&#231;&#227;o, pois foi elaborada longitudinalmente (gastos com medicamentos computados ao longo de 12 meses), mas tratada transversalmente (total acumulado ao longo dos 12 meses). O montante total de gastos com medicamento reflete um construto transversal, embora sua constru&#231;&#227;o considere os 12 meses de seguimento. Esta particularidade da vari&#225;vel dependente, somada ao fato das vari&#225;veis comportamentais terem sido coletadas em apenas dois momentos do estudo (linha de base e ao final de 12 meses), nos levou a criar os quatro modelos propostos no artigo, os quais caracterizam uma leitura transversal do problema (principalmente os modelos A [dados de linha de base] e B [ao final de 12 meses]). Infelizmente, a avalia&#231;&#227;o mensal das vari&#225;veis comportamentais n&#227;o foi uma op&#231;&#227;o metodol&#243;gica dispon&#237;vel.</p>
      <p>Em um modelo ideal, a vari&#225;vel dependente e as vari&#225;veis independentes deveriam ser coletadas mensalmente, possibilitando identificar o impacto de mudan&#231;as nas vari&#225;veis comportamentais sobre mudan&#231;as na trajet&#243;ria dos custos com medicamento ao longo do ano. Por&#233;m, repito, este n&#227;o foi o objetivo do referido artigo.<xref ref-type="bibr" rid="B6">1</xref> Para este tipo de an&#225;lise, modelos de equa&#231;&#227;o estrutural com modelagem espec&#237;fica (<italic>Latent Growth Curve Analysis</italic>) seriam mais interessantes (at&#233; mais que a regress&#227;o linear mista), pois possibilitariam analisar o impacto direto da taxa de mudan&#231;a ao longo do seguimento na vari&#225;vel independente sobre a taxa de mudan&#231;a da vari&#225;vel dependente.<xref ref-type="bibr" rid="B8">3</xref> As medidas de &#8220;impacto&#8221; geradas pelo modelo s&#227;o facilmente interpretadas, pois podem ser expressas em medidas de correla&#231;&#227;o, as quais adicionalmente fornecem medidas de tamanho de efeito.<xref ref-type="bibr" rid="B9">4</xref>
			</p>
      <p>Adicionalmente, a apresenta&#231;&#227;o da vari&#225;vel dependente da forma como foi apresentada (transversalmente, os gastos acumulados ao longo do seguimento) se fez necess&#225;ria devido a particularidades observadas na estrutura da mesma. Diferentemente de outras vari&#225;veis usualmente mensuradas nas diferentes &#225;reas das ci&#234;ncias da sa&#250;de (estatura, press&#227;o arterial, componentes do perfil lip&#237;dico), as quais n&#227;o apresentam o valor zero, os gastos com medicamentos ocorrem de maneira irregular, refletindo-se na elevada ocorr&#234;ncia de valores zero, ou seja, &#233; poss&#237;vel relatar gastos no primeiro m&#234;s de coleta e, ao longo dos demais meses, relatar nenhum gasto. Nesse pano de fundo, an&#225;lises considerando a vari&#225;vel m&#234;s a m&#234;s seriam problem&#225;ticas. Da mesma forma, a quest&#227;o da variabilidade intra indiv&#237;duo precisa ser considerada com cautela neste estudo devido ao fato do gasto com medicamento do m&#234;s anterior n&#227;o se repetir no m&#234;s seguinte, diferentemente do observado para vari&#225;veis como a estatura<xref ref-type="bibr" rid="B10">5</xref> que, mesmo sem ganho, o valor do m&#234;s anterior se repetir&#225; no m&#234;s seguinte.</p>
      <p>Por fim, a aus&#234;ncia de rela&#231;&#245;es significativas para obesidade e fumo n&#227;o chega a ser uma surpresa neste estudo, devido ao fato da amostra ser relativamente jovem, sem a presen&#231;a de doen&#231;as cr&#244;nicas e com baixa ocorr&#234;ncia de tabagismo.</p>
    </body>
  </sub-article>
</article>
