With few exceptions, research identifying and developing prognostic markers and risk scores has focused on the level or mean of baseline values. The prodigious success of this approach in prospective cohort studies has resulted in a voluminous literature on prognostic markers [1–3]. The availability of many prognostic markers resulted in development of risk scores that enhance the prognostic accuracy beyond that of any individual marker. Risk scores have grown in popularity since the Framingham Risk Score , and now a variety of cardiovascular risk scores, are available (e.g. SCORE project  and QRISK ) including adaptions to different outcomes (e.g. mortality ). The robust evidence base for numerous prognostic markers and the availability of high-quality risk scores prompt the question, ‘what's next?’. One logical progression is the development and testing of indicated prevention interventions for high-risk individuals. To better inform intervention development, future cohort studies must move beyond the mean to study change and variability in biomarkers. Cohort studies alone will never provide conclusive causal evidence. However, testing whether change predicts morbidity and mortality will provide stronger evidence for causality and more informative estimates of the potential effect size from intervention.
The emphasis from the wider cardiovascular disease literature on prognostic models built from baseline levels of biomarkers is echoed in albuminuria. A meta-analysis of over 100 000 patients demonstrated that albuminuria predicted all-cause and cardiovascular-specific mortality in the general population . A recent individual patient data meta-analysis of over 600 000 patients showed that albuminuria predicts cardiovascular mortality over and above existing common prognostic markers: age, sex, race/ethnicity, smoking, blood pressure, antihypertensive medication, diabetes, and cholesterol . These results suggest the hypothesis that lowering albuminuria may be protective for mortality.
In this issue of Journal of Hypertension, Romundstad et al. report the results of a study testing whether change in albuminuria predicts mortality. In a population-based cohort of adults without diabetes, they assessed change in albuminuria over 11 years, with an extended 9-year follow-up to assess mortality. They found that adjusting for baseline albuminuria, a trajectory of increasing albuminuria predicted higher mortality. However, the magnitude of this effect appears more modest than reported in the meta-analyses that synthesized literature using baseline levels of albuminuria to predict mortality [8,9]. The Chronic Kidney Disease Prognosis Consortium reported that compared with an albumin creatinine ratio (ACR) of 0.6 mg/mmol, having 0.5 or 2.8 mg/mmol higher ACR (corresponding to absolute values of 1.1 or 3.4 mg/mmol ACR) was associated with hazard ratios of 1.20 and 1.63, respectively . These effects were adjusted for several covariates including age and sex. In contrast, adjusted for age and sex, Romundstad et al. found a hazard ratio of 1.04 for a 1.97–2.91 mg/mmol increase in ACR compared with stable (−1.40 to +1.40 mg/mmol change in ACR). Indeed, on an adjusted basis, even the top quartile of change, at least 5.10 mg/mmol (a minimum of 3.70 mg/mmol higher than the highest change in the stable group), was associated with a hazard ratio of just 1.35. By comparison, in the meta-analysis, a smaller difference of 2.8 mg/mmol in absolute ACR had a larger hazard ratio of 1.63 .
Although not focused on mortality, a recent study reported a meta-analysis of the relation between albuminuria reduction in trials of antihypertensive treatment and cardiac events . Across six trials with an average 5-year follow-up, there was a relative risk of 0.49 for reduced/stable versus increasing albuminuria . Intriguingly, Romundstad et al. found that a decrease in albuminuria predicted higher mortality overall. On an adjusted basis, the increased risk associated with decreasing albuminuria was comparable in magnitude with the top quartile of increasing albuminuria. Further analysis revealed that decreasing albuminuria was associated with lower mortality over the first 4 years of follow-up, after which it was associated with increased mortality risk. Although these findings are consistent with clinical trial data on cardiac events for a comparable follow-up period, they highlight the added value of extended follow-up.
The current study has a number of limitations. There was considerable dropout across study waves, although this is common over 20 years of follow-up. The authors appropriately used multiple imputation, which is recommended for minimizing bias from missing data in clinical studies . As happens in studies without random assignment, inference regarding causality is limited. The findings that increasing and decreasing albuminuria, at least short term, predicted higher mortality risk may be explained by some third factor. These concerns are partially mitigated as Romundstad et al. adjusted for a robust set of plausible alternate explanatory factors. Nevertheless, it is never possible to adjust for all potential confounds. Thus it remains to be confirmed whether intervening to stabilize (or decrease) albuminuria in a general population without high risk results in short-term or long-term mortality benefits.
Following the success of developing accurate cardiovascular risk scores, researchers using cohort studies are challenged to leverage the strengths of observational designs and extended follow-ups to optimally inform intervention development. The study by Romundstad et al. in this issue provides an example of the future of cohort studies providing higher quality evidence by testing change in albuminuria over time on mortality from extended follow-up. Further studies could build on these findings by testing whether late emergence of high risk for decreasing albuminuria occurs for outcomes directly linked to albuminuria such as renal events. If not, it would suggest that late emerging risk may be unique to outcomes that integrate across systems, such as mortality or perhaps overall comorbidity.
Existing cohort studies are also ideally suited to explore novel dimensions of change such as variability. Variation within and across days in biomarkers has often been considered measurement error or noise. However, emerging evidence indicates that variability may be a hallmark or even driver of poor health. Higher variability in SBP measurements within visits predicts poorer health outcomes including all-cause mortality in the general population (e.g., ). These findings hold even after adjusting for mean SBP. Similarly, in patients with type 2 diabetes, higher variability in glycosylated hemoglobin (Hb) predicts all-cause mortality . There are plausible drivers of day-to-day variations beyond measurement error. Experimental sleep studies show effects of sleep on BP  and glucose metabolism . Furthermore, a recent systematic review found that sleep variability is associated with poorer health . Romundstad et al. pointed to the use of average ACR across 3 consecutive days at each visit as a strength of their study to reduce measurement error. However, the recent research on the utility of BP, glycosylated Hb, and sleep variability calls the decision to treat variation as noise into question. Urinary albumin excretion rate also has diurnal variation , suggesting it may well be influenced by sleep. Future studies should examine whether daily variation in albuminuria may be of clinical importance rather than assume it is noise.
The future of cardiovascular cohort studies is bright. There are numerous opportunities to advance upon traditional cross-sectional or prediction-only study designs using longitudinal assessments to study change and variability in biomarkers and other risk factors. To date, many cardiovascular cohort studies have focused on identifying who is at high risk. Employing stronger longitudinal designs, cardiovascular cohort studies may also play a role in guiding the development of interventions to reduce the burden of cardiovascular disease in these at-risk individuals.
Conflicts of interest
There are no conflicts of interest.
1. Prospective Studies Collaboration. Age-specific relevance of usual blood pressure to vascular mortality: a meta-analysis of individual data for one million adults in 61 prospective studies. Lancet
2. Khaw K-T, Wareham N, Luben R, Bingham S, Oakes S, Welch A, et al. Glycated haemoglobin, diabetes, and mortality in men in Norfolk cohort of European Prospective Investigation of Cancer and Nutrition (EPIC-Norfolk). BMJ
3. Gruenewald TL, Seeman TE, Ryff CD, Karlamangla AS, Singer BH. Combinations of biomarkers predictive of later life mortality. Proc Natl Acad Sci U S A
4. Wilson PW, D’Agostino RB, Levy D, Belanger AM, Silbershatz H, Kannel WB. Prediction of coronary heart disease using risk factor categories. Circulation
5. Conroy RM, Pyörälä K, Fitzgerald AP, Sans S, Menotti A, De Backer G, et al. Estimation of ten-year risk of fatal cardiovascular disease in Europe: the SCORE project. Eur Heart J
6. Hippisley-Cox J, Coupland C, Vinogradova Y, Robson J, May M, Brindle P. Derivation and validation of QRISK, a new cardiovascular disease risk score for the United Kingdom: prospective open cohort study. BMJ
7. Ankle Brachial Index Collaboration. Ankle brachial index combined with Framingham Risk Score to predict cardiovascular events and mortality: a meta-analysis. JAMA
8. Chronic Kidney Disease Prognosis Consortium. Association of estimated glomerular filtration rate and albuminuria with all-cause and cardiovascular mortality in general population cohorts: a collaborative meta-analysis. Lancet
9. Matsushita K, Coresh J, Sang Y, Chalmers J, Fox C, Guallar E, et al. Estimated glomerular filtration rate and albuminuria for prediction of cardiovascular outcomes: a collaborative meta-analysis of individual participant data. Lancet Diabetes Endocrinol
10. Romundstad S, Hatlen G, Hallan SI. Long-term changes in albuminuria: underlying causes and future mortality risk in a 20-year prospective cohort: the Nord-Trøndelag Health (HUNT) Study. J Hypertens
11. Viazzi F, Muiesan ML, Schillaci G, Salvetti M, Pucci G, Bonino B, et al. Changes in albuminuria and cardiovascular risk under antihypertensive treatment: a systematic review and meta-regression analysis. J Hypertens
; [Epub ahead of print].
12. Little RJ, D’Agostino R, Cohen ML, Dickersin K, Emerson SS, Farrar JT, et al. The prevention and treatment of missing data in clinical trials. N Engl J Med
13. Muntner P, Shimbo D, Tonelli M, Reynolds K, Arnett DK, Oparil S. The relationship between visit-to-visit variability in systolic blood pressure and all-cause mortality in the general population findings from NHANES III, 1988 to 1994. Hypertension
14. Hirakawa Y, Arima H, Zoungas S, Ninomiya T, Cooper M, Hamet P, et al. Impact of visit-to-visit glycemic variability on the risks of macrovascular and microvascular events and all-cause mortality in type 2 diabetes: the ADVANCE trial. Diabetes Care
15. Lusardi P, Zoppi A, Preti P, Pesce RM, Piazza E, Fogari R. Effects of insufficient sleep on blood pressure in hypertensive patients: a 24-h study. Am J Hypertens
16. Spiegel K, Tasali E, Leproult R, Van Cauter E. Effects of poor and short sleep on glucose metabolism and obesity risk. Nat Rev Endocrinol
17. Bei B, Wiley JF, Trinder J, Manber R. Beyond the mean: a systematic review on the correlates of daily intraindividual variability of sleep/wake patterns. Sleep Med Rev
18. Nishimura M, Uzu T, Fujii T, Kimura G. Disturbed circadian rhythm of urinary albumin excretion in nondipper type of essential hypertension. Am J Nephrol