Persistent symptoms following SARS-CoV-2 infection amongst children and young people: A meta-analysis of controlled and uncontrolled studies.

BACKGROUND: Data on the long-term impact of SARS-CoV-2 infection in children and young people (CYP) are conflicting. We assessed evidence on long-term post-COVID symptoms in CYP examining prevalence, risk factors, type and duration. METHODS: Systematic search of published and unpublished literature using 13 online databases between 01/12/2019 and 31/07/2021. Eligible studies reported CYP ≤19 years with confirmed or probable SARS-CoV-2 with any symptoms persisting beyond acute illness. Random effects meta-analyses estimated pooled risk difference in symptom prevalence (controlled studies only) and pooled prevalence (uncontrolled studies also included). Meta-regression examined study characteristics hypothesised to be associated with symptom prevalence. Prospectively registered: CRD42021233153. FINDINGS: Twenty two of 3357 unique studies were eligible, including 23,141 CYP. Median duration of follow-up was 125 days (IQR 99-231). Pooled risk difference in post-COVID cases compared to controls (5 studies) were significantly higher for cognitive difficulties (3% (95% CI 1, 4)), headache (5% (1, 8)), loss of smell (8%, (2, 15)), sore throat (2% (1, 2)) and sore eyes (2% (1, 3)) but not abdominal pain, cough, fatigue, myalgia, insomnia, diarrhoea, fever, dizziness or dyspnoea. Pooled prevalence of symptoms in post-COVID participants in 17 studies ranged from 15% (diarrhoea) to 47% (fatigue). Age was associated with higher prevalence of all symptoms except cough. Higher study quality was associated with lower prevalence of all symptoms, except loss of smell and cognitive symptoms. INTERPRETATION: The frequency of the majority of reported persistent symptoms was similar in SARS-CoV-2 positive cases and controls. This systematic review and meta-analysis highlights the critical importance of a control group in studies on CYP post SARS-CoV-2 infection.

Causal mediation analysis with multiple mediators.

In diverse fields of empirical research-including many in the biological sciences-attempts are made to decompose the effect of an exposure on an outcome into its effects via a number of different pathways. For example, we may wish to separate the effect of heavy alcohol consumption on systolic blood pressure (SBP) into effects via body mass index (BMI), via gamma-glutamyl transpeptidase (GGT), and via other pathways. Much progress has been made, mainly due to contributions from the field of causal inference, in understanding the precise nature of statistical estimands that capture such intuitive effects, the assumptions under which they can be identified, and statistical methods for doing so. These contributions have focused almost entirely on settings with a single mediator, or a set of mediators considered en bloc; in many applications, however, researchers attempt a much more ambitious decomposition into numerous path-specific effects through many mediators. In this article, we give counterfactual definitions of such path-specific estimands in settings with multiple mediators, when earlier mediators may affect later ones, showing that there are many ways in which decomposition can be done. We discuss the strong assumptions under which the effects are identified, suggesting a sensitivity analysis approach when a particular subset of the assumptions cannot be justified. These ideas are illustrated using data on alcohol consumption, SBP, BMI, and GGT from the Izhevsk Family Study. We aim to bridge the gap from "single mediator theory" to "multiple mediator practice," highlighting the ambitious nature of this endeavor and giving practical suggestions on how to proceed.

Methods for dealing with time-dependent confounding.

Longitudinal studies, where data are repeatedly collected on subjects over a period, are common in medical research. When estimating the effect of a time-varying treatment or exposure on an outcome of interest measured at a later time, standard methods fail to give consistent estimators in the presence of time-varying confounders if those confounders are themselves affected by the treatment. Robins and colleagues have proposed several alternative methods that, provided certain assumptions hold, avoid the problems associated with standard approaches. They include the g-computation formula, inverse probability weighted estimation of marginal structural models and g-estimation of structural nested models. In this tutorial, we give a description of each of these methods, exploring the links and differences between them and the reasons for choosing one over the others in different settings.

The measurement of household socio-economic position in tuberculosis prevalence surveys: a sensitivity analysis.

OBJECTIVE: To assess the robustness of socio-economic inequalities in tuberculosis (TB) prevalence surveys. DESIGN: Data were drawn from the TB prevalence survey conducted in Lusaka Province, Zambia, in 2005-2006. We compared TB socio-economic inequalities measured through an asset-based index (Index 0) using principal component analysis (PCA) with those observed using three alternative indices: Index 1 and Index 2 accounted respectively for the biases resulting from the inclusion of urban assets and food-related variables in Index 0. Index 3 was built using regression-based analysis instead of PCA to account for the effect of using a different assets weighting strategy. RESULTS: Household socio-economic position (SEP) was significantly associated with prevalent TB, regardless of the index used; however, the magnitude of inequalities did vary across indices. A strong association was found for Index 2, suggesting that the exclusion of food-related variables did not reduce the extent of association between SEP and prevalent TB. The weakest association was found for Index 1, indicating that the exclusion of urban assets did not lead to higher extent of TB inequalities. CONCLUSION: TB socio-economic inequalities seem to be robust to the choice of SEP indicator. The epidemiological meaning of the different extent of TB inequalities is unclear. Further studies are needed to confirm our conclusions.

BMI peak in infancy as a predictor for later BMI in the Uppsala Family Study.

BACKGROUND: The development of obesity through childhood, often characterized by using body mass index (BMI), has received much recent interest because of the rapidly increasing levels of obesity worldwide. However, the extent to which the BMI trajectory in the first year of life (the BMI 'peak' in particular) is associated with BMI in later childhood has received little attention. SUBJECTS: The Uppsala Family Study includes 602 families, comprising mother, father and two consecutive singleton offspring, both of whom were delivered at the Uppsala Academic Hospital, Sweden, between 1987 and 1995. The children's postnatal growth data, including serial measurements of height and weight (from which BMI was calculated), were obtained from health records. All children had a physical examination when they were aged between 5 and 13 years, at which height and weight were again recorded and used to calculate age- and sex-adjusted BMI z-scores. METHODS: Subject-specific growth curves were fitted to the infant BMI data using penalized splines with random coefficients, and from these the location of the BMI peak for each participant was estimated. A multilevel modelling approach was used to assess the relationships between the BMI peak and BMI z-score in later childhood. RESULTS: The BMI peak occurred, on average, slightly later in female children, with a higher BMI peak in male children. Considered separately, both age and BMI at BMI peak were positively associated with later BMI z-score. Considered jointly, both dimensions of BMI peak retained their positive associations. CONCLUSIONS: The growth trajectory associated with higher childhood BMI appears to include a later and/or higher BMI peak in infancy.

Assessing the contribution of fibrinogen in predicting risk of death in men with peripheral arterial disease.

BACKGROUND: Although fibrinogen is known to be an independent population-level risk factor for cardiovascular disease in healthy individuals, less is known about its value for individual-level risk prediction. OBJECTIVES: To assess the independent contribution of plasma fibrinogen to risk prediction in men with peripheral arterial disease. PATIENTS AND METHODS: We used data from the 785 men randomized to placebo in the Lower Extremity Arterial Disease Event Reduction (LEADER) trial. Men were followed at 6-monthly intervals up to 3 years, during which 116 patients died. Multivariable standard and pooled logistic regression were used to model odds of death in the next 3 years or in a 6-month interval. The c-statistic and predictiveness curves were used to assess improvement in predictive ability. RESULTS: Fibrinogen measured at baseline was an independent predictor of all-cause mortality risk (adjusted odds ratio [OR] 1.44, 95% confidence interval [CI] 1.02-1.94, for a 1 g L(-1) increase). Adding baseline fibrinogen to a set of other risk factors did not, however, substantially improve predictive ability. Similarly, fibrinogen measured at the start of a 6-month interval was independently associated with odds of death in the next 6 months (adjusted OR 1.65; 95% CI 0.96-2.73). Again, predictiveness curves with and without fibrinogen did not substantially differ, although the c-statistic increased by 0.011. CONCLUSIONS: Although fibrinogen was independently associated with both 6-month and 3-year mortality risk, individual-level risk prediction was not substantially improved by including fibrinogen in risk models.

Long-term effects of hemostatic variables on fatal coronary heart disease: 30-year results from the first prospective Northwick Park Heart Study (NPHS-I).

BACKGROUND: The long-term associations of established risk factors for coronary heart disease (CHD), for example cholesterol, are well known, but not for the less familiar hemostatic variables. OBJECTIVES: To establish whether associations between hemostatic variables and CHD first identified nearly three decades ago have persisted long-term. METHODS: The first Northwick Park Heart Study (NPHS-I) recruited 2167 white men and 941 white women, average age at entry 48 years, on whom measures of factor (F) VII activity (VIIc) and plasma fibrinogen were carried out, both at entry and at follow-up approximately 6 years later. RESULTS: During a median follow-up of 29 years, 231 male and 36 female CHD deaths were recorded from notifications by the Office for National Statistics. VIIc at recruitment was significantly related to CHD mortality, corrected rate ratio, RR, per 1 SD increase 1.56 (95% CI 1.29, 1.88) in men and RR 1.78 (95% CI 1.17, 2.72) in women. Recruitment fibrinogen was also strongly related to CHD mortality in men, RR 1.63 (95% CI 1.33, 1.99) but not in women, RR 0.75 (95% CI 0.40, 1.43). The associations persisted after controlling for confounders and were confirmed using 6-year follow-up measurements and in analyses omitting deaths within 10 years of recruitment. CONCLUSIONS: The hemostatic system contributes to CHD mortality, and its effect is stable over time. For VIIc, the effect was similar in men and women, while for fibrinogen it appeared to be present only in men.

Is the association of birth weight with premenopausal breast cancer risk mediated through childhood growth?

Several studies have found positive associations between birth weight and breast cancer risk at premenopausal ages. The mechanisms underlying this association are not known, but it is possible that it may be mediated through childhood growth. We examined data from a British cohort of 2176 women born in 1946 and for whom there were prospective measurements of birth weight and of body size throughout life. In all, 59 breast cancer cases occurred during follow-up, 21 of whom were known to be premenopausal. Women who weighed at least 4 kg at birth were five times (relative risk (RR)=5.03; 95% confidence interval=1.13, 22.5) more likely to develop premenopausal breast cancer than those who weighed less than 3 kg (P-value for linear trend=0.03). This corresponded to an RR of 2.31 (0.95, 5.64) per 1 kg increase in birth weight. Birth weight was also a predictor of postnatal growth, that is, women who were heavy at birth remained taller and heavier throughout their childhood and young adulthood. However, the effect of birth weight on premenopausal breast cancer risk was only reduced slightly after simultaneous adjustment for height and body mass index (BMI) at age 2 years and height and BMI velocities throughout childhood and adolescence (adjusted RR=1.94 (0.74, 5.14) per 1 kg increase in birth weight). The pathways through which birth weight is associated with premenopausal breast cancer risk seem to be largely independent of those underlying the relation of postnatal growth to risk.

Childhood growth and breast cancer.

Adult height is known to be positively associated with breast cancer risk. The mechanism underlying this association is complex, since adult height is positively correlated with age at menarche, which in turn is negatively associated with breast cancer risk. The authors used prospective data from a British cohort of 2,547 girls followed from birth in 1946 to the end of 1999 to examine breast cancer risk in relation to childhood growth. As expected, adult height was positively associated with age at menarche and breast cancer. In childhood, cases were taller and leaner, on average, than noncases. Significant predictors of breast cancer risk in models containing all components of growth were height velocity at age 4-7 years (for a one-standard-deviation increase, odds ratio (OR) = 1.54, 95% confidence interval (CI): 1.13, 2.09) and age 11-15 years (OR = 1.29, 95% CI: 0.97, 1.71) and body mass index velocity (weight (kg)/height (m)(2)/year) at age 2-4 years (OR = 0.63, 95% CI: 0.48, 0.83). The effects of these variables were particularly marked in women with early menarche (age <12.5 years). These findings suggest that women who grow faster in childhood and reach an adult height above the average for their menarche category are at particularly increased risk of breast cancer.

Life-course body size and perimenopausal mammographic parenchymal patterns in the MRC 1946 British birth cohort.

Dense mammographic parenchymal patterns are associated with an increased risk of breast cancer. Certain features of body size have been found to be associated with breast cancer risk, but less is known about their relation to breast density. We investigated the association of birth size, childhood growth and life-course changes in body size with Wolfe grade in 1298 perimenopausal women from a British cohort of women born in 1946. The cohort benefits from repeated measures of body size in childhood and adulthood. We obtained mammograms for 90% of women who at age 53 years reported having previously had a mammogram. We found no associations with birth weight or maximum attained height. Body mass index (BMI) at age 53 years and breast size were independently and inversely associated with Wolfe grade (P-value for trend <0.001 for both). Women who reached puberty later were at a greater odds of a higher Wolfe grade than women who had an earlier puberty (odds ratio associated with a 1 year delay in menarche 1.14, 95% CI: 1.01-1.27, adjusted for BMI and breast size at mammography). A higher BMI at any age during childhood or adult life was associated with a reduction in the odds of a higher Wolfe grade, after controlling for breast size and BMI at mammography, for example, standardised odds ratio for height at age 7 was 0.72 (95% CI: 0.64, 0.81). These findings reveal the importance of taking life-course changes in body size, and not just contemporaneous measures, into account when using mammographic density as an intermediate marker for risk of breast cancer.

Survival from breast cancer among South Asian and non-South Asian women resident in South East England.

Ethnic differences in breast cancer survival have been observed in the USA but have not been examined in Britain. We aimed to investigate such differences between South Asian (i.e. those with family roots in the Indian subcontinent) and non-South Asian (essentially British-native) women in England. Primary breast cancer cases incident in 1986 -1993 and resident in South East England were ascertained through the Thames Cancer and Registry and followed up to the end of 1997. Cases of South Asian ethnicity were identified on the basis of their names by using a previously validated computer algorithm. A total of 1037 South Asian and 50 201 non-South Asian breast cancer cases were included in the analysis; 30% of the South Asian (n=312) and 44% (n=22 201) of the non-South Asian cases died during follow-up. South Asian cases had a higher relative survival than non-South Asians throughout the follow-up period. The 10-year relative survival rates were 72.6% (95% confidence interval: 69.0, 75.9%) and 65.2% (64.5, 65.8%) for South Asians and non-South Asians, respectively. The excess mortality rates experienced by South Asians were 82% (72, 94%) of those experienced by non-South Asians (P=0.004). The magnitude of this effect was slightly reduced with adjustment for differences in age at diagnosis, but was strengthened with further adjustment for differences in stage at presentation and socioeconomic deprivation (excess mortality rates in South Asians relative to non-South Asians=72% (63, 82%), P&<0.001). These findings indicate that the higher survival from breast cancer in the first 10 years after diagnosis among South Asian was not due to differences in age at diagnosis, socioeconomic deprivation or disease stage at presentation.

Fetal growth and subsequent risk of breast cancer: results from long term follow up of Swedish cohort.

OBJECTIVE: To investigate whether size at birth and rate of fetal growth influence the risk of breast cancer in adulthood. DESIGN: Cohort identified from detailed birth records, with 97% follow up. SETTING: Uppsala Academic Hospital, Sweden. PARTICIPANTS: 5358 singleton females born during 1915-29, alive and traced to the 1960 census. MAIN OUTCOME MEASURES: Incidence of breast cancer before (at age <50 years) and after (> or = 50 years) the menopause. RESULTS: Size at birth was positively associated with rates of breast cancer in premenopausal women. In women who weighed > or =4000 g at birth rates of breast cancer were 3.5 times (95% confidence interval 1.3 to 9.3) those in women of similar gestational age who weighed <3000 g at birth. Rates in women in the top fifths of the distributions of birth length and head circumference were 3.4 (1.5 to 7.9) and 4.0 (1.6 to 10.0) times those in the lowest fifths (adjusted for gestational age). The effect of birth weight disappeared after adjustment for birth length or head circumference, whereas the effects of birth length and head circumference remained significant after adjustment for birth weight. For a given size at birth, gestational age was inversely associated with risk (P=0.03 for linear trend). Adjustment for markers of adult risk factors did not affect these findings. Birth size was not associated with rates of breast cancer in postmenopausal women. CONCLUSIONS: Size at birth, particularly length and head circumference, is associated with risk of breast cancer in women aged <50 years. Fetal growth rate, as measured by birth size adjusted for gestational age, rather than size at birth may be the aetiologically relevant factor in premenopausal breast cancer.

Urinary androgens and breast cancer risk: results from a long-term prospective study based in Guernsey.

Between 1961 and 1967 a cohort of over 5000 women volunteered for a prospective study to determine the relationship between the urinary androgen metabolites, androsterone (A) and aetiocholanolone (E), and risk of breast cancer. During the first 10 years of the study the concentration of urinary A and E was determined in 1887 of the urine specimens. In 1971 we reported that subnormal amounts of urinary A and E were associated with a significantly increased risk of breast cancer. The cohort has been followed regularly during the 37 years since inception of the study and, by May 1998, 248 women had been diagnosed with breast cancer. Urinary androgen metabolites had been measured in 116 of these cases. Analysis of these data confirmed that women diagnosed in the first decade of the study were more likely to have low levels of urinary androgen metabolites. In the following decades, however, those who developed breast cancer were more likely to have manifested an increased A and E excretion. The reversal in the relationship between androgen metabolite excretion and risk suggests that age, or probably more importantly, menopausal status at diagnosis is an important modifying factor. Dichotomizing at age 50 it was found that in the younger age group (predominantly premenopausal) the rate ratios in the lowest tertile of A or E excretion were two- to threefold greater than for those in the highest tertile (chi2(1) = 3.57; P = 0.06: chi2(1) = 4.70; P = 0.03 for A and E respectively). In contrast, in the older age group comprising predominantly post-menopausal women, the rate ratios associated with the lowest tertile of A or E were half that of those in the highest tertile (chi2(1) = 4.10; P = 0.04; chi2(1) = 8.72; P = 0.003 for A and E respectively). This suggests that there may be different endocrine promotional factors for pre-and post-menopausal breast cancer. Hormonal risk factors may vary during the lifetime of an individual woman and this may have profound consequences for prevention strategies.

Risk factors for testicular cancer: a case-control study in twins.

Early life and anthropometric risk factors for testicular cancer were examined in a case-control study in England and Wales in which affected male twins were compared with their unaffected male co-twins. Questionnaire data was obtained for 60 twin pairs. Significantly raised risk of testicular cancer occurred in twins who had longer arms and legs than their co-twin. There was a significant excess of testicular cancer reported in non-twin brothers, as well as in twin brothers, of cases. Risk was also significantly raised in relation to cryptorchidism. The results on limb length suggest that factors, perhaps nutritional, affecting growth before puberty, may be causes of testicular cancer. The results on risk in brothers add to evidence of a large genetic component in aetiology of the tumour. The risk associated with cryptorchidism in the twins accords with the hypothesis that cryptorchidism is causally associated with testicular cancer because it is a cause of the malignancy, rather than because the same maternal factors experienced in utero cause both conditions.

Risks of breast and testicular cancers in young adult twins in England and Wales: evidence on prenatal and genetic aetiology.

BACKGROUND: Aetiology of breast and testicular cancers may have prenatal factors, possibly exposure of the fetus to high concentrations of maternal oestrogen. Dizygotic twinning probably involves high hormone concentrations, and therefore, dizygotic twins might be at raised risk of these cancers. The aetiologies of breast and testicular cancers have genetic components, for breast cancer, especially at younger ages. Twins of these probands may, therefore, be at high risk. We investigated risk in twins of patients with breast cancer at young ages or with testicular cancer. METHODS: We identified twins with breast cancer incident at ages younger than 45 years and with incident testicular cancer in England and Wales during 1971-89 by cross-matching national cancer-registration and births records. We determined zygosity by questionnaires to the patients. The twins of probands were followed up for cancer incidence and death. We analysed risks of breast and testicular cancer in dizygotic twins compared with monozygotic twins, and in monozygotic and dizygotic twins of probands. FINDINGS: We identified 500 twins with breast cancer and 194 with testicular cancer. We found a non-significantly raised risk of breast cancer in dizygotic compared with monozygotic twins younger than 30 years (odds ratio 2.3 [95% CI 0.9-5.9]) but not older. The overall risk of testicular cancer was significantly higher in dizygotic twins than in monozygotic twins (1.5 [1.1-2.2]) consequent on a risk for seminomas was high (3.2 [1.6-6.5]; p = 0.001). Risk of breast cancer was significantly raised in female twins of probands (standardised incidence ratio 7.7 [4.9-12.2], p < 0.001). The relative risk of breast cancer was 34.7 (9.5-126.5) in monozygotic twins of women in whom breast cancer had occurred before age 35 years. The cumulative risk of breast cancer for these twins by age 40 years was 29% (13-56). The relative risk of testicular cancer was 37.5 (12.3-115.6) in twins of men with testicular cancer. The cumulative risk by age 40 years in monozygotic twins of men with testicular cancer was 14% (4-46). INTERPRETATION: The higher risks of these cancers in dizygotic than in monozygotic twins support a prenatal aetiology, and are compatible with aetiology related to raised maternal concentrations of free, unbound oestrogens. The results for twins of probands have implications for genetic aetiology; appropriate clinical action for monozygotic twins needs consideration.

Multilevel models for longitudinal variables prognostic for survival.

The issue of modelling the joint distribution of survival time and of prognostic variables measured periodically has recently become of interest in the AIDS literature but is of relevance in other applications. The focus of this paper is on clinical trials where follow-up measurements of potentially prognostic variables are often collected but not routinely used. These measurements can be used to study the biological evolution of the disease of interest; in particular the effect of an active treatment can be examined by comparing the time profiles of patients in the active and placebo group. It is proposed to use multilevel regression analysis to model the individual repeated observations as function of time and, possibly, treatment. To address the problem of informative drop-out--which may arise if deaths (or any other censoring events) are related to the unobserved values of the prognostic variables--we analyse sequentially overlapping portions of the follow-up information. An example arising from a randomized clinical trial for the treatment of primary biliary cirrhosis is examined in detail.

Review of survival analyses published in cancer journals.

Survival analysis has found widespread applications in medicine in the last 10-15 years. However, there has been no published review of the use and presentation of survival analyses. We have carried out a systematic review of the research papers published between October and December 1991 in five clinical oncology journals. A total of 132 papers were reviewed. We looked at several aspects of study design, data handling, analysis and presentation of the results. We found that almost half of the papers did not give any summary of length of follow-up; that in 62% of papers at least one end point was not clearly defined; and that both logrank and multivariate analyses were frequently reported at most only as P-values [63/84 (75%) and 22/47 (47%) respectively]. Furthermore, although many studies were small, uncertainty of the estimates was rarely indicated [in 13/84 (15%) logrank and 16/47 (34%) multivariate results]. The procedure for categorisation of continuous variables in logrank analyses was explained in only 8/49 (16%) papers. The quality of graphs was felt to be poor in 43/117 (37%) papers which included at least one survival curve. To address some of the presentational inadequacies found in this review we include new suggested guidelines for the presentation of survival analyses in medical journals. These would complement the statistical guidelines recommended by several clinical oncology journals.

Serum prolactin levels and their relationship to survival in women with operable breast cancer.

The prognostic value of serum prolactin levels was assessed in a sequential series of 739 patients who were initially treated at Guy's Hospital, London, between 1975 and 1980. Prolactin was measured in 472 patients 1 day before (Hpr1) and in 457 patients 10 days after (Hpr2) mastectomy. Follow-up of the patients was up to August 1992 giving 6139 women-years with a median follow-up time of 11.5 years (13.7 for patients still living and 5.1 for those dead). The association between the three prolactin variables and reproductive and clinical factors was examined before assessing the prognostic value of prolactin levels in terms of overall, disease-specific and disease-free survival. Multivariate survival models were used in order to adjust for the effect of other prognostic variables. These were found to be: tumour size, degree of nodal involvement, tumour grade and age at diagnosis. The results showed that high Hpr2 or high postoperative increase in prolactin (i.e. Hpr2-Hpr1) were significantly related to shorter disease-specific survival (p = 0.04 and 0.01, respectively) in postmenopausal women. In addition there was some indication, which did not attain formal significance, for this association to occur for disease-free survival. Thus the rise in blood prolactin levels after surgery may be a weak indicator of poor prognosis of breast cancer in postmenopausal women.

Bone density of normal women in relation to endogenous and exogenous oestrogens.

To evaluate the relationship between duration of hormone replacement therapy (HRT) and forearm bone density, a population-based study has been conducted of 4638 women comparing bone density of 1079 women who had taken HRT for varying times with those who had never taken it. Forearm bone mineral content (BMC) was measured by single photon absorptiometry. Additionally, free oestradiol and sex hormone binding globulin levels were measured in women with high and low bone mineral content. There were no detectable differences in BMC in those who took HRT for up to 24 months compared with those who did not. Thereafter there were significant positive regression coefficients for duration periods of 2 to 5 yr (beta = 0.77, P = 0.04 and beta = 1.76, P = 0.03) and more than 5 yr (beta = 1.66, P < 0.0005 and beta = 1.87, P = 0.009) in post-menopausal and oophorectomized women. The major determinants of BMC were height, and in post-menopausal or oophorectomized women, age. Weight and parity also had additional explanatory power for pre- and post-menopausal women. Those who were in the highest quintile of BMC but had never taken HRT showed significantly elevated percentage free oestradiol compared with those in the lowest quintile. Thus HRT is osteoprotective if taken after the menopause for more than 24 months.

Practical problems in fitting a proportional hazards model to data with updated measurements of the covariates.

We review and discuss the practical problems encountered when analysing the effect on survival of covariates which are measured repeatedly over time. Specific issues arise over and above those met with the standard proportional hazards model and concern all stages of data preparation, data analysis and interpretation of the results. Data from a randomized clinical trial of patients with primary biliary cirrhosis, on whom several measurements were taken at regular intervals after entry, are presented as an illustration.